steam-frame/LisherSong--ps5-web-file-manager
1
0 Fork 0
mirror of https://github.com/LisherSong/ps5-web-file-manager.git synced 2026-10-06 07:00:28 +02:00
Code Issues Packages Projects Releases 2 Wiki Activity

feat(7z): vendor LZMA SDK and add the 7z fixture/verification harness

Browse Source 
Groundwork for 7z support (the second of the six format x volume
combinations). No engine code yet: this lands the vendor subset, real
fixtures and the end-to-end driver, so the engine has something to be
built and checked against.

Vendor (third_party/7z, LZMA SDK 26.03, public domain):

- decoder-only subset: 7zArcIn/7zDec container, Lzma/Lzma2/Ppmd7/Bcj2/
  Bra/Delta codecs, Aes/Sha256 for the password channel to come
- the encoder half (LzmaEnc, Xz*, Sort, Threads, ...) is deliberately
  not vendored; the engine never encodes
- README.md records two SDK limitations that shape the engine design:
  * CSzFolder caps a folder at 4 coders / 3 bonds, so 7-Zip's own BCJ2
    chain (BCJ2 + 4xLZMA2 = 5 coders) is refused by SzAr_DecodeFolder
    even though SzArEx_Open parses it happily - the engine therefore
    parses folder blobs and drives the codec chain itself
  * the C decoder ships no 7zAES coder at all, so both -p and -mhe=on
    archives are rejected until the engine implements it on top of the
    vendored Aes.c / Sha256.c

Tests:

- make_sevenz_fixtures.py builds real archives with an actual 7-Zip
  binary (7za from the Extra package; falls back to the reduced 7zr,
  which has no PPMd encoder): store/lzma/lzma2/ppmd/bcj/delta/bcj2,
  AES with plain and with encrypted headers, a 7-part -v100k volume
  set, and UTF-8 entry names
- sevenz_e2e.c extracts an archive and validates every entry against
  the stored CRC; it calls the wide-character file APIs on Windows, so
  non-ASCII entry names are really created instead of silently failing
- run-sevenz-tests.sh builds the subset and the driver, then runs the
  fixture matrix with an explicit KNOWN_GAPS list that fails loudly
  when one of the gaps starts passing

Status: store, lzma, lzma2, ppmd, bcj, delta and utf8 all extract
byte-identical to the source tree (7/10). bcj2, aes, aeshe and
vol.7z.001 are the known gaps, each mapped to a specific piece of
engine work still to come.
This commit is contained in:
Songlx516 committed 2026-09-12 17:17:45 +08:00
1 parent da565ccb7c
commit 8bee84bd49
60 files changed
+19470 -2
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

.gitignore
+8
View File
@@ -17,6 +17,14 @@ web-file-mgr-linux
.build/*.c
.build/test-out.txt
# 7z work area: downloaded SDK + 7-Zip binaries and local build output.
.build/7zdl/
.build/7zh/
.build/sevenz-test/
# Generated 7z fixtures (tests/make_sevenz_fixtures.py rebuilds them).
tests/fixtures-7z/
# Editor / OS noise
.vscode/
.idea/
THIRD_PARTY_NOTICES
+18 -2
View File
@@ -2,7 +2,7 @@ Third-Party Notices
===================
This project vendors a minimal set of third-party source files under
`third_party/` to provide the ZIP and RAR extraction features (the
`third_party/` to provide the ZIP, RAR and 7z extraction features (the
`/api/extract` endpoint). Their full license texts are included
alongside the sources.
@@ -32,6 +32,21 @@ alongside the sources.
relaxed warnings; unrar_c_api.h is project-authored and
carries the project's license)
4. LZMA SDK (7z decoder)
----------------------
Version : 26.03 (2026-09-03)
Source : https://www.7-zip.org/sdk.html — release `lzma2603.7z` from
https://github.com/ip7z/7zip/releases
License : Public domain ("LZMA SDK is written and placed in the public
domain by Igor Pavlov", see third_party/7z/DOC/lzma-sdk.txt)
Files : third_party/7z/** (decoder-only subset, compiled with relaxed
warnings; see third_party/7z/README.md for the file list)
The LZMA SDK is the engine behind `src/sevenz_extract.c` (the v1.9.x 7z
support: LZMA/LZMA2/PPMd/Copy plus the BCJ, BCJ2 and Delta filters). Only the
C implementation is used — it builds with the plain PS5 C toolchain and does
not pull in the C++ runtime. The encoder half of the SDK is not vendored.
unrar is the engine behind `src/rar_extract.c` (the v1.9 RAR support: RAR4,
RAR5 including WinRAR 6/7 "v6" compression, and multi-volume archives; the
engine can also decrypt via RARSetPassword once a password channel is wired
@@ -46,5 +61,6 @@ replaced by the rarlab UnRAR source in v1.9; see git history under
Libraries in sections 1 and 2 are distributed under the zlib license, which
permits redistribution in source and binary form provided the copyright
notice and this list of conditions are retained. unrar is distributed under
its own freeware terms. See the individual LICENSE / license.txt files in
its own freeware terms. The LZMA SDK (section 4) is in the public domain and
carries no conditions. See the individual LICENSE / license.txt files in
each `third_party/` subdirectory for the complete terms.
tests/make_sevenz_fixtures.py
+253
View File
@@ -0,0 +1,253 @@
#!/usr/bin/env python3
"""Generate real .7z fixtures for the host test suite.
The fixtures are produced by an actual 7-Zip binary so that the archives the
engine has to read are genuine, not hand-rolled. The tool is looked up in this
order:
1. ``--tool <path>``
2. ``$SEVENZ_TOOL``
3. the full ``7za.exe`` from the "7-Zip Extra" package under ``.build/7zdl/``
4. the reduced ``7zr.exe`` under ``.build/7zdl/``
5. ``7z`` / ``7zr`` / ``7za`` on PATH
Both binaries come from https://github.com/ip7z/7zip/releases (7-Zip is public
domain). The reduced ``7zr.exe`` has **no PPMd encoder**, so that fixture is
skipped (with a warning) unless a full build is available.
Output goes to ``tests/fixtures-7z/``. Only files this script owns are removed
on a re-run (identified by an ``OWN_`` prefix list), never the whole directory.
python tests/make_sevenz_fixtures.py [--big]
"""
from __future__ import annotations
import argparse
import os
import random
import shutil
import subprocess
import sys
HERE = os.path.dirname(os.path.abspath(__file__))
ROOT = os.path.dirname(HERE)
OUT = os.path.join(HERE, "fixtures-7z")
SRC = os.path.join(OUT, "_src")
# Everything this script may delete on a re-run.
OWN_PREFIXES = (
"store", "lzma", "lzma2", "ppmd", "bcj", "delta", "aes", "vol", "utf8",
"big", "empty", "_src",
)
PASSWORD = "Secret123"
# --------------------------------------------------------------------------
# tool discovery / process helpers
# --------------------------------------------------------------------------
def find_tool(explicit: str | None) -> str:
if explicit:
if os.path.exists(explicit):
return explicit
raise SystemExit(f"7z tool not found: {explicit}")
env = os.environ.get("SEVENZ_TOOL")
if env and os.path.exists(env):
return env
# Prefer the full 7za.exe (has PPMd); fall back to the reduced 7zr.exe.
for rel in ("7zdl/extra/x64/7za.exe", "7zdl/extra/7za.exe", "7zdl/7zr.exe"):
local = os.path.join(ROOT, ".build", rel.replace("/", os.sep))
if os.path.exists(local):
return local
for name in ("7z", "7zr", "7za"):
found = shutil.which(name)
if found:
return found
raise SystemExit(
"no 7z tool found; pass --tool, set SEVENZ_TOOL, or drop 7zr.exe into "
".build/7zdl/"
)
def run(tool: str, args: list[str], optional: bool = False) -> bool:
"""Run the 7z tool. Returns True on success; raises unless ``optional``."""
proc = subprocess.run(
[tool] + args, cwd=ROOT, stdout=subprocess.PIPE, stderr=subprocess.STDOUT
)
if proc.returncode == 0:
return True
out = proc.stdout.decode("utf-8", "replace")
# 7-Zip writes its diagnostics in the OEM code page; try the common ones.
for enc in ("gbk", "cp936", "utf-8"):
try:
out = proc.stdout.decode(enc)
break
except UnicodeDecodeError:
continue
if optional:
print(f" !! skipped: {' '.join(args[:4])}\n {out.strip()[-200:]}")
return False
raise SystemExit(
f"7z failed ({proc.returncode}): {tool} {' '.join(args)}\n{out}"
)
def win(path: str) -> str:
"""7zr.exe is a native Windows binary: hand it a Windows path."""
return path.replace("/", "\\")
# --------------------------------------------------------------------------
# source tree
# --------------------------------------------------------------------------
def x86ish(size: int) -> bytes:
"""Bytes that look like x86 code so BCJ/BCJ2 filters are actually useful."""
out = bytearray()
rnd = random.Random(1234)
while len(out) < size:
op = rnd.choice((0xE8, 0xE9, 0xE8, 0xE9, 0x0F, 0x8B, 0xC3, 0x90))
out.append(op)
if op in (0xE8, 0xE9):
out += rnd.randrange(0, 1 << 24).to_bytes(4, "little")
elif op == 0x0F:
out.append(0x8B)
else:
out += bytes(rnd.randrange(256) for _ in range(rnd.randrange(1, 6)))
return bytes(out[:size])
def build_source(sizes: dict[str, int]) -> None:
os.makedirs(os.path.join(SRC, "sub"), exist_ok=True)
rnd = random.Random(7)
with open(os.path.join(SRC, "readme.txt"), "wb") as fh:
fh.write(b"sevenz fixture\n" * 20)
with open(os.path.join(SRC, "binary.bin"), "wb") as fh:
fh.write(bytes(rnd.randrange(256) for _ in range(sizes["binary"])))
# Highly compressible: exercises the LZMA/LZMA2/PPMd long-range paths.
with open(os.path.join(SRC, "zeros.bin"), "wb") as fh:
fh.write(b"\0" * sizes["zeros"])
with open(os.path.join(SRC, "sub", "nested.txt"), "wb") as fh:
fh.write(b"nested entry\n" * 100)
with open(os.path.join(SRC, "sub", "code.bin"), "wb") as fh:
fh.write(x86ish(sizes["code"]))
# Non-ASCII names must survive the UTF-16 name table round trip.
with open(os.path.join(SRC, "sub", "\u4e2d\u6587-\u30c6\u30b9\u30c8.txt"), "wb") as fh:
fh.write("unicode name\n".encode("utf-8") * 30)
def write_big_file(path: str, size: int) -> None:
"""A large, semi-compressible file: forces multi-chunk streaming decode."""
rnd = random.Random(99)
block = bytes(rnd.randrange(256) for _ in range(64 * 1024))
with open(path, "wb") as fh:
written = 0
while written < size:
n = min(len(block), size - written)
fh.write(block[:n])
written += n
# --------------------------------------------------------------------------
# archives
# --------------------------------------------------------------------------
# name -> (extra 7z arguments, takes password, required)
VARIANTS: list[tuple[str, list[str], bool, bool]] = [
("store", ["-mx0"], False, True),
("lzma2", ["-m0=lzma2", "-mx5"], False, True),
("lzma", ["-m0=lzma"], False, True),
# The reduced 7zr.exe has no PPMd encoder, so this one is best-effort.
("ppmd", ["-m0=ppmd"], False, False),
("bcj", ["-m0=bcj", "-m1=lzma2"], False, True),
("delta", ["-m0=delta:4", "-m1=lzma2"], False, True),
("bcj2", ["-m0=bcj2", "-m1=lzma2", "-m2=lzma2", "-m3=lzma2", "-m4=lzma2"], False, True),
("aes", ["-m0=lzma2", "-mx5"], True, True),
("utf8", ["-m0=lzma2", "-mx5"], False, True),
]
def build_archives(tool: str, big: bool) -> None:
for name, extra, secret, required in VARIANTS:
arch = os.path.join(OUT, f"{name}.7z")
if os.path.exists(arch):
os.remove(arch)
args = ["a", "-t7z", win(arch), win(SRC), "-y"] + extra
if secret:
args.append(f"-p{PASSWORD}")
if name == "aes":
# Header stays readable: only the streams are encrypted.
args.append("-mhe=off")
run(tool, args, optional=not required)
# Encrypted header (-mhe=on): the archive cannot even be listed without
# the password, so the engine must ask for it up front.
run(tool, ["a", "-t7z", win(os.path.join(OUT, "aeshe.7z")), win(SRC),
"-y", "-m0=lzma2", "-mx5", f"-p{PASSWORD}", "-mhe=on"])
# Multi-volume (-v): 100 KiB parts force the fixture across several files.
vols = os.path.join(OUT, "vol.7z")
for stale in sorted(os.listdir(OUT)):
if stale.startswith("vol.7z"):
os.remove(os.path.join(OUT, stale))
run(tool, ["a", "-t7z", win(vols), win(SRC), "-y", "-m0=lzma2", "-mx5",
"-v100k"])
if big:
bigsrc = os.path.join(OUT, "_big")
os.makedirs(bigsrc, exist_ok=True)
write_big_file(os.path.join(bigsrc, "big.bin"), 96 * 1024 * 1024)
run(tool, ["a", "-t7z", win(os.path.join(OUT, "big.7z")), win(bigsrc),
"-y", "-m0=lzma2", "-mx1"])
shutil.rmtree(bigsrc, ignore_errors=True)
# --------------------------------------------------------------------------
def clean() -> None:
os.makedirs(OUT, exist_ok=True)
for entry in os.listdir(OUT):
if entry.startswith(OWN_PREFIXES):
path = os.path.join(OUT, entry)
if os.path.isdir(path):
shutil.rmtree(path, ignore_errors=True)
else:
os.remove(path)
def main() -> int:
ap = argparse.ArgumentParser()
ap.add_argument("--tool", help="path to 7zr.exe / 7z.exe / 7zr")
ap.add_argument("--big", action="store_true",
help="also build a ~96 MiB single-file archive")
opts = ap.parse_args()
tool = find_tool(opts.tool)
print(f"7z tool: {tool}")
clean()
build_source({"binary": 300_000, "zeros": 2_000_000, "code": 400_000})
build_archives(tool, opts.big)
print(f"fixtures written to {OUT}")
for name in sorted(os.listdir(OUT)):
path = os.path.join(OUT, name)
if os.path.isfile(path):
print(f" {name} ({os.path.getsize(path)} bytes)")
return 0
if __name__ == "__main__":
sys.exit(main())
tests/run-sevenz-tests.sh
+113
View File
@@ -0,0 +1,113 @@
#!/usr/bin/env bash
# Host test runner for the 7z engine (vendor subset + fixture matrix).
#
# ./tests/run-sevenz-tests.sh
#
# Builds the vendored LZMA SDK subset and the end-to-end driver, generates real
# .7z fixtures with a 7-Zip binary, extracts every fixture and compares the
# result byte for byte against the source tree.
#
# On Windows this expects MinGW gcc in PATH and must be started with the MSYS
# bash explicitly (`/usr/bin/bash tests/run-sevenz-tests.sh`) -- a bare `bash`
# can resolve to C:\Windows\System32\bash.exe, i.e. the WSL launcher.
set -e
ROOT="$(cd "$(dirname "$0")/.." && pwd -W 2>/dev/null || pwd)"
BUILD="$ROOT/.build/sevenz-test"
SEVENZ_DIR="$ROOT/third_party/7z"
FIXTURES="$ROOT/tests/fixtures-7z"
PYTHON="${PYTHON:-python3}"
CC="${CC:-gcc}"
# Archives that the current engine cannot read yet. Each entry needs a reason;
# when one of them starts passing the script says so, so the list cannot rot.
KNOWN_GAPS="bcj2 aes aeshe vol.7z.001"
# bcj2 - folder has 5 coders; needs the engine's own chain driver
# aes - 7zAES coder not implemented yet
# aeshe - encrypted header (-mhe=on), same coder
# vol.7z.001 - multi-volume input; needs the volume-set stream
find "$BUILD" -maxdepth 1 -type f \( -name '*.o' -o -name '*.exe' \) -delete 2>/dev/null || true
mkdir -p "$BUILD"
# ---------------------------------------------------------------- vendor
CFLAGS_7Z=(-O2 -w -DZ7_PPMD_SUPPORT -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
-DNDEBUG -D_REENTRANT)
for src in "$SEVENZ_DIR"/*.c; do
name="$(basename "$src" .c)"
"$CC" -c "${CFLAGS_7Z[@]}" -o "$BUILD/$name.o" "$src"
done
# unrar-style extra libs are only needed by the Windows path of 7zFile.c.
EXTRA_LIBS=()
case "$(uname -s)" in
MINGW*|MSYS*|CYGWIN*) EXTRA_LIBS=(-lole32 -loleaut32 -luuid -ladvapi32 -luser32 -lshell32) ;;
esac
"$CC" -O2 -w -I"$SEVENZ_DIR" -o "$BUILD/sevenz_e2e" \
"$ROOT/tests/sevenz_e2e.c" "$BUILD"/*.o "${EXTRA_LIBS[@]}"
# ---------------------------------------------------------------- fixtures
if [ ! -f "$FIXTURES/lzma2.7z" ]; then
echo "== generating 7z fixtures =="
if ! "$PYTHON" "$ROOT/tests/make_sevenz_fixtures.py"; then
echo "SKIP: no 7z tool available, cannot build fixtures" >&2
exit 0
fi
fi
# ---------------------------------------------------------------- matrix
pass=0
fail=0
is_gap() {
for g in $KNOWN_GAPS; do
[ "$g" = "$1" ] && return 0
done
return 1
}
run_case() {
local name="$1" archive="$2"
local out="$BUILD/out/$name"
local log="$BUILD/out/$name.log"
rm -rf "$out"
mkdir -p "$out"
if "$BUILD/sevenz_e2e" "$archive" "$out" >"$log" 2>&1 &&
diff -r "$FIXTURES/_src" "$out/_src" >/dev/null 2>&1; then
if is_gap "$name"; then
printf ' %-12s GAP CLOSED (remove from KNOWN_GAPS)\n' "$name"
fail=$((fail + 1))
else
printf ' %-12s ok\n' "$name"
pass=$((pass + 1))
fi
return
fi
if is_gap "$name"; then
printf ' %-12s known gap (%s)\n' "$name" "$(head -1 "$log")"
pass=$((pass + 1))
else
printf ' %-12s FAIL\n' "$name"
sed -n '1,20p' "$log" | sed 's/^/ /'
fail=$((fail + 1))
fi
}
echo "== 7z fixture matrix =="
for a in store lzma2 lzma ppmd bcj delta utf8 bcj2 aes aeshe; do
[ -f "$FIXTURES/$a.7z" ] || continue
run_case "$a" "$FIXTURES/$a.7z"
done
if [ -f "$FIXTURES/vol.7z.001" ]; then
run_case "vol.7z.001" "$FIXTURES/vol.7z.001"
fi
echo
echo "$pass passed, $fail failed"
[ "$fail" -eq 0 ]
tests/sevenz_e2e.c
+302
View File
@@ -0,0 +1,302 @@
/*
* Standalone 7z driver for host end-to-end checks.
*
* sevenz_e2e <archive.7z> <out-dir>
*
* Lists every entry, extracts the archive into <out-dir> and prints the
* per-entry result. Exit status is non-zero if any entry fails.
*
* This is a *diagnostic* tool: it exercises the same container layer as the
* real extraction path, so it is the quickest way to tell whether a failure
* comes from the container/codec layer or from the web server around it.
*
* Entry names are UTF-16 inside a 7z archive; on a Windows host the file APIs
* must therefore be called with the wide-character variants or non-ASCII names
* silently fail to be created.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#if defined(_WIN32)
#include <direct.h>
#include <wchar.h>
#else
#include <unistd.h>
#endif
#include "7z.h"
#include "7zAlloc.h"
#include "7zCrc.h"
#include "7zFile.h"
#define INPUT_BUF_SIZE (1u << 18)
#define MAX_PATH_LEN 4096
static ISzAlloc g_alloc = { SzAlloc, SzFree };
static ISzAlloc g_temp = { SzAlloc, SzFree };
static int g_failures = 0;
/* ---------------------------------------------------------------- paths */
#if defined(_WIN32)
/* MinGW's wchar_t is 16 bit, matching UTF-16. */
static void utf8_to_utf16(const char *src, wchar_t *dst, size_t cap)
{
size_t out = 0;
while (*src && out + 2 < cap) {
unsigned char c = (unsigned char)*src++;
UInt32 cp;
if (c < 0x80) {
cp = c;
} else if ((c & 0xE0) == 0xC0 && (src[0] & 0xC0) == 0x80) {
cp = ((UInt32)(c & 0x1F) << 6) | (UInt32)(*src++ & 0x3F);
} else if ((c & 0xF0) == 0xE0 && (src[0] & 0xC0) == 0x80 && (src[1] & 0xC0) == 0x80) {
cp = ((UInt32)(c & 0x0F) << 12) | ((UInt32)(src[0] & 0x3F) << 6) | (UInt32)(src[1] & 0x3F);
src += 2;
} else if ((c & 0xF8) == 0xF0 && (src[0] & 0xC0) == 0x80 && (src[1] & 0xC0) == 0x80 && (src[2] & 0xC0) == 0x80) {
cp = ((UInt32)(c & 0x07) << 18) | ((UInt32)(src[0] & 0x3F) << 12) |
((UInt32)(src[1] & 0x3F) << 6) | (UInt32)(src[2] & 0x3F);
src += 3;
} else {
cp = '?';
}
if (cp >= 0x10000) {
cp -= 0x10000;
dst[out++] = (wchar_t)(0xD800 | (cp >> 10));
dst[out++] = (wchar_t)(0xDC00 | (cp & 0x3FF));
} else {
dst[out++] = (wchar_t)cp;
}
}
dst[out] = 0;
}
static void path_mkdir(const char *path)
{
wchar_t wide[MAX_PATH_LEN];
utf8_to_utf16(path, wide, MAX_PATH_LEN);
_wmkdir(wide);
}
static FILE *path_fopen_write(const char *path)
{
wchar_t wide[MAX_PATH_LEN];
utf8_to_utf16(path, wide, MAX_PATH_LEN);
return _wfopen(wide, L"wb");
}
#else /* POSIX: byte paths are UTF-8 already */
static void path_mkdir(const char *path) { mkdir(path, 0755); }
static FILE *path_fopen_write(const char *path) { return fopen(path, "wb"); }
#endif
static void make_dirs(const char *path)
{
char tmp[MAX_PATH_LEN];
size_t i, n = strlen(path);
if (n + 1 > sizeof(tmp))
return;
memcpy(tmp, path, n + 1);
for (i = 1; i < n; i++) {
if (tmp[i] == '/' || tmp[i] == '\\') {
char c = tmp[i];
tmp[i] = 0;
path_mkdir(tmp);
tmp[i] = c;
}
}
/* The final component counts too: callers hand us either a directory entry
or the parent directory of a file. */
path_mkdir(tmp);
}
static int write_file(const char *path, const Byte *data, size_t size)
{
FILE *fh = path_fopen_write(path);
if (!fh)
return -1;
if (size && fwrite(data, 1, size, fh) != size) {
fclose(fh);
return -1;
}
return fclose(fh) == 0 ? 0 : -1;
}
/* UTF-16 (LE, as stored by the 7z name table) to UTF-8. */
static void utf16_to_utf8(const UInt16 *src, char *dst, size_t dst_size)
{
size_t out = 0;
while (*src) {
UInt32 c = *src++;
if (c >= 0xD800 && c <= 0xDBFF && *src >= 0xDC00 && *src <= 0xDFFF)
c = 0x10000 + ((c - 0xD800) << 10) + (*src++ - 0xDC00);
if (out + 5 >= dst_size)
break;
if (c < 0x80) {
dst[out++] = (char)c;
} else if (c < 0x800) {
dst[out++] = (char)(0xC0 | (c >> 6));
dst[out++] = (char)(0x80 | (c & 0x3F));
} else if (c < 0x10000) {
dst[out++] = (char)(0xE0 | (c >> 12));
dst[out++] = (char)(0x80 | ((c >> 6) & 0x3F));
dst[out++] = (char)(0x80 | (c & 0x3F));
} else {
dst[out++] = (char)(0xF0 | (c >> 18));
dst[out++] = (char)(0x80 | ((c >> 12) & 0x3F));
dst[out++] = (char)(0x80 | ((c >> 6) & 0x3F));
dst[out++] = (char)(0x80 | (c & 0x3F));
}
}
dst[out] = 0;
}
/* ---------------------------------------------------------------- main */
int main(int argc, char **argv)
{
CFileInStream archive_stream;
CLookToRead2 look_stream;
CSzArEx db;
SRes res;
UInt16 *name16 = NULL;
size_t name16_cap = 0;
UInt32 i;
UInt32 block_index = 0xFFFFFFFF;
Byte *out_buffer = NULL;
size_t out_buffer_size = 0;
const char *out_dir;
if (argc < 3) {
fprintf(stderr, "usage: %s <archive.7z> <out-dir>\n", argv[0]);
return 2;
}
out_dir = argv[2];
if (InFile_Open(&archive_stream.file, argv[1]) != 0) {
fprintf(stderr, "cannot open %s\n", argv[1]);
return 1;
}
FileInStream_CreateVTable(&archive_stream);
archive_stream.wres = 0;
LookToRead2_CreateVTable(&look_stream, 0);
look_stream.buf = (Byte *)ISzAlloc_Alloc(&g_alloc, INPUT_BUF_SIZE);
if (!look_stream.buf) {
fprintf(stderr, "out of memory\n");
return 1;
}
look_stream.bufSize = INPUT_BUF_SIZE;
look_stream.realStream = &archive_stream.vt;
LookToRead2_INIT(&look_stream)
CrcGenerateTable();
SzArEx_Init(&db);
res = SzArEx_Open(&db, &look_stream.vt, &g_alloc, &g_temp);
if (res != SZ_OK) {
fprintf(stderr, "SzArEx_Open failed: res=%d\n", (int)res);
if (res == SZ_ERROR_UNSUPPORTED)
fprintf(stderr, " (unsupported coder - encrypted header or exotic method)\n");
return 1;
}
printf("entries: %u\n", (unsigned)db.NumFiles);
for (i = 0; i < db.NumFiles; i++) {
const int is_dir = SzArEx_IsDir(&db, i);
const size_t len = SzArEx_GetFileNameUtf16(&db, i, NULL);
char rel[MAX_PATH_LEN];
char full[MAX_PATH_LEN];
size_t offset = 0;
size_t processed = 0;
if (len + 1 > name16_cap) {
UInt16 *grown = (UInt16 *)realloc(name16, (len + 1) * sizeof(UInt16));
if (!grown) {
fprintf(stderr, "out of memory\n");
return 1;
}
name16 = grown;
name16_cap = len + 1;
}
SzArEx_GetFileNameUtf16(&db, i, name16);
utf16_to_utf8(name16, rel, sizeof(rel));
if (snprintf(full, sizeof(full), "%s/%s", out_dir, rel) >= (int)sizeof(full)) {
printf(" SKIP (path too long) %s\n", rel);
g_failures++;
continue;
}
if (is_dir) {
make_dirs(full);
printf(" dir %s\n", rel);
continue;
}
res = SzArEx_Extract(&db, &look_stream.vt, i, &block_index, &out_buffer,
&out_buffer_size, &offset, &processed, &g_alloc, &g_temp);
if (res != SZ_OK) {
printf(" FAIL %s (res=%d)\n", rel, (int)res);
g_failures++;
/* SzArEx_Extract leaves the block cache primed with a half-decoded
buffer when the folder decode fails; drop it so the next entry
reports its own error instead of a bogus CRC mismatch. */
block_index = 0xFFFFFFFF;
continue;
}
{
char *slash = strrchr(full, '/');
if (slash) {
*slash = 0;
make_dirs(full);
*slash = '/';
}
}
if (write_file(full, out_buffer + offset, processed) != 0) {
printf(" FAIL %s (write error)\n", rel);
g_failures++;
continue;
}
{
UInt32 crc = 0;
if (SzBitWithVals_Check(&db.CRCs, i)) {
crc = CrcCalc(out_buffer + offset, processed);
if (crc != db.CRCs.Vals[i]) {
printf(" FAIL %s (crc mismatch)\n", rel);
g_failures++;
continue;
}
}
}
printf(" ok %s (%lu bytes)\n", rel, (unsigned long)processed);
}
ISzAlloc_Free(&g_alloc, out_buffer);
ISzAlloc_Free(&g_alloc, look_stream.buf);
SzArEx_Free(&db, &g_alloc);
File_Close(&archive_stream.file);
free(name16);
printf("%s: %d failure(s)\n", argv[1], g_failures);
return g_failures ? 1 : 0;
}
third_party/7z/7z.h
Vendored
+204
View File
@@ -0,0 +1,204 @@
/* 7z.h -- 7z interface
2023-04-02 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_7Z_H
#define ZIP7_INC_7Z_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#define k7zStartHeaderSize 0x20
#define k7zSignatureSize 6
extern const Byte k7zSignature[k7zSignatureSize];
typedef struct
{
const Byte *Data;
size_t Size;
} CSzData;
/* CSzCoderInfo & CSzFolder support only default methods */
typedef struct
{
size_t PropsOffset;
UInt32 MethodID;
Byte NumStreams;
Byte PropsSize;
} CSzCoderInfo;
typedef struct
{
UInt32 InIndex;
UInt32 OutIndex;
} CSzBond;
#define SZ_NUM_CODERS_IN_FOLDER_MAX 4
#define SZ_NUM_BONDS_IN_FOLDER_MAX 3
#define SZ_NUM_PACK_STREAMS_IN_FOLDER_MAX 4
typedef struct
{
UInt32 NumCoders;
UInt32 NumBonds;
UInt32 NumPackStreams;
UInt32 UnpackStream;
UInt32 PackStreams[SZ_NUM_PACK_STREAMS_IN_FOLDER_MAX];
CSzBond Bonds[SZ_NUM_BONDS_IN_FOLDER_MAX];
CSzCoderInfo Coders[SZ_NUM_CODERS_IN_FOLDER_MAX];
} CSzFolder;
SRes SzGetNextFolderItem(CSzFolder *f, CSzData *sd);
typedef struct
{
UInt32 Low;
UInt32 High;
} CNtfsFileTime;
typedef struct
{
Byte *Defs; /* MSB 0 bit numbering */
UInt32 *Vals;
} CSzBitUi32s;
typedef struct
{
Byte *Defs; /* MSB 0 bit numbering */
// UInt64 *Vals;
CNtfsFileTime *Vals;
} CSzBitUi64s;
#define SzBitArray_Check(p, i) (((p)[(i) >> 3] & (0x80 >> ((i) & 7))) != 0)
#define SzBitWithVals_Check(p, i) ((p)->Defs && ((p)->Defs[(i) >> 3] & (0x80 >> ((i) & 7))) != 0)
typedef struct
{
UInt32 NumPackStreams;
UInt32 NumFolders;
UInt64 *PackPositions; // NumPackStreams + 1
CSzBitUi32s FolderCRCs; // NumFolders
size_t *FoCodersOffsets; // NumFolders + 1
UInt32 *FoStartPackStreamIndex; // NumFolders + 1
UInt32 *FoToCoderUnpackSizes; // NumFolders + 1
Byte *FoToMainUnpackSizeIndex; // NumFolders
UInt64 *CoderUnpackSizes; // for all coders in all folders
Byte *CodersData;
UInt64 RangeLimit;
} CSzAr;
UInt64 SzAr_GetFolderUnpackSize(const CSzAr *p, UInt32 folderIndex);
SRes SzAr_DecodeFolder(const CSzAr *p, UInt32 folderIndex,
ILookInStreamPtr stream, UInt64 startPos,
Byte *outBuffer, size_t outSize,
ISzAllocPtr allocMain);
typedef struct
{
CSzAr db;
UInt64 startPosAfterHeader;
UInt64 dataPos;
UInt32 NumFiles;
UInt64 *UnpackPositions; // NumFiles + 1
// Byte *IsEmptyFiles;
Byte *IsDirs;
CSzBitUi32s CRCs;
CSzBitUi32s Attribs;
// CSzBitUi32s Parents;
CSzBitUi64s MTime;
CSzBitUi64s CTime;
UInt32 *FolderToFile; // NumFolders + 1
UInt32 *FileToFolder; // NumFiles
size_t *FileNameOffsets; /* in 2-byte steps */
Byte *FileNames; /* UTF-16-LE */
} CSzArEx;
#define SzArEx_IsDir(p, i) (SzBitArray_Check((p)->IsDirs, i))
#define SzArEx_GetFileSize(p, i) ((p)->UnpackPositions[(i) + 1] - (p)->UnpackPositions[i])
void SzArEx_Init(CSzArEx *p);
void SzArEx_Free(CSzArEx *p, ISzAllocPtr alloc);
UInt64 SzArEx_GetFolderStreamPos(const CSzArEx *p, UInt32 folderIndex, UInt32 indexInFolder);
int SzArEx_GetFolderFullPackSize(const CSzArEx *p, UInt32 folderIndex, UInt64 *resSize);
/*
if dest == NULL, the return value specifies the required size of the buffer,
in 16-bit characters, including the null-terminating character.
if dest != NULL, the return value specifies the number of 16-bit characters that
are written to the dest, including the null-terminating character. */
size_t SzArEx_GetFileNameUtf16(const CSzArEx *p, size_t fileIndex, UInt16 *dest);
/*
size_t SzArEx_GetFullNameLen(const CSzArEx *p, size_t fileIndex);
UInt16 *SzArEx_GetFullNameUtf16_Back(const CSzArEx *p, size_t fileIndex, UInt16 *dest);
*/
/*
SzArEx_Extract extracts file from archive
*outBuffer must be 0 before first call for each new archive.
Extracting cache:
If you need to decompress more than one file, you can send
these values from previous call:
*blockIndex,
*outBuffer,
*outBufferSize
You can consider "*outBuffer" as cache of solid block. If your archive is solid,
it will increase decompression speed.
If you use external function, you can declare these 3 cache variables
(blockIndex, outBuffer, outBufferSize) as static in that external function.
Free *outBuffer and set *outBuffer to 0, if you want to flush cache.
*/
SRes SzArEx_Extract(
const CSzArEx *db,
ILookInStreamPtr inStream,
UInt32 fileIndex, /* index of file */
UInt32 *blockIndex, /* index of solid block */
Byte **outBuffer, /* pointer to pointer to output buffer (allocated with allocMain) */
size_t *outBufferSize, /* buffer size for output buffer */
size_t *offset, /* offset of stream for required file in *outBuffer */
size_t *outSizeProcessed, /* size of file in *outBuffer */
ISzAllocPtr allocMain,
ISzAllocPtr allocTemp);
/*
SzArEx_Open Errors:
SZ_ERROR_NO_ARCHIVE
SZ_ERROR_ARCHIVE
SZ_ERROR_UNSUPPORTED
SZ_ERROR_MEM
SZ_ERROR_CRC
SZ_ERROR_INPUT_EOF
SZ_ERROR_FAIL
*/
SRes SzArEx_Open(CSzArEx *p, ILookInStreamPtr inStream,
ISzAllocPtr allocMain, ISzAllocPtr allocTemp);
EXTERN_C_END
#endif
third_party/7z/7zAlloc.c
Vendored
+89
View File
@@ -0,0 +1,89 @@
/* 7zAlloc.c -- Allocation functions for 7z processing
2023-03-04 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include <stdlib.h>
#include "7zAlloc.h"
/* #define SZ_ALLOC_DEBUG */
/* use SZ_ALLOC_DEBUG to debug alloc/free operations */
#ifdef SZ_ALLOC_DEBUG
/*
#ifdef _WIN32
#include "7zWindows.h"
#endif
*/
#include <stdio.h>
static int g_allocCount = 0;
static int g_allocCountTemp = 0;
static void Print_Alloc(const char *s, size_t size, int *counter)
{
const unsigned size2 = (unsigned)size;
fprintf(stderr, "\n%s count = %10d : %10u bytes; ", s, *counter, size2);
(*counter)++;
}
static void Print_Free(const char *s, int *counter)
{
(*counter)--;
fprintf(stderr, "\n%s count = %10d", s, *counter);
}
#endif
void *SzAlloc(ISzAllocPtr p, size_t size)
{
UNUSED_VAR(p)
if (size == 0)
return 0;
#ifdef SZ_ALLOC_DEBUG
Print_Alloc("Alloc", size, &g_allocCount);
#endif
return malloc(size);
}
void SzFree(ISzAllocPtr p, void *address)
{
UNUSED_VAR(p)
#ifdef SZ_ALLOC_DEBUG
if (address)
Print_Free("Free ", &g_allocCount);
#endif
free(address);
}
void *SzAllocTemp(ISzAllocPtr p, size_t size)
{
UNUSED_VAR(p)
if (size == 0)
return 0;
#ifdef SZ_ALLOC_DEBUG
Print_Alloc("Alloc_temp", size, &g_allocCountTemp);
/*
#ifdef _WIN32
return HeapAlloc(GetProcessHeap(), 0, size);
#endif
*/
#endif
return malloc(size);
}
void SzFreeTemp(ISzAllocPtr p, void *address)
{
UNUSED_VAR(p)
#ifdef SZ_ALLOC_DEBUG
if (address)
Print_Free("Free_temp ", &g_allocCountTemp);
/*
#ifdef _WIN32
HeapFree(GetProcessHeap(), 0, address);
return;
#endif
*/
#endif
free(address);
}
third_party/7z/7zAlloc.h
Vendored
+19
View File
@@ -0,0 +1,19 @@
/* 7zAlloc.h -- Allocation functions
2023-03-04 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_7Z_ALLOC_H
#define ZIP7_INC_7Z_ALLOC_H
#include "7zTypes.h"
EXTERN_C_BEGIN
void *SzAlloc(ISzAllocPtr p, size_t size);
void SzFree(ISzAllocPtr p, void *address);
void *SzAllocTemp(ISzAllocPtr p, size_t size);
void SzFreeTemp(ISzAllocPtr p, void *address);
EXTERN_C_END
#endif
third_party/7z/7zArcIn.c
Vendored
+1701
View File
File diff suppressed because it is too large. Load diff
third_party/7z/7zBuf.c
Vendored
+36
View File
@@ -0,0 +1,36 @@
/* 7zBuf.c -- Byte Buffer
2017-04-03 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "7zBuf.h"
void Buf_Init(CBuf *p)
{
p->data = 0;
p->size = 0;
}
int Buf_Create(CBuf *p, size_t size, ISzAllocPtr alloc)
{
p->size = 0;
if (size == 0)
{
p->data = 0;
return 1;
}
p->data = (Byte *)ISzAlloc_Alloc(alloc, size);
if (p->data)
{
p->size = size;
return 1;
}
return 0;
}
void Buf_Free(CBuf *p, ISzAllocPtr alloc)
{
ISzAlloc_Free(alloc, p->data);
p->data = 0;
p->size = 0;
}
third_party/7z/7zBuf.h
Vendored
+35
View File
@@ -0,0 +1,35 @@
/* 7zBuf.h -- Byte Buffer
2023-03-04 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_7Z_BUF_H
#define ZIP7_INC_7Z_BUF_H
#include "7zTypes.h"
EXTERN_C_BEGIN
typedef struct
{
Byte *data;
size_t size;
} CBuf;
void Buf_Init(CBuf *p);
int Buf_Create(CBuf *p, size_t size, ISzAllocPtr alloc);
void Buf_Free(CBuf *p, ISzAllocPtr alloc);
typedef struct
{
Byte *data;
size_t size;
size_t pos;
} CDynBuf;
void DynBuf_Construct(CDynBuf *p);
void DynBuf_SeekToBeg(CDynBuf *p);
int DynBuf_Write(CDynBuf *p, const Byte *buf, size_t size, ISzAllocPtr alloc);
void DynBuf_Free(CDynBuf *p, ISzAllocPtr alloc);
EXTERN_C_END
#endif
third_party/7z/7zBuf2.c
Vendored
+52
View File
@@ -0,0 +1,52 @@
/* 7zBuf2.c -- Byte Buffer
2017-04-03 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include <string.h>
#include "7zBuf.h"
void DynBuf_Construct(CDynBuf *p)
{
p->data = 0;
p->size = 0;
p->pos = 0;
}
void DynBuf_SeekToBeg(CDynBuf *p)
{
p->pos = 0;
}
int DynBuf_Write(CDynBuf *p, const Byte *buf, size_t size, ISzAllocPtr alloc)
{
if (size > p->size - p->pos)
{
size_t newSize = p->pos + size;
Byte *data;
newSize += newSize / 4;
data = (Byte *)ISzAlloc_Alloc(alloc, newSize);
if (!data)
return 0;
p->size = newSize;
if (p->pos != 0)
memcpy(data, p->data, p->pos);
ISzAlloc_Free(alloc, p->data);
p->data = data;
}
if (size != 0)
{
memcpy(p->data + p->pos, buf, size);
p->pos += size;
}
return 1;
}
void DynBuf_Free(CDynBuf *p, ISzAllocPtr alloc)
{
ISzAlloc_Free(alloc, p->data);
p->data = 0;
p->size = 0;
p->pos = 0;
}
third_party/7z/7zCrc.c
Vendored
+420
View File
@@ -0,0 +1,420 @@
/* 7zCrc.c -- CRC32 calculation and init
2024-03-01 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "7zCrc.h"
#include "CpuArch.h"
// for debug:
// #define __ARM_FEATURE_CRC32 1
#ifdef __ARM_FEATURE_CRC32
// #pragma message("__ARM_FEATURE_CRC32")
#define Z7_CRC_HW_FORCE
#endif
// #define Z7_CRC_DEBUG_BE
#ifdef Z7_CRC_DEBUG_BE
#undef MY_CPU_LE
#define MY_CPU_BE
#endif
#ifdef Z7_CRC_HW_FORCE
#define Z7_CRC_NUM_TABLES_USE 1
#else
#ifdef Z7_CRC_NUM_TABLES
#define Z7_CRC_NUM_TABLES_USE Z7_CRC_NUM_TABLES
#else
#define Z7_CRC_NUM_TABLES_USE 12
#endif
#endif
#if Z7_CRC_NUM_TABLES_USE < 1
#error Stop_Compiling_Bad_Z7_CRC_NUM_TABLES
#endif
#if defined(MY_CPU_LE) || (Z7_CRC_NUM_TABLES_USE == 1)
#define Z7_CRC_NUM_TABLES_TOTAL Z7_CRC_NUM_TABLES_USE
#else
#define Z7_CRC_NUM_TABLES_TOTAL (Z7_CRC_NUM_TABLES_USE + 1)
#endif
#ifndef Z7_CRC_HW_FORCE
#if Z7_CRC_NUM_TABLES_USE == 1 \
|| (!defined(MY_CPU_LE) && !defined(MY_CPU_BE))
#define CRC_UPDATE_BYTE_2(crc, b) (table[((crc) ^ (b)) & 0xFF] ^ ((crc) >> 8))
#define Z7_CRC_UPDATE_T1_FUNC_NAME CrcUpdateGT1
static UInt32 Z7_FASTCALL Z7_CRC_UPDATE_T1_FUNC_NAME(UInt32 v, const void *data, size_t size)
{
const UInt32 *table = g_CrcTable;
const Byte *p = (const Byte *)data;
const Byte *lim = p + size;
for (; p != lim; p++)
v = CRC_UPDATE_BYTE_2(v, *p);
return v;
}
#endif
#if Z7_CRC_NUM_TABLES_USE != 1
#ifndef MY_CPU_BE
#define FUNC_NAME_LE_2(s) CrcUpdateT ## s
#define FUNC_NAME_LE_1(s) FUNC_NAME_LE_2(s)
#define FUNC_NAME_LE FUNC_NAME_LE_1(Z7_CRC_NUM_TABLES_USE)
UInt32 Z7_FASTCALL FUNC_NAME_LE (UInt32 v, const void *data, size_t size, const UInt32 *table);
#endif
#ifndef MY_CPU_LE
#define FUNC_NAME_BE_2(s) CrcUpdateT1_BeT ## s
#define FUNC_NAME_BE_1(s) FUNC_NAME_BE_2(s)
#define FUNC_NAME_BE FUNC_NAME_BE_1(Z7_CRC_NUM_TABLES_USE)
UInt32 Z7_FASTCALL FUNC_NAME_BE (UInt32 v, const void *data, size_t size, const UInt32 *table);
#endif
#endif
#endif // Z7_CRC_HW_FORCE
/* ---------- hardware CRC ---------- */
#ifdef MY_CPU_LE
#if defined(MY_CPU_ARM_OR_ARM64)
// #pragma message("ARM*")
#if (defined(__clang__) && (__clang_major__ >= 3)) \
|| defined(__GNUC__) && (__GNUC__ >= 6) && defined(MY_CPU_ARM64) \
|| defined(__GNUC__) && (__GNUC__ >= 8)
#if !defined(__ARM_FEATURE_CRC32)
// #pragma message("!defined(__ARM_FEATURE_CRC32)")
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define __ARM_FEATURE_CRC32 1
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#define Z7_ARM_FEATURE_CRC32_WAS_SET
#if defined(__clang__)
#if defined(MY_CPU_ARM64)
#define ATTRIB_CRC __attribute__((__target__("crc")))
#else
#define ATTRIB_CRC __attribute__((__target__("armv8-a,crc")))
#endif
#else
#if defined(MY_CPU_ARM64)
#if !defined(Z7_GCC_VERSION) || (Z7_GCC_VERSION >= 60000)
#define ATTRIB_CRC __attribute__((__target__("+crc")))
#endif
#else
#if !defined(Z7_GCC_VERSION) || (__GNUC__ >= 8)
#if defined(__ARM_FP) && __GNUC__ >= 8
// for -mfloat-abi=hard: similar to <arm_acle.h>
#define ATTRIB_CRC __attribute__((__target__("arch=armv8-a+crc+simd")))
#else
#define ATTRIB_CRC __attribute__((__target__("arch=armv8-a+crc")))
#endif
#endif
#endif
#endif
#endif
#if defined(__ARM_FEATURE_CRC32)
// #pragma message("<arm_acle.h>")
/*
arm_acle.h (GGC):
before Nov 17, 2017:
#ifdef __ARM_FEATURE_CRC32
Nov 17, 2017: gcc10.0 (gcc 9.2.0) checked"
#if __ARM_ARCH >= 8
#pragma GCC target ("arch=armv8-a+crc")
Aug 22, 2019: GCC 8.4?, 9.2.1, 10.1:
#ifdef __ARM_FEATURE_CRC32
#ifdef __ARM_FP
#pragma GCC target ("arch=armv8-a+crc+simd")
#else
#pragma GCC target ("arch=armv8-a+crc")
#endif
*/
#if defined(__ARM_ARCH) && __ARM_ARCH < 8
#if defined(Z7_GCC_VERSION) && (__GNUC__ == 8) && (Z7_GCC_VERSION < 80400) \
|| defined(Z7_GCC_VERSION) && (__GNUC__ == 9) && (Z7_GCC_VERSION < 90201) \
|| defined(Z7_GCC_VERSION) && (__GNUC__ == 10) && (Z7_GCC_VERSION < 100100)
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
// #pragma message("#define __ARM_ARCH 8")
#undef __ARM_ARCH
#define __ARM_ARCH 8
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif
#define Z7_CRC_HW_USE
#include <arm_acle.h>
#endif
#elif defined(_MSC_VER)
#if defined(MY_CPU_ARM64)
#if (_MSC_VER >= 1910)
#ifdef __clang__
// #define Z7_CRC_HW_USE
// #include <arm_acle.h>
#else
#define Z7_CRC_HW_USE
#include <intrin.h>
#endif
#endif
#endif
#endif
#else // non-ARM*
// #define Z7_CRC_HW_USE // for debug : we can test HW-branch of code
#ifdef Z7_CRC_HW_USE
#include "7zCrcEmu.h"
#endif
#endif // non-ARM*
#if defined(Z7_CRC_HW_USE)
// #pragma message("USE ARM HW CRC")
#ifdef MY_CPU_64BIT
#define CRC_HW_WORD_TYPE UInt64
#define CRC_HW_WORD_FUNC __crc32d
#else
#define CRC_HW_WORD_TYPE UInt32
#define CRC_HW_WORD_FUNC __crc32w
#endif
#define CRC_HW_UNROLL_BYTES (sizeof(CRC_HW_WORD_TYPE) * 4)
#ifdef ATTRIB_CRC
ATTRIB_CRC
#endif
Z7_NO_INLINE
#ifdef Z7_CRC_HW_FORCE
UInt32 Z7_FASTCALL CrcUpdate
#else
static UInt32 Z7_FASTCALL CrcUpdate_HW
#endif
(UInt32 v, const void *data, size_t size)
{
const Byte *p = (const Byte *)data;
for (; size != 0 && ((unsigned)(ptrdiff_t)p & (CRC_HW_UNROLL_BYTES - 1)) != 0; size--)
v = __crc32b(v, *p++);
if (size >= CRC_HW_UNROLL_BYTES)
{
const Byte *lim = p + size;
size &= CRC_HW_UNROLL_BYTES - 1;
lim -= size;
do
{
v = CRC_HW_WORD_FUNC(v, *(const CRC_HW_WORD_TYPE *)(const void *)(p));
v = CRC_HW_WORD_FUNC(v, *(const CRC_HW_WORD_TYPE *)(const void *)(p + sizeof(CRC_HW_WORD_TYPE)));
p += 2 * sizeof(CRC_HW_WORD_TYPE);
v = CRC_HW_WORD_FUNC(v, *(const CRC_HW_WORD_TYPE *)(const void *)(p));
v = CRC_HW_WORD_FUNC(v, *(const CRC_HW_WORD_TYPE *)(const void *)(p + sizeof(CRC_HW_WORD_TYPE)));
p += 2 * sizeof(CRC_HW_WORD_TYPE);
}
while (p != lim);
}
for (; size != 0; size--)
v = __crc32b(v, *p++);
return v;
}
#ifdef Z7_ARM_FEATURE_CRC32_WAS_SET
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#undef __ARM_FEATURE_CRC32
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#undef Z7_ARM_FEATURE_CRC32_WAS_SET
#endif
#endif // defined(Z7_CRC_HW_USE)
#endif // MY_CPU_LE
#ifndef Z7_CRC_HW_FORCE
#if defined(Z7_CRC_HW_USE) || defined(Z7_CRC_UPDATE_T1_FUNC_NAME)
/*
typedef UInt32 (Z7_FASTCALL *Z7_CRC_UPDATE_WITH_TABLE_FUNC)
(UInt32 v, const void *data, size_t size, const UInt32 *table);
Z7_CRC_UPDATE_WITH_TABLE_FUNC g_CrcUpdate;
*/
static unsigned g_Crc_Algo;
#if (!defined(MY_CPU_LE) && !defined(MY_CPU_BE))
static unsigned g_Crc_Be;
#endif
#endif // defined(Z7_CRC_HW_USE) || defined(Z7_CRC_UPDATE_T1_FUNC_NAME)
Z7_NO_INLINE
#ifdef Z7_CRC_HW_USE
static UInt32 Z7_FASTCALL CrcUpdate_Base
#else
UInt32 Z7_FASTCALL CrcUpdate
#endif
(UInt32 crc, const void *data, size_t size)
{
#if Z7_CRC_NUM_TABLES_USE == 1
return Z7_CRC_UPDATE_T1_FUNC_NAME(crc, data, size);
#else // Z7_CRC_NUM_TABLES_USE != 1
#ifdef Z7_CRC_UPDATE_T1_FUNC_NAME
if (g_Crc_Algo == 1)
return Z7_CRC_UPDATE_T1_FUNC_NAME(crc, data, size);
#endif
#ifdef MY_CPU_LE
return FUNC_NAME_LE(crc, data, size, g_CrcTable);
#elif defined(MY_CPU_BE)
return FUNC_NAME_BE(crc, data, size, g_CrcTable);
#else
if (g_Crc_Be)
return FUNC_NAME_BE(crc, data, size, g_CrcTable);
else
return FUNC_NAME_LE(crc, data, size, g_CrcTable);
#endif
#endif // Z7_CRC_NUM_TABLES_USE != 1
}
#ifdef Z7_CRC_HW_USE
Z7_NO_INLINE
UInt32 Z7_FASTCALL CrcUpdate(UInt32 crc, const void *data, size_t size)
{
if (g_Crc_Algo == 0)
return CrcUpdate_HW(crc, data, size);
return CrcUpdate_Base(crc, data, size);
}
#endif
#endif // !defined(Z7_CRC_HW_FORCE)
UInt32 Z7_FASTCALL CrcCalc(const void *data, size_t size)
{
return CrcUpdate(CRC_INIT_VAL, data, size) ^ CRC_INIT_VAL;
}
MY_ALIGN(64)
UInt32 g_CrcTable[256 * Z7_CRC_NUM_TABLES_TOTAL];
void Z7_FASTCALL CrcGenerateTable(void)
{
UInt32 i;
for (i = 0; i < 256; i++)
{
#if defined(Z7_CRC_HW_FORCE)
g_CrcTable[i] = __crc32b(i, 0);
#else
#define kCrcPoly 0xEDB88320
UInt32 r = i;
unsigned j;
for (j = 0; j < 8; j++)
r = (r >> 1) ^ (kCrcPoly & ((UInt32)0 - (r & 1)));
g_CrcTable[i] = r;
#endif
}
for (i = 256; i < 256 * Z7_CRC_NUM_TABLES_USE; i++)
{
const UInt32 r = g_CrcTable[(size_t)i - 256];
g_CrcTable[i] = g_CrcTable[r & 0xFF] ^ (r >> 8);
}
#if !defined(Z7_CRC_HW_FORCE) && \
(defined(Z7_CRC_HW_USE) || defined(Z7_CRC_UPDATE_T1_FUNC_NAME) || defined(MY_CPU_BE))
#if Z7_CRC_NUM_TABLES_USE <= 1
g_Crc_Algo = 1;
#else // Z7_CRC_NUM_TABLES_USE <= 1
#if defined(MY_CPU_LE)
g_Crc_Algo = Z7_CRC_NUM_TABLES_USE;
#else // !defined(MY_CPU_LE)
{
#ifndef MY_CPU_BE
UInt32 k = 0x01020304;
const Byte *p = (const Byte *)&k;
if (p[0] == 4 && p[1] == 3)
g_Crc_Algo = Z7_CRC_NUM_TABLES_USE;
else if (p[0] != 1 || p[1] != 2)
g_Crc_Algo = 1;
else
#endif // MY_CPU_BE
{
for (i = 256 * Z7_CRC_NUM_TABLES_TOTAL - 1; i >= 256; i--)
{
const UInt32 x = g_CrcTable[(size_t)i - 256];
g_CrcTable[i] = Z7_BSWAP32(x);
}
#if defined(Z7_CRC_UPDATE_T1_FUNC_NAME)
g_Crc_Algo = Z7_CRC_NUM_TABLES_USE;
#endif
#if (!defined(MY_CPU_LE) && !defined(MY_CPU_BE))
g_Crc_Be = 1;
#endif
}
}
#endif // !defined(MY_CPU_LE)
#ifdef MY_CPU_LE
#ifdef Z7_CRC_HW_USE
if (CPU_IsSupported_CRC32())
g_Crc_Algo = 0;
#endif // Z7_CRC_HW_USE
#endif // MY_CPU_LE
#endif // Z7_CRC_NUM_TABLES_USE <= 1
#endif // g_Crc_Algo was declared
}
Z7_CRC_UPDATE_FUNC z7_GetFunc_CrcUpdate(unsigned algo)
{
if (algo == 0)
return &CrcUpdate;
#if defined(Z7_CRC_HW_USE)
if (algo == sizeof(CRC_HW_WORD_TYPE) * 8)
{
#ifdef Z7_CRC_HW_FORCE
return &CrcUpdate;
#else
if (g_Crc_Algo == 0)
return &CrcUpdate_HW;
#endif
}
#endif
#ifndef Z7_CRC_HW_FORCE
if (algo == Z7_CRC_NUM_TABLES_USE)
return
#ifdef Z7_CRC_HW_USE
&CrcUpdate_Base;
#else
&CrcUpdate;
#endif
#endif
return NULL;
}
#undef kCrcPoly
#undef Z7_CRC_NUM_TABLES_USE
#undef Z7_CRC_NUM_TABLES_TOTAL
#undef CRC_UPDATE_BYTE_2
#undef FUNC_NAME_LE_2
#undef FUNC_NAME_LE_1
#undef FUNC_NAME_LE
#undef FUNC_NAME_BE_2
#undef FUNC_NAME_BE_1
#undef FUNC_NAME_BE
#undef CRC_HW_UNROLL_BYTES
#undef CRC_HW_WORD_FUNC
#undef CRC_HW_WORD_TYPE
third_party/7z/7zCrc.h
Vendored
+28
View File
@@ -0,0 +1,28 @@
/* 7zCrc.h -- CRC32 calculation
2024-01-22 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_7Z_CRC_H
#define ZIP7_INC_7Z_CRC_H
#include "7zTypes.h"
EXTERN_C_BEGIN
extern UInt32 g_CrcTable[];
/* Call CrcGenerateTable one time before other CRC functions */
void Z7_FASTCALL CrcGenerateTable(void);
#define CRC_INIT_VAL 0xFFFFFFFF
#define CRC_GET_DIGEST(crc) ((crc) ^ CRC_INIT_VAL)
#define CRC_UPDATE_BYTE(crc, b) (g_CrcTable[((crc) ^ (b)) & 0xFF] ^ ((crc) >> 8))
UInt32 Z7_FASTCALL CrcUpdate(UInt32 crc, const void *data, size_t size);
UInt32 Z7_FASTCALL CrcCalc(const void *data, size_t size);
typedef UInt32 (Z7_FASTCALL *Z7_CRC_UPDATE_FUNC)(UInt32 v, const void *data, size_t size);
Z7_CRC_UPDATE_FUNC z7_GetFunc_CrcUpdate(unsigned algo);
EXTERN_C_END
#endif
third_party/7z/7zCrcOpt.c
Vendored
+199
View File
@@ -0,0 +1,199 @@
/* 7zCrcOpt.c -- CRC32 calculation (optimized functions)
2023-12-07 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "CpuArch.h"
#if !defined(Z7_CRC_NUM_TABLES) || Z7_CRC_NUM_TABLES > 1
// for debug only : define Z7_CRC_DEBUG_BE to test big-endian code in little-endian cpu
// #define Z7_CRC_DEBUG_BE
#ifdef Z7_CRC_DEBUG_BE
#undef MY_CPU_LE
#define MY_CPU_BE
#endif
// the value Z7_CRC_NUM_TABLES_USE must be defined to same value as in 7zCrc.c
#ifdef Z7_CRC_NUM_TABLES
#define Z7_CRC_NUM_TABLES_USE Z7_CRC_NUM_TABLES
#else
#define Z7_CRC_NUM_TABLES_USE 12
#endif
#if Z7_CRC_NUM_TABLES_USE % 4 || \
Z7_CRC_NUM_TABLES_USE < 4 * 1 || \
Z7_CRC_NUM_TABLES_USE > 4 * 6
#error Stop_Compiling_Bad_Z7_CRC_NUM_TABLES
#endif
#ifndef MY_CPU_BE
#define CRC_UPDATE_BYTE_2(crc, b) (table[((crc) ^ (b)) & 0xFF] ^ ((crc) >> 8))
#define Q(n, d) \
( (table + ((n) * 4 + 3) * 0x100)[(Byte)(d)] \
^ (table + ((n) * 4 + 2) * 0x100)[((d) >> 1 * 8) & 0xFF] \
^ (table + ((n) * 4 + 1) * 0x100)[((d) >> 2 * 8) & 0xFF] \
^ (table + ((n) * 4 + 0) * 0x100)[((d) >> 3 * 8)] )
#define R(a) *((const UInt32 *)(const void *)p + (a))
#define CRC_FUNC_PRE_LE2(step) \
UInt32 Z7_FASTCALL CrcUpdateT ## step (UInt32 v, const void *data, size_t size, const UInt32 *table)
#define CRC_FUNC_PRE_LE(step) \
CRC_FUNC_PRE_LE2(step); \
CRC_FUNC_PRE_LE2(step)
CRC_FUNC_PRE_LE(Z7_CRC_NUM_TABLES_USE)
{
const Byte *p = (const Byte *)data;
const Byte *lim;
for (; size && ((unsigned)(ptrdiff_t)p & (7 - (Z7_CRC_NUM_TABLES_USE & 4))) != 0; size--, p++)
v = CRC_UPDATE_BYTE_2(v, *p);
lim = p + size;
if (size >= Z7_CRC_NUM_TABLES_USE)
{
lim -= Z7_CRC_NUM_TABLES_USE;
do
{
v ^= R(0);
{
#if Z7_CRC_NUM_TABLES_USE == 1 * 4
v = Q(0, v);
#else
#define U2(r, op) \
{ d = R(r); x op Q(Z7_CRC_NUM_TABLES_USE / 4 - 1 - (r), d); }
UInt32 d, x;
U2(1, =)
#if Z7_CRC_NUM_TABLES_USE >= 3 * 4
#define U(r) U2(r, ^=)
U(2)
#if Z7_CRC_NUM_TABLES_USE >= 4 * 4
U(3)
#if Z7_CRC_NUM_TABLES_USE >= 5 * 4
U(4)
#if Z7_CRC_NUM_TABLES_USE >= 6 * 4
U(5)
#if Z7_CRC_NUM_TABLES_USE >= 7 * 4
#error Stop_Compiling_Bad_Z7_CRC_NUM_TABLES
#endif
#endif
#endif
#endif
#endif
#undef U
#undef U2
v = x ^ Q(Z7_CRC_NUM_TABLES_USE / 4 - 1, v);
#endif
}
p += Z7_CRC_NUM_TABLES_USE;
}
while (p <= lim);
lim += Z7_CRC_NUM_TABLES_USE;
}
for (; p < lim; p++)
v = CRC_UPDATE_BYTE_2(v, *p);
return v;
}
#undef CRC_UPDATE_BYTE_2
#undef R
#undef Q
#undef CRC_FUNC_PRE_LE
#undef CRC_FUNC_PRE_LE2
#endif
#ifndef MY_CPU_LE
#define CRC_UPDATE_BYTE_2_BE(crc, b) (table[((crc) >> 24) ^ (b)] ^ ((crc) << 8))
#define Q(n, d) \
( (table + ((n) * 4 + 0) * 0x100)[((d)) & 0xFF] \
^ (table + ((n) * 4 + 1) * 0x100)[((d) >> 1 * 8) & 0xFF] \
^ (table + ((n) * 4 + 2) * 0x100)[((d) >> 2 * 8) & 0xFF] \
^ (table + ((n) * 4 + 3) * 0x100)[((d) >> 3 * 8)] )
#ifdef Z7_CRC_DEBUG_BE
#define R(a) GetBe32a((const UInt32 *)(const void *)p + (a))
#else
#define R(a) *((const UInt32 *)(const void *)p + (a))
#endif
#define CRC_FUNC_PRE_BE2(step) \
UInt32 Z7_FASTCALL CrcUpdateT1_BeT ## step (UInt32 v, const void *data, size_t size, const UInt32 *table)
#define CRC_FUNC_PRE_BE(step) \
CRC_FUNC_PRE_BE2(step); \
CRC_FUNC_PRE_BE2(step)
CRC_FUNC_PRE_BE(Z7_CRC_NUM_TABLES_USE)
{
const Byte *p = (const Byte *)data;
const Byte *lim;
table += 0x100;
v = Z7_BSWAP32(v);
for (; size && ((unsigned)(ptrdiff_t)p & (7 - (Z7_CRC_NUM_TABLES_USE & 4))) != 0; size--, p++)
v = CRC_UPDATE_BYTE_2_BE(v, *p);
lim = p + size;
if (size >= Z7_CRC_NUM_TABLES_USE)
{
lim -= Z7_CRC_NUM_TABLES_USE;
do
{
v ^= R(0);
{
#if Z7_CRC_NUM_TABLES_USE == 1 * 4
v = Q(0, v);
#else
#define U2(r, op) \
{ d = R(r); x op Q(Z7_CRC_NUM_TABLES_USE / 4 - 1 - (r), d); }
UInt32 d, x;
U2(1, =)
#if Z7_CRC_NUM_TABLES_USE >= 3 * 4
#define U(r) U2(r, ^=)
U(2)
#if Z7_CRC_NUM_TABLES_USE >= 4 * 4
U(3)
#if Z7_CRC_NUM_TABLES_USE >= 5 * 4
U(4)
#if Z7_CRC_NUM_TABLES_USE >= 6 * 4
U(5)
#if Z7_CRC_NUM_TABLES_USE >= 7 * 4
#error Stop_Compiling_Bad_Z7_CRC_NUM_TABLES
#endif
#endif
#endif
#endif
#endif
#undef U
#undef U2
v = x ^ Q(Z7_CRC_NUM_TABLES_USE / 4 - 1, v);
#endif
}
p += Z7_CRC_NUM_TABLES_USE;
}
while (p <= lim);
lim += Z7_CRC_NUM_TABLES_USE;
}
for (; p < lim; p++)
v = CRC_UPDATE_BYTE_2_BE(v, *p);
return Z7_BSWAP32(v);
}
#undef CRC_UPDATE_BYTE_2_BE
#undef R
#undef Q
#undef CRC_FUNC_PRE_BE
#undef CRC_FUNC_PRE_BE2
#endif
#undef Z7_CRC_NUM_TABLES_USE
#endif
third_party/7z/7zDec.c
Vendored
+672
View File
@@ -0,0 +1,672 @@
/* 7zDec.c -- Decoding from 7z folder
: Igor Pavlov : Public domain */
#include "Precomp.h"
#include <string.h>
/* #define Z7_PPMD_SUPPORT */
#include "7z.h"
#include "7zCrc.h"
#include "Bcj2.h"
#include "Bra.h"
#include "CpuArch.h"
#include "Delta.h"
#include "LzmaDec.h"
#include "Lzma2Dec.h"
#ifdef Z7_PPMD_SUPPORT
#include "Ppmd7.h"
#endif
#define k_Copy 0
#ifndef Z7_NO_METHOD_LZMA2
#define k_LZMA2 0x21
#endif
#define k_LZMA 0x30101
#define k_BCJ2 0x303011B
#if !defined(Z7_NO_METHODS_FILTERS)
#define Z7_USE_BRANCH_FILTER
#endif
#if !defined(Z7_NO_METHODS_FILTERS) || \
defined(Z7_USE_NATIVE_BRANCH_FILTER) && defined(MY_CPU_ARM64)
#define Z7_USE_FILTER_ARM64
#ifndef Z7_USE_BRANCH_FILTER
#define Z7_USE_BRANCH_FILTER
#endif
#define k_ARM64 0xa
#endif
#if !defined(Z7_NO_METHODS_FILTERS) || \
defined(Z7_USE_NATIVE_BRANCH_FILTER) && defined(MY_CPU_ARMT)
#define Z7_USE_FILTER_ARMT
#ifndef Z7_USE_BRANCH_FILTER
#define Z7_USE_BRANCH_FILTER
#endif
#define k_ARMT 0x3030701
#endif
#ifndef Z7_NO_METHODS_FILTERS
#define k_Delta 3
#define k_RISCV 0xb
#define k_BCJ 0x3030103
#define k_PPC 0x3030205
#define k_IA64 0x3030401
#define k_ARM 0x3030501
#define k_SPARC 0x3030805
#endif
#ifdef Z7_PPMD_SUPPORT
#define k_PPMD 0x30401
typedef struct
{
IByteIn vt;
const Byte *cur;
const Byte *end;
const Byte *begin;
UInt64 processed;
BoolInt extra;
SRes res;
ILookInStreamPtr inStream;
} CByteInToLook;
static Byte ReadByte(IByteInPtr pp)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CByteInToLook)
if (p->cur != p->end)
return *p->cur++;
if (p->res == SZ_OK)
{
size_t size = (size_t)(p->cur - p->begin);
p->processed += size;
p->res = ILookInStream_Skip(p->inStream, size);
size = (1 << 25);
p->res = ILookInStream_Look(p->inStream, (const void **)&p->begin, &size);
p->cur = p->begin;
p->end = p->begin + size;
if (size != 0)
return *p->cur++;
}
p->extra = True;
return 0;
}
static SRes SzDecodePpmd(const Byte *props, unsigned propsSize, UInt64 inSize, ILookInStreamPtr inStream,
Byte *outBuffer, SizeT outSize, ISzAllocPtr allocMain)
{
CPpmd7 *ppmd;
SRes res;
unsigned order;
UInt32 memSize;
if (propsSize != 5)
return SZ_ERROR_UNSUPPORTED;
order = props[0];
memSize = GetUi32(props + 1);
if (order < PPMD7_MIN_ORDER ||
order > PPMD7_MAX_ORDER ||
memSize < PPMD7_MIN_MEM_SIZE ||
memSize > PPMD7_MAX_MEM_SIZE)
return SZ_ERROR_UNSUPPORTED;
if ((ppmd = (CPpmd7 *)ISzAlloc_Alloc(allocMain, sizeof(CPpmd7))) == NULL)
return SZ_ERROR_MEM;
Ppmd7_Construct(ppmd);
res = SZ_ERROR_MEM;
if (Ppmd7_Alloc(ppmd, memSize, allocMain))
{
CByteInToLook s;
s.vt.Read = ReadByte;
s.inStream = inStream;
s.begin = s.end = s.cur = NULL;
s.extra = False;
s.res = SZ_OK;
s.processed = 0;
Ppmd7_Init(ppmd, order);
ppmd->rc.dec.Stream = &s.vt;
res = SZ_ERROR_DATA;
if (Ppmd7z_RangeDec_Init(&ppmd->rc.dec) && !s.extra)
{
Byte *buf = outBuffer;
const Byte *lim = buf + outSize;
for (; buf != lim; buf++)
{
int sym = Ppmd7z_DecodeSymbol(ppmd);
if (s.extra || sym < 0)
break;
*buf = (Byte)sym;
}
if (buf == lim)
if (Ppmd7z_RangeDec_IsFinishedOK(&ppmd->rc.dec)
// || (Ppmd7z_DecodeSymbol(&ppmd) == PPMD7_SYM_END && Ppmd7z_RangeDec_IsFinishedOK(&ppmd.rc.dec))
)
res = SZ_OK;
}
if (s.extra)
res = (s.res != SZ_OK ? s.res : SZ_ERROR_DATA);
else if (s.processed + (size_t)(s.cur - s.begin) != inSize)
res = SZ_ERROR_DATA;
Ppmd7_Free(ppmd, allocMain);
}
ISzAlloc_Free(allocMain, ppmd);
return res;
}
#endif
static SRes SzDecodeLzma(const Byte *props, unsigned propsSize, UInt64 inSize, ILookInStreamPtr inStream,
Byte *outBuffer, SizeT outSize, ISzAllocPtr allocMain)
{
CLzmaDec state;
SRes res = SZ_OK;
LzmaDec_CONSTRUCT(&state)
RINOK(LzmaDec_AllocateProbs(&state, props, propsSize, allocMain))
state.dic = outBuffer;
state.dicBufSize = outSize;
LzmaDec_Init(&state);
for (;;)
{
const void *inBuf = NULL;
size_t lookahead = (1 << 18);
if (lookahead > inSize)
lookahead = (size_t)inSize;
res = ILookInStream_Look(inStream, &inBuf, &lookahead);
if (res != SZ_OK)
break;
{
SizeT inProcessed = (SizeT)lookahead, dicPos = state.dicPos;
ELzmaStatus status;
res = LzmaDec_DecodeToDic(&state, outSize, (const Byte *)inBuf, &inProcessed, LZMA_FINISH_END, &status);
lookahead -= inProcessed;
inSize -= inProcessed;
if (res != SZ_OK)
break;
if (status == LZMA_STATUS_FINISHED_WITH_MARK)
{
if (outSize != state.dicPos || inSize != 0)
res = SZ_ERROR_DATA;
break;
}
if (outSize == state.dicPos && inSize == 0 && status == LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK)
break;
if (inProcessed == 0 && dicPos == state.dicPos)
{
res = SZ_ERROR_DATA;
break;
}
res = ILookInStream_Skip(inStream, inProcessed);
if (res != SZ_OK)
break;
}
}
LzmaDec_FreeProbs(&state, allocMain);
return res;
}
#ifndef Z7_NO_METHOD_LZMA2
static SRes SzDecodeLzma2(const Byte *props, unsigned propsSize, UInt64 inSize, ILookInStreamPtr inStream,
Byte *outBuffer, SizeT outSize, ISzAllocPtr allocMain)
{
CLzma2Dec state;
SRes res = SZ_OK;
Lzma2Dec_CONSTRUCT(&state)
if (propsSize != 1)
return SZ_ERROR_DATA;
RINOK(Lzma2Dec_AllocateProbs(&state, props[0], allocMain))
state.decoder.dic = outBuffer;
state.decoder.dicBufSize = outSize;
Lzma2Dec_Init(&state);
for (;;)
{
const void *inBuf = NULL;
size_t lookahead = (1 << 18);
if (lookahead > inSize)
lookahead = (size_t)inSize;
res = ILookInStream_Look(inStream, &inBuf, &lookahead);
if (res != SZ_OK)
break;
{
SizeT inProcessed = (SizeT)lookahead, dicPos = state.decoder.dicPos;
ELzmaStatus status;
res = Lzma2Dec_DecodeToDic(&state, outSize, (const Byte *)inBuf, &inProcessed, LZMA_FINISH_END, &status);
lookahead -= inProcessed;
inSize -= inProcessed;
if (res != SZ_OK)
break;
if (status == LZMA_STATUS_FINISHED_WITH_MARK)
{
if (outSize != state.decoder.dicPos || inSize != 0)
res = SZ_ERROR_DATA;
break;
}
if (inProcessed == 0 && dicPos == state.decoder.dicPos)
{
res = SZ_ERROR_DATA;
break;
}
res = ILookInStream_Skip(inStream, inProcessed);
if (res != SZ_OK)
break;
}
}
Lzma2Dec_FreeProbs(&state, allocMain);
return res;
}
#endif
static SRes SzDecodeCopy(UInt64 inSize, ILookInStreamPtr inStream, Byte *outBuffer)
{
while (inSize > 0)
{
const void *inBuf;
size_t curSize = (1 << 18);
if (curSize > inSize)
curSize = (size_t)inSize;
RINOK(ILookInStream_Look(inStream, &inBuf, &curSize))
if (curSize == 0)
return SZ_ERROR_INPUT_EOF;
memcpy(outBuffer, inBuf, curSize);
outBuffer += curSize;
inSize -= curSize;
RINOK(ILookInStream_Skip(inStream, curSize))
}
return SZ_OK;
}
static BoolInt IS_MAIN_METHOD(UInt32 m)
{
switch (m)
{
case k_Copy:
case k_LZMA:
#ifndef Z7_NO_METHOD_LZMA2
case k_LZMA2:
#endif
#ifdef Z7_PPMD_SUPPORT
case k_PPMD:
#endif
return True;
default:
return False;
}
}
static BoolInt IS_SUPPORTED_CODER(const CSzCoderInfo *c)
{
return
c->NumStreams == 1
/* && c->MethodID <= (UInt32)0xFFFFFFFF */
&& IS_MAIN_METHOD((UInt32)c->MethodID);
}
#define IS_BCJ2(c) ((c)->MethodID == k_BCJ2 && (c)->NumStreams == 4)
static SRes CheckSupportedFolder(const CSzFolder *f)
{
if (f->NumCoders < 1 || f->NumCoders > 4)
return SZ_ERROR_UNSUPPORTED;
if (!IS_SUPPORTED_CODER(&f->Coders[0]))
return SZ_ERROR_UNSUPPORTED;
if (f->NumCoders == 1)
{
if (f->NumPackStreams != 1 || f->PackStreams[0] != 0 || f->NumBonds != 0)
return SZ_ERROR_UNSUPPORTED;
return SZ_OK;
}
#if defined(Z7_USE_BRANCH_FILTER)
if (f->NumCoders == 2)
{
const CSzCoderInfo *c = &f->Coders[1];
if (
/* c->MethodID > (UInt32)0xFFFFFFFF || */
c->NumStreams != 1
|| f->NumPackStreams != 1
|| f->PackStreams[0] != 0
|| f->NumBonds != 1
|| f->Bonds[0].InIndex != 1
|| f->Bonds[0].OutIndex != 0)
return SZ_ERROR_UNSUPPORTED;
switch ((UInt32)c->MethodID)
{
#if !defined(Z7_NO_METHODS_FILTERS)
case k_Delta:
case k_BCJ:
case k_PPC:
case k_IA64:
case k_SPARC:
case k_ARM:
case k_RISCV:
#endif
#ifdef Z7_USE_FILTER_ARM64
case k_ARM64:
#endif
#ifdef Z7_USE_FILTER_ARMT
case k_ARMT:
#endif
break;
default:
return SZ_ERROR_UNSUPPORTED;
}
return SZ_OK;
}
#endif
if (f->NumCoders == 4)
{
if (!IS_SUPPORTED_CODER(&f->Coders[1])
|| !IS_SUPPORTED_CODER(&f->Coders[2])
|| !IS_BCJ2(&f->Coders[3]))
return SZ_ERROR_UNSUPPORTED;
if (f->NumPackStreams != 4
|| f->PackStreams[0] != 2
|| f->PackStreams[1] != 6
|| f->PackStreams[2] != 1
|| f->PackStreams[3] != 0
|| f->NumBonds != 3
|| f->Bonds[0].InIndex != 5 || f->Bonds[0].OutIndex != 0
|| f->Bonds[1].InIndex != 4 || f->Bonds[1].OutIndex != 1
|| f->Bonds[2].InIndex != 3 || f->Bonds[2].OutIndex != 2)
return SZ_ERROR_UNSUPPORTED;
return SZ_OK;
}
return SZ_ERROR_UNSUPPORTED;
}
static SRes SzFolder_Decode2(const CSzFolder *folder,
const Byte *propsData,
const UInt64 *unpackSizes,
const UInt64 *packPositions,
ILookInStreamPtr inStream, UInt64 startPos,
Byte *outBuffer, SizeT outSize, ISzAllocPtr allocMain,
Byte *tempBuf[])
{
UInt32 ci;
SizeT tempSizes[3] = { 0, 0, 0};
SizeT tempSize3 = 0;
Byte *tempBuf3 = 0;
RINOK(CheckSupportedFolder(folder))
for (ci = 0; ci < folder->NumCoders; ci++)
{
const CSzCoderInfo *coder = &folder->Coders[ci];
if (IS_MAIN_METHOD((UInt32)coder->MethodID))
{
UInt32 si = 0;
UInt64 offset;
UInt64 inSize;
Byte *outBufCur = outBuffer;
SizeT outSizeCur = outSize;
if (folder->NumCoders == 4)
{
const UInt32 indices[] = { 3, 2, 0 };
const UInt64 unpackSize = unpackSizes[ci];
si = indices[ci];
if (ci < 2)
{
Byte *temp;
outSizeCur = (SizeT)unpackSize;
if (outSizeCur != unpackSize)
return SZ_ERROR_MEM;
temp = (Byte *)ISzAlloc_Alloc(allocMain, outSizeCur);
if (!temp && outSizeCur != 0)
return SZ_ERROR_MEM;
outBufCur = tempBuf[1 - ci] = temp;
tempSizes[1 - ci] = outSizeCur;
}
else if (ci == 2)
{
if (unpackSize > outSize) /* check it */
return SZ_ERROR_PARAM;
tempBuf3 = outBufCur = outBuffer + (outSize - (size_t)unpackSize);
tempSize3 = outSizeCur = (SizeT)unpackSize;
}
else
return SZ_ERROR_UNSUPPORTED;
}
offset = packPositions[si];
inSize = packPositions[(size_t)si + 1] - offset;
RINOK(LookInStream_SeekTo(inStream, startPos + offset))
if (coder->MethodID == k_Copy)
{
if (inSize != outSizeCur) /* check it */
return SZ_ERROR_DATA;
RINOK(SzDecodeCopy(inSize, inStream, outBufCur))
}
else if (coder->MethodID == k_LZMA)
{
RINOK(SzDecodeLzma(propsData + coder->PropsOffset, coder->PropsSize, inSize, inStream, outBufCur, outSizeCur, allocMain))
}
#ifndef Z7_NO_METHOD_LZMA2
else if (coder->MethodID == k_LZMA2)
{
RINOK(SzDecodeLzma2(propsData + coder->PropsOffset, coder->PropsSize, inSize, inStream, outBufCur, outSizeCur, allocMain))
}
#endif
#ifdef Z7_PPMD_SUPPORT
else if (coder->MethodID == k_PPMD)
{
RINOK(SzDecodePpmd(propsData + coder->PropsOffset, coder->PropsSize, inSize, inStream, outBufCur, outSizeCur, allocMain))
}
#endif
else
return SZ_ERROR_UNSUPPORTED;
}
else if (coder->MethodID == k_BCJ2)
{
const UInt64 offset = packPositions[1];
const UInt64 s3Size = packPositions[2] - offset;
if (ci != 3)
return SZ_ERROR_UNSUPPORTED;
tempSizes[2] = (SizeT)s3Size;
if (tempSizes[2] != s3Size)
return SZ_ERROR_MEM;
tempBuf[2] = (Byte *)ISzAlloc_Alloc(allocMain, tempSizes[2]);
if (!tempBuf[2] && tempSizes[2] != 0)
return SZ_ERROR_MEM;
RINOK(LookInStream_SeekTo(inStream, startPos + offset))
RINOK(SzDecodeCopy(s3Size, inStream, tempBuf[2]))
if ((tempSizes[0] & 3) != 0 ||
(tempSizes[1] & 3) != 0 ||
tempSize3 + tempSizes[0] + tempSizes[1] != outSize)
return SZ_ERROR_DATA;
{
CBcj2Dec p;
p.bufs[0] = tempBuf3; p.lims[0] = tempBuf3 + tempSize3;
p.bufs[1] = tempBuf[0]; p.lims[1] = tempBuf[0] + tempSizes[0];
p.bufs[2] = tempBuf[1]; p.lims[2] = tempBuf[1] + tempSizes[1];
p.bufs[3] = tempBuf[2]; p.lims[3] = tempBuf[2] + tempSizes[2];
p.dest = outBuffer;
p.destLim = outBuffer + outSize;
Bcj2Dec_Init(&p);
RINOK(Bcj2Dec_Decode(&p))
{
unsigned i;
for (i = 0; i < 4; i++)
if (p.bufs[i] != p.lims[i])
return SZ_ERROR_DATA;
if (p.dest != p.destLim || !Bcj2Dec_IsMaybeFinished(&p))
return SZ_ERROR_DATA;
}
}
}
#if defined(Z7_USE_BRANCH_FILTER)
else if (ci == 1)
{
#if !defined(Z7_NO_METHODS_FILTERS)
if (coder->MethodID == k_Delta)
{
if (coder->PropsSize != 1)
return SZ_ERROR_UNSUPPORTED;
{
Byte state[DELTA_STATE_SIZE];
Delta_Init(state);
Delta_Decode(state, (unsigned)(propsData[coder->PropsOffset]) + 1, outBuffer, outSize);
}
continue;
}
#endif
#ifdef Z7_USE_FILTER_ARM64
if (coder->MethodID == k_ARM64)
{
UInt32 pc = 0;
if (coder->PropsSize == 4)
{
pc = GetUi32(propsData + coder->PropsOffset);
if (pc & 3)
return SZ_ERROR_UNSUPPORTED;
}
else if (coder->PropsSize != 0)
return SZ_ERROR_UNSUPPORTED;
z7_BranchConv_ARM64_Dec(outBuffer, outSize, pc);
continue;
}
#endif
#if !defined(Z7_NO_METHODS_FILTERS)
if (coder->MethodID == k_RISCV)
{
UInt32 pc = 0;
if (coder->PropsSize == 4)
{
pc = GetUi32(propsData + coder->PropsOffset);
if (pc & 1)
return SZ_ERROR_UNSUPPORTED;
}
else if (coder->PropsSize != 0)
return SZ_ERROR_UNSUPPORTED;
z7_BranchConv_RISCV_Dec(outBuffer, outSize, pc);
continue;
}
#endif
#if !defined(Z7_NO_METHODS_FILTERS) || defined(Z7_USE_FILTER_ARMT)
{
if (coder->PropsSize != 0)
return SZ_ERROR_UNSUPPORTED;
#define CASE_BRA_CONV(isa) case k_ ## isa: Z7_BRANCH_CONV_DEC(isa)(outBuffer, outSize, 0); break; // pc = 0;
switch (coder->MethodID)
{
#if !defined(Z7_NO_METHODS_FILTERS)
case k_BCJ:
{
UInt32 state = Z7_BRANCH_CONV_ST_X86_STATE_INIT_VAL;
z7_BranchConvSt_X86_Dec(outBuffer, outSize, 0, &state); // pc = 0
break;
}
case k_PPC: Z7_BRANCH_CONV_DEC_2(BranchConv_PPC)(outBuffer, outSize, 0); break; // pc = 0;
// CASE_BRA_CONV(PPC)
CASE_BRA_CONV(IA64)
CASE_BRA_CONV(SPARC)
CASE_BRA_CONV(ARM)
#endif
#if !defined(Z7_NO_METHODS_FILTERS) || defined(Z7_USE_FILTER_ARMT)
CASE_BRA_CONV(ARMT)
#endif
default:
return SZ_ERROR_UNSUPPORTED;
}
continue;
}
#endif
} // (c == 1)
#endif // Z7_USE_BRANCH_FILTER
else
return SZ_ERROR_UNSUPPORTED;
}
return SZ_OK;
}
SRes SzAr_DecodeFolder(const CSzAr *p, UInt32 folderIndex,
ILookInStreamPtr inStream, UInt64 startPos,
Byte *outBuffer, size_t outSize,
ISzAllocPtr allocMain)
{
SRes res;
CSzFolder folder;
CSzData sd;
const Byte *data = p->CodersData + p->FoCodersOffsets[folderIndex];
sd.Data = data;
sd.Size = p->FoCodersOffsets[(size_t)folderIndex + 1] - p->FoCodersOffsets[folderIndex];
res = SzGetNextFolderItem(&folder, &sd);
if (res != SZ_OK)
return res;
if (sd.Size != 0
|| folder.UnpackStream != p->FoToMainUnpackSizeIndex[folderIndex]
|| outSize != SzAr_GetFolderUnpackSize(p, folderIndex))
return SZ_ERROR_FAIL;
{
unsigned i;
Byte *tempBuf[3] = { 0, 0, 0};
res = SzFolder_Decode2(&folder, data,
&p->CoderUnpackSizes[p->FoToCoderUnpackSizes[folderIndex]],
p->PackPositions + p->FoStartPackStreamIndex[folderIndex],
inStream, startPos,
outBuffer, (SizeT)outSize, allocMain, tempBuf);
for (i = 0; i < 3; i++)
ISzAlloc_Free(allocMain, tempBuf[i]);
if (res == SZ_OK)
if (SzBitWithVals_Check(&p->FolderCRCs, folderIndex))
if (CrcCalc(outBuffer, outSize) != p->FolderCRCs.Vals[folderIndex])
res = SZ_ERROR_CRC;
return res;
}
}
third_party/7z/7zFile.c
Vendored
+443
View File
@@ -0,0 +1,443 @@
/* 7zFile.c -- File IO
2023-04-02 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "7zFile.h"
#ifndef USE_WINDOWS_FILE
#include <errno.h>
#ifndef USE_FOPEN
#include <stdio.h>
#include <fcntl.h>
#ifdef _WIN32
#include <io.h>
typedef int ssize_t;
typedef int off_t;
#else
#include <unistd.h>
#endif
#endif
#else
/*
ReadFile and WriteFile functions in Windows have BUG:
If you Read or Write 64MB or more (probably min_failure_size = 64MB - 32KB + 1)
from/to Network file, it returns ERROR_NO_SYSTEM_RESOURCES
(Insufficient system resources exist to complete the requested service).
Probably in some version of Windows there are problems with other sizes:
for 32 MB (maybe also for 16 MB).
And message can be "Network connection was lost"
*/
#endif
#define kChunkSizeMax (1 << 22)
void File_Construct(CSzFile *p)
{
#ifdef USE_WINDOWS_FILE
p->handle = INVALID_HANDLE_VALUE;
#elif defined(USE_FOPEN)
p->file = NULL;
#else
p->fd = -1;
#endif
}
#if !defined(UNDER_CE) || !defined(USE_WINDOWS_FILE)
static WRes File_Open(CSzFile *p, const char *name, int writeMode)
{
#ifdef USE_WINDOWS_FILE
p->handle = CreateFileA(name,
writeMode ? GENERIC_WRITE : GENERIC_READ,
FILE_SHARE_READ, NULL,
writeMode ? CREATE_ALWAYS : OPEN_EXISTING,
FILE_ATTRIBUTE_NORMAL, NULL);
return (p->handle != INVALID_HANDLE_VALUE) ? 0 : GetLastError();
#elif defined(USE_FOPEN)
p->file = fopen(name, writeMode ? "wb+" : "rb");
return (p->file != 0) ? 0 :
#ifdef UNDER_CE
2; /* ENOENT */
#else
errno;
#endif
#else
int flags = (writeMode ? (O_CREAT | O_EXCL | O_WRONLY) : O_RDONLY);
#ifdef O_BINARY
flags |= O_BINARY;
#endif
p->fd = open(name, flags, 0666);
return (p->fd != -1) ? 0 : errno;
#endif
}
WRes InFile_Open(CSzFile *p, const char *name) { return File_Open(p, name, 0); }
WRes OutFile_Open(CSzFile *p, const char *name)
{
#if defined(USE_WINDOWS_FILE) || defined(USE_FOPEN)
return File_Open(p, name, 1);
#else
p->fd = creat(name, 0666);
return (p->fd != -1) ? 0 : errno;
#endif
}
#endif
#ifdef USE_WINDOWS_FILE
static WRes File_OpenW(CSzFile *p, const WCHAR *name, int writeMode)
{
p->handle = CreateFileW(name,
writeMode ? GENERIC_WRITE : GENERIC_READ,
FILE_SHARE_READ, NULL,
writeMode ? CREATE_ALWAYS : OPEN_EXISTING,
FILE_ATTRIBUTE_NORMAL, NULL);
return (p->handle != INVALID_HANDLE_VALUE) ? 0 : GetLastError();
}
WRes InFile_OpenW(CSzFile *p, const WCHAR *name) { return File_OpenW(p, name, 0); }
WRes OutFile_OpenW(CSzFile *p, const WCHAR *name) { return File_OpenW(p, name, 1); }
#endif
WRes File_Close(CSzFile *p)
{
#ifdef USE_WINDOWS_FILE
if (p->handle != INVALID_HANDLE_VALUE)
{
if (!CloseHandle(p->handle))
return GetLastError();
p->handle = INVALID_HANDLE_VALUE;
}
#elif defined(USE_FOPEN)
if (p->file != NULL)
{
int res = fclose(p->file);
if (res != 0)
{
if (res == EOF)
return errno;
return res;
}
p->file = NULL;
}
#else
if (p->fd != -1)
{
if (close(p->fd) != 0)
return errno;
p->fd = -1;
}
#endif
return 0;
}
WRes File_Read(CSzFile *p, void *data, size_t *size)
{
size_t originalSize = *size;
*size = 0;
if (originalSize == 0)
return 0;
#ifdef USE_WINDOWS_FILE
do
{
const DWORD curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : (DWORD)originalSize;
DWORD processed = 0;
const BOOL res = ReadFile(p->handle, data, curSize, &processed, NULL);
data = (void *)((Byte *)data + processed);
originalSize -= processed;
*size += processed;
if (!res)
return GetLastError();
// debug : we can break here for partial reading mode
if (processed == 0)
break;
}
while (originalSize > 0);
#elif defined(USE_FOPEN)
do
{
const size_t curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : originalSize;
const size_t processed = fread(data, 1, curSize, p->file);
data = (void *)((Byte *)data + (size_t)processed);
originalSize -= processed;
*size += processed;
if (processed != curSize)
return ferror(p->file);
// debug : we can break here for partial reading mode
if (processed == 0)
break;
}
while (originalSize > 0);
#else
do
{
const size_t curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : originalSize;
const ssize_t processed = read(p->fd, data, curSize);
if (processed == -1)
return errno;
if (processed == 0)
break;
data = (void *)((Byte *)data + (size_t)processed);
originalSize -= (size_t)processed;
*size += (size_t)processed;
// debug : we can break here for partial reading mode
// break;
}
while (originalSize > 0);
#endif
return 0;
}
WRes File_Write(CSzFile *p, const void *data, size_t *size)
{
size_t originalSize = *size;
*size = 0;
if (originalSize == 0)
return 0;
#ifdef USE_WINDOWS_FILE
do
{
const DWORD curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : (DWORD)originalSize;
DWORD processed = 0;
const BOOL res = WriteFile(p->handle, data, curSize, &processed, NULL);
data = (const void *)((const Byte *)data + processed);
originalSize -= processed;
*size += processed;
if (!res)
return GetLastError();
if (processed == 0)
break;
}
while (originalSize > 0);
#elif defined(USE_FOPEN)
do
{
const size_t curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : originalSize;
const size_t processed = fwrite(data, 1, curSize, p->file);
data = (void *)((Byte *)data + (size_t)processed);
originalSize -= processed;
*size += processed;
if (processed != curSize)
return ferror(p->file);
if (processed == 0)
break;
}
while (originalSize > 0);
#else
do
{
const size_t curSize = (originalSize > kChunkSizeMax) ? kChunkSizeMax : originalSize;
const ssize_t processed = write(p->fd, data, curSize);
if (processed == -1)
return errno;
if (processed == 0)
break;
data = (const void *)((const Byte *)data + (size_t)processed);
originalSize -= (size_t)processed;
*size += (size_t)processed;
}
while (originalSize > 0);
#endif
return 0;
}
WRes File_Seek(CSzFile *p, Int64 *pos, ESzSeek origin)
{
#ifdef USE_WINDOWS_FILE
DWORD moveMethod;
UInt32 low = (UInt32)*pos;
LONG high = (LONG)((UInt64)*pos >> 16 >> 16); /* for case when UInt64 is 32-bit only */
// (int) to eliminate clang warning
switch ((int)origin)
{
case SZ_SEEK_SET: moveMethod = FILE_BEGIN; break;
case SZ_SEEK_CUR: moveMethod = FILE_CURRENT; break;
case SZ_SEEK_END: moveMethod = FILE_END; break;
default: return ERROR_INVALID_PARAMETER;
}
low = SetFilePointer(p->handle, (LONG)low, &high, moveMethod);
if (low == (UInt32)0xFFFFFFFF)
{
WRes res = GetLastError();
if (res != NO_ERROR)
return res;
}
*pos = ((Int64)high << 32) | low;
return 0;
#else
int moveMethod; // = origin;
switch ((int)origin)
{
case SZ_SEEK_SET: moveMethod = SEEK_SET; break;
case SZ_SEEK_CUR: moveMethod = SEEK_CUR; break;
case SZ_SEEK_END: moveMethod = SEEK_END; break;
default: return EINVAL;
}
#if defined(USE_FOPEN)
{
int res = fseek(p->file, (long)*pos, moveMethod);
if (res == -1)
return errno;
*pos = ftell(p->file);
if (*pos == -1)
return errno;
return 0;
}
#else
{
off_t res = lseek(p->fd, (off_t)*pos, moveMethod);
if (res == -1)
return errno;
*pos = res;
return 0;
}
#endif // USE_FOPEN
#endif // USE_WINDOWS_FILE
}
WRes File_GetLength(CSzFile *p, UInt64 *length)
{
#ifdef USE_WINDOWS_FILE
DWORD sizeHigh;
DWORD sizeLow = GetFileSize(p->handle, &sizeHigh);
if (sizeLow == 0xFFFFFFFF)
{
DWORD res = GetLastError();
if (res != NO_ERROR)
return res;
}
*length = (((UInt64)sizeHigh) << 32) + sizeLow;
return 0;
#elif defined(USE_FOPEN)
long pos = ftell(p->file);
int res = fseek(p->file, 0, SEEK_END);
*length = ftell(p->file);
fseek(p->file, pos, SEEK_SET);
return res;
#else
off_t pos;
*length = 0;
pos = lseek(p->fd, 0, SEEK_CUR);
if (pos != -1)
{
const off_t len2 = lseek(p->fd, 0, SEEK_END);
const off_t res2 = lseek(p->fd, pos, SEEK_SET);
if (len2 != -1)
{
*length = (UInt64)len2;
if (res2 != -1)
return 0;
}
}
return errno;
#endif
}
/* ---------- FileSeqInStream ---------- */
static SRes FileSeqInStream_Read(ISeqInStreamPtr pp, void *buf, size_t *size)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CFileSeqInStream)
const WRes wres = File_Read(&p->file, buf, size);
p->wres = wres;
return (wres == 0) ? SZ_OK : SZ_ERROR_READ;
}
void FileSeqInStream_CreateVTable(CFileSeqInStream *p)
{
p->vt.Read = FileSeqInStream_Read;
}
/* ---------- FileInStream ---------- */
static SRes FileInStream_Read(ISeekInStreamPtr pp, void *buf, size_t *size)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CFileInStream)
const WRes wres = File_Read(&p->file, buf, size);
p->wres = wres;
return (wres == 0) ? SZ_OK : SZ_ERROR_READ;
}
static SRes FileInStream_Seek(ISeekInStreamPtr pp, Int64 *pos, ESzSeek origin)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CFileInStream)
const WRes wres = File_Seek(&p->file, pos, origin);
p->wres = wres;
return (wres == 0) ? SZ_OK : SZ_ERROR_READ;
}
void FileInStream_CreateVTable(CFileInStream *p)
{
p->vt.Read = FileInStream_Read;
p->vt.Seek = FileInStream_Seek;
}
/* ---------- FileOutStream ---------- */
static size_t FileOutStream_Write(ISeqOutStreamPtr pp, const void *data, size_t size)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CFileOutStream)
const WRes wres = File_Write(&p->file, data, &size);
p->wres = wres;
return size;
}
void FileOutStream_CreateVTable(CFileOutStream *p)
{
p->vt.Write = FileOutStream_Write;
}
third_party/7z/7zFile.h
Vendored
+91
View File
@@ -0,0 +1,91 @@
/* 7zFile.h -- File IO
: Igor Pavlov : Public domain */
#ifndef ZIP7_INC_FILE_H
#define ZIP7_INC_FILE_H
#ifdef _WIN32
#define USE_WINDOWS_FILE
#endif
#ifdef USE_WINDOWS_FILE
#include "7zWindows.h"
#else
// note: USE_FOPEN mode is limited to 32-bit file size
// #define USE_FOPEN
// #include <stdio.h>
#endif
#include "7zTypes.h"
EXTERN_C_BEGIN
/* ---------- File ---------- */
typedef struct
{
#ifdef USE_WINDOWS_FILE
HANDLE handle;
#elif defined(USE_FOPEN)
FILE *file;
#else
int fd;
#endif
} CSzFile;
void File_Construct(CSzFile *p);
#if !defined(UNDER_CE) || !defined(USE_WINDOWS_FILE)
WRes InFile_Open(CSzFile *p, const char *name);
WRes OutFile_Open(CSzFile *p, const char *name);
#endif
#ifdef USE_WINDOWS_FILE
WRes InFile_OpenW(CSzFile *p, const WCHAR *name);
WRes OutFile_OpenW(CSzFile *p, const WCHAR *name);
#endif
WRes File_Close(CSzFile *p);
/* reads max(*size, remain file's size) bytes */
WRes File_Read(CSzFile *p, void *data, size_t *size);
/* writes *size bytes */
WRes File_Write(CSzFile *p, const void *data, size_t *size);
WRes File_Seek(CSzFile *p, Int64 *pos, ESzSeek origin);
WRes File_GetLength(CSzFile *p, UInt64 *length);
/* ---------- FileInStream ---------- */
typedef struct
{
ISeqInStream vt;
CSzFile file;
WRes wres;
} CFileSeqInStream;
void FileSeqInStream_CreateVTable(CFileSeqInStream *p);
typedef struct
{
ISeekInStream vt;
CSzFile file;
WRes wres;
} CFileInStream;
void FileInStream_CreateVTable(CFileInStream *p);
typedef struct
{
ISeqOutStream vt;
CSzFile file;
WRes wres;
} CFileOutStream;
void FileOutStream_CreateVTable(CFileOutStream *p);
EXTERN_C_END
#endif
third_party/7z/7zStream.c
Vendored
+199
View File
@@ -0,0 +1,199 @@
/* 7zStream.c -- 7z Stream functions
2023-04-02 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include <string.h>
#include "7zTypes.h"
SRes SeqInStream_ReadMax(ISeqInStreamPtr stream, void *buf, size_t *processedSize)
{
size_t size = *processedSize;
*processedSize = 0;
while (size != 0)
{
size_t cur = size;
const SRes res = ISeqInStream_Read(stream, buf, &cur);
*processedSize += cur;
buf = (void *)((Byte *)buf + cur);
size -= cur;
if (res != SZ_OK)
return res;
if (cur == 0)
return SZ_OK;
}
return SZ_OK;
}
/*
SRes SeqInStream_Read2(ISeqInStreamPtr stream, void *buf, size_t size, SRes errorType)
{
while (size != 0)
{
size_t processed = size;
RINOK(ISeqInStream_Read(stream, buf, &processed))
if (processed == 0)
return errorType;
buf = (void *)((Byte *)buf + processed);
size -= processed;
}
return SZ_OK;
}
SRes SeqInStream_Read(ISeqInStreamPtr stream, void *buf, size_t size)
{
return SeqInStream_Read2(stream, buf, size, SZ_ERROR_INPUT_EOF);
}
*/
SRes SeqInStream_ReadByte(ISeqInStreamPtr stream, Byte *buf)
{
size_t processed = 1;
RINOK(ISeqInStream_Read(stream, buf, &processed))
return (processed == 1) ? SZ_OK : SZ_ERROR_INPUT_EOF;
}
SRes LookInStream_SeekTo(ILookInStreamPtr stream, UInt64 offset)
{
Int64 t = (Int64)offset;
return ILookInStream_Seek(stream, &t, SZ_SEEK_SET);
}
SRes LookInStream_LookRead(ILookInStreamPtr stream, void *buf, size_t *size)
{
const void *lookBuf;
if (*size == 0)
return SZ_OK;
RINOK(ILookInStream_Look(stream, &lookBuf, size))
memcpy(buf, lookBuf, *size);
return ILookInStream_Skip(stream, *size);
}
SRes LookInStream_Read2(ILookInStreamPtr stream, void *buf, size_t size, SRes errorType)
{
while (size != 0)
{
size_t processed = size;
RINOK(ILookInStream_Read(stream, buf, &processed))
if (processed == 0)
return errorType;
buf = (void *)((Byte *)buf + processed);
size -= processed;
}
return SZ_OK;
}
SRes LookInStream_Read(ILookInStreamPtr stream, void *buf, size_t size)
{
return LookInStream_Read2(stream, buf, size, SZ_ERROR_INPUT_EOF);
}
#define GET_LookToRead2 Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CLookToRead2)
static SRes LookToRead2_Look_Lookahead(ILookInStreamPtr pp, const void **buf, size_t *size)
{
SRes res = SZ_OK;
GET_LookToRead2
size_t size2 = p->size - p->pos;
if (size2 == 0 && *size != 0)
{
p->pos = 0;
p->size = 0;
size2 = p->bufSize;
res = ISeekInStream_Read(p->realStream, p->buf, &size2);
p->size = size2;
}
if (*size > size2)
*size = size2;
*buf = p->buf + p->pos;
return res;
}
static SRes LookToRead2_Look_Exact(ILookInStreamPtr pp, const void **buf, size_t *size)
{
SRes res = SZ_OK;
GET_LookToRead2
size_t size2 = p->size - p->pos;
if (size2 == 0 && *size != 0)
{
p->pos = 0;
p->size = 0;
if (*size > p->bufSize)
*size = p->bufSize;
res = ISeekInStream_Read(p->realStream, p->buf, size);
size2 = p->size = *size;
}
if (*size > size2)
*size = size2;
*buf = p->buf + p->pos;
return res;
}
static SRes LookToRead2_Skip(ILookInStreamPtr pp, size_t offset)
{
GET_LookToRead2
p->pos += offset;
return SZ_OK;
}
static SRes LookToRead2_Read(ILookInStreamPtr pp, void *buf, size_t *size)
{
GET_LookToRead2
size_t rem = p->size - p->pos;
if (rem == 0)
return ISeekInStream_Read(p->realStream, buf, size);
if (rem > *size)
rem = *size;
memcpy(buf, p->buf + p->pos, rem);
p->pos += rem;
*size = rem;
return SZ_OK;
}
static SRes LookToRead2_Seek(ILookInStreamPtr pp, Int64 *pos, ESzSeek origin)
{
GET_LookToRead2
p->pos = p->size = 0;
return ISeekInStream_Seek(p->realStream, pos, origin);
}
void LookToRead2_CreateVTable(CLookToRead2 *p, int lookahead)
{
p->vt.Look = lookahead ?
LookToRead2_Look_Lookahead :
LookToRead2_Look_Exact;
p->vt.Skip = LookToRead2_Skip;
p->vt.Read = LookToRead2_Read;
p->vt.Seek = LookToRead2_Seek;
}
static SRes SecToLook_Read(ISeqInStreamPtr pp, void *buf, size_t *size)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CSecToLook)
return LookInStream_LookRead(p->realStream, buf, size);
}
void SecToLook_CreateVTable(CSecToLook *p)
{
p->vt.Read = SecToLook_Read;
}
static SRes SecToRead_Read(ISeqInStreamPtr pp, void *buf, size_t *size)
{
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(CSecToRead)
return ILookInStream_Read(p->realStream, buf, size);
}
void SecToRead_CreateVTable(CSecToRead *p)
{
p->vt.Read = SecToRead_Read;
}
third_party/7z/7zTypes.h
Vendored
+600
View File
@@ -0,0 +1,600 @@
/* 7zTypes.h -- Basic types
: Igor Pavlov : Public domain */
#ifndef ZIP7_7Z_TYPES_H
#define ZIP7_7Z_TYPES_H
#ifdef _WIN32
/* #include <windows.h> */
#else
#include <errno.h>
#endif
#include <stddef.h>
#ifndef EXTERN_C_BEGIN
#ifdef __cplusplus
#define EXTERN_C_BEGIN extern "C" {
#define EXTERN_C_END }
#else
#define EXTERN_C_BEGIN
#define EXTERN_C_END
#endif
#endif
EXTERN_C_BEGIN
#define SZ_OK 0
#define SZ_ERROR_DATA 1
#define SZ_ERROR_MEM 2
#define SZ_ERROR_CRC 3
#define SZ_ERROR_UNSUPPORTED 4
#define SZ_ERROR_PARAM 5
#define SZ_ERROR_INPUT_EOF 6
#define SZ_ERROR_OUTPUT_EOF 7
#define SZ_ERROR_READ 8
#define SZ_ERROR_WRITE 9
#define SZ_ERROR_PROGRESS 10
#define SZ_ERROR_FAIL 11
#define SZ_ERROR_THREAD 12
#define SZ_ERROR_ARCHIVE 16
#define SZ_ERROR_NO_ARCHIVE 17
typedef int SRes;
#ifdef _MSC_VER
#define MY_ALIGN_IN_STRUCT(n) __declspec(align(n))
#if _MSC_VER > 1200
#define MY_ALIGN(n) MY_ALIGN_IN_STRUCT(n)
#else
#define MY_ALIGN(n)
#endif
#else
/*
// C11/C++11:
#include <stdalign.h>
#define MY_ALIGN(n) alignas(n)
*/
#define MY_ALIGN(n) __attribute__ ((aligned(n)))
#define MY_ALIGN_IN_STRUCT(n) MY_ALIGN(n)
#endif
#ifdef _WIN32
/* typedef DWORD WRes; */
typedef unsigned WRes;
#define MY_SRes_HRESULT_FROM_WRes(x) HRESULT_FROM_WIN32(x)
// #define MY_HRES_ERROR_INTERNAL_ERROR MY_SRes_HRESULT_FROM_WRes(ERROR_INTERNAL_ERROR)
#else // _WIN32
// #define ENV_HAVE_LSTAT
typedef int WRes;
// (FACILITY_ERRNO = 0x800) is 7zip's FACILITY constant to represent (errno) errors in HRESULT
#define MY_FACILITY_ERRNO 0x800
#define MY_FACILITY_WIN32 7
#define MY_FACILITY_WRes MY_FACILITY_ERRNO
#define MY_HRESULT_FROM_errno_CONST_ERROR(x) ((HRESULT)( \
( (HRESULT)(x) & 0x0000FFFF) \
| (MY_FACILITY_WRes << 16) \
| (HRESULT)0x80000000 ))
#define MY_SRes_HRESULT_FROM_WRes(x) \
((HRESULT)(x) <= 0 ? ((HRESULT)(x)) : MY_HRESULT_FROM_errno_CONST_ERROR(x))
// we call macro HRESULT_FROM_WIN32 for system errors (WRes) that are (errno)
#define HRESULT_FROM_WIN32(x) MY_SRes_HRESULT_FROM_WRes(x)
/*
#define ERROR_FILE_NOT_FOUND 2L
#define ERROR_ACCESS_DENIED 5L
#define ERROR_NO_MORE_FILES 18L
#define ERROR_LOCK_VIOLATION 33L
#define ERROR_FILE_EXISTS 80L
#define ERROR_DISK_FULL 112L
#define ERROR_NEGATIVE_SEEK 131L
#define ERROR_ALREADY_EXISTS 183L
#define ERROR_DIRECTORY 267L
#define ERROR_TOO_MANY_POSTS 298L
#define ERROR_INTERNAL_ERROR 1359L
#define ERROR_INVALID_REPARSE_DATA 4392L
#define ERROR_REPARSE_TAG_INVALID 4393L
#define ERROR_REPARSE_TAG_MISMATCH 4394L
*/
// we use errno equivalents for some WIN32 errors:
#define ERROR_INVALID_PARAMETER EINVAL
#define ERROR_INVALID_FUNCTION EINVAL
#define ERROR_ALREADY_EXISTS EEXIST
#define ERROR_FILE_EXISTS EEXIST
#define ERROR_PATH_NOT_FOUND ENOENT
#define ERROR_FILE_NOT_FOUND ENOENT
#define ERROR_DISK_FULL ENOSPC
// #define ERROR_INVALID_HANDLE EBADF
// we use FACILITY_WIN32 for errors that has no errno equivalent
// Too many posts were made to a semaphore.
#define ERROR_TOO_MANY_POSTS ((HRESULT)0x8007012AL)
#define ERROR_INVALID_REPARSE_DATA ((HRESULT)0x80071128L)
#define ERROR_REPARSE_TAG_INVALID ((HRESULT)0x80071129L)
// if (MY_FACILITY_WRes != FACILITY_WIN32),
// we use FACILITY_WIN32 for COM errors:
#define E_OUTOFMEMORY ((HRESULT)0x8007000EL)
#define E_INVALIDARG ((HRESULT)0x80070057L)
#define MY_E_ERROR_NEGATIVE_SEEK ((HRESULT)0x80070083L)
/*
// we can use FACILITY_ERRNO for some COM errors, that have errno equivalents:
#define E_OUTOFMEMORY MY_HRESULT_FROM_errno_CONST_ERROR(ENOMEM)
#define E_INVALIDARG MY_HRESULT_FROM_errno_CONST_ERROR(EINVAL)
#define MY_E_ERROR_NEGATIVE_SEEK MY_HRESULT_FROM_errno_CONST_ERROR(EINVAL)
*/
#define TEXT(quote) quote
#define FILE_ATTRIBUTE_READONLY 0x0001
#define FILE_ATTRIBUTE_HIDDEN 0x0002
#define FILE_ATTRIBUTE_SYSTEM 0x0004
#define FILE_ATTRIBUTE_DIRECTORY 0x0010
#define FILE_ATTRIBUTE_ARCHIVE 0x0020
#define FILE_ATTRIBUTE_DEVICE 0x0040
#define FILE_ATTRIBUTE_NORMAL 0x0080
#define FILE_ATTRIBUTE_TEMPORARY 0x0100
#define FILE_ATTRIBUTE_SPARSE_FILE 0x0200
#define FILE_ATTRIBUTE_REPARSE_POINT 0x0400
#define FILE_ATTRIBUTE_COMPRESSED 0x0800
#define FILE_ATTRIBUTE_OFFLINE 0x1000
#define FILE_ATTRIBUTE_NOT_CONTENT_INDEXED 0x2000
#define FILE_ATTRIBUTE_ENCRYPTED 0x4000
#define FILE_ATTRIBUTE_UNIX_EXTENSION 0x8000 /* trick for Unix */
#endif
#ifndef RINOK
#define RINOK(x) { const int _result_ = (x); if (_result_ != 0) return _result_; }
#endif
#ifndef RINOK_WRes
#define RINOK_WRes(x) { const WRes _result_ = (x); if (_result_ != 0) return _result_; }
#endif
typedef unsigned char Byte;
typedef short Int16;
typedef unsigned short UInt16;
#ifdef Z7_DECL_Int32_AS_long
typedef long Int32;
typedef unsigned long UInt32;
#else
typedef int Int32;
typedef unsigned int UInt32;
#endif
#ifndef _WIN32
typedef int INT;
typedef Int32 INT32;
typedef unsigned int UINT;
typedef UInt32 UINT32;
typedef INT32 LONG; // LONG, ULONG and DWORD must be 32-bit for _WIN32 compatibility
typedef UINT32 ULONG;
#undef DWORD
typedef UINT32 DWORD;
#define VOID void
#define HRESULT LONG
typedef void *LPVOID;
// typedef void VOID;
// typedef ULONG_PTR DWORD_PTR, *PDWORD_PTR;
// gcc / clang on Unix : sizeof(long==sizeof(void*) in 32 or 64 bits)
typedef long INT_PTR;
typedef unsigned long UINT_PTR;
typedef long LONG_PTR;
typedef unsigned long DWORD_PTR;
typedef size_t SIZE_T;
#endif // _WIN32
#define MY_HRES_ERROR_INTERNAL_ERROR ((HRESULT)0x8007054FL)
#ifdef Z7_DECL_Int64_AS_long
typedef long Int64;
typedef unsigned long UInt64;
#else
#if (defined(_MSC_VER) || defined(__BORLANDC__)) && !defined(__clang__)
typedef __int64 Int64;
typedef unsigned __int64 UInt64;
#else
#if defined(__clang__) || defined(__GNUC__)
#include <stdint.h>
typedef int64_t Int64;
typedef uint64_t UInt64;
#else
typedef long long int Int64;
typedef unsigned long long int UInt64;
// #define UINT64_CONST(n) n ## ULL
#endif
#endif
#endif
#define UINT64_CONST(n) n
#ifdef Z7_DECL_SizeT_AS_unsigned_int
typedef unsigned int SizeT;
#else
typedef size_t SizeT;
#endif
/*
#if (defined(_MSC_VER) && _MSC_VER <= 1200)
typedef size_t MY_uintptr_t;
#else
#include <stdint.h>
typedef uintptr_t MY_uintptr_t;
#endif
*/
typedef int BoolInt;
/* typedef BoolInt Bool; */
#define True 1
#define False 0
#ifdef _WIN32
#define Z7_STDCALL __stdcall
#else
#define Z7_STDCALL
#endif
#ifdef _MSC_VER
#if _MSC_VER >= 1300
#define Z7_NO_INLINE __declspec(noinline)
#else
#define Z7_NO_INLINE
#endif
#define Z7_FORCE_INLINE __forceinline
#define Z7_CDECL __cdecl
#define Z7_FASTCALL __fastcall
#else // _MSC_VER
#if (defined(__GNUC__) && (__GNUC__ >= 4)) \
|| (defined(__clang__) && (__clang_major__ >= 4)) \
|| defined(__INTEL_COMPILER) \
|| defined(__xlC__)
#define Z7_NO_INLINE __attribute__((noinline))
#define Z7_FORCE_INLINE __attribute__((always_inline)) inline
#else
#define Z7_NO_INLINE
#define Z7_FORCE_INLINE
#endif
#define Z7_CDECL
#if defined(_M_IX86) \
|| defined(__i386__)
// #define Z7_FASTCALL __attribute__((fastcall))
// #define Z7_FASTCALL __attribute__((cdecl))
#define Z7_FASTCALL
#elif defined(MY_CPU_AMD64)
// #define Z7_FASTCALL __attribute__((ms_abi))
#define Z7_FASTCALL
#else
#define Z7_FASTCALL
#endif
#endif // _MSC_VER
/* The following interfaces use first parameter as pointer to structure */
// #define Z7_C_IFACE_CONST_QUAL
#define Z7_C_IFACE_CONST_QUAL const
#define Z7_C_IFACE_DECL(a) \
struct a ## _; \
typedef Z7_C_IFACE_CONST_QUAL struct a ## _ * a ## Ptr; \
typedef struct a ## _ a; \
struct a ## _
Z7_C_IFACE_DECL (IByteIn)
{
Byte (*Read)(IByteInPtr p); /* reads one byte, returns 0 in case of EOF or error */
};
#define IByteIn_Read(p) (p)->Read(p)
Z7_C_IFACE_DECL (IByteOut)
{
void (*Write)(IByteOutPtr p, Byte b);
};
#define IByteOut_Write(p, b) (p)->Write(p, b)
Z7_C_IFACE_DECL (ISeqInStream)
{
SRes (*Read)(ISeqInStreamPtr p, void *buf, size_t *size);
/* if (input(*size) != 0 && output(*size) == 0) means end_of_stream.
(output(*size) < input(*size)) is allowed */
};
#define ISeqInStream_Read(p, buf, size) (p)->Read(p, buf, size)
/* try to read as much as avail in stream and limited by (*processedSize) */
SRes SeqInStream_ReadMax(ISeqInStreamPtr stream, void *buf, size_t *processedSize);
/* it can return SZ_ERROR_INPUT_EOF */
// SRes SeqInStream_Read(ISeqInStreamPtr stream, void *buf, size_t size);
// SRes SeqInStream_Read2(ISeqInStreamPtr stream, void *buf, size_t size, SRes errorType);
SRes SeqInStream_ReadByte(ISeqInStreamPtr stream, Byte *buf);
Z7_C_IFACE_DECL (ISeqOutStream)
{
size_t (*Write)(ISeqOutStreamPtr p, const void *buf, size_t size);
/* Returns: result - the number of actually written bytes.
(result < size) means error */
};
#define ISeqOutStream_Write(p, buf, size) (p)->Write(p, buf, size)
typedef enum
{
SZ_SEEK_SET = 0,
SZ_SEEK_CUR = 1,
SZ_SEEK_END = 2
} ESzSeek;
Z7_C_IFACE_DECL (ISeekInStream)
{
SRes (*Read)(ISeekInStreamPtr p, void *buf, size_t *size); /* same as ISeqInStream::Read */
SRes (*Seek)(ISeekInStreamPtr p, Int64 *pos, ESzSeek origin);
};
#define ISeekInStream_Read(p, buf, size) (p)->Read(p, buf, size)
#define ISeekInStream_Seek(p, pos, origin) (p)->Seek(p, pos, origin)
Z7_C_IFACE_DECL (ILookInStream)
{
SRes (*Look)(ILookInStreamPtr p, const void **buf, size_t *size);
/* if (input(*size) != 0 && output(*size) == 0) means end_of_stream.
(output(*size) > input(*size)) is not allowed
(output(*size) < input(*size)) is allowed */
SRes (*Skip)(ILookInStreamPtr p, size_t offset);
/* offset must be <= output(*size) of Look */
SRes (*Read)(ILookInStreamPtr p, void *buf, size_t *size);
/* reads directly (without buffer). It's same as ISeqInStream::Read */
SRes (*Seek)(ILookInStreamPtr p, Int64 *pos, ESzSeek origin);
};
#define ILookInStream_Look(p, buf, size) (p)->Look(p, buf, size)
#define ILookInStream_Skip(p, offset) (p)->Skip(p, offset)
#define ILookInStream_Read(p, buf, size) (p)->Read(p, buf, size)
#define ILookInStream_Seek(p, pos, origin) (p)->Seek(p, pos, origin)
SRes LookInStream_LookRead(ILookInStreamPtr stream, void *buf, size_t *size);
SRes LookInStream_SeekTo(ILookInStreamPtr stream, UInt64 offset);
/* reads via ILookInStream::Read */
SRes LookInStream_Read2(ILookInStreamPtr stream, void *buf, size_t size, SRes errorType);
SRes LookInStream_Read(ILookInStreamPtr stream, void *buf, size_t size);
typedef struct
{
ILookInStream vt;
ISeekInStreamPtr realStream;
size_t pos;
size_t size; /* it's data size */
/* the following variables must be set outside */
Byte *buf;
size_t bufSize;
} CLookToRead2;
void LookToRead2_CreateVTable(CLookToRead2 *p, int lookahead);
#define LookToRead2_INIT(p) { (p)->pos = (p)->size = 0; }
typedef struct
{
ISeqInStream vt;
ILookInStreamPtr realStream;
} CSecToLook;
void SecToLook_CreateVTable(CSecToLook *p);
typedef struct
{
ISeqInStream vt;
ILookInStreamPtr realStream;
} CSecToRead;
void SecToRead_CreateVTable(CSecToRead *p);
Z7_C_IFACE_DECL (ICompressProgress)
{
SRes (*Progress)(ICompressProgressPtr p, UInt64 inSize, UInt64 outSize);
/* Returns: result. (result != SZ_OK) means break.
Value (UInt64)(Int64)-1 for size means unknown value. */
};
#define ICompressProgress_Progress(p, inSize, outSize) (p)->Progress(p, inSize, outSize)
typedef struct ISzAlloc ISzAlloc;
typedef const ISzAlloc * ISzAllocPtr;
struct ISzAlloc
{
void *(*Alloc)(ISzAllocPtr p, size_t size);
void (*Free)(ISzAllocPtr p, void *address); /* address can be 0 */
};
#define ISzAlloc_Alloc(p, size) (p)->Alloc(p, size)
#define ISzAlloc_Free(p, a) (p)->Free(p, a)
/* deprecated */
#define IAlloc_Alloc(p, size) ISzAlloc_Alloc(p, size)
#define IAlloc_Free(p, a) ISzAlloc_Free(p, a)
#ifndef MY_offsetof
#ifdef offsetof
#define MY_offsetof(type, m) offsetof(type, m)
/*
#define MY_offsetof(type, m) FIELD_OFFSET(type, m)
*/
#else
#define MY_offsetof(type, m) ((size_t)&(((type *)0)->m))
#endif
#endif
#ifndef Z7_container_of
/*
#define Z7_container_of(ptr, type, m) container_of(ptr, type, m)
#define Z7_container_of(ptr, type, m) CONTAINING_RECORD(ptr, type, m)
#define Z7_container_of(ptr, type, m) ((type *)((char *)(ptr) - offsetof(type, m)))
#define Z7_container_of(ptr, type, m) (&((type *)0)->m == (ptr), ((type *)(((char *)(ptr)) - MY_offsetof(type, m))))
*/
/*
GCC shows warning: "perhaps the 'offsetof' macro was used incorrectly"
GCC 3.4.4 : classes with constructor
GCC 4.8.1 : classes with non-public variable members"
*/
#define Z7_container_of(ptr, type, m) \
((type *)(void *)((char *)(void *) \
(1 ? (ptr) : &((type *)NULL)->m) - MY_offsetof(type, m)))
#define Z7_container_of_CONST(ptr, type, m) \
((const type *)(const void *)((const char *)(const void *) \
(1 ? (ptr) : &((type *)NULL)->m) - MY_offsetof(type, m)))
/*
#define Z7_container_of_NON_CONST_FROM_CONST(ptr, type, m) \
((type *)(void *)(const void *)((const char *)(const void *) \
(1 ? (ptr) : &((type *)NULL)->m) - MY_offsetof(type, m)))
*/
#endif
#define Z7_CONTAINER_FROM_VTBL_SIMPLE(ptr, type, m) ((type *)(void *)(ptr))
// #define Z7_CONTAINER_FROM_VTBL(ptr, type, m) Z7_CONTAINER_FROM_VTBL_SIMPLE(ptr, type, m)
#define Z7_CONTAINER_FROM_VTBL(ptr, type, m) Z7_container_of(ptr, type, m)
// #define Z7_CONTAINER_FROM_VTBL(ptr, type, m) Z7_container_of_NON_CONST_FROM_CONST(ptr, type, m)
#define Z7_CONTAINER_FROM_VTBL_CONST(ptr, type, m) Z7_container_of_CONST(ptr, type, m)
#define Z7_CONTAINER_FROM_VTBL_CLS(ptr, type, m) Z7_CONTAINER_FROM_VTBL_SIMPLE(ptr, type, m)
/*
#define Z7_CONTAINER_FROM_VTBL_CLS(ptr, type, m) Z7_CONTAINER_FROM_VTBL(ptr, type, m)
*/
#if defined (__clang__) || defined(__GNUC__)
#define Z7_DIAGNOSTIC_IGNORE_BEGIN_CAST_QUAL \
_Pragma("GCC diagnostic push") \
_Pragma("GCC diagnostic ignored \"-Wcast-qual\"")
#define Z7_DIAGNOSTIC_IGNORE_END_CAST_QUAL \
_Pragma("GCC diagnostic pop")
#else
#define Z7_DIAGNOSTIC_IGNORE_BEGIN_CAST_QUAL
#define Z7_DIAGNOSTIC_IGNORE_END_CAST_QUAL
#endif
#define Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR(ptr, type, m, p) \
Z7_DIAGNOSTIC_IGNORE_BEGIN_CAST_QUAL \
type *p = Z7_CONTAINER_FROM_VTBL(ptr, type, m); \
Z7_DIAGNOSTIC_IGNORE_END_CAST_QUAL
#define Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR_pp_vt_p(type) \
Z7_CONTAINER_FROM_VTBL_TO_DECL_VAR(pp, type, vt, p)
// #define ZIP7_DECLARE_HANDLE(name) typedef void *name;
#define Z7_DECLARE_HANDLE(name) struct name##_dummy{int unused;}; typedef struct name##_dummy *name;
#define Z7_memset_0_ARRAY(a) memset((a), 0, sizeof(a))
#define Z7_memset_0_VAR(a) memset(&(a), 0, sizeof(a))
#ifndef Z7_ARRAY_SIZE
#define Z7_ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
#endif
#ifdef _WIN32
#define CHAR_PATH_SEPARATOR '\\'
#define WCHAR_PATH_SEPARATOR L'\\'
#define STRING_PATH_SEPARATOR "\\"
#define WSTRING_PATH_SEPARATOR L"\\"
#else
#define CHAR_PATH_SEPARATOR '/'
#define WCHAR_PATH_SEPARATOR L'/'
#define STRING_PATH_SEPARATOR "/"
#define WSTRING_PATH_SEPARATOR L"/"
#endif
#define k_PropVar_TimePrec_0 0
#define k_PropVar_TimePrec_Unix 1
#define k_PropVar_TimePrec_DOS 2
#define k_PropVar_TimePrec_HighPrec 3
#define k_PropVar_TimePrec_Base 16
#define k_PropVar_TimePrec_100ns (k_PropVar_TimePrec_Base + 7)
#define k_PropVar_TimePrec_1ns (k_PropVar_TimePrec_Base + 9)
EXTERN_C_END
#endif
/*
#ifndef Z7_ST
#ifdef _7ZIP_ST
#define Z7_ST
#endif
#endif
*/
third_party/7z/7zVersion.h
Vendored
+27
View File
@@ -0,0 +1,27 @@
#define MY_VER_MAJOR 26
#define MY_VER_MINOR 3
#define MY_VER_BUILD 0
#define MY_VERSION_NUMBERS "26.03"
#define MY_VERSION MY_VERSION_NUMBERS
#ifdef MY_CPU_NAME
#define MY_VERSION_CPU MY_VERSION " (" MY_CPU_NAME ")"
#else
#define MY_VERSION_CPU MY_VERSION
#endif
#define MY_DATE "2026-09-03"
#undef MY_COPYRIGHT
#undef MY_VERSION_COPYRIGHT_DATE
#define MY_AUTHOR_NAME "Igor Pavlov"
#define MY_COPYRIGHT_PD "Igor Pavlov : Public domain"
#define MY_COPYRIGHT_CR "Copyright (c) 1999-2026 Igor Pavlov"
#ifdef USE_COPYRIGHT_CR
#define MY_COPYRIGHT MY_COPYRIGHT_CR
#else
#define MY_COPYRIGHT MY_COPYRIGHT_PD
#endif
#define MY_COPYRIGHT_DATE MY_COPYRIGHT " : " MY_DATE
#define MY_VERSION_COPYRIGHT_DATE MY_VERSION_CPU " : " MY_COPYRIGHT " : " MY_DATE
third_party/7z/7zWindows.h
Vendored
+107
View File
@@ -0,0 +1,107 @@
/* 7zWindows.h -- Windows.h and related code
Igor Pavlov : Public domain */
#ifndef ZIP7_INC_7Z_WINDOWS_H
#define ZIP7_INC_7Z_WINDOWS_H
#ifdef _WIN32
#if defined(_MSC_VER) && _MSC_VER >= 1950 && !defined(__clang__) // VS2026
// <Windows.h> and some another windows files need that option
// VS2026: wtypesbase.h: warning C4865: 'tagCLSCTX': the underlying type will change from 'int' to 'unsigned int' when '/Zc:enumTypes' is specified on the command line
#pragma warning(disable : 4865)
#endif
#if defined(__clang__)
# pragma clang diagnostic push
#endif
#if defined(_MSC_VER)
#pragma warning(push)
#pragma warning(disable : 4668) // '_WIN32_WINNT' is not defined as a preprocessor macro, replacing with '0' for '#if/#elif'
#if _MSC_VER == 1900
// for old kit10 versions
// #pragma warning(disable : 4255) // winuser.h(13979): warning C4255: 'GetThreadDpiAwarenessContext':
#endif
// win10 Windows Kit:
#endif // _MSC_VER
#if defined(_MSC_VER) && _MSC_VER <= 1200 && !defined(_WIN64)
// for msvc6 without sdk2003
#define RPC_NO_WINDOWS_H
#endif
#if defined(__MINGW32__) || defined(__MINGW64__)
// #if defined(__GNUC__) && !defined(__clang__)
#include <windows.h>
#else
#include <Windows.h>
#endif
// #include <basetsd.h>
// #include <wtypes.h>
// but if precompiled with clang-cl then we need
// #include <windows.h>
#if defined(_MSC_VER)
#pragma warning(pop)
#endif
#if defined(__clang__)
# pragma clang diagnostic pop
#endif
#if defined(_MSC_VER) && _MSC_VER <= 1200 && !defined(_WIN64)
#ifndef _W64
typedef long LONG_PTR, *PLONG_PTR;
typedef unsigned long ULONG_PTR, *PULONG_PTR;
typedef ULONG_PTR DWORD_PTR, *PDWORD_PTR;
#define Z7_OLD_WIN_SDK
#endif // _W64
#endif // _MSC_VER == 1200
#ifdef Z7_OLD_WIN_SDK
#ifndef INVALID_FILE_ATTRIBUTES
#define INVALID_FILE_ATTRIBUTES ((DWORD)-1)
#endif
#ifndef INVALID_SET_FILE_POINTER
#define INVALID_SET_FILE_POINTER ((DWORD)-1)
#endif
#ifndef FILE_SPECIAL_ACCESS
#define FILE_SPECIAL_ACCESS (FILE_ANY_ACCESS)
#endif
// ShlObj.h:
// #define BIF_NEWDIALOGSTYLE 0x0040
#pragma warning(disable : 4201)
// #pragma warning(disable : 4115)
#undef VARIANT_TRUE
#define VARIANT_TRUE ((VARIANT_BOOL)-1)
#endif
#endif // Z7_OLD_WIN_SDK
#ifdef UNDER_CE
#undef VARIANT_TRUE
#define VARIANT_TRUE ((VARIANT_BOOL)-1)
#endif
#if defined(_MSC_VER)
#if _MSC_VER >= 1400 && _MSC_VER <= 1600
// BaseTsd.h(148) : 'HandleToULong' : unreferenced inline function has been removed
// string.h
// #pragma warning(disable : 4514)
#endif
#endif
/* #include "7zTypes.h" */
#endif
third_party/7z/7zip_gcc_c.mak
Vendored
+360
View File
@@ -0,0 +1,360 @@
MY_ARCH_2 = $(MY_ARCH)
MY_ASM = jwasm
MY_ASM = asmc
ifndef RC
#RC=windres.exe --target=pe-x86-64
#RC=windres.exe -F pe-i386
RC=windres.exe
endif
PROGPATH = $(O)/$(PROG)
PROGPATH_STATIC = $(O)/$(PROG)s
ifneq ($(CC), xlc)
CFLAGS_WARN_WALL = -Wall -Werror -Wextra
endif
# for object file
CFLAGS_BASE_LIST = -c
# for ASM file
# CFLAGS_BASE_LIST = -S
FLAGS_FLTO = -flto
FLAGS_FLTO =
CFLAGS_BASE = $(MY_ARCH_2) -O2 $(CFLAGS_BASE_LIST) $(CFLAGS_WARN_WALL) $(CFLAGS_WARN) \
-DNDEBUG -D_REENTRANT -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
ifdef SystemDrive
IS_MINGW = 1
else
ifdef SYSTEMDRIVE
# ifdef OS
IS_MINGW = 1
endif
endif
ifdef IS_MINGW
LDFLAGS_STATIC_2 = -static
else
ifndef DEF_FILE
ifndef IS_NOT_STANDALONE
ifndef MY_DYNAMIC_LINK
ifneq ($(CC), clang)
LDFLAGS_STATIC_2 =
# -static
# -static-libstdc++ -static-libgcc
endif
endif
endif
endif
endif
LDFLAGS_STATIC = -DNDEBUG $(LDFLAGS_STATIC_2)
ifdef DEF_FILE
ifdef IS_MINGW
SHARED_EXT=.dll
LDFLAGS = -shared -DEF $(DEF_FILE) $(LDFLAGS_STATIC)
else
SHARED_EXT=.so
LDFLAGS = -shared -fPIC $(LDFLAGS_STATIC)
CC_SHARED=-fPIC
endif
else
LDFLAGS = $(LDFLAGS_STATIC)
# -s is not required for clang, do we need it for GGC ???
# -s
#-static -static-libgcc -static-libstdc++
ifdef IS_MINGW
SHARED_EXT=.exe
else
SHARED_EXT=
endif
endif
PROGPATH = $(O)/$(PROG)$(SHARED_EXT)
PROGPATH_STATIC = $(O)/$(PROG)s$(SHARED_EXT)
ifndef O
O=_o
endif
ifdef IS_MINGW
ifdef MSYSTEM
RM = rm -f
MY_MKDIR=mkdir -p
DEL_OBJ_EXE = -$(RM) $(PROGPATH) $(PROGPATH_STATIC) $(OBJS)
else
RM = del
MY_MKDIR=mkdir
DEL_OBJ_EXE = -$(RM) $(O)\*.o $(O)\$(PROG).exe $(O)\$(PROG).dll
endif
LIB2 = -lole32 -loleaut32 -luuid -ladvapi32 -luser32 -lshell32
CFLAGS_EXTRA = -DUNICODE -D_UNICODE
# -Wno-delete-non-virtual-dtor
else
RM = rm -f
MY_MKDIR=mkdir -p
# CFLAGS_BASE := $(CFLAGS_BASE) -DZ7_ST
# CFLAGS_EXTRA = -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
# LOCAL_LIBS=-lpthread
# LOCAL_LIBS_DLL=$(LOCAL_LIBS) -ldl
LIB2 = -lpthread -ldl
DEL_OBJ_EXE = -$(RM) $(PROGPATH) $(PROGPATH_STATIC) $(OBJS)
endif
ifdef IS_X64
AFLAGS_ABI = -elf64 -DABI_LINUX
else
AFLAGS_ABI = -elf -DABI_LINUX -DABI_CDECL
# -DABI_CDECL
# -DABI_LINUX
# -DABI_CDECL
endif
AFLAGS = $(AFLAGS_ABI) -Fo$(O)/
C_WARN_FLAGS =
CFLAGS = $(LOCAL_FLAGS) $(CFLAGS_BASE2) $(CFLAGS_BASE) $(CFLAGS_EXTRA) $(C_WARN_FLAGS) $(FLAGS_FLTO) $(CC_SHARED) -o $@
STATIC_TARGET=
ifdef COMPL_STATIC
STATIC_TARGET=$(PROGPATH_STATIC)
endif
all: $(O) $(PROGPATH) $(STATIC_TARGET)
$(O):
$(MY_MKDIR) $(O)
ifneq ($(CC), $(CROSS_COMPILE)clang)
LFLAGS_STRIP = -s
endif
LFLAGS_ALL = $(LFLAGS_STRIP) $(MY_ARCH_2) $(LDFLAGS) $(FLAGS_FLTO) $(LD_arch) $(OBJS) $(MY_LIBS) $(LIB2)
$(PROGPATH): $(OBJS)
$(CC) -o $(PROGPATH) $(LFLAGS_ALL)
$(PROGPATH_STATIC): $(OBJS)
$(CC) -static -o $(PROGPATH_STATIC) $(LFLAGS_ALL)
ifndef NO_DEFAULT_RES
# old mingw without -FO
# windres.exe $(RFLAGS) resource.rc $O/resource.o
$O/resource.o: resource.rc
$(RC) $(RFLAGS) resource.rc $(O)/resource.o
endif
# windres.exe $(RFLAGS) resource.rc $(O)\resource.o
# windres.exe $(RFLAGS) resource.rc -FO $(O)/resource.o
# $(RC) $(RFLAGS) resource.rc -FO $(O)/resource.o
$O/7zAlloc.o: ../../../C/7zAlloc.c
$(CC) $(CFLAGS) $<
$O/7zArcIn.o: ../../../C/7zArcIn.c
$(CC) $(CFLAGS) $<
$O/7zBuf.o: ../../../C/7zBuf.c
$(CC) $(CFLAGS) $<
$O/7zBuf2.o: ../../../C/7zBuf2.c
$(CC) $(CFLAGS) $<
$O/7zCrc.o: ../../../C/7zCrc.c
$(CC) $(CFLAGS) $<
$O/7zDec.o: ../../../C/7zDec.c
$(CC) $(CFLAGS) $<
$O/7zFile.o: ../../../C/7zFile.c
$(CC) $(CFLAGS) $<
$O/7zStream.o: ../../../C/7zStream.c
$(CC) $(CFLAGS) $<
$O/Aes.o: ../../../C/Aes.c
$(CC) $(CFLAGS) $<
$O/Alloc.o: ../../../C/Alloc.c
$(CC) $(CFLAGS) $<
$O/Bcj2.o: ../../../C/Bcj2.c
$(CC) $(CFLAGS) $<
$O/Bcj2Enc.o: ../../../C/Bcj2Enc.c
$(CC) $(CFLAGS) $<
$O/Blake2s.o: ../../../C/Blake2s.c
$(CC) $(CFLAGS) $<
$O/Bra.o: ../../../C/Bra.c
$(CC) $(CFLAGS) $<
$O/Bra86.o: ../../../C/Bra86.c
$(CC) $(CFLAGS) $<
$O/BraIA64.o: ../../../C/BraIA64.c
$(CC) $(CFLAGS) $<
$O/BwtSort.o: ../../../C/BwtSort.c
$(CC) $(CFLAGS) $<
$O/CpuArch.o: ../../../C/CpuArch.c
$(CC) $(CFLAGS) $<
$O/Delta.o: ../../../C/Delta.c
$(CC) $(CFLAGS) $<
$O/DllSecur.o: ../../../C/DllSecur.c
$(CC) $(CFLAGS) $<
$O/HuffEnc.o: ../../../C/HuffEnc.c
$(CC) $(CFLAGS) $<
$O/LzFind.o: ../../../C/LzFind.c
$(CC) $(CFLAGS) $<
# ifdef MT_FILES
$O/LzFindMt.o: ../../../C/LzFindMt.c
$(CC) $(CFLAGS) $<
$O/LzFindOpt.o: ../../../C/LzFindOpt.c
$(CC) $(CFLAGS) $<
$O/Threads.o: ../../../C/Threads.c
$(CC) $(CFLAGS) $<
# endif
$O/LzmaEnc.o: ../../../C/LzmaEnc.c
$(CC) $(CFLAGS) $<
$O/Lzma86Dec.o: ../../../C/Lzma86Dec.c
$(CC) $(CFLAGS) $<
$O/Lzma86Enc.o: ../../../C/Lzma86Enc.c
$(CC) $(CFLAGS) $<
$O/Lzma2Dec.o: ../../../C/Lzma2Dec.c
$(CC) $(CFLAGS) $<
$O/Lzma2DecMt.o: ../../../C/Lzma2DecMt.c
$(CC) $(CFLAGS) $<
$O/Lzma2Enc.o: ../../../C/Lzma2Enc.c
$(CC) $(CFLAGS) $<
$O/LzmaLib.o: ../../../C/LzmaLib.c
$(CC) $(CFLAGS) $<
$O/MtCoder.o: ../../../C/MtCoder.c
$(CC) $(CFLAGS) $<
$O/MtDec.o: ../../../C/MtDec.c
$(CC) $(CFLAGS) $<
$O/Ppmd7.o: ../../../C/Ppmd7.c
$(CC) $(CFLAGS) $<
$O/Ppmd7aDec.o: ../../../C/Ppmd7aDec.c
$(CC) $(CFLAGS) $<
$O/Ppmd7Dec.o: ../../../C/Ppmd7Dec.c
$(CC) $(CFLAGS) $<
$O/Ppmd7Enc.o: ../../../C/Ppmd7Enc.c
$(CC) $(CFLAGS) $<
$O/Ppmd8.o: ../../../C/Ppmd8.c
$(CC) $(CFLAGS) $<
$O/Ppmd8Dec.o: ../../../C/Ppmd8Dec.c
$(CC) $(CFLAGS) $<
$O/Ppmd8Enc.o: ../../../C/Ppmd8Enc.c
$(CC) $(CFLAGS) $<
$O/Sha1.o: ../../../C/Sha1.c
$(CC) $(CFLAGS) $<
$O/Sha256.o: ../../../C/Sha256.c
$(CC) $(CFLAGS) $<
$O/Sort.o: ../../../C/Sort.c
$(CC) $(CFLAGS) $<
$O/SwapBytes.o: ../../../C/SwapBytes.c
$(CC) $(CFLAGS) $<
$O/Xz.o: ../../../C/Xz.c
$(CC) $(CFLAGS) $<
$O/XzCrc64.o: ../../../C/XzCrc64.c
$(CC) $(CFLAGS) $<
$O/XzDec.o: ../../../C/XzDec.c
$(CC) $(CFLAGS) $<
$O/XzEnc.o: ../../../C/XzEnc.c
$(CC) $(CFLAGS) $<
$O/XzIn.o: ../../../C/XzIn.c
$(CC) $(CFLAGS) $<
ifdef USE_ASM
ifdef IS_X64
USE_X86_ASM=1
else
ifdef IS_X86
USE_X86_ASM=1
endif
endif
endif
ifdef USE_X86_ASM
$O/7zCrcOpt.o: ../../../Asm/x86/7zCrcOpt.asm
$(MY_ASM) $(AFLAGS) $<
$O/XzCrc64Opt.o: ../../../Asm/x86/XzCrc64Opt.asm
$(MY_ASM) $(AFLAGS) $<
$O/AesOpt.o: ../../../Asm/x86/AesOpt.asm
$(MY_ASM) $(AFLAGS) $<
$O/Sha1Opt.o: ../../../Asm/x86/Sha1Opt.asm
$(MY_ASM) $(AFLAGS) $<
$O/Sha256Opt.o: ../../../Asm/x86/Sha256Opt.asm
$(MY_ASM) $(AFLAGS) $<
else
$O/7zCrcOpt.o: ../../7zCrcOpt.c
$(CC) $(CFLAGS) $<
$O/XzCrc64Opt.o: ../../XzCrc64Opt.c
$(CC) $(CFLAGS) $<
$O/Sha1Opt.o: ../../Sha1Opt.c
$(CC) $(CFLAGS) $<
$O/Sha256Opt.o: ../../Sha256Opt.c
$(CC) $(CFLAGS) $<
$O/AesOpt.o: ../../AesOpt.c
$(CC) $(CFLAGS) $<
endif
ifdef USE_LZMA_DEC_ASM
ifdef IS_X64
$O/LzmaDecOpt.o: ../../../Asm/x86/LzmaDecOpt.asm
$(MY_ASM) $(AFLAGS) $<
endif
ifdef IS_ARM64
$O/LzmaDecOpt.o: ../../../Asm/arm64/LzmaDecOpt.S ../../../Asm/arm64/7zAsm.S
$(CC) $(CFLAGS) $(ASM_FLAGS) $<
endif
$O/LzmaDec.o: ../../LzmaDec.c
$(CC) $(CFLAGS) -DZ7_LZMA_DEC_OPT $<
else
$O/LzmaDec.o: ../../LzmaDec.c
$(CC) $(CFLAGS) $<
endif
$O/7zMain.o: ../../../C/Util/7z/7zMain.c
$(CC) $(CFLAGS) $<
$O/7zipInstall.o: ../../../C/Util/7zipInstall/7zipInstall.c
$(CC) $(CFLAGS) $<
$O/7zipUninstall.o: ../../../C/Util/7zipUninstall/7zipUninstall.c
$(CC) $(CFLAGS) $<
$O/LzmaUtil.o: ../../../C/Util/Lzma/LzmaUtil.c
$(CC) $(CFLAGS) $<
$O/XzUtil.o: ../../../C/Util/Xz/XzUtil.c
$(CC) $(CFLAGS) $<
clean:
-$(DEL_OBJ_EXE)
third_party/7z/Aes.c
Vendored
+429
View File
@@ -0,0 +1,429 @@
/* Aes.c -- AES encryption / decryption
2024-03-01 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "CpuArch.h"
#include "Aes.h"
AES_CODE_FUNC g_AesCbc_Decode;
#ifndef Z7_SFX
AES_CODE_FUNC g_AesCbc_Encode;
AES_CODE_FUNC g_AesCtr_Code;
UInt32 g_Aes_SupportedFunctions_Flags;
#endif
MY_ALIGN(64)
static UInt32 T[256 * 4];
MY_ALIGN(64)
static const Byte Sbox[256] = {
0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16};
MY_ALIGN(64)
static UInt32 D[256 * 4];
MY_ALIGN(64)
static Byte InvS[256];
#define xtime(x) ((((x) << 1) ^ (((x) & 0x80) != 0 ? 0x1B : 0)) & 0xFF)
#define Ui32(a0, a1, a2, a3) ((UInt32)(a0) | ((UInt32)(a1) << 8) | ((UInt32)(a2) << 16) | ((UInt32)(a3) << 24))
#define gb0(x) ( (x) & 0xFF)
#define gb1(x) (((x) >> ( 8)) & 0xFF)
#define gb2(x) (((x) >> (16)) & 0xFF)
#define gb3(x) (((x) >> (24)))
#define gb(n, x) gb ## n(x)
#define TT(x) (T + (x << 8))
#define DD(x) (D + (x << 8))
// #define Z7_SHOW_AES_STATUS
#ifdef MY_CPU_X86_OR_AMD64
#if defined(__INTEL_COMPILER)
#if (__INTEL_COMPILER >= 1110)
#define USE_HW_AES
#if (__INTEL_COMPILER >= 1900)
#define USE_HW_VAES
#endif
#endif
#elif defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40400)
#define USE_HW_AES
#if defined(__clang__) && (__clang_major__ >= 8) \
|| defined(__GNUC__) && (__GNUC__ >= 8)
#define USE_HW_VAES
#endif
#elif defined(_MSC_VER)
#define USE_HW_AES
#define USE_HW_VAES
#endif
#elif defined(MY_CPU_ARM_OR_ARM64) && defined(MY_CPU_LE)
#if defined(__ARM_FEATURE_AES) \
|| defined(__ARM_FEATURE_CRYPTO)
#define USE_HW_AES
#else
#if defined(MY_CPU_ARM64) \
|| defined(__ARM_ARCH) && (__ARM_ARCH >= 4) \
|| defined(Z7_MSC_VER_ORIGINAL)
#if defined(__ARM_FP) && \
( defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(__GNUC__) && (__GNUC__ >= 6) \
) \
|| defined(Z7_MSC_VER_ORIGINAL) && (_MSC_VER >= 1910)
#if defined(MY_CPU_ARM64) \
|| !defined(Z7_CLANG_VERSION) \
|| defined(__ARM_NEON) && \
(Z7_CLANG_VERSION < 170000 || \
Z7_CLANG_VERSION > 170001)
#define USE_HW_AES
#endif
#endif
#endif
#endif
#endif
#ifdef USE_HW_AES
// #pragma message("=== Aes.c USE_HW_AES === ")
#ifdef Z7_SHOW_AES_STATUS
#include <stdio.h>
#define PRF(x) x
#else
#define PRF(x)
#endif
#endif
void AesGenTables(void)
{
unsigned i;
for (i = 0; i < 256; i++)
InvS[Sbox[i]] = (Byte)i;
for (i = 0; i < 256; i++)
{
{
const UInt32 a1 = Sbox[i];
const UInt32 a2 = xtime(a1);
const UInt32 a3 = a2 ^ a1;
TT(0)[i] = Ui32(a2, a1, a1, a3);
TT(1)[i] = Ui32(a3, a2, a1, a1);
TT(2)[i] = Ui32(a1, a3, a2, a1);
TT(3)[i] = Ui32(a1, a1, a3, a2);
}
{
const UInt32 a1 = InvS[i];
const UInt32 a2 = xtime(a1);
const UInt32 a4 = xtime(a2);
const UInt32 a8 = xtime(a4);
const UInt32 a9 = a8 ^ a1;
const UInt32 aB = a8 ^ a2 ^ a1;
const UInt32 aD = a8 ^ a4 ^ a1;
const UInt32 aE = a8 ^ a4 ^ a2;
DD(0)[i] = Ui32(aE, a9, aD, aB);
DD(1)[i] = Ui32(aB, aE, a9, aD);
DD(2)[i] = Ui32(aD, aB, aE, a9);
DD(3)[i] = Ui32(a9, aD, aB, aE);
}
}
{
AES_CODE_FUNC d = AesCbc_Decode;
#ifndef Z7_SFX
AES_CODE_FUNC e = AesCbc_Encode;
AES_CODE_FUNC c = AesCtr_Code;
UInt32 flags = 0;
#endif
#ifdef USE_HW_AES
if (CPU_IsSupported_AES())
{
// #pragma message ("AES HW")
PRF(printf("\n===AES HW\n"));
d = AesCbc_Decode_HW;
#ifndef Z7_SFX
e = AesCbc_Encode_HW;
c = AesCtr_Code_HW;
flags = k_Aes_SupportedFunctions_HW;
#endif
#ifdef MY_CPU_X86_OR_AMD64
#ifdef USE_HW_VAES
if (CPU_IsSupported_VAES_AVX2())
{
PRF(printf("\n===vaes avx2\n"));
d = AesCbc_Decode_HW_256;
#ifndef Z7_SFX
c = AesCtr_Code_HW_256;
flags |= k_Aes_SupportedFunctions_HW_256;
#endif
}
#endif
#endif
}
#endif
g_AesCbc_Decode = d;
#ifndef Z7_SFX
g_AesCbc_Encode = e;
g_AesCtr_Code = c;
g_Aes_SupportedFunctions_Flags = flags;
#endif
}
}
#define HT(i, x, s) TT(x)[gb(x, s[(i + x) & 3])]
#define HT4(m, i, s, p) m[i] = \
HT(i, 0, s) ^ \
HT(i, 1, s) ^ \
HT(i, 2, s) ^ \
HT(i, 3, s) ^ w[p + i]
#define HT16(m, s, p) \
HT4(m, 0, s, p); \
HT4(m, 1, s, p); \
HT4(m, 2, s, p); \
HT4(m, 3, s, p); \
#define FT(i, x) Sbox[gb(x, m[(i + x) & 3])]
#define FT4(i) dest[i] = Ui32(FT(i, 0), FT(i, 1), FT(i, 2), FT(i, 3)) ^ w[i];
#define HD(i, x, s) DD(x)[gb(x, s[(i - x) & 3])]
#define HD4(m, i, s, p) m[i] = \
HD(i, 0, s) ^ \
HD(i, 1, s) ^ \
HD(i, 2, s) ^ \
HD(i, 3, s) ^ w[p + i];
#define HD16(m, s, p) \
HD4(m, 0, s, p); \
HD4(m, 1, s, p); \
HD4(m, 2, s, p); \
HD4(m, 3, s, p); \
#define FD(i, x) InvS[gb(x, m[(i - x) & 3])]
#define FD4(i) dest[i] = Ui32(FD(i, 0), FD(i, 1), FD(i, 2), FD(i, 3)) ^ w[i];
void Z7_FASTCALL Aes_SetKey_Enc(UInt32 *w, const Byte *key, unsigned keySize)
{
unsigned i, m;
const UInt32 *wLim;
UInt32 t;
UInt32 rcon = 1;
keySize /= 4;
w[0] = ((UInt32)keySize / 2) + 3;
w += 4;
for (i = 0; i < keySize; i++, key += 4)
w[i] = GetUi32(key);
t = w[(size_t)keySize - 1];
wLim = w + (size_t)keySize * 3 + 28;
m = 0;
do
{
if (m == 0)
{
t = Ui32(Sbox[gb1(t)] ^ rcon, Sbox[gb2(t)], Sbox[gb3(t)], Sbox[gb0(t)]);
rcon <<= 1;
if (rcon & 0x100)
rcon = 0x1b;
m = keySize;
}
else if (m == 4 && keySize > 6)
t = Ui32(Sbox[gb0(t)], Sbox[gb1(t)], Sbox[gb2(t)], Sbox[gb3(t)]);
m--;
t ^= w[0];
w[keySize] = t;
}
while (++w != wLim);
}
void Z7_FASTCALL Aes_SetKey_Dec(UInt32 *w, const Byte *key, unsigned keySize)
{
unsigned i, num;
Aes_SetKey_Enc(w, key, keySize);
num = keySize + 20;
w += 8;
for (i = 0; i < num; i++)
{
UInt32 r = w[i];
w[i] =
DD(0)[Sbox[gb0(r)]] ^
DD(1)[Sbox[gb1(r)]] ^
DD(2)[Sbox[gb2(r)]] ^
DD(3)[Sbox[gb3(r)]];
}
}
/* Aes_Encode and Aes_Decode functions work with little-endian words.
src and dest are pointers to 4 UInt32 words.
src and dest can point to same block */
// Z7_FORCE_INLINE
static void Aes_Encode(const UInt32 *w, UInt32 *dest, const UInt32 *src)
{
UInt32 s[4];
UInt32 m[4];
UInt32 numRounds2 = w[0];
w += 4;
s[0] = src[0] ^ w[0];
s[1] = src[1] ^ w[1];
s[2] = src[2] ^ w[2];
s[3] = src[3] ^ w[3];
w += 4;
for (;;)
{
HT16(m, s, 0)
if (--numRounds2 == 0)
break;
HT16(s, m, 4)
w += 8;
}
w += 4;
FT4(0)
FT4(1)
FT4(2)
FT4(3)
}
Z7_FORCE_INLINE
static void Aes_Decode(const UInt32 *w, UInt32 *dest, const UInt32 *src)
{
UInt32 s[4];
UInt32 m[4];
UInt32 numRounds2 = w[0];
w += 4 + numRounds2 * 8;
s[0] = src[0] ^ w[0];
s[1] = src[1] ^ w[1];
s[2] = src[2] ^ w[2];
s[3] = src[3] ^ w[3];
for (;;)
{
w -= 8;
HD16(m, s, 4)
if (--numRounds2 == 0)
break;
HD16(s, m, 0)
}
FD4(0)
FD4(1)
FD4(2)
FD4(3)
}
void AesCbc_Init(UInt32 *p, const Byte *iv)
{
unsigned i;
for (i = 0; i < 4; i++)
p[i] = GetUi32(iv + i * 4);
}
void Z7_FASTCALL AesCbc_Encode(UInt32 *p, Byte *data, size_t numBlocks)
{
for (; numBlocks != 0; numBlocks--, data += AES_BLOCK_SIZE)
{
p[0] ^= GetUi32(data);
p[1] ^= GetUi32(data + 4);
p[2] ^= GetUi32(data + 8);
p[3] ^= GetUi32(data + 12);
Aes_Encode(p + 4, p, p);
SetUi32(data, p[0])
SetUi32(data + 4, p[1])
SetUi32(data + 8, p[2])
SetUi32(data + 12, p[3])
}
}
void Z7_FASTCALL AesCbc_Decode(UInt32 *p, Byte *data, size_t numBlocks)
{
UInt32 in[4], out[4];
for (; numBlocks != 0; numBlocks--, data += AES_BLOCK_SIZE)
{
in[0] = GetUi32(data);
in[1] = GetUi32(data + 4);
in[2] = GetUi32(data + 8);
in[3] = GetUi32(data + 12);
Aes_Decode(p + 4, out, in);
SetUi32(data, p[0] ^ out[0])
SetUi32(data + 4, p[1] ^ out[1])
SetUi32(data + 8, p[2] ^ out[2])
SetUi32(data + 12, p[3] ^ out[3])
p[0] = in[0];
p[1] = in[1];
p[2] = in[2];
p[3] = in[3];
}
}
void Z7_FASTCALL AesCtr_Code(UInt32 *p, Byte *data, size_t numBlocks)
{
for (; numBlocks != 0; numBlocks--)
{
UInt32 temp[4];
unsigned i;
if (++p[0] == 0)
p[1]++;
Aes_Encode(p + 4, temp, p);
for (i = 0; i < 4; i++, data += 4)
{
const UInt32 t = temp[i];
#ifdef MY_CPU_LE_UNALIGN
*((UInt32 *)(void *)data) ^= t;
#else
data[0] = (Byte)(data[0] ^ (t & 0xFF));
data[1] = (Byte)(data[1] ^ ((t >> 8) & 0xFF));
data[2] = (Byte)(data[2] ^ ((t >> 16) & 0xFF));
data[3] = (Byte)(data[3] ^ ((t >> 24)));
#endif
}
}
}
#undef xtime
#undef Ui32
#undef gb0
#undef gb1
#undef gb2
#undef gb3
#undef gb
#undef TT
#undef DD
#undef USE_HW_AES
#undef PRF
third_party/7z/Aes.h
Vendored
+60
View File
@@ -0,0 +1,60 @@
/* Aes.h -- AES encryption / decryption
2023-04-02 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_AES_H
#define ZIP7_INC_AES_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#define AES_BLOCK_SIZE 16
/* Call AesGenTables one time before other AES functions */
void AesGenTables(void);
/* UInt32 pointers must be 16-byte aligned */
/* 16-byte (4 * 32-bit words) blocks: 1 (IV) + 1 (keyMode) + 15 (AES-256 roundKeys) */
#define AES_NUM_IVMRK_WORDS ((1 + 1 + 15) * 4)
/* aes - 16-byte aligned pointer to keyMode+roundKeys sequence */
/* keySize = 16 or 24 or 32 (bytes) */
typedef void (Z7_FASTCALL *AES_SET_KEY_FUNC)(UInt32 *aes, const Byte *key, unsigned keySize);
void Z7_FASTCALL Aes_SetKey_Enc(UInt32 *aes, const Byte *key, unsigned keySize);
void Z7_FASTCALL Aes_SetKey_Dec(UInt32 *aes, const Byte *key, unsigned keySize);
/* ivAes - 16-byte aligned pointer to iv+keyMode+roundKeys sequence: UInt32[AES_NUM_IVMRK_WORDS] */
void AesCbc_Init(UInt32 *ivAes, const Byte *iv); /* iv size is AES_BLOCK_SIZE */
/* data - 16-byte aligned pointer to data */
/* numBlocks - the number of 16-byte blocks in data array */
typedef void (Z7_FASTCALL *AES_CODE_FUNC)(UInt32 *ivAes, Byte *data, size_t numBlocks);
extern AES_CODE_FUNC g_AesCbc_Decode;
#ifndef Z7_SFX
extern AES_CODE_FUNC g_AesCbc_Encode;
extern AES_CODE_FUNC g_AesCtr_Code;
#define k_Aes_SupportedFunctions_HW (1 << 2)
#define k_Aes_SupportedFunctions_HW_256 (1 << 3)
extern UInt32 g_Aes_SupportedFunctions_Flags;
#endif
#define Z7_DECLARE_AES_CODE_FUNC(funcName) \
void Z7_FASTCALL funcName(UInt32 *ivAes, Byte *data, size_t numBlocks);
Z7_DECLARE_AES_CODE_FUNC (AesCbc_Encode)
Z7_DECLARE_AES_CODE_FUNC (AesCbc_Decode)
Z7_DECLARE_AES_CODE_FUNC (AesCtr_Code)
Z7_DECLARE_AES_CODE_FUNC (AesCbc_Encode_HW)
Z7_DECLARE_AES_CODE_FUNC (AesCbc_Decode_HW)
Z7_DECLARE_AES_CODE_FUNC (AesCtr_Code_HW)
Z7_DECLARE_AES_CODE_FUNC (AesCbc_Decode_HW_256)
Z7_DECLARE_AES_CODE_FUNC (AesCtr_Code_HW_256)
EXTERN_C_END
#endif
third_party/7z/AesOpt.c
Vendored
+1002
View File
File diff suppressed because it is too large. Load diff
third_party/7z/Alloc.c
Vendored
+774
View File
@@ -0,0 +1,774 @@
/* Alloc.c -- Memory allocation functions
: Igor Pavlov : Public domain */
#include "Precomp.h"
#ifdef _WIN32
#include "7zWindows.h"
#endif
#include <stdlib.h>
#include "Alloc.h"
#if defined(Z7_LARGE_PAGES) && defined(_WIN32) && \
(!defined(Z7_WIN32_WINNT_MIN) || Z7_WIN32_WINNT_MIN < 0x0502) // < Win2003 (xp-64)
#define Z7_USE_DYN_GetLargePageMinimum
#endif
// for debug:
#if 0
#if defined(__CHERI__) && defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 16)
// #pragma message("=== Z7_ALLOC_NO_OFFSET_ALLOCATOR === ")
#define Z7_ALLOC_NO_OFFSET_ALLOCATOR
#endif
#endif
// #define SZ_ALLOC_DEBUG
/* use SZ_ALLOC_DEBUG to debug alloc/free operations */
#ifdef SZ_ALLOC_DEBUG
#include <string.h>
#include <stdio.h>
static int g_allocCount = 0;
#ifdef _WIN32
static int g_allocCountMid = 0;
#ifdef Z7_LARGE_PAGES
static int g_allocCountBig = 0;
#endif
#endif
#define CONVERT_INT_TO_STR(charType, tempSize) \
char temp[tempSize]; unsigned i = 0; \
while (val >= 10) { temp[i++] = (char)('0' + (unsigned)(val % 10)); val /= 10; } \
*s++ = (charType)('0' + (unsigned)val); \
while (i != 0) { i--; *s++ = temp[i]; } \
*s = 0;
static void ConvertUInt64ToString(UInt64 val, char *s)
{
CONVERT_INT_TO_STR(char, 24)
}
#define GET_HEX_CHAR(t) ((char)(((t < 10) ? ('0' + t) : ('A' + (t - 10)))))
static void ConvertUInt64ToHex(UInt64 val, char *s)
{
UInt64 v = val;
unsigned i;
for (i = 1;; i++)
{
v >>= 4;
if (v == 0)
break;
}
s[i] = 0;
do
{
unsigned t = (unsigned)(val & 0xF);
val >>= 4;
s[--i] = GET_HEX_CHAR(t);
}
while (i);
}
#define DEBUG_OUT_STREAM stderr
static void Print(const char *s)
{
fputs(s, DEBUG_OUT_STREAM);
}
static void PrintAligned(const char *s, size_t align)
{
size_t len = strlen(s);
for(;;)
{
fputc(' ', DEBUG_OUT_STREAM);
if (len >= align)
break;
++len;
}
Print(s);
}
static void PrintLn(void)
{
Print("\n");
}
static void PrintHex(UInt64 v, size_t align)
{
char s[32];
ConvertUInt64ToHex(v, s);
PrintAligned(s, align);
}
static void PrintDec(int v, size_t align)
{
char s[32];
ConvertUInt64ToString((unsigned)v, s);
PrintAligned(s, align);
}
static void PrintAddr(void *p)
{
PrintHex((UInt64)(size_t)(ptrdiff_t)p, 12);
}
#define PRINT_REALLOC(name, cnt, size, ptr) { \
Print(name " "); \
if (!ptr) PrintDec(cnt++, 10); \
PrintHex(size, 10); \
PrintAddr(ptr); \
PrintLn(); }
#define PRINT_ALLOC(name, cnt, size, ptr) { \
Print(name " "); \
PrintDec(cnt++, 10); \
PrintHex(size, 10); \
PrintAddr(ptr); \
PrintLn(); }
#define PRINT_FREE(name, cnt, ptr) if (ptr) { \
Print(name " "); \
PrintDec(--cnt, 10); \
PrintAddr(ptr); \
PrintLn(); }
#else
#ifdef _WIN32
#ifdef Z7_LARGE_PAGES
#define PRINT_ALLOC(name, cnt, size, ptr)
#endif
#endif
#define PRINT_FREE(name, cnt, ptr)
#define Print(s)
#define PrintLn()
#ifndef Z7_ALLOC_NO_OFFSET_ALLOCATOR
#define PrintHex(v, align)
#endif
#define PrintAddr(p)
#endif
/*
by specification:
malloc(non_NULL, 0) : returns NULL or a unique pointer value that can later be successfully passed to free()
realloc(NULL, size) : the call is equivalent to malloc(size)
realloc(non_NULL, 0) : the call is equivalent to free(ptr)
in main compilers:
malloc(0) : returns non_NULL
realloc(NULL, 0) : returns non_NULL
realloc(non_NULL, 0) : returns NULL
*/
void *MyAlloc(size_t size)
{
if (size == 0)
return NULL;
// PRINT_ALLOC("Alloc ", g_allocCount, size, NULL)
#ifdef SZ_ALLOC_DEBUG
{
void *p = malloc(size);
if (p)
{
PRINT_ALLOC("Alloc ", g_allocCount, size, p)
}
return p;
}
#else
return malloc(size);
#endif
}
void MyFree(void *address)
{
PRINT_FREE("Free ", g_allocCount, address)
free(address);
}
void *MyRealloc(void *address, size_t size)
{
if (size == 0)
{
MyFree(address);
return NULL;
}
// PRINT_REALLOC("Realloc ", g_allocCount, size, address)
#ifdef SZ_ALLOC_DEBUG
{
void *p = realloc(address, size);
if (p)
{
PRINT_REALLOC("Realloc ", g_allocCount, size, address)
}
return p;
}
#else
return realloc(address, size);
#endif
}
#ifdef _WIN32
void *MidAlloc(size_t size)
{
if (size == 0)
return NULL;
#ifdef SZ_ALLOC_DEBUG
{
void *p = VirtualAlloc(NULL, size, MEM_COMMIT, PAGE_READWRITE);
if (p)
{
PRINT_ALLOC("Alloc-Mid", g_allocCountMid, size, p)
}
return p;
}
#else
return VirtualAlloc(NULL, size, MEM_COMMIT, PAGE_READWRITE);
#endif
}
void MidFree(void *address)
{
PRINT_FREE("Free-Mid", g_allocCountMid, address)
if (!address)
return;
VirtualFree(address, 0, MEM_RELEASE);
}
#ifdef Z7_LARGE_PAGES
// #pragma message("Z7_LARGE_PAGES")
#ifdef MEM_LARGE_PAGES
#define MY_MEM_LARGE_PAGES MEM_LARGE_PAGES
#else
#define MY_MEM_LARGE_PAGES 0x20000000
#endif
extern
size_t g_LargePageSize;
size_t g_LargePageSize = 0;
extern
size_t g_LargePageThresholdMin;
size_t g_LargePageThresholdMin = 0;
extern
UInt32 g_LargePageFlags;
UInt32 g_LargePageFlags = 0;
void *BigAlloc(size_t size)
{
if (size == 0)
return NULL;
PRINT_ALLOC("Alloc-Big", g_allocCountBig, size, NULL)
#ifdef Z7_LARGE_PAGES
{
const size_t ps = g_LargePageSize - 1;
if (ps < (1u << 30) && size > g_LargePageThresholdMin)
{
const size_t size2 = (size + ps) & ~ps;
if (size2 >= size)
{
void *p = VirtualAlloc(NULL, size2, MEM_COMMIT | MY_MEM_LARGE_PAGES, PAGE_READWRITE);
if (p)
{
PRINT_ALLOC("Alloc-BM ", g_allocCountMid, size2, p)
return p;
}
if (g_LargePageFlags & Z7_LARGE_PAGES_FLAG_FAIL_STOP)
return p;
}
}
}
#endif
return MidAlloc(size);
}
void BigFree(void *address)
{
PRINT_FREE("Free-Big", g_allocCountBig, address)
MidFree(address);
}
#endif // Z7_LARGE_PAGES
#endif // _WIN32
static void *SzAlloc(ISzAllocPtr p, size_t size) { UNUSED_VAR(p) return MyAlloc(size); }
static void SzFree(ISzAllocPtr p, void *address) { UNUSED_VAR(p) MyFree(address); }
const ISzAlloc g_Alloc = { SzAlloc, SzFree };
#ifdef _WIN32
static void *SzMidAlloc(ISzAllocPtr p, size_t size) { UNUSED_VAR(p) return MidAlloc(size); }
static void SzMidFree(ISzAllocPtr p, void *address) { UNUSED_VAR(p) MidFree(address); }
const ISzAlloc g_MidAlloc = { SzMidAlloc, SzMidFree };
#endif
#if defined(Z7_LARGE_PAGES)
static void *SzBigAlloc(ISzAllocPtr p, size_t size) { UNUSED_VAR(p) return BigAlloc(size); }
static void SzBigFree(ISzAllocPtr p, void *address) { UNUSED_VAR(p) BigFree(address); }
const ISzAlloc g_BigAlloc = { SzBigAlloc, SzBigFree };
#endif
#ifndef Z7_ALLOC_NO_OFFSET_ALLOCATOR
#define ADJUST_ALLOC_SIZE 0
/*
#define ADJUST_ALLOC_SIZE (sizeof(void *) - 1)
*/
/*
Use (ADJUST_ALLOC_SIZE = (sizeof(void *) - 1)), if
MyAlloc() can return address that is NOT multiple of sizeof(void *).
*/
/*
uintptr_t : <stdint.h> C99 (optional)
: unsupported in VS6
*/
typedef
#ifdef _WIN32
UINT_PTR
#elif 1
uintptr_t
#else
ptrdiff_t
#endif
MY_uintptr_t;
#if 0 \
|| (defined(__CHERI__) \
|| defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ > 8))
// for 128-bit pointers (cheri):
#define MY_ALIGN_PTR_DOWN(p, align) \
((void *)((char *)(p) - ((size_t)(MY_uintptr_t)(p) & ((align) - 1))))
#else
#define MY_ALIGN_PTR_DOWN(p, align) \
((void *)((((MY_uintptr_t)(p)) & ~((MY_uintptr_t)(align) - 1))))
#endif
#endif
#ifndef _WIN32
#include <unistd.h> // for _POSIX_ADVISORY_INFO : for some linux
#if (defined(Z7_ALLOC_NO_OFFSET_ALLOCATOR) \
|| defined(_POSIX_C_SOURCE) && (_POSIX_C_SOURCE >= 200112L) \
|| defined(_POSIX_ADVISORY_INFO) && (_POSIX_ADVISORY_INFO >= 200112L) \
|| defined(__APPLE__) \
/* || defined(__linux__) */)
#define USE_posix_memalign
// #pragma message("USE_posix_memalign")
#endif
#endif
#ifndef USE_posix_memalign
#define MY_ALIGN_PTR_UP_PLUS(p, align) MY_ALIGN_PTR_DOWN(((char *)(p) + (align) + ADJUST_ALLOC_SIZE), align)
#endif
/*
This posix_memalign() is for test purposes only.
We also need special Free() function instead of free(),
if this posix_memalign() is used.
*/
/*
static int posix_memalign(void **ptr, size_t align, size_t size)
{
size_t newSize = size + align;
void *p;
void *pAligned;
*ptr = NULL;
if (newSize < size)
return 12; // ENOMEM
p = MyAlloc(newSize);
if (!p)
return 12; // ENOMEM
pAligned = MY_ALIGN_PTR_UP_PLUS(p, align);
((void **)pAligned)[-1] = p;
*ptr = pAligned;
return 0;
}
*/
/*
ALLOC_ALIGN_SIZE >= sizeof(void *)
ALLOC_ALIGN_SIZE >= cache_line_size
*/
#define ALLOC_ALIGN_SIZE ((size_t)1 << 7)
void *z7_AlignedAlloc(size_t size)
{
#ifndef USE_posix_memalign
void *p;
void *pAligned;
size_t newSize;
/* also we can allocate additional dummy ALLOC_ALIGN_SIZE bytes after aligned
block to prevent cache line sharing with another allocated blocks */
newSize = size + ALLOC_ALIGN_SIZE * 1 + ADJUST_ALLOC_SIZE;
if (newSize < size)
return NULL;
p = MyAlloc(newSize);
if (!p)
return NULL;
pAligned = MY_ALIGN_PTR_UP_PLUS(p, ALLOC_ALIGN_SIZE);
Print(" size="); PrintHex(size, 8);
Print(" a_size="); PrintHex(newSize, 8);
Print(" ptr="); PrintAddr(p);
Print(" a_ptr="); PrintAddr(pAligned);
PrintLn();
((void **)pAligned)[-1] = p;
return pAligned;
#else
void *p;
if (posix_memalign(&p, ALLOC_ALIGN_SIZE, size))
return NULL;
Print(" posix_memalign="); PrintAddr(p);
PrintLn();
return p;
#endif
}
void z7_AlignedFree(void *address)
{
#ifndef USE_posix_memalign
if (address)
MyFree(((void **)address)[-1]);
#else
free(address);
#endif
}
static void *SzAlignedAlloc(ISzAllocPtr pp, size_t size)
{
UNUSED_VAR(pp)
return z7_AlignedAlloc(size);
}
static void SzAlignedFree(ISzAllocPtr pp, void *address)
{
UNUSED_VAR(pp)
#ifndef USE_posix_memalign
if (address)
MyFree(((void **)address)[-1]);
#else
free(address);
#endif
}
#ifndef _WIN32
#ifdef Z7_LARGE_PAGES
#if 0 // 1 for debug
#include <stdio.h>
#include <string.h> // for strerror()
#define PRF(x) x
#else
#define PRF(x)
#endif
#ifdef USE_posix_memalign
/* madvise():
glibc <= 2.19 : _BSD_SOURCE
glibc > 2.19 : _DEFAULT_SOURCE
*/
/* && (defined(_DEFAULT_SOURCE) || defined(_BSD_SOURCE)) */
#if 1 && !defined(Z7_NO_MADVISE) && \
(defined(__linux__) || defined(__unix__) || defined(__APPLE__))
#include <sys/mman.h> // for madvise
// #pragma message("sys/mman.h")
#if (defined(MADV_HUGEPAGE) && defined(MADV_NOHUGEPAGE))
#define Z7_USE_BIG_ALLOC_MADVISE
// #pragma message("Z7_USE_BIG_ALLOC_MADVISE")
#endif
#endif
#endif // USE_posix_memalign
#ifdef Z7_USE_BIG_ALLOC_MADVISE
#define LARGE_PAGE_SIZE_DEFAULT (1 << 21)
#else
#define LARGE_PAGE_SIZE_DEFAULT 0
#endif
extern
size_t g_LargePageSize;
size_t g_LargePageSize = LARGE_PAGE_SIZE_DEFAULT;
extern
size_t g_LargePageThresholdMin;
size_t g_LargePageThresholdMin = LARGE_PAGE_SIZE_DEFAULT / 2;
extern
UInt32 g_LargePageFlags;
UInt32 g_LargePageFlags = 0;
void *BigAlloc(size_t size)
{
if (size == 0)
return NULL;
#ifdef USE_posix_memalign
{
const size_t pageSize = g_LargePageSize;
void *buf = NULL; // on Linux (and other systems), posix_memalign() does not modify memptr on failure (POSIX.1-2008 TC2).
PRF(printf("\nBigAlloc 0x%08x=%5uMB", (unsigned)(size), (unsigned)(size >> 20));)
if (pageSize && size > g_LargePageThresholdMin)
{
int res;
const size_t mask = pageSize - 1;
/* we can allocate aligned size, so data at the end of buffer also will use huge page
if (size2 for madvise() is not aligned for huge page size)
{ Last data block will use small pages. It reduces memory allocation,
but last data block with small pages can work slower.
It's useful, if we have very large HUGE_PAGE: 32MB or 512MB. }
*/
size_t size2 = (size + mask) & ~mask;
if (size2 < size || (size & mask) <= g_LargePageThresholdMin)
size2 = size;
res = posix_memalign(&buf, pageSize, size2);
PRF(printf(" posix_memalign size=0x%08x=%5uMB align=%u",
(unsigned)(size2), (unsigned)(size2 >> 20), (unsigned)pageSize);)
PRF(printf(" buf=%p", (void *)buf);)
if (res == 0)
{
#ifdef Z7_USE_BIG_ALLOC_MADVISE
if ((g_LargePageFlags & Z7_LARGE_PAGES_FLAG_NO_MADVISE) == 0)
{
// Advise the kernel to use huge pages for this memory range
// MADV_HUGEPAGE / MADV_NOHUGEPAGE : since Linux 2.6.38
// madvise() only operates on whole pages, therefore addr must be page-aligned (4KB/8KB/16KB/64KB).
// The value of size is rounded up to a multiple of page size.
PRF(printf(" madvise g_LargePageFlags=%x", (unsigned)g_LargePageFlags);)
res = madvise(buf, size2, (g_LargePageFlags & Z7_LARGE_PAGES_FLAG_NO_HUGEPAGE) ? MADV_NOHUGEPAGE : MADV_HUGEPAGE);
if (res)
{
PRF(printf("\nERROR res=%d, errno=%d=%s\n", res, (int)errno, strerror(errno));)
if (g_LargePageFlags & Z7_LARGE_PAGES_FLAG_FAIL_STOP)
{
free(buf);
return NULL;
}
}
}
#endif // Z7_USE_BIG_ALLOC_MADVISE
PRF(printf("\n");)
return buf;
}
PRF(printf("\nERROR res=%d=%s\n", res, strerror(res));)
if (g_LargePageFlags & Z7_LARGE_PAGES_FLAG_FAIL_STOP)
return NULL;
// (res == ENOMEM) "Out of memory" is possible, if pageSize is too big.
// so we do second attempt with smaller alignment
}
}
#endif // !USE_posix_memalign
PRF(printf(" z7_AlignedAlloc size=0x%08x=%5uMB\n", (unsigned)(size), (unsigned)(size >> 20));)
return z7_AlignedAlloc(size);
}
void BigFree(void *address)
{
z7_AlignedFree(address);
}
#endif // Z7_LARGE_PAGES
#endif // !_WIN32
#ifdef Z7_LARGE_PAGES
void z7_LargePage_Set(UInt32 flags, size_t pageSize, size_t threshold)
{
g_LargePageFlags = flags;
#ifdef _WIN32
if ((flags & Z7_LARGE_PAGES_FLAG_USE_HUGEPAGE) == 0)
{
g_LargePageSize = 0;
g_LargePageThresholdMin = 0;
}
else
{
if ((flags & Z7_LARGE_PAGES_FLAG_DIRECT_PAGE_SIZE) == 0)
{
#ifdef Z7_USE_DYN_GetLargePageMinimum
Z7_DIAGNOSTIC_IGNORE_CAST_FUNCTION
typedef SIZE_T (WINAPI *Func_GetLargePageMinimum)(VOID);
const
Func_GetLargePageMinimum fn =
(Func_GetLargePageMinimum) Z7_CAST_FUNC_C GetProcAddress(GetModuleHandle(TEXT("kernel32.dll")),
"GetLargePageMinimum");
if (fn)
pageSize = fn();
else
pageSize = 0;
#else
pageSize = GetLargePageMinimum();
#endif
if (pageSize & (pageSize - 1))
pageSize = 0;
}
g_LargePageSize = pageSize;
if ((flags & Z7_LARGE_PAGES_FLAG_DIRECT_THRESHOLD) == 0)
threshold = pageSize / 2;
g_LargePageThresholdMin = threshold;
}
#else // !_WIN32
if (flags & Z7_LARGE_PAGES_FLAG_NO_PAGECODE)
{
g_LargePageSize = 0;
g_LargePageThresholdMin = 0;
}
else
{
if ((flags & Z7_LARGE_PAGES_FLAG_DIRECT_PAGE_SIZE) == 0)
pageSize = LARGE_PAGE_SIZE_DEFAULT;
g_LargePageSize = pageSize;
if ((flags & Z7_LARGE_PAGES_FLAG_DIRECT_THRESHOLD) == 0)
threshold = pageSize / 2;
g_LargePageThresholdMin = threshold;
}
// PRF(printf("\ng_LargePageSize=%x g_LargePageThresholdMin = %x g_LargePageFlags = %x", (unsigned)g_LargePageSize, (unsigned)g_LargePageThresholdMin, (unsigned)g_LargePageFlags);)
#endif // !_WIN32
}
#endif // Z7_LARGE_PAGES
const ISzAlloc g_AlignedAlloc = { SzAlignedAlloc, SzAlignedFree };
/* we align ptr to support cases where CAlignOffsetAlloc::offset is not multiply of sizeof(void *) */
#ifndef Z7_ALLOC_NO_OFFSET_ALLOCATOR
#if 1
#define MY_ALIGN_PTR_DOWN_1(p) MY_ALIGN_PTR_DOWN(p, sizeof(void *))
#define REAL_BLOCK_PTR_VAR(p) ((void **)MY_ALIGN_PTR_DOWN_1(p))[-1]
#else
// we can use this simplified code,
// if (CAlignOffsetAlloc::offset == (k * sizeof(void *))
#define REAL_BLOCK_PTR_VAR(p) (((void **)(p))[-1])
#endif
#endif
#if 0
#ifndef Z7_ALLOC_NO_OFFSET_ALLOCATOR
#include <stdio.h>
static void PrintPtr(const char *s, const void *p)
{
const Byte *p2 = (const Byte *)&p;
unsigned i;
printf("%s %p ", s, p);
for (i = sizeof(p); i != 0;)
{
i--;
printf("%02x", p2[i]);
}
printf("\n");
}
#endif
#endif
static void *AlignOffsetAlloc_Alloc(ISzAllocPtr pp, size_t size)
{
#if defined(Z7_ALLOC_NO_OFFSET_ALLOCATOR)
UNUSED_VAR(pp)
return z7_AlignedAlloc(size);
#else
const CAlignOffsetAlloc *p = Z7_CONTAINER_FROM_VTBL_CONST(pp, CAlignOffsetAlloc, vt);
void *adr;
void *pAligned;
size_t newSize;
size_t extra;
size_t alignSize = (size_t)1 << p->numAlignBits;
if (alignSize < sizeof(void *))
alignSize = sizeof(void *);
if (p->offset >= alignSize)
return NULL;
/* also we can allocate additional dummy ALLOC_ALIGN_SIZE bytes after aligned
block to prevent cache line sharing with another allocated blocks */
extra = p->offset & (sizeof(void *) - 1);
newSize = size + alignSize + extra + ADJUST_ALLOC_SIZE;
if (newSize < size)
return NULL;
adr = ISzAlloc_Alloc(p->baseAlloc, newSize);
if (!adr)
return NULL;
pAligned = (char *)MY_ALIGN_PTR_DOWN((char *)adr +
alignSize - p->offset + extra + ADJUST_ALLOC_SIZE, alignSize) + p->offset;
#if 0
printf("\nalignSize = %6x, offset=%6x, size=%8x \n", (unsigned)alignSize, (unsigned)p->offset, (unsigned)size);
PrintPtr("base", adr);
PrintPtr("alig", pAligned);
#endif
PrintLn();
Print("- Aligned: ");
Print(" size="); PrintHex(size, 8);
Print(" a_size="); PrintHex(newSize, 8);
Print(" ptr="); PrintAddr(adr);
Print(" a_ptr="); PrintAddr(pAligned);
PrintLn();
REAL_BLOCK_PTR_VAR(pAligned) = adr;
return pAligned;
#endif
}
static void AlignOffsetAlloc_Free(ISzAllocPtr pp, void *address)
{
#if defined(Z7_ALLOC_NO_OFFSET_ALLOCATOR)
UNUSED_VAR(pp)
z7_AlignedFree(address);
#else
if (address)
{
const CAlignOffsetAlloc *p = Z7_CONTAINER_FROM_VTBL_CONST(pp, CAlignOffsetAlloc, vt);
PrintLn();
Print("- Aligned Free: ");
PrintLn();
ISzAlloc_Free(p->baseAlloc, REAL_BLOCK_PTR_VAR(address));
}
#endif
}
void AlignOffsetAlloc_CreateVTable(CAlignOffsetAlloc *p)
{
p->vt.Alloc = AlignOffsetAlloc_Alloc;
p->vt.Free = AlignOffsetAlloc_Free;
}
third_party/7z/Alloc.h
Vendored
+76
View File
@@ -0,0 +1,76 @@
/* Alloc.h -- Memory allocation functions
: Igor Pavlov : Public domain */
#ifndef ZIP7_INC_ALLOC_H
#define ZIP7_INC_ALLOC_H
#include "7zTypes.h"
EXTERN_C_BEGIN
/*
MyFree(NULL) : is allowed, as free(NULL)
MyAlloc(0) : returns NULL : but malloc(0) is allowed to return NULL or non_NULL
MyRealloc(NULL, 0) : returns NULL : but realloc(NULL, 0) is allowed to return NULL or non_NULL
MyRealloc() is similar to realloc() for the following cases:
MyRealloc(non_NULL, 0) : returns NULL and always calls MyFree(ptr)
MyRealloc(NULL, non_ZERO) : returns NULL, if allocation failed
MyRealloc(non_NULL, non_ZERO) : returns NULL, if reallocation failed
*/
void *MyAlloc(size_t size);
void MyFree(void *address);
void *MyRealloc(void *address, size_t size);
void *z7_AlignedAlloc(size_t size);
void z7_AlignedFree(void *p);
extern const ISzAlloc g_Alloc;
extern const ISzAlloc g_AlignedAlloc;
#ifdef _WIN32
void *MidAlloc(size_t size);
void MidFree(void *address);
extern const ISzAlloc g_MidAlloc;
#else
#define MidAlloc(size) z7_AlignedAlloc(size)
#define MidFree(address) z7_AlignedFree(address)
#define g_MidAlloc g_AlignedAlloc
#endif
#ifdef Z7_LARGE_PAGES
#define Z7_LARGE_PAGES_FLAG_USE_HUGEPAGE (1 << 0) // PAGE_ALIGNED / MADV_HUGEPAGE
#define Z7_LARGE_PAGES_FLAG_NO_PAGECODE (1 << 1) // no PAGE_ALIGNED / no madvise
#define Z7_LARGE_PAGES_FLAG_NO_MADVISE (1 << 2) // PAGE_ALIGNED / no madvise : for THP=always
#define Z7_LARGE_PAGES_FLAG_NO_HUGEPAGE (1 << 3) // PAGE_ALIGNED / MADV_NOHUGEPAGE
#define Z7_LARGE_PAGES_FLAG_FAIL_STOP (1 << 15) // for benchmarks
#define Z7_LARGE_PAGES_FLAG_DIRECT_PAGE_SIZE (1 << 16)
#define Z7_LARGE_PAGES_FLAG_DIRECT_THRESHOLD (1 << 17)
void z7_LargePage_Set(UInt32 flags, size_t pageSize, size_t threshold);
void *BigAlloc(size_t size);
void BigFree(void *address);
extern const ISzAlloc g_BigAlloc;
#else
#define BigAlloc(size) MidAlloc(size)
#define BigFree(address) MidFree(address)
#define g_BigAlloc g_MidAlloc
#endif
typedef struct
{
ISzAlloc vt;
ISzAllocPtr baseAlloc;
unsigned numAlignBits; /* ((1 << numAlignBits) >= sizeof(void *)) */
size_t offset; /* (offset == (k * sizeof(void *)) && offset < (1 << numAlignBits) */
} CAlignOffsetAlloc;
void AlignOffsetAlloc_CreateVTable(CAlignOffsetAlloc *p);
EXTERN_C_END
#endif
third_party/7z/Bcj2.c
Vendored
+290
View File
@@ -0,0 +1,290 @@
/* Bcj2.c -- BCJ2 Decoder (Converter for x86 code)
2023-03-01 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Bcj2.h"
#include "CpuArch.h"
#define kTopValue ((UInt32)1 << 24)
#define kNumBitModelTotalBits 11
#define kBitModelTotal (1 << kNumBitModelTotalBits)
#define kNumMoveBits 5
// UInt32 bcj2_stats[256 + 2][2];
void Bcj2Dec_Init(CBcj2Dec *p)
{
unsigned i;
p->state = BCJ2_STREAM_RC; // BCJ2_DEC_STATE_OK;
p->ip = 0;
p->temp = 0;
p->range = 0;
p->code = 0;
for (i = 0; i < sizeof(p->probs) / sizeof(p->probs[0]); i++)
p->probs[i] = kBitModelTotal >> 1;
}
SRes Bcj2Dec_Decode(CBcj2Dec *p)
{
UInt32 v = p->temp;
// const Byte *src;
if (p->range <= 5)
{
UInt32 code = p->code;
p->state = BCJ2_DEC_STATE_ERROR; /* for case if we return SZ_ERROR_DATA; */
for (; p->range != 5; p->range++)
{
if (p->range == 1 && code != 0)
return SZ_ERROR_DATA;
if (p->bufs[BCJ2_STREAM_RC] == p->lims[BCJ2_STREAM_RC])
{
p->state = BCJ2_STREAM_RC;
return SZ_OK;
}
code = (code << 8) | *(p->bufs[BCJ2_STREAM_RC])++;
p->code = code;
}
if (code == 0xffffffff)
return SZ_ERROR_DATA;
p->range = 0xffffffff;
}
// else
{
unsigned state = p->state;
// we check BCJ2_IS_32BIT_STREAM() here instead of check in the main loop
if (BCJ2_IS_32BIT_STREAM(state))
{
const Byte *cur = p->bufs[state];
if (cur == p->lims[state])
return SZ_OK;
p->bufs[state] = cur + 4;
{
const UInt32 ip = p->ip + 4;
v = GetBe32a(cur) - ip;
p->ip = ip;
}
state = BCJ2_DEC_STATE_ORIG_0;
}
if ((unsigned)(state - BCJ2_DEC_STATE_ORIG_0) < 4)
{
Byte *dest = p->dest;
for (;;)
{
if (dest == p->destLim)
{
p->state = state;
p->temp = v;
return SZ_OK;
}
*dest++ = (Byte)v;
p->dest = dest;
if (++state == BCJ2_DEC_STATE_ORIG_3 + 1)
break;
v >>= 8;
}
}
}
// src = p->bufs[BCJ2_STREAM_MAIN];
for (;;)
{
/*
if (BCJ2_IS_32BIT_STREAM(p->state))
p->state = BCJ2_DEC_STATE_OK;
else
*/
{
if (p->range < kTopValue)
{
if (p->bufs[BCJ2_STREAM_RC] == p->lims[BCJ2_STREAM_RC])
{
p->state = BCJ2_STREAM_RC;
p->temp = v;
return SZ_OK;
}
p->range <<= 8;
p->code = (p->code << 8) | *(p->bufs[BCJ2_STREAM_RC])++;
}
{
const Byte *src = p->bufs[BCJ2_STREAM_MAIN];
const Byte *srcLim;
Byte *dest = p->dest;
{
const SizeT rem = (SizeT)(p->lims[BCJ2_STREAM_MAIN] - src);
SizeT num = (SizeT)(p->destLim - dest);
if (num >= rem)
num = rem;
#define NUM_ITERS 4
#if (NUM_ITERS & (NUM_ITERS - 1)) == 0
num &= ~((SizeT)NUM_ITERS - 1); // if (NUM_ITERS == (1 << x))
#else
num -= num % NUM_ITERS; // if (NUM_ITERS != (1 << x))
#endif
srcLim = src + num;
}
#define NUM_SHIFT_BITS 24
#define ONE_ITER(indx) { \
const unsigned b = src[indx]; \
*dest++ = (Byte)b; \
v = (v << NUM_SHIFT_BITS) | b; \
if (((b + (0x100 - 0xe8)) & 0xfe) == 0) break; \
if (((v - (((UInt32)0x0f << (NUM_SHIFT_BITS)) + 0x80)) & \
((((UInt32)1 << (4 + NUM_SHIFT_BITS)) - 0x1) << 4)) == 0) break; \
/* ++dest */; /* v = b; */ }
if (src != srcLim)
for (;;)
{
/* The dependency chain of 2-cycle for (v) calculation is not big problem here.
But we can remove dependency chain with v = b in the end of loop. */
ONE_ITER(0)
#if (NUM_ITERS > 1)
ONE_ITER(1)
#if (NUM_ITERS > 2)
ONE_ITER(2)
#if (NUM_ITERS > 3)
ONE_ITER(3)
#if (NUM_ITERS > 4)
ONE_ITER(4)
#if (NUM_ITERS > 5)
ONE_ITER(5)
#if (NUM_ITERS > 6)
ONE_ITER(6)
#if (NUM_ITERS > 7)
ONE_ITER(7)
#endif
#endif
#endif
#endif
#endif
#endif
#endif
src += NUM_ITERS;
if (src == srcLim)
break;
}
if (src == srcLim)
#if (NUM_ITERS > 1)
for (;;)
#endif
{
#if (NUM_ITERS > 1)
if (src == p->lims[BCJ2_STREAM_MAIN] || dest == p->destLim)
#endif
{
const SizeT num = (SizeT)(src - p->bufs[BCJ2_STREAM_MAIN]);
p->bufs[BCJ2_STREAM_MAIN] = src;
p->dest = dest;
p->ip += (UInt32)num;
/* state BCJ2_STREAM_MAIN has more priority than BCJ2_STATE_ORIG */
p->state =
src == p->lims[BCJ2_STREAM_MAIN] ?
(unsigned)BCJ2_STREAM_MAIN :
(unsigned)BCJ2_DEC_STATE_ORIG;
p->temp = v;
return SZ_OK;
}
#if (NUM_ITERS > 1)
ONE_ITER(0)
src++;
#endif
}
{
const SizeT num = (SizeT)(dest - p->dest);
p->dest = dest; // p->dest += num;
p->bufs[BCJ2_STREAM_MAIN] += num; // = src;
p->ip += (UInt32)num;
}
{
UInt32 bound, ttt;
CBcj2Prob *prob; // unsigned index;
/*
prob = p->probs + (unsigned)((Byte)v == 0xe8 ?
2 + (Byte)(v >> 8) :
((v >> 5) & 1)); // ((Byte)v < 0xe8 ? 0 : 1));
*/
{
const unsigned c = ((v + 0x17) >> 6) & 1;
prob = p->probs + (unsigned)
(((0 - c) & (Byte)(v >> NUM_SHIFT_BITS)) + c + ((v >> 5) & 1));
// (Byte)
// 8x->0 : e9->1 : xxe8->xx+2
// 8x->0x100 : e9->0x101 : xxe8->xx
// (((0x100 - (e & ~v)) & (0x100 | (v >> 8))) + (e & v));
// (((0x101 + (~e | v)) & (0x100 | (v >> 8))) + (e & v));
}
ttt = *prob;
bound = (p->range >> kNumBitModelTotalBits) * ttt;
if (p->code < bound)
{
// bcj2_stats[prob - p->probs][0]++;
p->range = bound;
*prob = (CBcj2Prob)(ttt + ((kBitModelTotal - ttt) >> kNumMoveBits));
continue;
}
{
// bcj2_stats[prob - p->probs][1]++;
p->range -= bound;
p->code -= bound;
*prob = (CBcj2Prob)(ttt - (ttt >> kNumMoveBits));
}
}
}
}
{
/* (v == 0xe8 ? 0 : 1) uses setcc instruction with additional zero register usage in x64 MSVC. */
// const unsigned cj = ((Byte)v == 0xe8) ? BCJ2_STREAM_CALL : BCJ2_STREAM_JUMP;
const unsigned cj = (((v + 0x57) >> 6) & 1) + BCJ2_STREAM_CALL;
const Byte *cur = p->bufs[cj];
Byte *dest;
SizeT rem;
if (cur == p->lims[cj])
{
p->state = cj;
break;
}
v = GetBe32a(cur);
p->bufs[cj] = cur + 4;
{
const UInt32 ip = p->ip + 4;
v -= ip;
p->ip = ip;
}
dest = p->dest;
rem = (SizeT)(p->destLim - dest);
if (rem < 4)
{
if ((unsigned)rem > 0) { dest[0] = (Byte)v; v >>= 8;
if ((unsigned)rem > 1) { dest[1] = (Byte)v; v >>= 8;
if ((unsigned)rem > 2) { dest[2] = (Byte)v; v >>= 8; }}}
p->temp = v;
p->dest = dest + rem;
p->state = BCJ2_DEC_STATE_ORIG_0 + (unsigned)rem;
break;
}
SetUi32(dest, v)
v >>= 24;
p->dest = dest + 4;
}
}
if (p->range < kTopValue && p->bufs[BCJ2_STREAM_RC] != p->lims[BCJ2_STREAM_RC])
{
p->range <<= 8;
p->code = (p->code << 8) | *(p->bufs[BCJ2_STREAM_RC])++;
}
return SZ_OK;
}
#undef NUM_ITERS
#undef ONE_ITER
#undef NUM_SHIFT_BITS
#undef kTopValue
#undef kNumBitModelTotalBits
#undef kBitModelTotal
#undef kNumMoveBits
third_party/7z/Bcj2.h
Vendored
+332
View File
@@ -0,0 +1,332 @@
/* Bcj2.h -- BCJ2 converter for x86 code (Branch CALL/JUMP variant2)
2023-03-02 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_BCJ2_H
#define ZIP7_INC_BCJ2_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#define BCJ2_NUM_STREAMS 4
enum
{
BCJ2_STREAM_MAIN,
BCJ2_STREAM_CALL,
BCJ2_STREAM_JUMP,
BCJ2_STREAM_RC
};
enum
{
BCJ2_DEC_STATE_ORIG_0 = BCJ2_NUM_STREAMS,
BCJ2_DEC_STATE_ORIG_1,
BCJ2_DEC_STATE_ORIG_2,
BCJ2_DEC_STATE_ORIG_3,
BCJ2_DEC_STATE_ORIG,
BCJ2_DEC_STATE_ERROR /* after detected data error */
};
enum
{
BCJ2_ENC_STATE_ORIG = BCJ2_NUM_STREAMS,
BCJ2_ENC_STATE_FINISHED /* it's state after fully encoded stream */
};
/* #define BCJ2_IS_32BIT_STREAM(s) ((s) == BCJ2_STREAM_CALL || (s) == BCJ2_STREAM_JUMP) */
#define BCJ2_IS_32BIT_STREAM(s) ((unsigned)((unsigned)(s) - (unsigned)BCJ2_STREAM_CALL) < 2)
/*
CBcj2Dec / CBcj2Enc
bufs sizes:
BUF_SIZE(n) = lims[n] - bufs[n]
bufs sizes for BCJ2_STREAM_CALL and BCJ2_STREAM_JUMP must be multiply of 4:
(BUF_SIZE(BCJ2_STREAM_CALL) & 3) == 0
(BUF_SIZE(BCJ2_STREAM_JUMP) & 3) == 0
*/
// typedef UInt32 CBcj2Prob;
typedef UInt16 CBcj2Prob;
/*
BCJ2 encoder / decoder internal requirements:
- If last bytes of stream contain marker (e8/e8/0f8x), then
there is also encoded symbol (0 : no conversion) in RC stream.
- One case of overlapped instructions is supported,
if last byte of converted instruction is (0f) and next byte is (8x):
marker [xx xx xx 0f] 8x
then the pair (0f 8x) is treated as marker.
*/
/* ---------- BCJ2 Decoder ---------- */
/*
CBcj2Dec:
(dest) is allowed to overlap with bufs[BCJ2_STREAM_MAIN], with the following conditions:
bufs[BCJ2_STREAM_MAIN] >= dest &&
bufs[BCJ2_STREAM_MAIN] - dest >=
BUF_SIZE(BCJ2_STREAM_CALL) +
BUF_SIZE(BCJ2_STREAM_JUMP)
reserve = bufs[BCJ2_STREAM_MAIN] - dest -
( BUF_SIZE(BCJ2_STREAM_CALL) +
BUF_SIZE(BCJ2_STREAM_JUMP) )
and additional conditions:
if (it's first call of Bcj2Dec_Decode() after Bcj2Dec_Init())
{
(reserve != 1) : if (ver < v23.00)
}
else // if there are more than one calls of Bcj2Dec_Decode() after Bcj2Dec_Init())
{
(reserve >= 6) : if (ver < v23.00)
(reserve >= 4) : if (ver >= v23.00)
We need that (reserve) because after first call of Bcj2Dec_Decode(),
CBcj2Dec::temp can contain up to 4 bytes for writing to (dest).
}
(reserve == 0) is allowed, if we decode full stream via single call of Bcj2Dec_Decode().
(reserve == 0) also is allowed in case of multi-call, if we use fixed buffers,
and (reserve) is calculated from full (final) sizes of all streams before first call.
*/
typedef struct
{
const Byte *bufs[BCJ2_NUM_STREAMS];
const Byte *lims[BCJ2_NUM_STREAMS];
Byte *dest;
const Byte *destLim;
unsigned state; /* BCJ2_STREAM_MAIN has more priority than BCJ2_STATE_ORIG */
UInt32 ip; /* property of starting base for decoding */
UInt32 temp; /* Byte temp[4]; */
UInt32 range;
UInt32 code;
CBcj2Prob probs[2 + 256];
} CBcj2Dec;
/* Note:
Bcj2Dec_Init() sets (CBcj2Dec::ip = 0)
if (ip != 0) property is required, the caller must set CBcj2Dec::ip after Bcj2Dec_Init()
*/
void Bcj2Dec_Init(CBcj2Dec *p);
/* Bcj2Dec_Decode():
returns:
SZ_OK
SZ_ERROR_DATA : if data in 5 starting bytes of BCJ2_STREAM_RC stream are not correct
*/
SRes Bcj2Dec_Decode(CBcj2Dec *p);
/* To check that decoding was finished you can compare
sizes of processed streams with sizes known from another sources.
You must do at least one mandatory check from the two following options:
- the check for size of processed output (ORIG) stream.
- the check for size of processed input (MAIN) stream.
additional optional checks:
- the checks for processed sizes of all input streams (MAIN, CALL, JUMP, RC)
- the checks Bcj2Dec_IsMaybeFinished*()
also before actual decoding you can check that the
following condition is met for stream sizes:
( size(ORIG) == size(MAIN) + size(CALL) + size(JUMP) )
*/
/* (state == BCJ2_STREAM_MAIN) means that decoder is ready for
additional input data in BCJ2_STREAM_MAIN stream.
Note that (state == BCJ2_STREAM_MAIN) is allowed for non-finished decoding.
*/
#define Bcj2Dec_IsMaybeFinished_state_MAIN(_p_) ((_p_)->state == BCJ2_STREAM_MAIN)
/* if the stream decoding was finished correctly, then range decoder
part of CBcj2Dec also was finished, and then (CBcj2Dec::code == 0).
Note that (CBcj2Dec::code == 0) is allowed for non-finished decoding.
*/
#define Bcj2Dec_IsMaybeFinished_code(_p_) ((_p_)->code == 0)
/* use Bcj2Dec_IsMaybeFinished() only as additional check
after at least one mandatory check from the two following options:
- the check for size of processed output (ORIG) stream.
- the check for size of processed input (MAIN) stream.
*/
#define Bcj2Dec_IsMaybeFinished(_p_) ( \
Bcj2Dec_IsMaybeFinished_state_MAIN(_p_) && \
Bcj2Dec_IsMaybeFinished_code(_p_))
/* ---------- BCJ2 Encoder ---------- */
typedef enum
{
BCJ2_ENC_FINISH_MODE_CONTINUE,
BCJ2_ENC_FINISH_MODE_END_BLOCK,
BCJ2_ENC_FINISH_MODE_END_STREAM
} EBcj2Enc_FinishMode;
/*
BCJ2_ENC_FINISH_MODE_CONTINUE:
process non finished encoding.
It notifies the encoder that additional further calls
can provide more input data (src) than provided by current call.
In that case the CBcj2Enc encoder still can move (src) pointer
up to (srcLim), but CBcj2Enc encoder can store some of the last
processed bytes (up to 4 bytes) from src to internal CBcj2Enc::temp[] buffer.
at return:
(CBcj2Enc::src will point to position that includes
processed data and data copied to (temp[]) buffer)
That data from (temp[]) buffer will be used in further calls.
BCJ2_ENC_FINISH_MODE_END_BLOCK:
finish encoding of current block (ended at srcLim) without RC flushing.
at return: if (CBcj2Enc::state == BCJ2_ENC_STATE_ORIG) &&
CBcj2Enc::src == CBcj2Enc::srcLim)
: it shows that block encoding was finished. And the encoder is
ready for new (src) data or for stream finish operation.
finished block means
{
CBcj2Enc has completed block encoding up to (srcLim).
(1 + 4 bytes) or (2 + 4 bytes) CALL/JUMP cortages will
not cross block boundary at (srcLim).
temporary CBcj2Enc buffer for (ORIG) src data is empty.
3 output uncompressed streams (MAIN, CALL, JUMP) were flushed.
RC stream was not flushed. And RC stream will cross block boundary.
}
Note: some possible implementation of BCJ2 encoder could
write branch marker (e8/e8/0f8x) in one call of Bcj2Enc_Encode(),
and it could calculate symbol for RC in another call of Bcj2Enc_Encode().
BCJ2 encoder uses ip/fileIp/fileSize/relatLimit values to calculate RC symbol.
And these CBcj2Enc variables can have different values in different Bcj2Enc_Encode() calls.
So caller must finish each block with BCJ2_ENC_FINISH_MODE_END_BLOCK
to ensure that RC symbol is calculated and written in proper block.
BCJ2_ENC_FINISH_MODE_END_STREAM
finish encoding of stream (ended at srcLim) fully including RC flushing.
at return: if (CBcj2Enc::state == BCJ2_ENC_STATE_FINISHED)
: it shows that stream encoding was finished fully,
and all output streams were flushed fully.
also Bcj2Enc_IsFinished() can be called.
*/
/*
32-bit relative offset in JUMP/CALL commands is
- (mod 4 GiB) for 32-bit x86 code
- signed Int32 for 64-bit x86-64 code
BCJ2 encoder also does internal relative to absolute address conversions.
And there are 2 possible ways to do it:
before v23: we used 32-bit variables and (mod 4 GiB) conversion
since v23: we use 64-bit variables and (signed Int32 offset) conversion.
The absolute address condition for conversion in v23:
((UInt64)((Int64)ip64 - (Int64)fileIp64 + 5 + (Int32)offset) < (UInt64)fileSize64)
note that if (fileSize64 > 2 GiB). there is difference between
old (mod 4 GiB) way (v22) and new (signed Int32 offset) way (v23).
And new (v23) way is more suitable to encode 64-bit x86-64 code for (fileSize64 > 2 GiB) cases.
*/
/*
// for old (v22) way for conversion:
typedef UInt32 CBcj2Enc_ip_unsigned;
typedef Int32 CBcj2Enc_ip_signed;
#define BCJ2_ENC_FileSize_MAX ((UInt32)1 << 31)
*/
typedef UInt64 CBcj2Enc_ip_unsigned;
typedef Int64 CBcj2Enc_ip_signed;
/* maximum size of file that can be used for conversion condition */
#define BCJ2_ENC_FileSize_MAX ((CBcj2Enc_ip_unsigned)0 - 2)
/* default value of fileSize64_minus1 variable that means
that absolute address limitation will not be used */
#define BCJ2_ENC_FileSizeField_UNLIMITED ((CBcj2Enc_ip_unsigned)0 - 1)
/* calculate value that later can be set to CBcj2Enc::fileSize64_minus1 */
#define BCJ2_ENC_GET_FileSizeField_VAL_FROM_FileSize(fileSize) \
((CBcj2Enc_ip_unsigned)(fileSize) - 1)
/* set CBcj2Enc::fileSize64_minus1 variable from size of file */
#define Bcj2Enc_SET_FileSize(p, fileSize) \
(p)->fileSize64_minus1 = BCJ2_ENC_GET_FileSizeField_VAL_FROM_FileSize(fileSize);
typedef struct
{
Byte *bufs[BCJ2_NUM_STREAMS];
const Byte *lims[BCJ2_NUM_STREAMS];
const Byte *src;
const Byte *srcLim;
unsigned state;
EBcj2Enc_FinishMode finishMode;
Byte context;
Byte flushRem;
Byte isFlushState;
Byte cache;
UInt32 range;
UInt64 low;
UInt64 cacheSize;
// UInt32 context; // for marker version, it can include marker flag.
/* (ip64) and (fileIp64) correspond to virtual source stream position
that doesn't include data in temp[] */
CBcj2Enc_ip_unsigned ip64; /* current (ip) position */
CBcj2Enc_ip_unsigned fileIp64; /* start (ip) position of current file */
CBcj2Enc_ip_unsigned fileSize64_minus1; /* size of current file (for conversion limitation) */
UInt32 relatLimit; /* (relatLimit <= ((UInt32)1 << 31)) : 0 means disable_conversion */
// UInt32 relatExcludeBits;
UInt32 tempTarget;
unsigned tempPos; /* the number of bytes that were copied to temp[] buffer
(tempPos <= 4) outside of Bcj2Enc_Encode() */
// Byte temp[4]; // for marker version
Byte temp[8];
CBcj2Prob probs[2 + 256];
} CBcj2Enc;
void Bcj2Enc_Init(CBcj2Enc *p);
/*
Bcj2Enc_Encode(): at exit:
p->State < BCJ2_NUM_STREAMS : we need more buffer space for output stream
(bufs[p->State] == lims[p->State])
p->State == BCJ2_ENC_STATE_ORIG : we need more data in input src stream
(src == srcLim)
p->State == BCJ2_ENC_STATE_FINISHED : after fully encoded stream
*/
void Bcj2Enc_Encode(CBcj2Enc *p);
/* Bcj2Enc encoder can look ahead for up 4 bytes of source stream.
CBcj2Enc::tempPos : is the number of bytes that were copied from input stream to temp[] buffer.
(CBcj2Enc::src) after Bcj2Enc_Encode() is starting position after
fully processed data and after data copied to temp buffer.
So if the caller needs to get real number of fully processed input
bytes (without look ahead data in temp buffer),
the caller must subtruct (CBcj2Enc::tempPos) value from processed size
value that is calculated based on current (CBcj2Enc::src):
cur_processed_pos = Calc_Big_Processed_Pos(enc.src)) -
Bcj2Enc_Get_AvailInputSize_in_Temp(&enc);
*/
/* get the size of input data that was stored in temp[] buffer: */
#define Bcj2Enc_Get_AvailInputSize_in_Temp(p) ((p)->tempPos)
#define Bcj2Enc_IsFinished(p) ((p)->flushRem == 0)
/* Note : the decoder supports overlapping of marker (0f 80).
But we can eliminate such overlapping cases by setting
the limit for relative offset conversion as
CBcj2Enc::relatLimit <= (0x0f << 24) == (240 MiB)
*/
/* default value for CBcj2Enc::relatLimit */
#define BCJ2_ENC_RELAT_LIMIT_DEFAULT ((UInt32)0x0f << 24)
#define BCJ2_ENC_RELAT_LIMIT_MAX ((UInt32)1 << 31)
// #define BCJ2_RELAT_EXCLUDE_NUM_BITS 5
EXTERN_C_END
#endif
third_party/7z/Bra.c
Vendored
+709
View File
@@ -0,0 +1,709 @@
/* Bra.c -- Branch converters for RISC code
2024-01-20 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Bra.h"
#include "RotateDefs.h"
#include "CpuArch.h"
#if defined(MY_CPU_SIZEOF_POINTER) \
&& ( MY_CPU_SIZEOF_POINTER == 4 \
|| MY_CPU_SIZEOF_POINTER == 8)
#define BR_CONV_USE_OPT_PC_PTR
#endif
#ifdef BR_CONV_USE_OPT_PC_PTR
#define BR_PC_INIT pc -= (UInt32)(SizeT)p;
#define BR_PC_GET (pc + (UInt32)(SizeT)p)
#else
#define BR_PC_INIT pc += (UInt32)size;
#define BR_PC_GET (pc - (UInt32)(SizeT)(lim - p))
// #define BR_PC_INIT
// #define BR_PC_GET (pc + (UInt32)(SizeT)(p - data))
#endif
#define BR_CONVERT_VAL(v, c) if (encoding) v += c; else v -= c;
// #define BR_CONVERT_VAL(v, c) if (!encoding) c = (UInt32)0 - c; v += c;
#define Z7_BRANCH_CONV(name) z7_ ## name
#define Z7_BRANCH_FUNC_MAIN(name) \
static \
Z7_FORCE_INLINE \
Z7_ATTRIB_NO_VECTOR \
Byte *Z7_BRANCH_CONV(name)(Byte *p, SizeT size, UInt32 pc, int encoding)
#define Z7_BRANCH_FUNC_IMP(name, m, encoding) \
Z7_NO_INLINE \
Z7_ATTRIB_NO_VECTOR \
Byte *m(name)(Byte *data, SizeT size, UInt32 pc) \
{ return Z7_BRANCH_CONV(name)(data, size, pc, encoding); } \
#ifdef Z7_EXTRACT_ONLY
#define Z7_BRANCH_FUNCS_IMP(name) \
Z7_BRANCH_FUNC_IMP(name, Z7_BRANCH_CONV_DEC_2, 0)
#else
#define Z7_BRANCH_FUNCS_IMP(name) \
Z7_BRANCH_FUNC_IMP(name, Z7_BRANCH_CONV_DEC_2, 0) \
Z7_BRANCH_FUNC_IMP(name, Z7_BRANCH_CONV_ENC_2, 1)
#endif
#if defined(__clang__)
#define BR_EXTERNAL_FOR
#define BR_NEXT_ITERATION continue;
#else
#define BR_EXTERNAL_FOR for (;;)
#define BR_NEXT_ITERATION break;
#endif
#if defined(__clang__) && (__clang_major__ >= 8) \
|| defined(__GNUC__) && (__GNUC__ >= 1000) \
// GCC is not good for __builtin_expect() here
/* || defined(_MSC_VER) && (_MSC_VER >= 1920) */
// #define Z7_unlikely [[unlikely]]
// #define Z7_LIKELY(x) (__builtin_expect((x), 1))
#define Z7_UNLIKELY(x) (__builtin_expect((x), 0))
// #define Z7_likely [[likely]]
#else
// #define Z7_LIKELY(x) (x)
#define Z7_UNLIKELY(x) (x)
// #define Z7_likely
#endif
Z7_BRANCH_FUNC_MAIN(BranchConv_ARM64)
{
// Byte *p = data;
const Byte *lim;
const UInt32 flag = (UInt32)1 << (24 - 4);
const UInt32 mask = ((UInt32)1 << 24) - (flag << 1);
size &= ~(SizeT)3;
// if (size == 0) return p;
lim = p + size;
BR_PC_INIT
pc -= 4; // because (p) will point to next instruction
BR_EXTERNAL_FOR
{
// Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
for (;;)
{
UInt32 v;
if Z7_UNLIKELY(p == lim)
return p;
v = GetUi32a(p);
p += 4;
if Z7_UNLIKELY(((v - 0x94000000) & 0xfc000000) == 0)
{
UInt32 c = BR_PC_GET >> 2;
BR_CONVERT_VAL(v, c)
v &= 0x03ffffff;
v |= 0x94000000;
SetUi32a(p - 4, v)
BR_NEXT_ITERATION
}
// v = rotlFixed(v, 8); v += (flag << 8) - 0x90; if Z7_UNLIKELY((v & ((mask << 8) + 0x9f)) == 0)
v -= 0x90000000; if Z7_UNLIKELY((v & 0x9f000000) == 0)
{
UInt32 z, c;
// v = rotrFixed(v, 8);
v += flag; if Z7_UNLIKELY(v & mask) continue;
z = (v & 0xffffffe0) | (v >> 26);
c = (BR_PC_GET >> (12 - 3)) & ~(UInt32)7;
BR_CONVERT_VAL(z, c)
v &= 0x1f;
v |= 0x90000000;
v |= z << 26;
v |= 0x00ffffe0 & ((z & (((flag << 1) - 1))) - flag);
SetUi32a(p - 4, v)
}
}
}
}
Z7_BRANCH_FUNCS_IMP(BranchConv_ARM64)
Z7_BRANCH_FUNC_MAIN(BranchConv_ARM)
{
// Byte *p = data;
const Byte *lim;
size &= ~(SizeT)3;
lim = p + size;
BR_PC_INIT
/* in ARM: branch offset is relative to the +2 instructions from current instruction.
(p) will point to next instruction */
pc += 8 - 4;
for (;;)
{
for (;;)
{
if Z7_UNLIKELY(p >= lim) { return p; } p += 4; if Z7_UNLIKELY(p[-1] == 0xeb) break;
if Z7_UNLIKELY(p >= lim) { return p; } p += 4; if Z7_UNLIKELY(p[-1] == 0xeb) break;
}
{
UInt32 v = GetUi32a(p - 4);
UInt32 c = BR_PC_GET >> 2;
BR_CONVERT_VAL(v, c)
v &= 0x00ffffff;
v |= 0xeb000000;
SetUi32a(p - 4, v)
}
}
}
Z7_BRANCH_FUNCS_IMP(BranchConv_ARM)
Z7_BRANCH_FUNC_MAIN(BranchConv_PPC)
{
// Byte *p = data;
const Byte *lim;
size &= ~(SizeT)3;
lim = p + size;
BR_PC_INIT
pc -= 4; // because (p) will point to next instruction
for (;;)
{
UInt32 v;
for (;;)
{
if Z7_UNLIKELY(p == lim)
return p;
// v = GetBe32a(p);
v = *(UInt32 *)(void *)p;
p += 4;
// if ((v & 0xfc000003) == 0x48000001) break;
// if ((p[-4] & 0xFC) == 0x48 && (p[-1] & 3) == 1) break;
if Z7_UNLIKELY(
((v - Z7_CONV_BE_TO_NATIVE_CONST32(0x48000001))
& Z7_CONV_BE_TO_NATIVE_CONST32(0xfc000003)) == 0) break;
}
{
v = Z7_CONV_NATIVE_TO_BE_32(v);
{
UInt32 c = BR_PC_GET;
BR_CONVERT_VAL(v, c)
}
v &= 0x03ffffff;
v |= 0x48000000;
SetBe32a(p - 4, v)
}
}
}
Z7_BRANCH_FUNCS_IMP(BranchConv_PPC)
#ifdef Z7_CPU_FAST_ROTATE_SUPPORTED
#define BR_SPARC_USE_ROTATE
#endif
Z7_BRANCH_FUNC_MAIN(BranchConv_SPARC)
{
// Byte *p = data;
const Byte *lim;
const UInt32 flag = (UInt32)1 << 22;
size &= ~(SizeT)3;
lim = p + size;
BR_PC_INIT
pc -= 4; // because (p) will point to next instruction
for (;;)
{
UInt32 v;
for (;;)
{
if Z7_UNLIKELY(p == lim)
return p;
/* // the code without GetBe32a():
{ const UInt32 v = GetUi16a(p) & 0xc0ff; p += 4; if (v == 0x40 || v == 0xc07f) break; }
*/
v = GetBe32a(p);
p += 4;
#ifdef BR_SPARC_USE_ROTATE
v = rotlFixed(v, 2);
v += (flag << 2) - 1;
if Z7_UNLIKELY((v & (3 - (flag << 3))) == 0)
#else
v += (UInt32)5 << 29;
v ^= (UInt32)7 << 29;
v += flag;
if Z7_UNLIKELY((v & (0 - (flag << 1))) == 0)
#endif
break;
}
{
// UInt32 v = GetBe32a(p - 4);
#ifndef BR_SPARC_USE_ROTATE
v <<= 2;
#endif
{
UInt32 c = BR_PC_GET;
BR_CONVERT_VAL(v, c)
}
v &= (flag << 3) - 1;
#ifdef BR_SPARC_USE_ROTATE
v -= (flag << 2) - 1;
v = rotrFixed(v, 2);
#else
v -= (flag << 2);
v >>= 2;
v |= (UInt32)1 << 30;
#endif
SetBe32a(p - 4, v)
}
}
}
Z7_BRANCH_FUNCS_IMP(BranchConv_SPARC)
Z7_BRANCH_FUNC_MAIN(BranchConv_ARMT)
{
// Byte *p = data;
Byte *lim;
size &= ~(SizeT)1;
// if (size == 0) return p;
if (size <= 2) return p;
size -= 2;
lim = p + size;
BR_PC_INIT
/* in ARM: branch offset is relative to the +2 instructions from current instruction.
(p) will point to the +2 instructions from current instruction */
// pc += 4 - 4;
// if (encoding) pc -= 0xf800 << 1; else pc += 0xf800 << 1;
// #define ARMT_TAIL_PROC { goto armt_tail; }
#define ARMT_TAIL_PROC { return p; }
do
{
/* in MSVC 32-bit x86 compilers:
UInt32 version : it loads value from memory with movzx
Byte version : it loads value to 8-bit register (AL/CL)
movzx version is slightly faster in some cpus
*/
unsigned b1;
// Byte / unsigned
b1 = p[1];
// optimized version to reduce one (p >= lim) check:
// unsigned a1 = p[1]; b1 = p[3]; p += 2; if Z7_LIKELY((b1 & (a1 ^ 8)) < 0xf8)
for (;;)
{
unsigned b3; // Byte / UInt32
/* (Byte)(b3) normalization can use low byte computations in MSVC.
It gives smaller code, and no loss of speed in some compilers/cpus.
But new MSVC 32-bit x86 compilers use more slow load
from memory to low byte register in that case.
So we try to use full 32-bit computations for faster code.
*/
// if (p >= lim) { ARMT_TAIL_PROC } b3 = b1 + 8; b1 = p[3]; p += 2; if ((b3 & b1) >= 0xf8) break;
if Z7_UNLIKELY(p >= lim) { ARMT_TAIL_PROC } b3 = p[3]; p += 2; if Z7_UNLIKELY((b3 & (b1 ^ 8)) >= 0xf8) break;
if Z7_UNLIKELY(p >= lim) { ARMT_TAIL_PROC } b1 = p[3]; p += 2; if Z7_UNLIKELY((b1 & (b3 ^ 8)) >= 0xf8) break;
}
{
/* we can adjust pc for (0xf800) to rid of (& 0x7FF) operation.
But gcc/clang for arm64 can use bfi instruction for full code here */
UInt32 v =
((UInt32)GetUi16a(p - 2) << 11) |
((UInt32)GetUi16a(p) & 0x7FF);
/*
UInt32 v =
((UInt32)p[1 - 2] << 19)
+ (((UInt32)p[1] & 0x7) << 8)
+ (((UInt32)p[-2] << 11))
+ (p[0]);
*/
p += 2;
{
UInt32 c = BR_PC_GET >> 1;
BR_CONVERT_VAL(v, c)
}
SetUi16a(p - 4, (UInt16)(((v >> 11) & 0x7ff) | 0xf000))
SetUi16a(p - 2, (UInt16)(v | 0xf800))
/*
p[-4] = (Byte)(v >> 11);
p[-3] = (Byte)(0xf0 | ((v >> 19) & 0x7));
p[-2] = (Byte)v;
p[-1] = (Byte)(0xf8 | (v >> 8));
*/
}
}
while (p < lim);
return p;
// armt_tail:
// if ((Byte)((lim[1] & 0xf8)) != 0xf0) { lim += 2; } return lim;
// return (Byte *)(lim + ((Byte)((lim[1] ^ 0xf0) & 0xf8) == 0 ? 0 : 2));
// return (Byte *)(lim + (((lim[1] ^ ~0xfu) & ~7u) == 0 ? 0 : 2));
// return (Byte *)(lim + 2 - (((((unsigned)lim[1] ^ 8) + 8) >> 7) & 2));
}
Z7_BRANCH_FUNCS_IMP(BranchConv_ARMT)
// #define BR_IA64_NO_INLINE
Z7_BRANCH_FUNC_MAIN(BranchConv_IA64)
{
// Byte *p = data;
const Byte *lim;
size &= ~(SizeT)15;
lim = p + size;
pc -= 1 << 4;
pc >>= 4 - 1;
// pc -= 1 << 1;
for (;;)
{
unsigned m;
for (;;)
{
if Z7_UNLIKELY(p == lim)
return p;
m = (unsigned)((UInt32)0x334b0000 >> (*p & 0x1e));
p += 16;
pc += 1 << 1;
if (m &= 3)
break;
}
{
p += (ptrdiff_t)m * 5 - 20; // negative value is expected here.
do
{
const UInt32 t =
#if defined(MY_CPU_X86_OR_AMD64)
// we use 32-bit load here to reduce code size on x86:
GetUi32(p);
#else
GetUi16(p);
#endif
UInt32 z = GetUi32(p + 1) >> m;
p += 5;
if (((t >> m) & (0x70 << 1)) == 0
&& ((z - (0x5000000 << 1)) & (0xf000000 << 1)) == 0)
{
UInt32 v = (UInt32)((0x8fffff << 1) | 1) & z;
z ^= v;
#ifdef BR_IA64_NO_INLINE
v |= (v & ((UInt32)1 << (23 + 1))) >> 3;
{
UInt32 c = pc;
BR_CONVERT_VAL(v, c)
}
v &= (0x1fffff << 1) | 1;
#else
{
if (encoding)
{
// pc &= ~(0xc00000 << 1); // we just need to clear at least 2 bits
pc &= (0x1fffff << 1) | 1;
v += pc;
}
else
{
// pc |= 0xc00000 << 1; // we need to set at least 2 bits
pc |= ~(UInt32)((0x1fffff << 1) | 1);
v -= pc;
}
}
v &= ~(UInt32)(0x600000 << 1);
#endif
v += (0x700000 << 1);
v &= (0x8fffff << 1) | 1;
z |= v;
z <<= m;
SetUi32(p + 1 - 5, z)
}
m++;
}
while (m &= 3); // while (m < 4);
}
}
}
Z7_BRANCH_FUNCS_IMP(BranchConv_IA64)
#define BR_CONVERT_VAL_ENC(v) v += BR_PC_GET;
#define BR_CONVERT_VAL_DEC(v) v -= BR_PC_GET;
#if 1 && defined(MY_CPU_LE_UNALIGN)
#define RISCV_USE_UNALIGNED_LOAD
#endif
#ifdef RISCV_USE_UNALIGNED_LOAD
#define RISCV_GET_UI32(p) GetUi32(p)
#define RISCV_SET_UI32(p, v) { SetUi32(p, v) }
#else
#define RISCV_GET_UI32(p) \
((UInt32)GetUi16a(p) + \
((UInt32)GetUi16a((p) + 2) << 16))
#define RISCV_SET_UI32(p, v) { \
SetUi16a(p, (UInt16)(v)) \
SetUi16a((p) + 2, (UInt16)(v >> 16)) }
#endif
#if 1 && defined(MY_CPU_LE)
#define RISCV_USE_16BIT_LOAD
#endif
#ifdef RISCV_USE_16BIT_LOAD
#define RISCV_LOAD_VAL(p) GetUi16a(p)
#else
#define RISCV_LOAD_VAL(p) (*(p))
#endif
#define RISCV_INSTR_SIZE 2
#define RISCV_STEP_1 (4 + RISCV_INSTR_SIZE)
#define RISCV_STEP_2 4
#define RISCV_REG_VAL (2 << 7)
#define RISCV_CMD_VAL 3
#if 1
// for code size optimization:
#define RISCV_DELTA_7F 0x7f
#else
#define RISCV_DELTA_7F 0
#endif
#define RISCV_CHECK_1(v, b) \
(((((b) - RISCV_CMD_VAL) ^ ((v) << 8)) & (0xf8000 + RISCV_CMD_VAL)) == 0)
#if 1
#define RISCV_CHECK_2(v, r) \
((((v) - ((RISCV_CMD_VAL << 12) | RISCV_REG_VAL | 8)) \
<< 18) \
< ((r) & 0x1d))
#else
// this branch gives larger code, because
// compilers generate larger code for big constants.
#define RISCV_CHECK_2(v, r) \
((((v) - ((RISCV_CMD_VAL << 12) | RISCV_REG_VAL)) \
& ((RISCV_CMD_VAL << 12) | RISCV_REG_VAL)) \
< ((r) & 0x1d))
#endif
#define RISCV_SCAN_LOOP \
Byte *lim; \
size &= ~(SizeT)(RISCV_INSTR_SIZE - 1); \
if (size <= 6) return p; \
size -= 6; \
lim = p + size; \
BR_PC_INIT \
for (;;) \
{ \
UInt32 a, v; \
/* Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE */ \
for (;;) \
{ \
if Z7_UNLIKELY(p >= lim) { return p; } \
a = (RISCV_LOAD_VAL(p) ^ 0x10u) + 1; \
if ((a & 0x77) == 0) break; \
a = (RISCV_LOAD_VAL(p + RISCV_INSTR_SIZE) ^ 0x10u) + 1; \
p += RISCV_INSTR_SIZE * 2; \
if ((a & 0x77) == 0) \
{ \
p -= RISCV_INSTR_SIZE; \
if Z7_UNLIKELY(p >= lim) { return p; } \
break; \
} \
}
// (xx6f ^ 10) + 1 = xx7f + 1 = xx80 : JAL
// (xxef ^ 10) + 1 = xxff + 1 = xx00 + 100 : JAL
// (xx17 ^ 10) + 1 = xx07 + 1 = xx08 : AUIPC
// (xx97 ^ 10) + 1 = xx87 + 1 = xx88 : AUIPC
Byte * Z7_BRANCH_CONV_ENC(RISCV)(Byte *p, SizeT size, UInt32 pc)
{
RISCV_SCAN_LOOP
v = a;
a = RISCV_GET_UI32(p);
#ifndef RISCV_USE_16BIT_LOAD
v += (UInt32)p[1] << 8;
#endif
if ((v & 8) == 0) // JAL
{
if ((v - (0x100 /* - RISCV_DELTA_7F */)) & 0xd80)
{
p += RISCV_INSTR_SIZE;
continue;
}
{
v = ((a & 1u << 31) >> 11)
| ((a & 0x3ff << 21) >> 20)
| ((a & 1 << 20) >> 9)
| (a & 0xff << 12);
BR_CONVERT_VAL_ENC(v)
// ((v & 1) == 0)
// v: bits [1 : 20] contain offset bits
#if 0 && defined(RISCV_USE_UNALIGNED_LOAD)
a &= 0xfff;
a |= ((UInt32)(v << 23))
| ((UInt32)(v << 7) & ((UInt32)0xff << 16))
| ((UInt32)(v >> 5) & ((UInt32)0xf0 << 8));
RISCV_SET_UI32(p, a)
#else // aligned
#if 0
SetUi16a(p, (UInt16)(((v >> 5) & 0xf000) | (a & 0xfff)))
#else
p[1] = (Byte)(((v >> 13) & 0xf0) | ((a >> 8) & 0xf));
#endif
#if 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED) && defined(MY_CPU_LE)
v <<= 15;
v = Z7_BSWAP32(v);
SetUi16a(p + 2, (UInt16)v)
#else
p[2] = (Byte)(v >> 9);
p[3] = (Byte)(v >> 1);
#endif
#endif // aligned
}
p += 4;
continue;
} // JAL
{
// AUIPC
if (v & 0xe80) // (not x0) and (not x2)
{
const UInt32 b = RISCV_GET_UI32(p + 4);
if (RISCV_CHECK_1(v, b))
{
{
const UInt32 temp = (b << 12) | (0x17 + RISCV_REG_VAL);
RISCV_SET_UI32(p, temp)
}
a &= 0xfffff000;
{
#if 1
const int t = -1 >> 1;
if (t != -1)
a += (b >> 20) - ((b >> 19) & 0x1000); // arithmetic right shift emulation
else
#endif
a += (UInt32)((Int32)b >> 20); // arithmetic right shift (sign-extension).
}
BR_CONVERT_VAL_ENC(a)
#if 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED) && defined(MY_CPU_LE)
a = Z7_BSWAP32(a);
RISCV_SET_UI32(p + 4, a)
#else
SetBe32(p + 4, a)
#endif
p += 8;
}
else
p += RISCV_STEP_1;
}
else
{
UInt32 r = a >> 27;
if (RISCV_CHECK_2(v, r))
{
v = RISCV_GET_UI32(p + 4);
r = (r << 7) + 0x17 + (v & 0xfffff000);
a = (a >> 12) | (v << 20);
RISCV_SET_UI32(p, r)
RISCV_SET_UI32(p + 4, a)
p += 8;
}
else
p += RISCV_STEP_2;
}
}
} // for
}
Byte * Z7_BRANCH_CONV_DEC(RISCV)(Byte *p, SizeT size, UInt32 pc)
{
RISCV_SCAN_LOOP
#ifdef RISCV_USE_16BIT_LOAD
if ((a & 8) == 0)
{
#else
v = a;
a += (UInt32)p[1] << 8;
if ((v & 8) == 0)
{
#endif
// JAL
a -= 0x100 - RISCV_DELTA_7F;
if (a & 0xd80)
{
p += RISCV_INSTR_SIZE;
continue;
}
{
const UInt32 a_old = (a + (0xef - RISCV_DELTA_7F)) & 0xfff;
#if 0 // unaligned
a = GetUi32(p);
v = (UInt32)(a >> 23) & ((UInt32)0xff << 1)
| (UInt32)(a >> 7) & ((UInt32)0xff << 9)
#elif 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED) && defined(MY_CPU_LE)
v = GetUi16a(p + 2);
v = Z7_BSWAP32(v) >> 15
#else
v = (UInt32)p[3] << 1
| (UInt32)p[2] << 9
#endif
| (UInt32)((a & 0xf000) << 5);
BR_CONVERT_VAL_DEC(v)
a = a_old
| (v << 11 & 1u << 31)
| (v << 20 & 0x3ff << 21)
| (v << 9 & 1 << 20)
| (v & 0xff << 12);
RISCV_SET_UI32(p, a)
}
p += 4;
continue;
} // JAL
{
// AUIPC
v = a;
#if 1 && defined(RISCV_USE_UNALIGNED_LOAD)
a = GetUi32(p);
#else
a |= (UInt32)GetUi16a(p + 2) << 16;
#endif
if ((v & 0xe80) == 0) // x0/x2
{
const UInt32 r = a >> 27;
if (RISCV_CHECK_2(v, r))
{
UInt32 b;
#if 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED) && defined(MY_CPU_LE)
b = RISCV_GET_UI32(p + 4);
b = Z7_BSWAP32(b);
#else
b = GetBe32(p + 4);
#endif
v = a >> 12;
BR_CONVERT_VAL_DEC(b)
a = (r << 7) + 0x17;
a += (b + 0x800) & 0xfffff000;
v |= b << 20;
RISCV_SET_UI32(p, a)
RISCV_SET_UI32(p + 4, v)
p += 8;
}
else
p += RISCV_STEP_2;
}
else
{
const UInt32 b = RISCV_GET_UI32(p + 4);
if (!RISCV_CHECK_1(v, b))
p += RISCV_STEP_1;
else
{
v = (a & 0xfffff000) | (b >> 20);
a = (b << 12) | (0x17 + RISCV_REG_VAL);
RISCV_SET_UI32(p, a)
RISCV_SET_UI32(p + 4, v)
p += 8;
}
}
}
} // for
}
third_party/7z/Bra.h
Vendored
+105
View File
@@ -0,0 +1,105 @@
/* Bra.h -- Branch converters for executables
2024-01-20 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_BRA_H
#define ZIP7_INC_BRA_H
#include "7zTypes.h"
EXTERN_C_BEGIN
/* #define PPC BAD_PPC_11 // for debug */
#define Z7_BRANCH_CONV_DEC_2(name) z7_ ## name ## _Dec
#define Z7_BRANCH_CONV_ENC_2(name) z7_ ## name ## _Enc
#define Z7_BRANCH_CONV_DEC(name) Z7_BRANCH_CONV_DEC_2(BranchConv_ ## name)
#define Z7_BRANCH_CONV_ENC(name) Z7_BRANCH_CONV_ENC_2(BranchConv_ ## name)
#define Z7_BRANCH_CONV_ST_DEC(name) z7_BranchConvSt_ ## name ## _Dec
#define Z7_BRANCH_CONV_ST_ENC(name) z7_BranchConvSt_ ## name ## _Enc
#define Z7_BRANCH_CONV_DECL(name) Byte * name(Byte *data, SizeT size, UInt32 pc)
#define Z7_BRANCH_CONV_ST_DECL(name) Byte * name(Byte *data, SizeT size, UInt32 pc, UInt32 *state)
typedef Z7_BRANCH_CONV_DECL( (*z7_Func_BranchConv));
typedef Z7_BRANCH_CONV_ST_DECL((*z7_Func_BranchConvSt));
#define Z7_BRANCH_CONV_ST_X86_STATE_INIT_VAL 0
Z7_BRANCH_CONV_ST_DECL (Z7_BRANCH_CONV_ST_DEC(X86));
Z7_BRANCH_CONV_ST_DECL (Z7_BRANCH_CONV_ST_ENC(X86));
#define Z7_BRANCH_FUNCS_DECL(name) \
Z7_BRANCH_CONV_DECL (Z7_BRANCH_CONV_DEC_2(name)); \
Z7_BRANCH_CONV_DECL (Z7_BRANCH_CONV_ENC_2(name));
Z7_BRANCH_FUNCS_DECL (BranchConv_ARM64)
Z7_BRANCH_FUNCS_DECL (BranchConv_ARM)
Z7_BRANCH_FUNCS_DECL (BranchConv_ARMT)
Z7_BRANCH_FUNCS_DECL (BranchConv_PPC)
Z7_BRANCH_FUNCS_DECL (BranchConv_SPARC)
Z7_BRANCH_FUNCS_DECL (BranchConv_IA64)
Z7_BRANCH_FUNCS_DECL (BranchConv_RISCV)
/*
These functions convert data that contain CPU instructions.
Each such function converts relative addresses to absolute addresses in some
branch instructions: CALL (in all converters) and JUMP (X86 converter only).
Such conversion allows to increase compression ratio, if we compress that data.
There are 2 types of converters:
Byte * Conv_RISC (Byte *data, SizeT size, UInt32 pc);
Byte * ConvSt_X86(Byte *data, SizeT size, UInt32 pc, UInt32 *state);
Each Converter supports 2 versions: one for encoding
and one for decoding (_Enc/_Dec postfixes in function name).
In params:
data : data buffer
size : size of data
pc : current virtual Program Counter (Instruction Pointer) value
In/Out param:
state : pointer to state variable (for X86 converter only)
Return:
The pointer to position in (data) buffer after last byte that was processed.
If the caller calls converter again, it must call it starting with that position.
But the caller is allowed to move data in buffer. So pointer to
current processed position also will be changed for next call.
Also the caller must increase internal (pc) value for next call.
Each converter has some characteristics: Endian, Alignment, LookAhead.
Type Endian Alignment LookAhead
X86 little 1 4
ARMT little 2 2
RISCV little 2 6
ARM little 4 0
ARM64 little 4 0
PPC big 4 0
SPARC big 4 0
IA64 little 16 0
(data) must be aligned for (Alignment).
processed size can be calculated as:
SizeT processed = Conv(data, size, pc) - data;
if (processed == 0)
it means that converter needs more data for processing.
If (size < Alignment + LookAhead)
then (processed == 0) is allowed.
Example code for conversion in loop:
UInt32 pc = 0;
size = 0;
for (;;)
{
size += Load_more_input_data(data + size);
SizeT processed = Conv(data, size, pc) - data;
if (processed == 0 && no_more_input_data_after_size)
break; // we stop convert loop
data += processed;
size -= processed;
pc += processed;
}
*/
EXTERN_C_END
#endif
third_party/7z/Bra86.c
Vendored
+187
View File
@@ -0,0 +1,187 @@
/* Bra86.c -- Branch converter for X86 code (BCJ)
2023-04-02 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Bra.h"
#include "CpuArch.h"
#if defined(MY_CPU_SIZEOF_POINTER) \
&& ( MY_CPU_SIZEOF_POINTER == 4 \
|| MY_CPU_SIZEOF_POINTER == 8)
#define BR_CONV_USE_OPT_PC_PTR
#endif
#ifdef BR_CONV_USE_OPT_PC_PTR
#define BR_PC_INIT pc -= (UInt32)(SizeT)p; // (MY_uintptr_t)
#define BR_PC_GET (pc + (UInt32)(SizeT)p)
#else
#define BR_PC_INIT pc += (UInt32)size;
#define BR_PC_GET (pc - (UInt32)(SizeT)(lim - p))
// #define BR_PC_INIT
// #define BR_PC_GET (pc + (UInt32)(SizeT)(p - data))
#endif
#define BR_CONVERT_VAL(v, c) if (encoding) v += c; else v -= c;
// #define BR_CONVERT_VAL(v, c) if (!encoding) c = (UInt32)0 - c; v += c;
#define Z7_BRANCH_CONV_ST(name) z7_BranchConvSt_ ## name
#define BR86_NEED_CONV_FOR_MS_BYTE(b) ((((b) + 1) & 0xfe) == 0)
#ifdef MY_CPU_LE_UNALIGN
#define BR86_PREPARE_BCJ_SCAN const UInt32 v = GetUi32(p) ^ 0xe8e8e8e8;
#define BR86_IS_BCJ_BYTE(n) ((v & ((UInt32)0xfe << (n) * 8)) == 0)
#else
#define BR86_PREPARE_BCJ_SCAN
// bad for MSVC X86 (partial write to byte reg):
#define BR86_IS_BCJ_BYTE(n) ((p[n - 4] & 0xfe) == 0xe8)
// bad for old MSVC (partial write to byte reg):
// #define BR86_IS_BCJ_BYTE(n) (((*p ^ 0xe8) & 0xfe) == 0)
#endif
static
Z7_FORCE_INLINE
Z7_ATTRIB_NO_VECTOR
Byte *Z7_BRANCH_CONV_ST(X86)(Byte *p, SizeT size, UInt32 pc, UInt32 *state, int encoding)
{
if (size < 5)
return p;
{
// Byte *p = data;
const Byte *lim = p + size - 4;
unsigned mask = (unsigned)*state; // & 7;
#ifdef BR_CONV_USE_OPT_PC_PTR
/* if BR_CONV_USE_OPT_PC_PTR is defined: we need to adjust (pc) for (+4),
because call/jump offset is relative to the next instruction.
if BR_CONV_USE_OPT_PC_PTR is not defined : we don't need to adjust (pc) for (+4),
because BR_PC_GET uses (pc - (lim - p)), and lim was adjusted for (-4) before.
*/
pc += 4;
#endif
BR_PC_INIT
goto start;
for (;; mask |= 4)
{
// cont: mask |= 4;
start:
if (p >= lim)
goto fin;
{
BR86_PREPARE_BCJ_SCAN
p += 4;
if (BR86_IS_BCJ_BYTE(0)) { goto m0; } mask >>= 1;
if (BR86_IS_BCJ_BYTE(1)) { goto m1; } mask >>= 1;
if (BR86_IS_BCJ_BYTE(2)) { goto m2; } mask = 0;
if (BR86_IS_BCJ_BYTE(3)) { goto a3; }
}
goto main_loop;
m0: p--;
m1: p--;
m2: p--;
if (mask == 0)
goto a3;
if (p > lim)
goto fin_p;
// if (((0x17u >> mask) & 1) == 0)
if (mask > 4 || mask == 3)
{
mask >>= 1;
continue; // goto cont;
}
mask >>= 1;
if (BR86_NEED_CONV_FOR_MS_BYTE(p[mask]))
continue; // goto cont;
// if (!BR86_NEED_CONV_FOR_MS_BYTE(p[3])) continue; // goto cont;
{
UInt32 v = GetUi32(p);
UInt32 c;
v += (1 << 24); if (v & 0xfe000000) continue; // goto cont;
c = BR_PC_GET;
BR_CONVERT_VAL(v, c)
{
mask <<= 3;
if (BR86_NEED_CONV_FOR_MS_BYTE(v >> mask))
{
v ^= (((UInt32)0x100 << mask) - 1);
#ifdef MY_CPU_X86
// for X86 : we can recalculate (c) to reduce register pressure
c = BR_PC_GET;
#endif
BR_CONVERT_VAL(v, c)
}
mask = 0;
}
// v = (v & ((1 << 24) - 1)) - (v & (1 << 24));
v &= (1 << 25) - 1; v -= (1 << 24);
SetUi32(p, v)
p += 4;
goto main_loop;
}
main_loop:
if (p >= lim)
goto fin;
for (;;)
{
BR86_PREPARE_BCJ_SCAN
p += 4;
if (BR86_IS_BCJ_BYTE(0)) { goto a0; }
if (BR86_IS_BCJ_BYTE(1)) { goto a1; }
if (BR86_IS_BCJ_BYTE(2)) { goto a2; }
if (BR86_IS_BCJ_BYTE(3)) { goto a3; }
if (p >= lim)
goto fin;
}
a0: p--;
a1: p--;
a2: p--;
a3:
if (p > lim)
goto fin_p;
// if (!BR86_NEED_CONV_FOR_MS_BYTE(p[3])) continue; // goto cont;
{
UInt32 v = GetUi32(p);
UInt32 c;
v += (1 << 24); if (v & 0xfe000000) continue; // goto cont;
c = BR_PC_GET;
BR_CONVERT_VAL(v, c)
// v = (v & ((1 << 24) - 1)) - (v & (1 << 24));
v &= (1 << 25) - 1; v -= (1 << 24);
SetUi32(p, v)
p += 4;
goto main_loop;
}
}
fin_p:
p--;
fin:
// the following processing for tail is optional and can be commented
/*
lim += 4;
for (; p < lim; p++, mask >>= 1)
if ((*p & 0xfe) == 0xe8)
break;
*/
*state = (UInt32)mask;
return p;
}
}
#define Z7_BRANCH_CONV_ST_FUNC_IMP(name, m, encoding) \
Z7_NO_INLINE \
Z7_ATTRIB_NO_VECTOR \
Byte *m(name)(Byte *data, SizeT size, UInt32 pc, UInt32 *state) \
{ return Z7_BRANCH_CONV_ST(name)(data, size, pc, state, encoding); }
Z7_BRANCH_CONV_ST_FUNC_IMP(X86, Z7_BRANCH_CONV_ST_DEC, 0)
#ifndef Z7_EXTRACT_ONLY
Z7_BRANCH_CONV_ST_FUNC_IMP(X86, Z7_BRANCH_CONV_ST_ENC, 1)
#endif
third_party/7z/BraIA64.c
Vendored
+14
View File
@@ -0,0 +1,14 @@
/* BraIA64.c -- Converter for IA-64 code
2023-02-20 : Igor Pavlov : Public domain */
#include "Precomp.h"
// the code was moved to Bra.c
#ifdef _MSC_VER
#pragma warning(disable : 4206) // nonstandard extension used : translation unit is empty
#endif
#if defined(__clang__)
#pragma GCC diagnostic ignored "-Wempty-translation-unit"
#endif
third_party/7z/Compiler.h
Vendored
+287
View File
@@ -0,0 +1,287 @@
/* Compiler.h : Compiler specific defines and pragmas
: Igor Pavlov : Public domain */
#ifndef ZIP7_INC_COMPILER_H
#define ZIP7_INC_COMPILER_H
#if defined(__clang__)
# define Z7_CLANG_VERSION (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__)
#endif
#if defined(__clang__) && defined(__apple_build_version__)
# define Z7_APPLE_CLANG_VERSION Z7_CLANG_VERSION
#elif defined(__clang__)
# define Z7_LLVM_CLANG_VERSION Z7_CLANG_VERSION
#elif defined(__GNUC__)
# define Z7_GCC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
#endif
#ifdef _MSC_VER
#if !defined(__clang__) && !defined(__GNUC__)
#define Z7_MSC_VER_ORIGINAL _MSC_VER
#endif
#endif
#if defined(__MINGW32__) || defined(__MINGW64__)
#define Z7_MINGW
#endif
#if defined(__LCC__) && (defined(__MCST__) || defined(__e2k__))
#define Z7_MCST_LCC
#define Z7_MCST_LCC_VERSION (__LCC__ * 100 + __LCC_MINOR__)
#endif
/*
#if defined(__AVX2__) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40900) \
|| defined(Z7_APPLE_CLANG_VERSION) && (Z7_APPLE_CLANG_VERSION >= 40600) \
|| defined(Z7_LLVM_CLANG_VERSION) && (Z7_LLVM_CLANG_VERSION >= 30100) \
|| defined(Z7_MSC_VER_ORIGINAL) && (Z7_MSC_VER_ORIGINAL >= 1800) \
|| defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1400)
#define Z7_COMPILER_AVX2_SUPPORTED
#endif
#endif
*/
// #pragma GCC diagnostic ignored "-Wunknown-pragmas"
#ifdef __clang__
// padding size of '' with 4 bytes to alignment boundary
#pragma GCC diagnostic ignored "-Wpadded"
#if defined(Z7_LLVM_CLANG_VERSION) && (__clang_major__ == 13) \
&& defined(__FreeBSD__)
// freebsd:
#pragma GCC diagnostic ignored "-Wexcess-padding"
#endif
#if defined(Z7_APPLE_CLANG_VERSION) && __clang_major__ >= 21
// warning: function MyAlloc might be an allocator wrapper
// clang in xcode: clang 21.0.0
#pragma GCC diagnostic ignored "-Wallocator-wrappers"
#endif
#if __clang_major__ >= 16
#pragma GCC diagnostic ignored "-Wunsafe-buffer-usage"
#endif
#if __clang_major__ == 13
#if defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 16)
// cheri
#pragma GCC diagnostic ignored "-Wcapability-to-integer-cast"
#endif
#endif
#if __clang_major__ == 13
// for <arm_neon.h>
#pragma GCC diagnostic ignored "-Wreserved-identifier"
#endif
#if __clang_major__ >= 22
// for "StdAfx.h" and "Precomp.h"
#pragma GCC diagnostic ignored "-Wshadow-header"
#endif
#endif // __clang__
#if defined(__clang__) && __clang_major__ >= 16
// #pragma GCC diagnostic ignored "-Wcast-function-type-strict"
#define Z7_DIAGNOSTIC_IGNORE_CAST_FUNCTION \
_Pragma("GCC diagnostic ignored \"-Wcast-function-type-strict\"")
#else
#define Z7_DIAGNOSTIC_IGNORE_CAST_FUNCTION
#endif
typedef void (*Z7_void_Function)(void);
#if defined(__clang__) || defined(__GNUC__)
#define Z7_CAST_FUNC_C (Z7_void_Function)
#elif defined(_MSC_VER) && _MSC_VER > 1920
#define Z7_CAST_FUNC_C (void *)
// #pragma warning(disable : 4191) // 'type cast': unsafe conversion from 'FARPROC' to 'void (__cdecl *)()'
#else
#define Z7_CAST_FUNC_C
#endif
/*
#if (defined(__GNUC__) && (__GNUC__ >= 8)) || defined(__clang__)
// #pragma GCC diagnostic ignored "-Wcast-function-type"
#endif
*/
#ifdef __GNUC__
#if defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40000) && (Z7_GCC_VERSION < 70000)
#pragma GCC diagnostic ignored "-Wstrict-aliasing"
#endif
#endif
#ifdef _MSC_VER
#ifdef UNDER_CE
#define RPC_NO_WINDOWS_H
/* #pragma warning(disable : 4115) // '_RPC_ASYNC_STATE' : named type definition in parentheses */
#pragma warning(disable : 4201) // nonstandard extension used : nameless struct/union
#pragma warning(disable : 4214) // nonstandard extension used : bit field types other than int
#endif
#if defined(_MSC_VER) && _MSC_VER >= 1800
#pragma warning(disable : 4464) // relative include path contains '..'
#endif
// #define Z7_ANALYZE_MODE
#if defined(Z7_ANALYZE_MODE) // && defined(Z7_MSC_VER_ORIGINAL)
// for -analyze switch:
#if _MSC_VER >= 1900
#pragma warning(disable : 6001) // Using uninitialized memory 'ps'.
#pragma warning(disable : 6031) // Return value ignored: 'CoInitialize'
#pragma warning(disable : 6553) // The annotation for function 'RegOpenKeyExW' on _Param_(3) does not apply to a value type.
#pragma warning(disable : 28159) // Random.cpp(21) : Consider using 'GetTickCount64' instead of 'GetTickCount'
#pragma warning(disable : 28160) // ProcessUtils.h(92) : Error annotation: Passing MEM_RELEASE and a non-zero dwSize parameter to VirtualFree is not allowed. This results in the failure of this call
#pragma warning(disable : 28251) // Inconsistent annotation for 'WinMain': this instance has no annotations. See c:\program files (x86)\windows kits\10\include\10.0.26100.0\um\winbase.h(1062).
#pragma warning(disable : 28286) // um\consoleapi.h(236) : For function 'ReadConsoleW', error near the end of 'SAL_readableTo(byteCount(*__formal(3,lpNumberOfCharsRead) * sizeof(TCHAR%)))'.
#pragma warning(disable : 28301) // No annotations for first declaration of 'PNTSTATUS'. See c:\program files (x86)\windows kits\10\include\10.0.26100.0\shared\bcrypt.h(40).
#endif
#if _MSC_VER >= 1600
#pragma warning(disable : 6011) // Dereferencing NULL pointer 'p->IsDirs'
#pragma warning(disable : 6255) // `_alloca` indicates failure by raising a stack overflow exception. Consider using _malloca instead.
#pragma warning(disable : 6258) // Using TerminateThread does not allow proper thread clean up.
// #pragma warning(disable : 6262) // SfxSetup.c : Function uses '32808' bytes of stack.
#pragma warning(disable : 6320) // Threads.c : Exception-filter expression is the constant EXCEPTION_EXECUTE_HANDLER. This might mask exceptions that were not intended to be handled.
#pragma warning(disable : 6385) // createcoder.cpp : Invalid data: accessing 'struct CCodecInfo const * * g_Codecs', the readable size is '256' bytes, but '4194244' bytes might be read: Lines: 338, 339, 340, 342, 344
#pragma warning(disable : 6387) // MyString.cpp : 's' could be '0': this does not adhere to the specification for the function 'wmemcpy'
#endif
#if _MSC_VER == 1600
#pragma warning(disable : 6200) // compressdialog.cpp : Index '31' is out of valid index range '0' to '9' for non-stack buffer 'g_Levels'
#pragma warning(disable : 6204) // Possible buffer overrun in call to 'memset': use of unchecked parameter 'size'
#pragma warning(disable : 6309) // Argument '1' is null: this does not adhere to function specification of 'GetProcAddress'
#pragma warning(disable : 6386) // myvector.h : Buffer overrun: accessing 'argument 1', the writable size is 'newCapacity*4' bytes, but '8' bytes might be written
#endif
#endif // Z7_ANALYZE_MODE
// == 1200 : -O1 : for __forceinline
// >= 1900 : -O1 : for printf
#pragma warning(disable : 4710) // function not inlined
#if _MSC_VER < 1900
// winnt.h: 'Int64ShllMod32'
#pragma warning(disable : 4514) // unreferenced inline function has been removed
#endif
#if _MSC_VER < 1300
// #pragma warning(disable : 4702) // unreachable code
// Bra.c : -O1:
#pragma warning(disable : 4714) // function marked as __forceinline not inlined
#endif
/*
#if _MSC_VER > 1400 && _MSC_VER <= 1900
// strcat: This function or variable may be unsafe
// sysinfoapi.h: kit10: GetVersion was declared deprecated
#pragma warning(disable : 4996)
#endif
*/
#if _MSC_VER > 1200
// -Wall warnings
#pragma warning(disable : 4711) // function selected for automatic inline expansion
#pragma warning(disable : 4820) // '2' bytes padding added after data member
#if _MSC_VER >= 1400 && _MSC_VER < 1920
// 1400: string.h: _DBG_MEMCPY_INLINE_
// 1600 - 191x : smmintrin.h __cplusplus'
// is not defined as a preprocessor macro, replacing with '0' for '#if/#elif'
#pragma warning(disable : 4668)
// 1400 - 1600 : WinDef.h : 'FARPROC' :
// 1900 - 191x : immintrin.h: _readfsbase_u32
// no function prototype given : converting '()' to '(void)'
#pragma warning(disable : 4255)
#endif
#if _MSC_VER >= 1914
// Compiler will insert Spectre mitigation for memory load if /Qspectre switch specified
#pragma warning(disable : 5045)
#endif
#endif // _MSC_VER > 1200
#endif // _MSC_VER
#if defined(__clang__) && (__clang_major__ >= 4)
#define Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE \
_Pragma("clang loop unroll(disable)") \
_Pragma("clang loop vectorize(disable)")
#define Z7_ATTRIB_NO_VECTORIZE
#elif defined(__GNUC__) && (__GNUC__ >= 5) \
&& (!defined(Z7_MCST_LCC_VERSION) || (Z7_MCST_LCC_VERSION >= 12610))
#define Z7_ATTRIB_NO_VECTORIZE __attribute__((optimize("no-tree-vectorize")))
// __attribute__((optimize("no-unroll-loops")));
#define Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
#elif defined(_MSC_VER) && (_MSC_VER >= 1920)
#define Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE \
_Pragma("loop( no_vector )")
#define Z7_ATTRIB_NO_VECTORIZE
#else
#define Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
#define Z7_ATTRIB_NO_VECTORIZE
#endif
#if defined(Z7_MSC_VER_ORIGINAL) && (Z7_MSC_VER_ORIGINAL >= 1920)
#define Z7_PRAGMA_OPTIMIZE_FOR_CODE_SIZE _Pragma("optimize ( \"s\", on )")
#define Z7_PRAGMA_OPTIMIZE_DEFAULT _Pragma("optimize ( \"\", on )")
#else
#define Z7_PRAGMA_OPTIMIZE_FOR_CODE_SIZE
#define Z7_PRAGMA_OPTIMIZE_DEFAULT
#endif
#if defined(MY_CPU_X86_OR_AMD64) && ( \
defined(__clang__) && (__clang_major__ >= 4) \
|| defined(__GNUC__) && (__GNUC__ >= 5))
#define Z7_ATTRIB_NO_SSE __attribute__((__target__("no-sse")))
#else
#define Z7_ATTRIB_NO_SSE
#endif
#define Z7_ATTRIB_NO_VECTOR \
Z7_ATTRIB_NO_VECTORIZE \
Z7_ATTRIB_NO_SSE
#if defined(__clang__) && (__clang_major__ >= 8) \
|| defined(__GNUC__) && (__GNUC__ >= 1000) \
/* || defined(_MSC_VER) && (_MSC_VER >= 1920) */
// GCC is not good for __builtin_expect()
#define Z7_LIKELY(x) (__builtin_expect((x), 1))
#define Z7_UNLIKELY(x) (__builtin_expect((x), 0))
// #define Z7_unlikely [[unlikely]]
// #define Z7_likely [[likely]]
#else
#define Z7_LIKELY(x) (x)
#define Z7_UNLIKELY(x) (x)
// #define Z7_likely
#endif
#if (defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30600))
#if (Z7_CLANG_VERSION < 130000)
#define Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER \
_Pragma("GCC diagnostic push") \
_Pragma("GCC diagnostic ignored \"-Wreserved-id-macro\"")
#else
#define Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER \
_Pragma("GCC diagnostic push") \
_Pragma("GCC diagnostic ignored \"-Wreserved-macro-identifier\"")
#endif
#define Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER \
_Pragma("GCC diagnostic pop")
#else
#define Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#define UNUSED_VAR(x) (void)x;
/* #define UNUSED_VAR(x) x=x; */
#endif
third_party/7z/CpuArch.c
Vendored
+970
View File
@@ -0,0 +1,970 @@
/* CpuArch.c -- CPU specific code
Igor Pavlov : Public domain */
#include "Precomp.h"
// #include <stdio.h>
#include "CpuArch.h"
#ifdef MY_CPU_X86_OR_AMD64
#undef NEED_CHECK_FOR_CPUID
#if !defined(MY_CPU_AMD64)
#define NEED_CHECK_FOR_CPUID
#endif
/*
cpuid instruction supports (subFunction) parameter in ECX,
that is used only with some specific (function) parameter values.
most functions use only (subFunction==0).
*/
/*
__cpuid(): MSVC and GCC/CLANG use same function/macro name
but parameters are different.
We use MSVC __cpuid() parameters style for our z7_x86_cpuid() function.
*/
#if defined(__GNUC__) /* && (__GNUC__ >= 10) */ \
|| defined(__clang__) /* && (__clang_major__ >= 10) */
/* there was some CLANG/GCC compilers that have issues with
rbx(ebx) handling in asm blocks in -fPIC mode (__PIC__ is defined).
compiler's <cpuid.h> contains the macro __cpuid() that is similar to our code.
The history of __cpuid() changes in CLANG/GCC:
GCC:
2007: it preserved ebx for (__PIC__ && __i386__)
2013: it preserved rbx and ebx for __PIC__
2014: it doesn't preserves rbx and ebx anymore
we suppose that (__GNUC__ >= 5) fixed that __PIC__ ebx/rbx problem.
CLANG:
2014+: it preserves rbx, but only for 64-bit code. No __PIC__ check.
Why CLANG cares about 64-bit mode only, and doesn't care about ebx (in 32-bit)?
Do we need __PIC__ test for CLANG or we must care about rbx even if
__PIC__ is not defined?
*/
#define ASM_LN "\n"
#if defined(MY_CPU_AMD64) && defined(__PIC__) \
&& ((defined (__GNUC__) && (__GNUC__ < 5)) || defined(__clang__))
/* "=&r" selects free register. It can select even rbx, if that register is free.
"=&D" for (RDI) also works, but the code can be larger with "=&D"
"2"(subFun) : 2 is (zero-based) index in the output constraint list "=c" (ECX). */
#define x86_cpuid_MACRO_2(p, func, subFunc) { \
__asm__ __volatile__ ( \
ASM_LN "mov %%rbx, %q1" \
ASM_LN "cpuid" \
ASM_LN "xchg %%rbx, %q1" \
: "=a" ((p)[0]), "=&r" ((p)[1]), "=c" ((p)[2]), "=d" ((p)[3]) : "0" (func), "2"(subFunc)); }
#elif defined(MY_CPU_X86) && defined(__PIC__) \
&& ((defined (__GNUC__) && (__GNUC__ < 5)) || defined(__clang__))
#define x86_cpuid_MACRO_2(p, func, subFunc) { \
__asm__ __volatile__ ( \
ASM_LN "mov %%ebx, %k1" \
ASM_LN "cpuid" \
ASM_LN "xchg %%ebx, %k1" \
: "=a" ((p)[0]), "=&r" ((p)[1]), "=c" ((p)[2]), "=d" ((p)[3]) : "0" (func), "2"(subFunc)); }
#else
#define x86_cpuid_MACRO_2(p, func, subFunc) { \
__asm__ __volatile__ ( \
ASM_LN "cpuid" \
: "=a" ((p)[0]), "=b" ((p)[1]), "=c" ((p)[2]), "=d" ((p)[3]) : "0" (func), "2"(subFunc)); }
#endif
#define x86_cpuid_MACRO(p, func) x86_cpuid_MACRO_2(p, func, 0)
void Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
{
x86_cpuid_MACRO(p, func)
}
static
void Z7_FASTCALL z7_x86_cpuid_subFunc(UInt32 p[4], UInt32 func, UInt32 subFunc)
{
x86_cpuid_MACRO_2(p, func, subFunc)
}
Z7_NO_INLINE
UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
{
#if defined(NEED_CHECK_FOR_CPUID)
#define EFALGS_CPUID_BIT 21
UInt32 a;
__asm__ __volatile__ (
ASM_LN "pushf"
ASM_LN "pushf"
ASM_LN "pop %0"
// ASM_LN "movl %0, %1"
// ASM_LN "xorl $0x200000, %0"
ASM_LN "btc %1, %0"
ASM_LN "push %0"
ASM_LN "popf"
ASM_LN "pushf"
ASM_LN "pop %0"
ASM_LN "xorl (%%esp), %0"
ASM_LN "popf"
ASM_LN
: "=&r" (a) // "=a"
: "i" (EFALGS_CPUID_BIT)
);
if ((a & (1 << EFALGS_CPUID_BIT)) == 0)
return 0;
#endif
{
UInt32 p[4];
x86_cpuid_MACRO(p, 0)
return p[0];
}
}
#undef ASM_LN
#elif !defined(_MSC_VER)
/*
// for gcc/clang and other: we can try to use __cpuid macro:
#include <cpuid.h>
void Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
{
__cpuid(func, p[0], p[1], p[2], p[3]);
}
UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
{
return (UInt32)__get_cpuid_max(0, NULL);
}
*/
// for unsupported cpuid:
void Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
{
UNUSED_VAR(func)
p[0] = p[1] = p[2] = p[3] = 0;
}
UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
{
return 0;
}
#else // _MSC_VER
#if !defined(MY_CPU_AMD64)
UInt32 __declspec(naked) Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
{
#if defined(NEED_CHECK_FOR_CPUID)
#define EFALGS_CPUID_BIT 21
__asm pushfd
__asm pushfd
/*
__asm pop eax
// __asm mov edx, eax
__asm btc eax, EFALGS_CPUID_BIT
__asm push eax
*/
__asm btc dword ptr [esp], EFALGS_CPUID_BIT
__asm popfd
__asm pushfd
__asm pop eax
// __asm xor eax, edx
__asm xor eax, [esp]
// __asm push edx
__asm popfd
__asm and eax, (1 shl EFALGS_CPUID_BIT)
__asm jz end_func
#endif
__asm push ebx
__asm xor eax, eax // func
__asm xor ecx, ecx // subFunction (optional) for (func == 0)
__asm cpuid
__asm pop ebx
#if defined(NEED_CHECK_FOR_CPUID)
end_func:
#endif
__asm ret 0
}
void __declspec(naked) Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
{
UNUSED_VAR(p)
UNUSED_VAR(func)
__asm push ebx
__asm push edi
__asm mov edi, ecx // p
__asm mov eax, edx // func
__asm xor ecx, ecx // subfunction (optional) for (func == 0)
__asm cpuid
__asm mov [edi ], eax
__asm mov [edi + 4], ebx
__asm mov [edi + 8], ecx
__asm mov [edi + 12], edx
__asm pop edi
__asm pop ebx
__asm ret 0
}
static
void __declspec(naked) Z7_FASTCALL z7_x86_cpuid_subFunc(UInt32 p[4], UInt32 func, UInt32 subFunc)
{
UNUSED_VAR(p)
UNUSED_VAR(func)
UNUSED_VAR(subFunc)
__asm push ebx
__asm push edi
__asm mov edi, ecx // p
__asm mov eax, edx // func
__asm mov ecx, [esp + 12] // subFunc
__asm cpuid
__asm mov [edi ], eax
__asm mov [edi + 4], ebx
__asm mov [edi + 8], ecx
__asm mov [edi + 12], edx
__asm pop edi
__asm pop ebx
__asm ret 4
}
#else // MY_CPU_AMD64
#if _MSC_VER >= 1600
#include <intrin.h>
#define MY_cpuidex __cpuidex
static
void Z7_FASTCALL z7_x86_cpuid_subFunc(UInt32 p[4], UInt32 func, UInt32 subFunc)
{
__cpuidex((int *)p, func, subFunc);
}
#else
/*
__cpuid (func == (0 or 7)) requires subfunction number in ECX.
MSDN: The __cpuid intrinsic clears the ECX register before calling the cpuid instruction.
__cpuid() in new MSVC clears ECX.
__cpuid() in old MSVC (14.00) x64 doesn't clear ECX
We still can use __cpuid for low (func) values that don't require ECX,
but __cpuid() in old MSVC will be incorrect for some func values: (func == 7).
So here we use the hack for old MSVC to send (subFunction) in ECX register to cpuid instruction,
where ECX value is first parameter for FASTCALL / NO_INLINE func.
So the caller of MY_cpuidex_HACK() sets ECX as subFunction, and
old MSVC for __cpuid() doesn't change ECX and cpuid instruction gets (subFunction) value.
DON'T remove Z7_NO_INLINE and Z7_FASTCALL for MY_cpuidex_HACK(): !!!
*/
static
Z7_NO_INLINE void Z7_FASTCALL MY_cpuidex_HACK(Int32 subFunction, Int32 func, Int32 *CPUInfo)
{
UNUSED_VAR(subFunction)
__cpuid(CPUInfo, func);
}
#define MY_cpuidex(info, func, func2) MY_cpuidex_HACK(func2, func, info)
#pragma message("======== MY_cpuidex_HACK WAS USED ========")
static
void Z7_FASTCALL z7_x86_cpuid_subFunc(UInt32 p[4], UInt32 func, UInt32 subFunc)
{
MY_cpuidex_HACK(subFunc, func, (Int32 *)p);
}
#endif // _MSC_VER >= 1600
#if !defined(MY_CPU_AMD64)
/* inlining for __cpuid() in MSVC x86 (32-bit) produces big ineffective code,
so we disable inlining here */
Z7_NO_INLINE
#endif
void Z7_FASTCALL z7_x86_cpuid(UInt32 p[4], UInt32 func)
{
MY_cpuidex((Int32 *)p, (Int32)func, 0);
}
Z7_NO_INLINE
UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void)
{
Int32 a[4];
MY_cpuidex(a, 0, 0);
return a[0];
}
#endif // MY_CPU_AMD64
#endif // _MSC_VER
#if defined(NEED_CHECK_FOR_CPUID)
#define CHECK_CPUID_IS_SUPPORTED { if (z7_x86_cpuid_GetMaxFunc() == 0) return 0; }
#else
#define CHECK_CPUID_IS_SUPPORTED
#endif
#undef NEED_CHECK_FOR_CPUID
static
BoolInt x86cpuid_Func_1(UInt32 *p)
{
CHECK_CPUID_IS_SUPPORTED
z7_x86_cpuid(p, 1);
return True;
}
/*
static const UInt32 kVendors[][1] =
{
{ 0x756E6547 }, // , 0x49656E69, 0x6C65746E },
{ 0x68747541 }, // , 0x69746E65, 0x444D4163 },
{ 0x746E6543 } // , 0x48727561, 0x736C7561 }
};
*/
/*
typedef struct
{
UInt32 maxFunc;
UInt32 vendor[3];
UInt32 ver;
UInt32 b;
UInt32 c;
UInt32 d;
} Cx86cpuid;
enum
{
CPU_FIRM_INTEL,
CPU_FIRM_AMD,
CPU_FIRM_VIA
};
int x86cpuid_GetFirm(const Cx86cpuid *p);
#define x86cpuid_ver_GetFamily(ver) (((ver >> 16) & 0xff0) | ((ver >> 8) & 0xf))
#define x86cpuid_ver_GetModel(ver) (((ver >> 12) & 0xf0) | ((ver >> 4) & 0xf))
#define x86cpuid_ver_GetStepping(ver) (ver & 0xf)
int x86cpuid_GetFirm(const Cx86cpuid *p)
{
unsigned i;
for (i = 0; i < sizeof(kVendors) / sizeof(kVendors[0]); i++)
{
const UInt32 *v = kVendors[i];
if (v[0] == p->vendor[0]
// && v[1] == p->vendor[1]
// && v[2] == p->vendor[2]
)
return (int)i;
}
return -1;
}
BoolInt CPU_Is_InOrder()
{
Cx86cpuid p;
UInt32 family, model;
if (!x86cpuid_CheckAndRead(&p))
return True;
family = x86cpuid_ver_GetFamily(p.ver);
model = x86cpuid_ver_GetModel(p.ver);
switch (x86cpuid_GetFirm(&p))
{
case CPU_FIRM_INTEL: return (family < 6 || (family == 6 && (
// In-Order Atom CPU
model == 0x1C // 45 nm, N4xx, D4xx, N5xx, D5xx, 230, 330
|| model == 0x26 // 45 nm, Z6xx
|| model == 0x27 // 32 nm, Z2460
|| model == 0x35 // 32 nm, Z2760
|| model == 0x36 // 32 nm, N2xxx, D2xxx
)));
case CPU_FIRM_AMD: return (family < 5 || (family == 5 && (model < 6 || model == 0xA)));
case CPU_FIRM_VIA: return (family < 6 || (family == 6 && model < 0xF));
}
return False; // v23 : unknown processors are not In-Order
}
*/
#ifdef _WIN32
#include "7zWindows.h"
#endif
#if !defined(MY_CPU_AMD64) && defined(_WIN32)
/* for legacy SSE ia32: there is no user-space cpu instruction to check
that OS supports SSE register storing/restoring on context switches.
So we need some OS-specific function to check that it's safe to use SSE registers.
*/
Z7_FORCE_INLINE
static BoolInt CPU_Sys_Is_SSE_Supported(void)
{
#ifdef _MSC_VER
#pragma warning(push)
#pragma warning(disable : 4996) // `GetVersion': was declared deprecated
#endif
/* low byte is major version of Windows
We suppose that any Windows version since
Windows2000 (major == 5) supports SSE registers */
return (Byte)GetVersion() >= 5;
#if defined(_MSC_VER)
#pragma warning(pop)
#endif
}
#define CHECK_SYS_SSE_SUPPORT if (!CPU_Sys_Is_SSE_Supported()) return False;
#else
#define CHECK_SYS_SSE_SUPPORT
#endif
#if !defined(MY_CPU_AMD64)
BoolInt CPU_IsSupported_CMOV(void)
{
UInt32 a[4];
if (!x86cpuid_Func_1(&a[0]))
return 0;
return (BoolInt)(a[3] >> 15) & 1;
}
BoolInt CPU_IsSupported_SSE(void)
{
UInt32 a[4];
CHECK_SYS_SSE_SUPPORT
if (!x86cpuid_Func_1(&a[0]))
return 0;
return (BoolInt)(a[3] >> 25) & 1;
}
BoolInt CPU_IsSupported_SSE2(void)
{
UInt32 a[4];
CHECK_SYS_SSE_SUPPORT
if (!x86cpuid_Func_1(&a[0]))
return 0;
return (BoolInt)(a[3] >> 26) & 1;
}
#endif
static UInt32 x86cpuid_Func_1_ECX(void)
{
UInt32 a[4];
CHECK_SYS_SSE_SUPPORT
if (!x86cpuid_Func_1(&a[0]))
return 0;
return a[2];
}
BoolInt CPU_IsSupported_AES(void)
{
return (BoolInt)(x86cpuid_Func_1_ECX() >> 25) & 1;
}
BoolInt CPU_IsSupported_SSSE3(void)
{
return (BoolInt)(x86cpuid_Func_1_ECX() >> 9) & 1;
}
BoolInt CPU_IsSupported_SSE41(void)
{
return (BoolInt)(x86cpuid_Func_1_ECX() >> 19) & 1;
}
BoolInt CPU_IsSupported_SHA(void)
{
CHECK_SYS_SSE_SUPPORT
if (z7_x86_cpuid_GetMaxFunc() < 7)
return False;
{
UInt32 d[4];
z7_x86_cpuid(d, 7);
return (BoolInt)(d[1] >> 29) & 1;
}
}
BoolInt CPU_IsSupported_SHA512(void)
{
if (!CPU_IsSupported_AVX2()) return False; // maybe CPU_IsSupported_AVX() is enough here
if (z7_x86_cpuid_GetMaxFunc() < 7)
return False;
{
UInt32 d[4];
z7_x86_cpuid_subFunc(d, 7, 0);
if (d[0] < 1) // d[0] - is max supported subleaf value
return False;
z7_x86_cpuid_subFunc(d, 7, 1);
return (BoolInt)(d[0]) & 1;
}
}
/*
MSVC: _xgetbv() intrinsic is available since VS2010SP1.
MSVC also defines (_XCR_XFEATURE_ENABLED_MASK) macro in
<immintrin.h> that we can use or check.
For any 32-bit x86 we can use asm code in MSVC,
but MSVC asm code is huge after compilation.
So _xgetbv() is better
ICC: _xgetbv() intrinsic is available (in what version of ICC?)
ICC defines (__GNUC___) and it supports gnu assembler
also ICC supports MASM style code with -use-msasm switch.
but ICC doesn't support __attribute__((__target__))
GCC/CLANG 9:
_xgetbv() is macro that works via __builtin_ia32_xgetbv()
and we need __attribute__((__target__("xsave")).
But with __target__("xsave") the function will be not
inlined to function that has no __target__("xsave") attribute.
If we want _xgetbv() call inlining, then we should use asm version
instead of calling _xgetbv().
Note:intrinsic is broke before GCC 8.2:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85684
*/
#if defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1100) \
|| defined(_MSC_VER) && (_MSC_VER >= 1600) && (_MSC_FULL_VER >= 160040219) \
|| defined(__GNUC__) && (__GNUC__ >= 9) \
|| defined(__clang__) && (__clang_major__ >= 9)
// we define ATTRIB_XGETBV, if we want to use predefined _xgetbv() from compiler
#if defined(__INTEL_COMPILER)
#define ATTRIB_XGETBV
#elif defined(__GNUC__) || defined(__clang__)
// we don't define ATTRIB_XGETBV here, because asm version is better for inlining.
// #define ATTRIB_XGETBV __attribute__((__target__("xsave")))
#else
#define ATTRIB_XGETBV
#endif
#endif
#if defined(ATTRIB_XGETBV)
#include <immintrin.h>
#endif
// XFEATURE_ENABLED_MASK/XCR0
#define MY_XCR_XFEATURE_ENABLED_MASK 0
#if defined(ATTRIB_XGETBV)
ATTRIB_XGETBV
#endif
static UInt64 x86_xgetbv_0(UInt32 num)
{
#if defined(ATTRIB_XGETBV)
{
return
#if (defined(_MSC_VER))
_xgetbv(num);
#else
__builtin_ia32_xgetbv(
#if !defined(__clang__)
(int)
#endif
num);
#endif
}
#elif defined(__GNUC__) || defined(__clang__) || defined(__SUNPRO_CC)
UInt32 a, d;
#if defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 4))
__asm__
(
"xgetbv"
: "=a"(a), "=d"(d) : "c"(num) : "cc"
);
#else // is old gcc
__asm__
(
".byte 0x0f, 0x01, 0xd0" "\n\t"
: "=a"(a), "=d"(d) : "c"(num) : "cc"
);
#endif
return ((UInt64)d << 32) | a;
// return a;
#elif defined(_MSC_VER) && !defined(MY_CPU_AMD64)
UInt32 a, d;
__asm {
push eax
push edx
push ecx
mov ecx, num;
// xor ecx, ecx // = MY_XCR_XFEATURE_ENABLED_MASK
_emit 0x0f
_emit 0x01
_emit 0xd0
mov a, eax
mov d, edx
pop ecx
pop edx
pop eax
}
return ((UInt64)d << 32) | a;
// return a;
#else // it's unknown compiler
// #error "Need xgetbv function"
UNUSED_VAR(num)
// for MSVC-X64 we could call external function from external file.
/* Actually we had checked OSXSAVE/AVX in cpuid before.
So it's expected that OS supports at least AVX and below. */
// if (num != MY_XCR_XFEATURE_ENABLED_MASK) return 0; // if not XCR0
return
// (1 << 0) | // x87
(1 << 1) // SSE
| (1 << 2); // AVX
#endif
}
#ifdef _WIN32
/*
Windows versions do not know about new ISA extensions that
can be introduced. But we still can use new extensions,
even if Windows doesn't report about supporting them,
But we can use new extensions, only if Windows knows about new ISA extension
that changes the number or size of registers: SSE, AVX/XSAVE, AVX512
So it's enough to check
MY_PF_AVX_INSTRUCTIONS_AVAILABLE
instead of
MY_PF_AVX2_INSTRUCTIONS_AVAILABLE
*/
#define MY_PF_XSAVE_ENABLED 17
// #define MY_PF_SSSE3_INSTRUCTIONS_AVAILABLE 36
// #define MY_PF_SSE4_1_INSTRUCTIONS_AVAILABLE 37
// #define MY_PF_SSE4_2_INSTRUCTIONS_AVAILABLE 38
// #define MY_PF_AVX_INSTRUCTIONS_AVAILABLE 39
// #define MY_PF_AVX2_INSTRUCTIONS_AVAILABLE 40
// #define MY_PF_AVX512F_INSTRUCTIONS_AVAILABLE 41
#endif
BoolInt CPU_IsSupported_AVX(void)
{
#ifdef _WIN32
if (!IsProcessorFeaturePresent(MY_PF_XSAVE_ENABLED))
return False;
/* PF_AVX_INSTRUCTIONS_AVAILABLE probably is supported starting from
some latest Win10 revisions. But we need AVX in older Windows also.
So we don't use the following check: */
/*
if (!IsProcessorFeaturePresent(MY_PF_AVX_INSTRUCTIONS_AVAILABLE))
return False;
*/
#endif
/*
OS must use new special XSAVE/XRSTOR instructions to save
AVX registers when it required for context switching.
At OS statring:
OS sets CR4.OSXSAVE flag to signal the processor that OS supports the XSAVE extensions.
Also OS sets bitmask in XCR0 register that defines what
registers will be processed by XSAVE instruction:
XCR0.SSE[bit 0] - x87 registers and state
XCR0.SSE[bit 1] - SSE registers and state
XCR0.AVX[bit 2] - AVX registers and state
CR4.OSXSAVE is reflected to CPUID.1:ECX.OSXSAVE[bit 27].
So we can read that bit in user-space.
XCR0 is available for reading in user-space by new XGETBV instruction.
*/
{
const UInt32 c = x86cpuid_Func_1_ECX();
if (0 == (1
& (c >> 28) // AVX instructions are supported by hardware
& (c >> 27))) // OSXSAVE bit: XSAVE and related instructions are enabled by OS.
return False;
}
/* also we can check
CPUID.1:ECX.XSAVE [bit 26] : that shows that
XSAVE, XRESTOR, XSETBV, XGETBV instructions are supported by hardware.
But that check is redundant, because if OSXSAVE bit is set, then XSAVE is also set */
/* If OS have enabled XSAVE extension instructions (OSXSAVE == 1),
in most cases we expect that OS also will support storing/restoring
for AVX and SSE states at least.
But to be ensure for that we call user-space instruction
XGETBV(0) to get XCR0 value that contains bitmask that defines
what exact states(registers) OS have enabled for storing/restoring.
*/
{
const UInt32 bm = (UInt32)x86_xgetbv_0(MY_XCR_XFEATURE_ENABLED_MASK);
// printf("\n=== XGetBV=0x%x\n", bm);
return 1
& (BoolInt)(bm >> 1) // SSE state is supported (set by OS) for storing/restoring
& (BoolInt)(bm >> 2); // AVX state is supported (set by OS) for storing/restoring
}
// since Win7SP1: we can use GetEnabledXStateFeatures();
}
BoolInt CPU_IsSupported_AVX2(void)
{
if (!CPU_IsSupported_AVX())
return False;
if (z7_x86_cpuid_GetMaxFunc() < 7)
return False;
{
UInt32 d[4];
z7_x86_cpuid(d, 7);
// printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
return 1
& (BoolInt)(d[1] >> 5); // avx2
}
}
#if 0
BoolInt CPU_IsSupported_AVX512F_AVX512VL(void)
{
if (!CPU_IsSupported_AVX())
return False;
if (z7_x86_cpuid_GetMaxFunc() < 7)
return False;
{
UInt32 d[4];
BoolInt v;
z7_x86_cpuid(d, 7);
// printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
v = 1
& (BoolInt)(d[1] >> 16) // avx512f
& (BoolInt)(d[1] >> 31); // avx512vl
if (!v)
return False;
}
{
const UInt32 bm = (UInt32)x86_xgetbv_0(MY_XCR_XFEATURE_ENABLED_MASK);
// printf("\n=== XGetBV=0x%x\n", bm);
return 1
& (BoolInt)(bm >> 5) // OPMASK
& (BoolInt)(bm >> 6) // ZMM upper 256-bit
& (BoolInt)(bm >> 7); // ZMM16 ... ZMM31
}
}
#endif
BoolInt CPU_IsSupported_VAES_AVX2(void)
{
if (!CPU_IsSupported_AVX())
return False;
if (z7_x86_cpuid_GetMaxFunc() < 7)
return False;
{
UInt32 d[4];
z7_x86_cpuid(d, 7);
// printf("\ncpuid(7): ebx=%8x ecx=%8x\n", d[1], d[2]);
return 1
& (BoolInt)(d[1] >> 5) // avx2
// & (d[1] >> 31) // avx512vl
& (BoolInt)(d[2] >> 9); // vaes // VEX-256/EVEX
}
}
BoolInt CPU_IsSupported_PageGB(void)
{
CHECK_CPUID_IS_SUPPORTED
{
UInt32 d[4];
z7_x86_cpuid(d, 0x80000000);
if (d[0] < 0x80000001)
return False;
z7_x86_cpuid(d, 0x80000001);
return (BoolInt)(d[3] >> 26) & 1;
}
}
#elif defined(MY_CPU_ARM_OR_ARM64)
#ifdef _WIN32
#include "7zWindows.h"
BoolInt CPU_IsSupported_CRC32(void) { return IsProcessorFeaturePresent(PF_ARM_V8_CRC32_INSTRUCTIONS_AVAILABLE) ? 1 : 0; }
BoolInt CPU_IsSupported_CRYPTO(void) { return IsProcessorFeaturePresent(PF_ARM_V8_CRYPTO_INSTRUCTIONS_AVAILABLE) ? 1 : 0; }
BoolInt CPU_IsSupported_NEON(void) { return IsProcessorFeaturePresent(PF_ARM_NEON_INSTRUCTIONS_AVAILABLE) ? 1 : 0; }
#else
#if defined(__APPLE__)
/*
#include <stdio.h>
#include <string.h>
static void Print_sysctlbyname(const char *name)
{
size_t bufSize = 256;
char buf[256];
int res = sysctlbyname(name, &buf, &bufSize, NULL, 0);
{
int i;
printf("\nres = %d : %s : '%s' : bufSize = %d, numeric", res, name, buf, (unsigned)bufSize);
for (i = 0; i < 20; i++)
printf(" %2x", (unsigned)(Byte)buf[i]);
}
}
*/
/*
Print_sysctlbyname("hw.pagesize");
Print_sysctlbyname("machdep.cpu.brand_string");
*/
static BoolInt z7_sysctlbyname_Get_BoolInt(const char *name)
{
UInt32 val = 0;
if (z7_sysctlbyname_Get_UInt32(name, &val) == 0 && val == 1)
return 1;
return 0;
}
BoolInt CPU_IsSupported_CRC32(void)
{
return z7_sysctlbyname_Get_BoolInt("hw.optional.armv8_crc32");
}
BoolInt CPU_IsSupported_NEON(void)
{
return z7_sysctlbyname_Get_BoolInt("hw.optional.neon");
}
BoolInt CPU_IsSupported_SHA512(void)
{
return z7_sysctlbyname_Get_BoolInt("hw.optional.armv8_2_sha512");
}
/*
BoolInt CPU_IsSupported_SHA3(void)
{
return z7_sysctlbyname_Get_BoolInt("hw.optional.armv8_2_sha3");
}
*/
#ifdef MY_CPU_ARM64
#define APPLE_CRYPTO_SUPPORT_VAL 1
#else
#define APPLE_CRYPTO_SUPPORT_VAL 0
#endif
BoolInt CPU_IsSupported_SHA1(void) { return APPLE_CRYPTO_SUPPORT_VAL; }
BoolInt CPU_IsSupported_SHA2(void) { return APPLE_CRYPTO_SUPPORT_VAL; }
BoolInt CPU_IsSupported_AES (void) { return APPLE_CRYPTO_SUPPORT_VAL; }
#else // __APPLE__
#if defined(__GLIBC__) && (__GLIBC__ * 100 + __GLIBC_MINOR__ >= 216)
#define Z7_GETAUXV_AVAILABLE
#elif !defined(__QNXNTO__)
// #pragma message("=== is not NEW GLIBC === ")
#if defined __has_include
#if __has_include (<sys/auxv.h>)
// #pragma message("=== sys/auxv.h is avail=== ")
#define Z7_GETAUXV_AVAILABLE
#endif
#endif
#endif
#ifdef Z7_GETAUXV_AVAILABLE
// #pragma message("=== Z7_GETAUXV_AVAILABLE === ")
#include <sys/auxv.h>
#define USE_HWCAP
#endif
#ifdef USE_HWCAP
#if defined(__FreeBSD__) || defined(__OpenBSD__)
static unsigned long MY_getauxval(int aux)
{
unsigned long val;
if (elf_aux_info(aux, &val, sizeof(val)))
return 0;
return val;
}
#else
#define MY_getauxval getauxval
#if defined __has_include
#if __has_include (<asm/hwcap.h>)
#include <asm/hwcap.h>
#endif
#endif
#endif
#define MY_HWCAP_CHECK_FUNC_2(name1, name2) \
BoolInt CPU_IsSupported_ ## name1(void) { return (MY_getauxval(AT_HWCAP) & (HWCAP_ ## name2)); }
#ifdef MY_CPU_ARM64
#define MY_HWCAP_CHECK_FUNC(name) \
MY_HWCAP_CHECK_FUNC_2(name, name)
#if 1 || defined(__ARM_NEON)
BoolInt CPU_IsSupported_NEON(void) { return True; }
#else
MY_HWCAP_CHECK_FUNC_2(NEON, ASIMD)
#endif
// MY_HWCAP_CHECK_FUNC (ASIMD)
#elif defined(MY_CPU_ARM)
#define MY_HWCAP_CHECK_FUNC(name) \
BoolInt CPU_IsSupported_ ## name(void) { return (MY_getauxval(AT_HWCAP2) & (HWCAP2_ ## name)); }
MY_HWCAP_CHECK_FUNC_2(NEON, NEON)
#endif
#else // USE_HWCAP
#define MY_HWCAP_CHECK_FUNC(name) \
BoolInt CPU_IsSupported_ ## name(void) { return 0; }
#if defined(__ARM_NEON)
BoolInt CPU_IsSupported_NEON(void) { return True; }
#else
MY_HWCAP_CHECK_FUNC(NEON)
#endif
#endif // USE_HWCAP
MY_HWCAP_CHECK_FUNC (CRC32)
MY_HWCAP_CHECK_FUNC (SHA1)
MY_HWCAP_CHECK_FUNC (SHA2)
MY_HWCAP_CHECK_FUNC (AES)
#ifdef MY_CPU_ARM64
// <hwcap.h> supports HWCAP_SHA512 and HWCAP_SHA3 since 2017.
// we define them here, if they are not defined
#ifndef HWCAP_SHA3
// #define HWCAP_SHA3 (1 << 17)
#endif
#ifndef HWCAP_SHA512
// #pragma message("=== HWCAP_SHA512 define === ")
#define HWCAP_SHA512 (1 << 21)
#endif
MY_HWCAP_CHECK_FUNC (SHA512)
// MY_HWCAP_CHECK_FUNC (SHA3)
#endif
#endif // __APPLE__
#endif // _WIN32
#endif // MY_CPU_ARM_OR_ARM64
#ifdef __APPLE__
#include <sys/sysctl.h>
int z7_sysctlbyname_Get(const char *name, void *buf, size_t *bufSize)
{
return sysctlbyname(name, buf, bufSize, NULL, 0);
}
int z7_sysctlbyname_Get_UInt32(const char *name, UInt32 *val)
{
size_t bufSize = sizeof(*val);
const int res = z7_sysctlbyname_Get(name, val, &bufSize);
if (res == 0 && bufSize != sizeof(*val))
return EFAULT;
return res;
}
#endif
third_party/7z/CpuArch.h
Vendored
+715
View File
@@ -0,0 +1,715 @@
/* CpuArch.h -- CPU specific code
Igor Pavlov : Public domain */
#ifndef ZIP7_INC_CPU_ARCH_H
#define ZIP7_INC_CPU_ARCH_H
#include "7zTypes.h"
EXTERN_C_BEGIN
/*
MY_CPU_LE means that CPU is LITTLE ENDIAN.
MY_CPU_BE means that CPU is BIG ENDIAN.
If MY_CPU_LE and MY_CPU_BE are not defined, we don't know about ENDIANNESS of platform.
MY_CPU_LE_UNALIGN means that CPU is LITTLE ENDIAN and CPU supports unaligned memory accesses.
MY_CPU_64BIT means that processor can work with 64-bit registers.
MY_CPU_64BIT can be used to select fast code branch
MY_CPU_64BIT doesn't mean that (sizeof(void *) == 8)
*/
#if !defined(_M_ARM64EC)
#if defined(_M_X64) \
|| defined(_M_AMD64) \
|| defined(__x86_64__) \
|| defined(__AMD64__) \
|| defined(__amd64__)
#define MY_CPU_AMD64
#ifdef __ILP32__
#define MY_CPU_NAME "x32"
#define MY_CPU_SIZEOF_POINTER 4
#else
#if defined(__APX_EGPR__) || defined(__EGPR__)
#define MY_CPU_NAME "x64-apx"
#define MY_CPU_AMD64_APX
#else
#define MY_CPU_NAME "x64"
#endif
#define MY_CPU_SIZEOF_POINTER 8
#endif
#define MY_CPU_64BIT
#endif
#endif
#if defined(_M_IX86) \
|| defined(__i386__)
#define MY_CPU_X86
#define MY_CPU_NAME "x86"
/* #define MY_CPU_32BIT */
#define MY_CPU_SIZEOF_POINTER 4
#endif
#if defined(__SSE2__) \
|| defined(MY_CPU_AMD64) \
|| defined(_M_IX86_FP) && (_M_IX86_FP >= 2)
#define MY_CPU_SSE2
#endif
#if defined(_M_ARM64) \
|| defined(_M_ARM64EC) \
|| defined(__AARCH64EL__) \
|| defined(__AARCH64EB__) \
|| defined(__aarch64__)
#define MY_CPU_ARM64
#if defined(__ILP32__) \
|| defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 4)
#define MY_CPU_NAME "arm64-32"
#define MY_CPU_SIZEOF_POINTER 4
#elif defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 16)
#define MY_CPU_NAME "arm64-128"
#define MY_CPU_SIZEOF_POINTER 16
#else
#if defined(_M_ARM64EC)
#define MY_CPU_NAME "arm64ec"
#else
#define MY_CPU_NAME "arm64"
#endif
#define MY_CPU_SIZEOF_POINTER 8
#endif
#define MY_CPU_64BIT
#endif
#if defined(_M_ARM) \
|| defined(_M_ARM_NT) \
|| defined(_M_ARMT) \
|| defined(__arm__) \
|| defined(__thumb__) \
|| defined(__ARMEL__) \
|| defined(__ARMEB__) \
|| defined(__THUMBEL__) \
|| defined(__THUMBEB__)
#define MY_CPU_ARM
#if defined(__thumb__) || defined(__THUMBEL__) || defined(_M_ARMT)
#define MY_CPU_ARMT
#define MY_CPU_NAME "armt"
#else
#define MY_CPU_ARM32
#define MY_CPU_NAME "arm"
#endif
/* #define MY_CPU_32BIT */
#define MY_CPU_SIZEOF_POINTER 4
#endif
#if defined(_M_IA64) \
|| defined(__ia64__)
#define MY_CPU_IA64
#define MY_CPU_NAME "ia64"
#define MY_CPU_64BIT
#endif
#if defined(__mips64) \
|| defined(__mips64__) \
|| (defined(__mips) && (__mips == 64 || __mips == 4 || __mips == 3))
#define MY_CPU_NAME "mips64"
#define MY_CPU_64BIT
#elif defined(__mips__)
#define MY_CPU_NAME "mips"
/* #define MY_CPU_32BIT */
#endif
#if defined(__ppc64__) \
|| defined(__powerpc64__) \
|| defined(__ppc__) \
|| defined(__powerpc__) \
|| defined(__PPC__) \
|| defined(_POWER)
#define MY_CPU_PPC_OR_PPC64
#if defined(__ppc64__) \
|| defined(__powerpc64__) \
|| defined(_LP64) \
|| defined(__64BIT__)
#ifdef __ILP32__
#define MY_CPU_NAME "ppc64-32"
#define MY_CPU_SIZEOF_POINTER 4
#else
#define MY_CPU_NAME "ppc64"
#define MY_CPU_SIZEOF_POINTER 8
#endif
#define MY_CPU_64BIT
#else
#define MY_CPU_NAME "ppc"
#define MY_CPU_SIZEOF_POINTER 4
/* #define MY_CPU_32BIT */
#endif
#endif
#if defined(__sparc__) \
|| defined(__sparc)
#define MY_CPU_SPARC
#if defined(__LP64__) \
|| defined(_LP64) \
|| defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 8)
#define MY_CPU_NAME "sparcv9"
#define MY_CPU_SIZEOF_POINTER 8
#define MY_CPU_64BIT
#elif defined(__sparc_v9__) \
|| defined(__sparcv9)
#define MY_CPU_64BIT
#if defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 4)
#define MY_CPU_NAME "sparcv9-32"
#else
#define MY_CPU_NAME "sparcv9m"
#endif
#elif defined(__sparc_v8__) \
|| defined(__sparcv8)
#define MY_CPU_NAME "sparcv8"
#define MY_CPU_SIZEOF_POINTER 4
#else
#define MY_CPU_NAME "sparc"
#endif
#endif
#if defined(__riscv) \
|| defined(__riscv__)
#define MY_CPU_RISCV
#if __riscv_xlen == 32
#define MY_CPU_NAME "riscv32"
#elif __riscv_xlen == 64
#define MY_CPU_NAME "riscv64"
#else
#define MY_CPU_NAME "riscv"
#endif
#endif
#if defined(__loongarch__)
#define MY_CPU_LOONGARCH
#if defined(__loongarch64) || defined(__loongarch_grlen) && (__loongarch_grlen == 64)
#define MY_CPU_64BIT
#endif
#if defined(__loongarch64)
#define MY_CPU_NAME "loongarch64"
#define MY_CPU_LOONGARCH64
#else
#define MY_CPU_NAME "loongarch"
#endif
#endif
#if defined(__s390__)
#if defined(__s390x__) || defined(__LP64__) || (defined(__SIZEOF_POINTER__) && __SIZEOF_POINTER__ == 8)
#define MY_CPU_NAME "s390x"
#define MY_CPU_SIZEOF_POINTER 8
#else
#define MY_CPU_NAME "s390"
#define MY_CPU_SIZEOF_POINTER 4
#endif
#endif
// #undef MY_CPU_NAME
// #undef MY_CPU_SIZEOF_POINTER
// #define __e2k__
// #define __SIZEOF_POINTER__ 4
#if defined(__e2k__)
#define MY_CPU_E2K
#if defined(__ILP32__) || defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 4)
#define MY_CPU_NAME "e2k-32"
#define MY_CPU_SIZEOF_POINTER 4
#else
#define MY_CPU_NAME "e2k"
#if defined(__LP64__) || defined(__SIZEOF_POINTER__) && (__SIZEOF_POINTER__ == 8)
#define MY_CPU_SIZEOF_POINTER 8
#endif
#endif
#define MY_CPU_64BIT
#endif
#if defined(MY_CPU_X86) || defined(MY_CPU_AMD64)
#define MY_CPU_X86_OR_AMD64
#endif
#if defined(MY_CPU_ARM) || defined(MY_CPU_ARM64)
#define MY_CPU_ARM_OR_ARM64
#endif
#ifdef _WIN32
#ifdef MY_CPU_ARM
#define MY_CPU_ARM_LE
#endif
#ifdef MY_CPU_ARM64
#define MY_CPU_ARM64_LE
#endif
#ifdef _M_IA64
#define MY_CPU_IA64_LE
#endif
#endif
// _LITTLE_ENDIAN macro can be defined for big-endian platform with some compilers
#if defined(MY_CPU_X86_OR_AMD64) \
|| defined(MY_CPU_ARM_LE) \
|| defined(MY_CPU_ARM64_LE) \
|| defined(MY_CPU_IA64_LE) \
|| defined(__LITTLE_ENDIAN__) \
|| defined(__ARMEL__) \
|| defined(__THUMBEL__) \
|| defined(__AARCH64EL__) \
|| defined(__MIPSEL__) \
|| defined(__MIPSEL) \
|| defined(_MIPSEL) \
|| defined(__BFIN__) \
|| (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__))
#define MY_CPU_LE
#endif
#if defined(__BIG_ENDIAN__) \
|| defined(__ARMEB__) \
|| defined(__THUMBEB__) \
|| defined(__AARCH64EB__) \
|| defined(__MIPSEB__) \
|| defined(__MIPSEB) \
|| defined(_MIPSEB) \
|| defined(__m68k__) \
|| defined(__s390__) \
|| defined(__s390x__) \
|| defined(__zarch__) \
|| (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
#define MY_CPU_BE
#endif
#if defined(MY_CPU_LE) && defined(MY_CPU_BE)
#error Stop_Compiling_Bad_Endian
#endif
#if !defined(MY_CPU_LE) && !defined(MY_CPU_BE)
#error Stop_Compiling_CPU_ENDIAN_must_be_detected_at_compile_time
#endif
#if defined(MY_CPU_32BIT) && defined(MY_CPU_64BIT)
#error Stop_Compiling_Bad_32_64_BIT
#endif
#ifdef __SIZEOF_POINTER__
#ifdef MY_CPU_SIZEOF_POINTER
#if MY_CPU_SIZEOF_POINTER != __SIZEOF_POINTER__
#error Stop_Compiling_Bad_MY_CPU_PTR_SIZE
#endif
#else
#define MY_CPU_SIZEOF_POINTER __SIZEOF_POINTER__
#endif
#endif
#if defined(MY_CPU_SIZEOF_POINTER) && (MY_CPU_SIZEOF_POINTER == 4)
#if defined (_LP64)
#error Stop_Compiling_Bad_MY_CPU_PTR_SIZE
#endif
#endif
#ifdef _MSC_VER
#if _MSC_VER >= 1300
#define MY_CPU_pragma_pack_push_1 __pragma(pack(push, 1))
#define MY_CPU_pragma_pop __pragma(pack(pop))
#else
#define MY_CPU_pragma_pack_push_1
#define MY_CPU_pragma_pop
#endif
#else
#ifdef __xlC__
#define MY_CPU_pragma_pack_push_1 _Pragma("pack(1)")
#define MY_CPU_pragma_pop _Pragma("pack()")
#else
#define MY_CPU_pragma_pack_push_1 _Pragma("pack(push, 1)")
#define MY_CPU_pragma_pop _Pragma("pack(pop)")
#endif
#endif
#ifndef MY_CPU_NAME
// #define MY_CPU_IS_UNKNOWN
#ifdef MY_CPU_LE
#define MY_CPU_NAME "LE"
#elif defined(MY_CPU_BE)
#define MY_CPU_NAME "BE"
#else
/*
#define MY_CPU_NAME ""
*/
#endif
#endif
#ifdef __has_builtin
#define Z7_has_builtin(x) __has_builtin(x)
#else
#define Z7_has_builtin(x) 0
#endif
#define Z7_BSWAP32_CONST(v) \
( (((UInt32)(v) << 24) ) \
| (((UInt32)(v) << 8) & (UInt32)0xff0000) \
| (((UInt32)(v) >> 8) & (UInt32)0xff00 ) \
| (((UInt32)(v) >> 24) ))
#if defined(_MSC_VER) && (_MSC_VER >= 1300)
#include <stdlib.h>
/* Note: these macros will use bswap instruction (486), that is unsupported in 386 cpu */
#pragma intrinsic(_byteswap_ushort)
#pragma intrinsic(_byteswap_ulong)
#pragma intrinsic(_byteswap_uint64)
#define Z7_BSWAP16(v) _byteswap_ushort(v)
#define Z7_BSWAP32(v) _byteswap_ulong (v)
#define Z7_BSWAP64(v) _byteswap_uint64(v)
#define Z7_CPU_FAST_BSWAP_SUPPORTED
/* GCC can generate slow code that calls function for __builtin_bswap32() for:
- GCC for RISCV, if Zbb/XTHeadBb extension is not used.
- GCC for SPARC.
The code from CLANG for SPARC also is not fastest.
So we don't define Z7_CPU_FAST_BSWAP_SUPPORTED in some cases.
*/
#elif (!defined(MY_CPU_RISCV) || defined (__riscv_zbb) || defined(__riscv_xtheadbb)) \
&& !defined(MY_CPU_SPARC) \
&& ( \
(defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3))) \
|| (defined(__clang__) && Z7_has_builtin(__builtin_bswap16)) \
)
#define Z7_BSWAP16(v) __builtin_bswap16(v)
#define Z7_BSWAP32(v) __builtin_bswap32(v)
#define Z7_BSWAP64(v) __builtin_bswap64(v)
#define Z7_CPU_FAST_BSWAP_SUPPORTED
#else
#define Z7_BSWAP16(v) ((UInt16) \
( ((UInt32)(v) << 8) \
| ((UInt32)(v) >> 8) \
))
#define Z7_BSWAP32(v) Z7_BSWAP32_CONST(v)
#define Z7_BSWAP64(v) \
( ( ( (UInt64)(v) ) << 8 * 7 ) \
| ( ( (UInt64)(v) & ((UInt32)0xff << 8 * 1) ) << 8 * 5 ) \
| ( ( (UInt64)(v) & ((UInt32)0xff << 8 * 2) ) << 8 * 3 ) \
| ( ( (UInt64)(v) & ((UInt32)0xff << 8 * 3) ) << 8 * 1 ) \
| ( ( (UInt64)(v) >> 8 * 1 ) & ((UInt32)0xff << 8 * 3) ) \
| ( ( (UInt64)(v) >> 8 * 3 ) & ((UInt32)0xff << 8 * 2) ) \
| ( ( (UInt64)(v) >> 8 * 5 ) & ((UInt32)0xff << 8 * 1) ) \
| ( ( (UInt64)(v) >> 8 * 7 ) ) \
)
#endif
#ifdef MY_CPU_LE
#if defined(MY_CPU_X86_OR_AMD64) \
|| defined(MY_CPU_ARM64) \
|| defined(MY_CPU_RISCV) && defined(__riscv_misaligned_fast) \
|| defined(MY_CPU_E2K) && defined(__iset__) && (__iset__ >= 6)
#define MY_CPU_LE_UNALIGN
#define MY_CPU_LE_UNALIGN_64
#elif defined(__ARM_FEATURE_UNALIGNED)
/* === ALIGNMENT on 32-bit arm and LDRD/STRD/LDM/STM instructions.
Description of problems:
problem-1 : 32-bit ARM architecture:
multi-access (pair of 32-bit accesses) instructions (LDRD/STRD/LDM/STM)
require 32-bit (WORD) alignment (by 32-bit ARM architecture).
So there is "Alignment fault exception", if data is not aligned for 32-bit.
problem-2 : 32-bit kernels and arm64 kernels:
32-bit linux kernels provide fixup for these "paired" instruction "Alignment fault exception".
So unaligned paired-access instructions work via exception handler in kernel in 32-bit linux.
But some arm64 kernels do not handle these faults in 32-bit programs.
So we have unhandled exception for such instructions.
Probably some new arm64 kernels have fixed it, and unaligned
paired-access instructions work in new kernels?
problem-3 : compiler for 32-bit arm:
Compilers use LDRD/STRD/LDM/STM for UInt64 accesses
and for another cases where two 32-bit accesses are fused
to one multi-access instruction.
So UInt64 variables must be aligned for 32-bit, and each
32-bit access must be aligned for 32-bit, if we want to
avoid "Alignment fault" exception (handled or unhandled).
problem-4 : performace:
Even if unaligned access is handled by kernel, it will be slow.
So if we allow unaligned access, we can get fast unaligned
single-access, and slow unaligned paired-access.
We don't allow unaligned access on 32-bit arm, because compiler
genarates paired-access instructions that require 32-bit alignment,
and some arm64 kernels have no handler for these instructions.
Also unaligned paired-access instructions will be slow, if kernel handles them.
*/
// it must be disabled:
// #define MY_CPU_LE_UNALIGN
#endif
#endif
#ifdef MY_CPU_LE_UNALIGN
#define GetUi16(p) (*(const UInt16 *)(const void *)(p))
#define GetUi32(p) (*(const UInt32 *)(const void *)(p))
#ifdef MY_CPU_LE_UNALIGN_64
#define GetUi64(p) (*(const UInt64 *)(const void *)(p))
#define SetUi64(p, v) { *(UInt64 *)(void *)(p) = (v); }
#endif
#define SetUi16(p, v) { *(UInt16 *)(void *)(p) = (v); }
#define SetUi32(p, v) { *(UInt32 *)(void *)(p) = (v); }
#else
#define GetUi16(p) ( (UInt16) ( \
((const Byte *)(p))[0] | \
((UInt16)((const Byte *)(p))[1] << 8) ))
#define GetUi32(p) ( \
((const Byte *)(p))[0] | \
((UInt32)((const Byte *)(p))[1] << 8) | \
((UInt32)((const Byte *)(p))[2] << 16) | \
((UInt32)((const Byte *)(p))[3] << 24))
#define SetUi16(p, v) { Byte *_ppp_ = (Byte *)(p); UInt32 _vvv_ = (v); \
_ppp_[0] = (Byte)_vvv_; \
_ppp_[1] = (Byte)(_vvv_ >> 8); }
#define SetUi32(p, v) { Byte *_ppp_ = (Byte *)(p); UInt32 _vvv_ = (v); \
_ppp_[0] = (Byte)_vvv_; \
_ppp_[1] = (Byte)(_vvv_ >> 8); \
_ppp_[2] = (Byte)(_vvv_ >> 16); \
_ppp_[3] = (Byte)(_vvv_ >> 24); }
#endif
#ifndef GetUi64
#define GetUi64(p) (GetUi32(p) | ((UInt64)GetUi32(((const Byte *)(p)) + 4) << 32))
#endif
#ifndef SetUi64
#define SetUi64(p, v) { Byte *_ppp2_ = (Byte *)(p); UInt64 _vvv2_ = (v); \
SetUi32(_ppp2_ , (UInt32)_vvv2_) \
SetUi32(_ppp2_ + 4, (UInt32)(_vvv2_ >> 32)) }
#endif
#if defined(MY_CPU_LE_UNALIGN) && defined(Z7_CPU_FAST_BSWAP_SUPPORTED)
#if 0
// Z7_BSWAP16 can be slow for x86-msvc
#define GetBe16_to32(p) (Z7_BSWAP16 (*(const UInt16 *)(const void *)(p)))
#else
#define GetBe16_to32(p) (Z7_BSWAP32 (*(const UInt16 *)(const void *)(p)) >> 16)
#endif
#define GetBe32(p) Z7_BSWAP32 (*(const UInt32 *)(const void *)(p))
#define SetBe32(p, v) { (*(UInt32 *)(void *)(p)) = Z7_BSWAP32(v); }
#if defined(MY_CPU_LE_UNALIGN_64)
#define GetBe64(p) Z7_BSWAP64 (*(const UInt64 *)(const void *)(p))
#define SetBe64(p, v) { (*(UInt64 *)(void *)(p)) = Z7_BSWAP64(v); }
#endif
#else
#define GetBe32(p) ( \
((UInt32)((const Byte *)(p))[0] << 24) | \
((UInt32)((const Byte *)(p))[1] << 16) | \
((UInt32)((const Byte *)(p))[2] << 8) | \
((const Byte *)(p))[3] )
#define SetBe32(p, v) { Byte *_ppp_ = (Byte *)(p); UInt32 _vvv_ = (v); \
_ppp_[0] = (Byte)(_vvv_ >> 24); \
_ppp_[1] = (Byte)(_vvv_ >> 16); \
_ppp_[2] = (Byte)(_vvv_ >> 8); \
_ppp_[3] = (Byte)_vvv_; }
#endif
#ifndef GetBe64
#define GetBe64(p) (((UInt64)GetBe32(p) << 32) | GetBe32(((const Byte *)(p)) + 4))
#endif
#ifndef SetBe64
#define SetBe64(p, v) { Byte *_ppp_ = (Byte *)(p); UInt64 _vvv_ = (v); \
_ppp_[0] = (Byte)(_vvv_ >> 56); \
_ppp_[1] = (Byte)(_vvv_ >> 48); \
_ppp_[2] = (Byte)(_vvv_ >> 40); \
_ppp_[3] = (Byte)(_vvv_ >> 32); \
_ppp_[4] = (Byte)(_vvv_ >> 24); \
_ppp_[5] = (Byte)(_vvv_ >> 16); \
_ppp_[6] = (Byte)(_vvv_ >> 8); \
_ppp_[7] = (Byte)_vvv_; }
#endif
#ifndef GetBe16
#ifdef GetBe16_to32
#define GetBe16(p) ( (UInt16) GetBe16_to32(p))
#else
#define GetBe16(p) ( (UInt16) ( \
((UInt16)((const Byte *)(p))[0] << 8) | \
((const Byte *)(p))[1] ))
#endif
#endif
#if defined(MY_CPU_BE)
#define Z7_CONV_BE_TO_NATIVE_CONST32(v) (v)
#define Z7_CONV_LE_TO_NATIVE_CONST32(v) Z7_BSWAP32_CONST(v)
#define Z7_CONV_NATIVE_TO_BE_32(v) (v)
// #define Z7_GET_NATIVE16_FROM_2_BYTES(b0, b1) ((b1) | ((b0) << 8))
#elif defined(MY_CPU_LE)
#define Z7_CONV_BE_TO_NATIVE_CONST32(v) Z7_BSWAP32_CONST(v)
#define Z7_CONV_LE_TO_NATIVE_CONST32(v) (v)
#define Z7_CONV_NATIVE_TO_BE_32(v) Z7_BSWAP32(v)
// #define Z7_GET_NATIVE16_FROM_2_BYTES(b0, b1) ((b0) | ((b1) << 8))
#else
#error Stop_Compiling_Unknown_Endian_CONV
#endif
#if defined(MY_CPU_BE)
#define GetBe64a(p) (*(const UInt64 *)(const void *)(p))
#define GetBe32a(p) (*(const UInt32 *)(const void *)(p))
#define GetBe16a(p) (*(const UInt16 *)(const void *)(p))
#define SetBe32a(p, v) { *(UInt32 *)(void *)(p) = (v); }
#define SetBe16a(p, v) { *(UInt16 *)(void *)(p) = (v); }
// gcc and clang for powerpc can transform load byte access to load reverse word access.
// sp we can use byte access instead of word access. Z7_BSWAP64 cab be slow
#if 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED) && defined(MY_CPU_64BIT)
#define GetUi64a(p) Z7_BSWAP64 (*(const UInt64 *)(const void *)(p))
#else
#define GetUi64a(p) GetUi64(p)
#endif
#if 1 && defined(Z7_CPU_FAST_BSWAP_SUPPORTED)
#define GetUi32a(p) Z7_BSWAP32 (*(const UInt32 *)(const void *)(p))
#else
#define GetUi32a(p) GetUi32(p)
#endif
#define GetUi16a(p) GetUi16(p)
#define SetUi32a(p, v) SetUi32(p, v)
#define SetUi16a(p, v) SetUi16(p, v)
#elif defined(MY_CPU_LE)
#define GetUi64a(p) (*(const UInt64 *)(const void *)(p))
#define GetUi32a(p) (*(const UInt32 *)(const void *)(p))
#define GetUi16a(p) (*(const UInt16 *)(const void *)(p))
#define SetUi32a(p, v) { *(UInt32 *)(void *)(p) = (v); }
#define SetUi16a(p, v) { *(UInt16 *)(void *)(p) = (v); }
#define GetBe64a(p) GetBe64(p)
#define GetBe32a(p) GetBe32(p)
#define GetBe16a(p) GetBe16(p)
#define SetBe32a(p, v) SetBe32(p, v)
#define SetBe16a(p, v) SetBe16(p, v)
#else
#error Stop_Compiling_Unknown_Endian_CPU_a
#endif
#ifndef GetBe16_to32
#define GetBe16_to32(p) GetBe16(p)
#endif
#if defined(MY_CPU_X86_OR_AMD64) \
|| defined(MY_CPU_ARM_OR_ARM64) \
|| defined(MY_CPU_PPC_OR_PPC64)
#define Z7_CPU_FAST_ROTATE_SUPPORTED
#endif
#ifdef MY_CPU_X86_OR_AMD64
void Z7_FASTCALL z7_x86_cpuid(UInt32 a[4], UInt32 function);
UInt32 Z7_FASTCALL z7_x86_cpuid_GetMaxFunc(void);
#if defined(MY_CPU_AMD64)
#define Z7_IF_X86_CPUID_SUPPORTED
#else
#define Z7_IF_X86_CPUID_SUPPORTED if (z7_x86_cpuid_GetMaxFunc())
#endif
BoolInt CPU_IsSupported_AES(void);
BoolInt CPU_IsSupported_AVX(void);
BoolInt CPU_IsSupported_AVX2(void);
BoolInt CPU_IsSupported_AVX512F_AVX512VL(void);
BoolInt CPU_IsSupported_VAES_AVX2(void);
BoolInt CPU_IsSupported_CMOV(void);
BoolInt CPU_IsSupported_SSE(void);
BoolInt CPU_IsSupported_SSE2(void);
BoolInt CPU_IsSupported_SSSE3(void);
BoolInt CPU_IsSupported_SSE41(void);
BoolInt CPU_IsSupported_SHA(void);
BoolInt CPU_IsSupported_SHA512(void);
BoolInt CPU_IsSupported_PageGB(void);
#elif defined(MY_CPU_ARM_OR_ARM64)
BoolInt CPU_IsSupported_CRC32(void);
BoolInt CPU_IsSupported_NEON(void);
#if defined(_WIN32)
BoolInt CPU_IsSupported_CRYPTO(void);
#define CPU_IsSupported_SHA1 CPU_IsSupported_CRYPTO
#define CPU_IsSupported_SHA2 CPU_IsSupported_CRYPTO
#define CPU_IsSupported_AES CPU_IsSupported_CRYPTO
#else
BoolInt CPU_IsSupported_SHA1(void);
BoolInt CPU_IsSupported_SHA2(void);
BoolInt CPU_IsSupported_AES(void);
#endif
BoolInt CPU_IsSupported_SHA512(void);
#endif
#if defined(__APPLE__)
int z7_sysctlbyname_Get(const char *name, void *buf, size_t *bufSize);
int z7_sysctlbyname_Get_UInt32(const char *name, UInt32 *val);
#endif
EXTERN_C_END
#endif
third_party/7z/DOC/7zFormat.txt
Vendored
+469
View File
@@ -0,0 +1,469 @@
7z Format description (18.06)
----------------------------
This file contains description of 7z archive format.
7z archive can contain files compressed with any method.
See "Methods.txt" for description for defined compressing methods.
Format structure Overview
-------------------------
Some fields can be optional.
Archive structure
~~~~~~~~~~~~~~~~~
SignatureHeader
[PackedStreams]
[PackedStreamsForHeaders]
[
Header
or
{
Packed Header
HeaderInfo
}
]
Header structure
~~~~~~~~~~~~~~~~
{
ArchiveProperties
AdditionalStreams
{
PackInfo
{
PackPos
NumPackStreams
Sizes[NumPackStreams]
CRCs[NumPackStreams]
}
CodersInfo
{
NumFolders
Folders[NumFolders]
{
NumCoders
CodersInfo[NumCoders]
{
ID
NumInStreams;
NumOutStreams;
PropertiesSize
Properties[PropertiesSize]
}
NumBindPairs
BindPairsInfo[NumBindPairs]
{
InIndex;
OutIndex;
}
PackedIndices
}
UnPackSize[Folders][Folders.NumOutstreams]
CRCs[NumFolders]
}
SubStreamsInfo
{
NumUnPackStreamsInFolders[NumFolders];
UnPackSizes[]
CRCs[]
}
}
MainStreamsInfo
{
(Same as in AdditionalStreams)
}
FilesInfo
{
NumFiles
Properties[]
{
ID
Size
Data
}
}
}
HeaderInfo structure
~~~~~~~~~~~~~~~~~~~~
{
(Same as in AdditionalStreams)
}
Notes about Notation and encoding
---------------------------------
7z uses little endian encoding.
7z archive format has optional headers that are marked as
[]
Header
[]
REAL_UINT64 means real UINT64.
UINT64 means real UINT64 encoded with the following scheme:
Size of encoding sequence depends from first byte:
First_Byte Extra_Bytes Value
(binary)
0xxxxxxx : ( xxxxxxx )
10xxxxxx BYTE y[1] : ( xxxxxx << (8 * 1)) + y
110xxxxx BYTE y[2] : ( xxxxx << (8 * 2)) + y
...
1111110x BYTE y[6] : ( x << (8 * 6)) + y
11111110 BYTE y[7] : y
11111111 BYTE y[8] : y
Property IDs
------------
0x00 = kEnd
0x01 = kHeader
0x02 = kArchiveProperties
0x03 = kAdditionalStreamsInfo
0x04 = kMainStreamsInfo
0x05 = kFilesInfo
0x06 = kPackInfo
0x07 = kUnPackInfo
0x08 = kSubStreamsInfo
0x09 = kSize
0x0A = kCRC
0x0B = kFolder
0x0C = kCodersUnPackSize
0x0D = kNumUnPackStream
0x0E = kEmptyStream
0x0F = kEmptyFile
0x10 = kAnti
0x11 = kName
0x12 = kCTime
0x13 = kATime
0x14 = kMTime
0x15 = kWinAttributes
0x16 = kComment
0x17 = kEncodedHeader
0x18 = kStartPos
0x19 = kDummy
7z format headers
-----------------
SignatureHeader
~~~~~~~~~~~~~~~
BYTE kSignature[6] = {'7', 'z', 0xBC, 0xAF, 0x27, 0x1C};
ArchiveVersion
{
BYTE Major; // now = 0
BYTE Minor; // now = 4
};
UINT32 StartHeaderCRC;
StartHeader
{
REAL_UINT64 NextHeaderOffset
REAL_UINT64 NextHeaderSize
UINT32 NextHeaderCRC
}
...........................
ArchiveProperties
~~~~~~~~~~~~~~~~~
BYTE NID::kArchiveProperties (0x02)
for (;;)
{
BYTE PropertyType;
if (aType == 0)
break;
UINT64 PropertySize;
BYTE PropertyData[PropertySize];
}
Digests (NumStreams)
~~~~~~~~~~~~~~~~~~~~~
BYTE AllAreDefined
if (AllAreDefined == 0)
{
for(NumStreams)
BIT Defined
}
UINT32 CRCs[NumDefined]
PackInfo
~~~~~~~~~~~~
BYTE NID::kPackInfo (0x06)
UINT64 PackPos
UINT64 NumPackStreams
[]
BYTE NID::kSize (0x09)
UINT64 PackSizes[NumPackStreams]
[]
[]
BYTE NID::kCRC (0x0A)
PackStreamDigests[NumPackStreams]
[]
BYTE NID::kEnd
Folder
~~~~~~
UINT64 NumCoders;
for (NumCoders)
{
BYTE
{
0:3 CodecIdSize
4: Is Complex Coder
5: There Are Attributes
6: Reserved
7: There are more alternative methods. (Not used anymore, must be 0).
}
BYTE CodecId[CodecIdSize]
if (Is Complex Coder)
{
UINT64 NumInStreams;
UINT64 NumOutStreams;
}
if (There Are Attributes)
{
UINT64 PropertiesSize
BYTE Properties[PropertiesSize]
}
}
NumBindPairs = NumOutStreamsTotal - 1;
for (NumBindPairs)
{
UINT64 InIndex;
UINT64 OutIndex;
}
NumPackedStreams = NumInStreamsTotal - NumBindPairs;
if (NumPackedStreams > 1)
for(NumPackedStreams)
{
UINT64 Index;
};
Coders Info
~~~~~~~~~~~
BYTE NID::kUnPackInfo (0x07)
BYTE NID::kFolder (0x0B)
UINT64 NumFolders
BYTE External
switch(External)
{
case 0:
Folders[NumFolders]
case 1:
UINT64 DataStreamIndex
}
BYTE ID::kCodersUnPackSize (0x0C)
for(Folders)
for(Folder.NumOutStreams)
UINT64 UnPackSize;
[]
BYTE NID::kCRC (0x0A)
UnPackDigests[NumFolders]
[]
BYTE NID::kEnd
SubStreams Info
~~~~~~~~~~~~~~
BYTE NID::kSubStreamsInfo; (0x08)
[]
BYTE NID::kNumUnPackStream; (0x0D)
UINT64 NumUnPackStreamsInFolders[NumFolders];
[]
[]
BYTE NID::kSize (0x09)
UINT64 UnPackSizes[]
[]
[]
BYTE NID::kCRC (0x0A)
Digests[Number of streams with unknown CRC]
[]
BYTE NID::kEnd
Streams Info
~~~~~~~~~~~~
[]
PackInfo
[]
[]
CodersInfo
[]
[]
SubStreamsInfo
[]
BYTE NID::kEnd
FilesInfo
~~~~~~~~~
BYTE NID::kFilesInfo; (0x05)
UINT64 NumFiles
for (;;)
{
BYTE PropertyType;
if (aType == 0)
break;
UINT64 Size;
switch(PropertyType)
{
kEmptyStream: (0x0E)
for(NumFiles)
BIT IsEmptyStream
kEmptyFile: (0x0F)
for(EmptyStreams)
BIT IsEmptyFile
kAnti: (0x10)
for(EmptyStreams)
BIT IsAntiFile
case kCTime: (0x12)
case kATime: (0x13)
case kMTime: (0x14)
BYTE AllAreDefined
if (AllAreDefined == 0)
{
for(NumFiles)
BIT TimeDefined
}
BYTE External;
if(External != 0)
UINT64 DataIndex
[]
for(Defined Items)
REAL_UINT64 Time
[]
kNames: (0x11)
BYTE External;
if(External != 0)
UINT64 DataIndex
[]
for(Files)
{
wchar_t Names[NameSize];
wchar_t 0;
}
[]
kAttributes: (0x15)
BYTE AllAreDefined
if (AllAreDefined == 0)
{
for(NumFiles)
BIT AttributesAreDefined
}
BYTE External;
if(External != 0)
UINT64 DataIndex
[]
for(Defined Attributes)
UINT32 Attributes
[]
}
}
Header
~~~~~~
BYTE NID::kHeader (0x01)
[]
ArchiveProperties
[]
[]
BYTE NID::kAdditionalStreamsInfo; (0x03)
StreamsInfo
[]
[]
BYTE NID::kMainStreamsInfo; (0x04)
StreamsInfo
[]
[]
FilesInfo
[]
BYTE NID::kEnd
HeaderInfo
~~~~~~~~~~
[]
BYTE NID::kEncodedHeader; (0x17)
StreamsInfo for Encoded Header
[]
---
End of document
third_party/7z/DOC/Methods.txt
Vendored
+177
View File
@@ -0,0 +1,177 @@
7-Zip method IDs for 7z and xz archives
---------------------------------------
Version: 24.02
Date: 2024-03-22
Each compression or crypto method in 7z is associated with unique binary value (ID).
The length of ID in bytes is arbitrary but it can not exceed 63 bits (8 bytes).
xz and 7z formats use same ID map.
If you want to add some new ID, you have two ways:
1) Write request for allocating IDs to 7-Zip developers.
2) Generate 8-bytes ID:
3F ZZ ZZ ZZ ZZ ZZ MM MM
3F - Prefix for random IDs (1 byte)
ZZ ZZ ZZ ZZ ZZ - Developer ID (5 bytes). Use real random bytes.
MM MM - Method ID (2 bytes)
You can notify 7-Zip developers about your Developer ID / Method ID.
Note: Use new ID, if old codec can not decode data encoded with new version.
List of defined IDs
-------------------
00 - Copy
03 - Delta
04 - BCJ (x86)
05 - PPC (big-endian)
06 - IA64
07 - ARM (little-endian)
08 - ARMT (little-endian)
09 - SPARC
0A - ARM64
0B - RISCV
21 - LZMA2
02.. - Common
03 [Swap]
- 2 Swap2
- 4 Swap4
03.. - 7z
01 -
01 - LZMA
03 - [Branch Codecs]
01 - [x86 Codecs]
03 - BCJ
1B - BCJ2 (4 packed streams)
02 -
05 - PPC (big-endian)
03 -
01 - Alpha
04 -
01 - IA64
05 -
01 - ARM (little-endian)
06 -
05 - M68 (big-endian)
07 -
01 - ARMT (little-endian)
08 -
05 - SPARC
04 -
01 - PPMD
7F -
01 - experimental method.
04.. - Misc codecs
00 - Reserved
01 - [Zip]
00 - Copy (not used. Use {00} instead)
01 - Shrink
06 - Implode
08 - Deflate
09 - Deflate64
0A - Imploding
0C - BZip2 (not used. Use {040202} instead)
0E - LZMA (LZMA-zip)
5D - ZSTD
5F - xz
60 - Jpeg
61 - WavPack
62 - PPMd (PPMd-zip)
63 - wzAES
02 -
02 - BZip2
03 - [Rar]
01 - Rar1
02 - Rar2
03 - Rar3
05 - Rar5
04 - [Arj]
01 - Arj(1,2,3)
02 - Arj4
05 - [Z]
06 - [Lzh]
07 - Reserved for 7z
08 - [Cab]
09 - [NSIS]
01 - DeflateNSIS
02 - BZip2NSIS
F7 - External codecs (that are not included to 7-Zip)
0x xx - reserved
10 xx - reserved (LZHAM)
01 - LZHAM
11 xx - reserved (Tino Reichardt)
01 - ZSTD
02 - BROTLI
04 - LZ4
05 - LZ5
06 - LIZARD
12 xx - reserved (Denis Anisimov)
01 - WavPack2
FE - eSplitter
FF - RawSplitter
06.. - Crypto
F0 - Ciphers without hashing algo
01 - [AES]
0x - AES-128
4x - AES-192
8x - AES-256
Cx - AES
x0 - ECB
x1 - CBC
x2 - CFB
x3 - OFB
x4 - CTR
F1 - Combine Ciphers
01 - [Zip]
01 - ZipCrypto (Main Zip crypto algo)
03 - [RAR]
02 -
03 - Rar29AES (AES-128 + modified SHA-1)
07 - [7z]
01 - 7zAES (AES-256 + SHA-256)
---
End of document
third_party/7z/DOC/lzma-sdk.txt
Vendored
+437
View File
@@ -0,0 +1,437 @@
LZMA SDK 26.03
--------------
LZMA SDK provides the documentation, samples, header files,
libraries, and tools you need to develop applications that
use 7z / LZMA / LZMA2 / XZ compression.
LZMA is an improved version of famous LZ77 compression algorithm.
It was improved in way of maximum increasing of compression ratio,
keeping high decompression speed and low memory requirements for
decompressing.
LZMA2 is a LZMA based compression method. LZMA2 provides better
multithreading support for compression than LZMA and some other improvements.
7z is a file format for data compression and file archiving.
7z is a main file format for 7-Zip compression program (www.7-zip.org).
7z format supports different compression methods: LZMA, LZMA2 and others.
7z also supports AES-256 based encryption.
XZ is a file format for data compression that uses LZMA2 compression.
XZ format provides additional features: SHA/CRC check, filters for
improved compression ratio, splitting to blocks and streams,
LICENSE
-------
LZMA SDK is written and placed in the public domain by Igor Pavlov.
Some code in LZMA SDK is based on public domain code from another developers:
1) PPMd var.H (2001): Dmitry Shkarin
2) SHA-256: Wei Dai (Crypto++ library)
Anyone is free to copy, modify, publish, use, compile, sell, or distribute the
original LZMA SDK code, either in source code form or as a compiled binary, for
any purpose, commercial or non-commercial, and by any means.
LZMA SDK code is compatible with open source licenses, for example, you can
include it to GNU GPL or GNU LGPL code.
LZMA SDK Contents
-----------------
Source code:
- C / C++ / C# / Java - LZMA compression and decompression
- C / C++ - LZMA2 compression and decompression
- C / C++ - XZ compression and decompression
- C - 7z decompression
- C++ - 7z compression and decompression
- C - small SFXs for installers (7z decompression)
- C++ - SFXs and SFXs for installers (7z decompression)
Precomiled binaries:
- console programs for lzma / 7z / xz compression and decompression
- SFX modules for installers.
How to compile with makefile in Windows
---------------------------------------
Some macronames can be defined for compiling with makefile:
PLATFORM
with possible values: x64, x86, arm64, arm, ia64
OLD_COMPILER
for old VC compiler, like MSCV 6.0.
MY_DYNAMIC_LINK
for dynamic linking to the run-time library (msvcrt.dll).
The default makefile option is static linking to the run-time library.
To compile 7zr.exe file for x64 with Visual Studio 2022,
use the following command sequence:
cd SRC\CPP\7zip\Bundles\Alone7z\
%comspec% /k "C:\Program Files\VS2022\VC\Auxiliary\Build\vcvars64.bat"
nmake
You can use other "vcvars*.bat" files from the "VS2022\VC\Auxiliary\Build" directory
to compile for other platforms:
vcvars64.bat
vcvarsamd64_arm64.bat
vcvarsamd64_x86.bat
UNIX/Linux version
------------------
There are several otpions to compile 7-Zip with different compilers: gcc and clang.
Also 7-Zip code contains two versions for some critical parts of code: in C and in Assembeler.
So if you compile the version with Assembeler code, you will get faster 7-Zip binary.
7-Zip's assembler code uses the following syntax for different platforms:
1) x86 and x86-64 (AMD64): MASM syntax.
There are 2 programs that supports MASM syntax in Linux.
' 'Asmc Macro Assembler and JWasm. But JWasm now doesn't support some
cpu instructions used in 7-Zip.
So you must install Asmc Macro Assembler in Linux, if you want to compile fastest version
of 7-Zip x86 and x86-64:
https://github.com/nidud/asmc
2) arm64: GNU assembler for ARM64 with preprocessor.
That systax of that arm64 assembler code in 7-Zip is supported by GCC and CLANG for ARM64.
There are different binaries that can be compiled from 7-Zip source.
There are 2 main files in folder for compiling:
makefile - that can be used for compiling Windows version of 7-Zip with nmake command
makefile.gcc - that can be used for compiling Linux/macOS versions of 7-Zip with make command
At first you must change the current folder to folder that contains `makefile.gcc`:
cd CPP/7zip/Bundles/Alone7z
Then you can compile `makefile.gcc` with the command:
make -j -f makefile.gcc
Also there are additional "*.mak" files in folder "CPP/7zip/" that can be used to compile
7-Zip binaries with optimized code and optimzing options.
To compile with GCC without assembler:
cd CPP/7zip/Bundles/Alone7z
make -j -f ../../cmpl_gcc.mak
To compile with CLANG without assembler:
make -j -f ../../cmpl_clang.mak
To compile 7-Zip for x86-64 with asmc assembler:
make -j -f ../../cmpl_gcc_x64.mak
To compile 7-Zip for arm64 with assembler:
make -j -f ../../cmpl_gcc_arm64.mak
To compile 7-Zip for arm64 for macOS:
make -j -f ../../cmpl_mac_arm64.mak
Also you can change some compiler options in the mak files:
cmpl_gcc.mak
var_gcc.mak
warn_gcc.mak
Also you can use p7zip (port of 7-Zip for POSIX systems like Unix or Linux):
http://p7zip.sourceforge.net/
Files
-----
DOC/7zC.txt - 7z ANSI-C Decoder description
DOC/7zFormat.txt - 7z Format description
DOC/installer.txt - information about 7-Zip for installers
DOC/lzma.txt - LZMA compression description
DOC/lzma-sdk.txt - LZMA SDK description (this file)
DOC/lzma-history.txt - history of LZMA SDK
DOC/lzma-specification.txt - Specification of LZMA
DOC/Methods.txt - Compression method IDs for .7z
bin/installer/ - example script to create installer that uses SFX module,
bin/7zdec.exe - simplified 7z archive decoder (x86 32-bit version)
bin/7zr.exe - 7-Zip console program (reduced version) (x86 32-bit version)
bin/x64/7zr.exe - 7-Zip console program (reduced version) (x64 version)
bin/x64/7zdec.exe - simplified 7z archive decoder (x64 version)
bin/arm64/7zr.exe - 7-Zip console program (reduced version) (arm64 version)
bin/arm64/7zdec.exe - simplified 7z archive decoder (arm64 version)
bin/lzma.exe - file->file LZMA encoder/decoder for Windows
bin/7zS2.sfx - small SFX module for installers (GUI version)
bin/7zS2con.sfx - small SFX module for installers (Console version)
bin/7zSD.sfx - SFX module for installers.
7zDec.exe
---------
7zDec.exe is simplified 7z archive decoder.
It supports only LZMA, LZMA2, and PPMd methods.
7zDec decodes whole solid block from 7z archive to RAM.
The RAM consumption can be high.
Source code structure
---------------------
Asm/ - asm files (optimized code for CRC calculation and Intel-AES encryption)
C/ - C files (compression / decompression and other)
Util/
7z - 7z decoder program (decoding 7z files)
Lzma - LZMA program (file->file LZMA encoder/decoder).
LzmaLib - LZMA library (.DLL for Windows)
SfxSetup - small SFX module for installers
CPP/ -- CPP files
Common - common files for C++ projects
Windows - common files for Windows related code
7zip - files related to 7-Zip
Archive - files related to archiving
Common - common files for archive handling
7z - 7z C++ Encoder/Decoder
Bundles - Modules that are bundles of other modules (files)
Alone7z - 7zr.exe: Standalone 7-Zip console program (reduced version)
Format7zExtractR - 7zxr.dll: Reduced version of 7z DLL: extracting from 7z/LZMA/BCJ/BCJ2.
Format7zR - 7zr.dll: Reduced version of 7z DLL: extracting/compressing to 7z/LZMA/BCJ/BCJ2
LzmaCon - lzma.exe: LZMA compression/decompression
LzmaSpec - example code for LZMA Specification
SFXCon - 7zCon.sfx: Console 7z SFX module
SFXSetup - 7zS.sfx: 7z SFX module for installers
SFXWin - 7z.sfx: GUI 7z SFX module
Common - common files for 7-Zip
Compress - files for compression/decompression
Crypto - files for encryption / decompression
UI - User Interface files
Client7z - Test application for 7za.dll, 7zr.dll, 7zxr.dll
Common - Common UI files
Console - Code for console program (7z.exe)
Explorer - Some code from 7-Zip Shell extension
FileManager - Some GUI code from 7-Zip File Manager
GUI - Some GUI code from 7-Zip
CS/ - C# files
7zip
Common - some common files for 7-Zip
Compress - files related to compression/decompression
LZ - files related to LZ (Lempel-Ziv) compression algorithm
LZMA - LZMA compression/decompression
LzmaAlone - file->file LZMA compression/decompression
RangeCoder - Range Coder (special code of compression/decompression)
Java/ - Java files
SevenZip
Compression - files related to compression/decompression
LZ - files related to LZ (Lempel-Ziv) compression algorithm
LZMA - LZMA compression/decompression
RangeCoder - Range Coder (special code of compression/decompression)
Note:
Asm / C / C++ source code of LZMA SDK is part of 7-Zip's source code.
7-Zip's source code can be downloaded from 7-Zip's SourceForge page:
http://sourceforge.net/projects/sevenzip/
LZMA features
-------------
- Variable dictionary size (up to 4 GB)
- Estimated compressing speed: about 2 MB/s on 2 GHz CPU
- Estimated decompressing speed:
- 20-30 MB/s on modern 2 GHz cpu
- 1-2 MB/s on 200 MHz simple RISC cpu: (ARM, MIPS, PowerPC)
- Small memory requirements for decompressing (16 KB + DictionarySize)
- Small code size for decompressing: 5-8 KB
LZMA decoder uses only integer operations and can be
implemented in any modern 32-bit CPU (or on 16-bit CPU with some conditions).
Some critical operations that affect the speed of LZMA decompression:
1) 32*16 bit integer multiply
2) Mispredicted branches (penalty mostly depends from pipeline length)
3) 32-bit shift and arithmetic operations
The speed of LZMA decompressing mostly depends from CPU speed.
Memory speed has no big meaning. But if your CPU has small data cache,
overall weight of memory speed will slightly increase.
How To Use
----------
Using LZMA encoder/decoder executable
--------------------------------------
Usage: LZMA <e|d> inputFile outputFile [<switches>...]
e: encode file
d: decode file
b: Benchmark. There are two tests: compressing and decompressing
with LZMA method. Benchmark shows rating in MIPS (million
instructions per second). Rating value is calculated from
measured speed and it is normalized with Intel's Core 2 results.
Also Benchmark checks possible hardware errors (RAM
errors in most cases). Benchmark uses these settings:
(-a1, -d21, -fb32, -mfbt4). You can change only -d parameter.
Also you can change the number of iterations. Example for 30 iterations:
LZMA b 30
Default number of iterations is 10.
<Switches>
-a{N}: set compression mode 0 = fast, 1 = normal
default: 1 (normal)
d{N}: Sets Dictionary size - [0, 31], default: N=24 (32 MB)
The maximum value for dictionary size is N=31 (2 GB).
Dictionary size is calculated as DictionarySize = 2^N bytes.
For decompressing file compressed by LZMA method with dictionary
size D = 2^N you need about D bytes of memory (RAM).
-fb{N}: set number of fast bytes - [5, 273], default: 128
Usually big number gives a little bit better compression ratio
and slower compression process.
-lc{N}: set number of literal context bits - [0, 8], default: 3
Sometimes lc=4 gives gain for big files.
-lp{N}: set number of literal pos bits - [0, 4], default: 0
lp switch is intended for periodical data when period is
equal 2^N. For example, for 32-bit (4 bytes)
periodical data you can use lp=2. Often it's better to set lc0,
if you change lp switch.
-pb{N}: set number of pos bits - [0, 4], default: 2
pb switch is intended for periodical data
when period is equal 2^N.
-mf{MF_ID}: set Match Finder. Default: bt4.
Algorithms from hc* group doesn't provide good compression
ratio, but they often works pretty fast in combination with
fast mode (-a0).
Memory requirements depend from dictionary size
(parameter "d" in table below).
MF_ID Memory Description
bt2 d * 9.5 + 4MB Binary Tree with 2 bytes hashing.
bt3 d * 11.5 + 4MB Binary Tree with 3 bytes hashing.
bt4 d * 11.5 + 4MB Binary Tree with 4 bytes hashing.
bt5 d * 11.5 + 4MB Binary Tree with 5 bytes hashing.
hc4 d * 7.5 + 4MB Hash Chain with 4 bytes hashing.
hc5 d * 7.5 + 4MB Hash Chain with 5 bytes hashing.
-eos: write End Of Stream marker. By default LZMA doesn't write
eos marker, since LZMA decoder knows uncompressed size
stored in .lzma file header.
-si: Read data from stdin (it will write End Of Stream marker).
-so: Write data to stdout
Examples:
1) LZMA e file.bin file.lzma -d16 -lc0
compresses file.bin to file.lzma with 64 KB dictionary (2^16=64K)
and 0 literal context bits. -lc0 allows to reduce memory requirements
for decompression.
2) LZMA e file.bin file.lzma -lc0 -lp2
compresses file.bin to file.lzma with settings suitable
for 32-bit periodical data (for example, ARM or MIPS code).
3) LZMA d file.lzma file.bin
decompresses file.lzma to file.bin.
Compression ratio hints
-----------------------
Recommendations
---------------
To increase the compression ratio for LZMA compressing it's desirable
to have aligned data (if it's possible) and also it's desirable to locate
data in such order, where code is grouped in one place and data is
grouped in other place (it's better than such mixing: code, data, code,
data, ...).
Filters
-------
You can increase the compression ratio for some data types, using
special filters before compressing. For example, it's possible to
increase the compression ratio on 5-10% for code for those CPU ISAs:
x86, IA-64, ARM, ARM-Thumb, PowerPC, SPARC.
You can find C source code of such filters in C/Bra*.* files
You can check the compression ratio gain of these filters with such
7-Zip commands (example for ARM code):
No filter:
7z a a1.7z a.bin -m0=lzma
With filter for little-endian ARM code:
7z a a2.7z a.bin -m0=arm -m1=lzma
It works in such manner:
Compressing = Filter_encoding + LZMA_encoding
Decompressing = LZMA_decoding + Filter_decoding
Compressing and decompressing speed of such filters is very high,
so it will not increase decompressing time too much.
Moreover, it reduces decompression time for LZMA_decoding,
since compression ratio with filtering is higher.
These filters convert CALL (calling procedure) instructions
from relative offsets to absolute addresses, so such data becomes more
compressible.
For some ISAs (for example, for MIPS) it's impossible to get gain from such filter.
---
http://www.7-zip.org
http://www.7-zip.org/sdk.html
http://www.7-zip.org/support.html
third_party/7z/Delta.c
Vendored
+169
View File
@@ -0,0 +1,169 @@
/* Delta.c -- Delta converter
2021-02-09 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Delta.h"
void Delta_Init(Byte *state)
{
unsigned i;
for (i = 0; i < DELTA_STATE_SIZE; i++)
state[i] = 0;
}
void Delta_Encode(Byte *state, unsigned delta, Byte *data, SizeT size)
{
Byte temp[DELTA_STATE_SIZE];
if (size == 0)
return;
{
unsigned i = 0;
do
temp[i] = state[i];
while (++i != delta);
}
if (size <= delta)
{
unsigned i = 0, k;
do
{
Byte b = *data;
*data++ = (Byte)(b - temp[i]);
temp[i] = b;
}
while (++i != size);
k = 0;
do
{
if (i == delta)
i = 0;
state[k] = temp[i++];
}
while (++k != delta);
return;
}
{
Byte *p = data + size - delta;
{
unsigned i = 0;
do
state[i] = *p++;
while (++i != delta);
}
{
const Byte *lim = data + delta;
ptrdiff_t dif = -(ptrdiff_t)delta;
if (((ptrdiff_t)size + dif) & 1)
{
--p; *p = (Byte)(*p - p[dif]);
}
while (p != lim)
{
--p; *p = (Byte)(*p - p[dif]);
--p; *p = (Byte)(*p - p[dif]);
}
dif = -dif;
do
{
--p; *p = (Byte)(*p - temp[--dif]);
}
while (dif != 0);
}
}
}
void Delta_Decode(Byte *state, unsigned delta, Byte *data, SizeT size)
{
unsigned i;
const Byte *lim;
if (size == 0)
return;
i = 0;
lim = data + size;
if (size <= delta)
{
do
*data = (Byte)(*data + state[i++]);
while (++data != lim);
for (; delta != i; state++, delta--)
*state = state[i];
data -= i;
}
else
{
/*
#define B(n) b ## n
#define I(n) Byte B(n) = state[n];
#define U(n) { B(n) = (Byte)((B(n)) + *data++); data[-1] = (B(n)); }
#define F(n) if (data != lim) { U(n) }
if (delta == 1)
{
I(0)
if ((lim - data) & 1) { U(0) }
while (data != lim) { U(0) U(0) }
data -= 1;
}
else if (delta == 2)
{
I(0) I(1)
lim -= 1; while (data < lim) { U(0) U(1) }
lim += 1; F(0)
data -= 2;
}
else if (delta == 3)
{
I(0) I(1) I(2)
lim -= 2; while (data < lim) { U(0) U(1) U(2) }
lim += 2; F(0) F(1)
data -= 3;
}
else if (delta == 4)
{
I(0) I(1) I(2) I(3)
lim -= 3; while (data < lim) { U(0) U(1) U(2) U(3) }
lim += 3; F(0) F(1) F(2)
data -= 4;
}
else
*/
{
do
{
*data = (Byte)(*data + state[i++]);
data++;
}
while (i != delta);
{
ptrdiff_t dif = -(ptrdiff_t)delta;
do
*data = (Byte)(*data + data[dif]);
while (++data != lim);
data += dif;
}
}
}
do
*state++ = *data;
while (++data != lim);
}
third_party/7z/Delta.h
Vendored
+19
View File
@@ -0,0 +1,19 @@
/* Delta.h -- Delta converter
2023-03-03 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_DELTA_H
#define ZIP7_INC_DELTA_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#define DELTA_STATE_SIZE 256
void Delta_Init(Byte *state);
void Delta_Encode(Byte *state, unsigned delta, Byte *data, SizeT size);
void Delta_Decode(Byte *state, unsigned delta, Byte *data, SizeT size);
EXTERN_C_END
#endif
third_party/7z/DllSecur.c
Vendored
+99
View File
@@ -0,0 +1,99 @@
/* DllSecur.c -- DLL loading security
: Igor Pavlov : Public domain */
#include "Precomp.h"
#ifdef _WIN32
#include "7zWindows.h"
#include "DllSecur.h"
#ifndef UNDER_CE
Z7_DIAGNOSTIC_IGNORE_CAST_FUNCTION
typedef BOOL (WINAPI *Func_SetDefaultDllDirectories)(DWORD DirectoryFlags);
#define MY_LOAD_LIBRARY_SEARCH_USER_DIRS 0x400
#define MY_LOAD_LIBRARY_SEARCH_SYSTEM32 0x800
#define DELIM "\0"
static const char * const g_Dlls =
"userenv"
DELIM "setupapi"
DELIM "apphelp"
DELIM "propsys"
DELIM "dwmapi"
DELIM "cryptbase"
DELIM "oleacc"
DELIM "clbcatq"
DELIM "version"
#ifndef _CONSOLE
DELIM "uxtheme"
#endif
DELIM;
#endif
#ifdef __clang__
#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
#endif
#if defined (_MSC_VER) && _MSC_VER >= 1900
// sysinfoapi.h: kit10: GetVersion was declared deprecated
#pragma warning(disable : 4996)
#endif
#define IF_NON_VISTA_SET_DLL_DIRS_AND_RETURN \
if ((UInt16)GetVersion() != 6) { \
const \
Func_SetDefaultDllDirectories setDllDirs = \
(Func_SetDefaultDllDirectories) Z7_CAST_FUNC_C GetProcAddress(GetModuleHandle(TEXT("kernel32.dll")), \
"SetDefaultDllDirectories"); \
if (setDllDirs) if (setDllDirs(MY_LOAD_LIBRARY_SEARCH_SYSTEM32 | MY_LOAD_LIBRARY_SEARCH_USER_DIRS)) return; }
void My_SetDefaultDllDirectories(void)
{
#ifndef UNDER_CE
IF_NON_VISTA_SET_DLL_DIRS_AND_RETURN
#endif
}
void LoadSecurityDlls(void)
{
#ifndef UNDER_CE
// at Vista (ver 6.0) : CoCreateInstance(CLSID_ShellLink, ...) doesn't work after SetDefaultDllDirectories() : Check it ???
IF_NON_VISTA_SET_DLL_DIRS_AND_RETURN
{
WCHAR buf[MAX_PATH + 100];
const char *dll;
unsigned pos = GetSystemDirectoryW(buf, MAX_PATH + 2);
if (pos == 0 || pos > MAX_PATH)
return;
if (buf[pos - 1] != '\\')
buf[pos++] = '\\';
for (dll = g_Dlls; *dll != 0;)
{
WCHAR *dest = &buf[pos];
for (;;)
{
const char c = *dll++;
if (c == 0)
break;
*dest++ = (Byte)c;
}
dest[0] = '.';
dest[1] = 'd';
dest[2] = 'l';
dest[3] = 'l';
dest[4] = 0;
// lstrcatW(buf, L".dll");
LoadLibraryExW(buf, NULL, LOAD_WITH_ALTERED_SEARCH_PATH);
}
}
#endif
}
#endif // _WIN32
third_party/7z/DllSecur.h
Vendored
+20
View File
@@ -0,0 +1,20 @@
/* DllSecur.h -- DLL loading for security
2023-03-03 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_DLL_SECUR_H
#define ZIP7_INC_DLL_SECUR_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#ifdef _WIN32
void My_SetDefaultDllDirectories(void);
void LoadSecurityDlls(void);
#endif
EXTERN_C_END
#endif
third_party/7z/Lzma2Dec.c
Vendored
+493
View File
@@ -0,0 +1,493 @@
/* Lzma2Dec.c -- LZMA2 Decoder
2024-03-01 : Igor Pavlov : Public domain */
/* #define SHOW_DEBUG_INFO */
#include "Precomp.h"
#ifdef SHOW_DEBUG_INFO
#include <stdio.h>
#endif
#include <string.h>
#include "Lzma2Dec.h"
/*
00000000 - End of data
00000001 U U - Uncompressed, reset dic, need reset state and set new prop
00000010 U U - Uncompressed, no reset
100uuuuu U U P P - LZMA, no reset
101uuuuu U U P P - LZMA, reset state
110uuuuu U U P P S - LZMA, reset state + set new prop
111uuuuu U U P P S - LZMA, reset state + set new prop, reset dic
u, U - Unpack Size
P - Pack Size
S - Props
*/
#define LZMA2_CONTROL_COPY_RESET_DIC 1
#define LZMA2_IS_UNCOMPRESSED_STATE(p) (((p)->control & (1 << 7)) == 0)
#define LZMA2_LCLP_MAX 4
#define LZMA2_DIC_SIZE_FROM_PROP(p) (((UInt32)2 | ((p) & 1)) << ((p) / 2 + 11))
#ifdef SHOW_DEBUG_INFO
#define PRF(x) x
#else
#define PRF(x)
#endif
typedef enum
{
LZMA2_STATE_CONTROL,
LZMA2_STATE_UNPACK0,
LZMA2_STATE_UNPACK1,
LZMA2_STATE_PACK0,
LZMA2_STATE_PACK1,
LZMA2_STATE_PROP,
LZMA2_STATE_DATA,
LZMA2_STATE_DATA_CONT,
LZMA2_STATE_FINISHED,
LZMA2_STATE_ERROR
} ELzma2State;
static SRes Lzma2Dec_GetOldProps(Byte prop, Byte *props)
{
UInt32 dicSize;
if (prop > 40)
return SZ_ERROR_UNSUPPORTED;
dicSize = (prop == 40) ? 0xFFFFFFFF : LZMA2_DIC_SIZE_FROM_PROP(prop);
props[0] = (Byte)LZMA2_LCLP_MAX;
props[1] = (Byte)(dicSize);
props[2] = (Byte)(dicSize >> 8);
props[3] = (Byte)(dicSize >> 16);
props[4] = (Byte)(dicSize >> 24);
return SZ_OK;
}
SRes Lzma2Dec_AllocateProbs(CLzma2Dec *p, Byte prop, ISzAllocPtr alloc)
{
Byte props[LZMA_PROPS_SIZE];
RINOK(Lzma2Dec_GetOldProps(prop, props))
return LzmaDec_AllocateProbs(&p->decoder, props, LZMA_PROPS_SIZE, alloc);
}
SRes Lzma2Dec_Allocate(CLzma2Dec *p, Byte prop, ISzAllocPtr alloc)
{
Byte props[LZMA_PROPS_SIZE];
RINOK(Lzma2Dec_GetOldProps(prop, props))
return LzmaDec_Allocate(&p->decoder, props, LZMA_PROPS_SIZE, alloc);
}
void Lzma2Dec_Init(CLzma2Dec *p)
{
p->state = LZMA2_STATE_CONTROL;
p->needInitLevel = 0xE0;
p->isExtraMode = False;
p->unpackSize = 0;
// p->decoder.dicPos = 0; // we can use it instead of full init
LzmaDec_Init(&p->decoder);
}
// ELzma2State
static unsigned Lzma2Dec_UpdateState(CLzma2Dec *p, Byte b)
{
switch (p->state)
{
case LZMA2_STATE_CONTROL:
p->isExtraMode = False;
p->control = b;
PRF(printf("\n %8X", (unsigned)p->decoder.dicPos));
PRF(printf(" %02X", (unsigned)b));
if (b == 0)
return LZMA2_STATE_FINISHED;
if (LZMA2_IS_UNCOMPRESSED_STATE(p))
{
if (b == LZMA2_CONTROL_COPY_RESET_DIC)
p->needInitLevel = 0xC0;
else if (b > 2 || p->needInitLevel == 0xE0)
return LZMA2_STATE_ERROR;
}
else
{
if (b < p->needInitLevel)
return LZMA2_STATE_ERROR;
p->needInitLevel = 0;
p->unpackSize = (UInt32)(b & 0x1F) << 16;
}
return LZMA2_STATE_UNPACK0;
case LZMA2_STATE_UNPACK0:
p->unpackSize |= (UInt32)b << 8;
return LZMA2_STATE_UNPACK1;
case LZMA2_STATE_UNPACK1:
p->unpackSize |= (UInt32)b;
p->unpackSize++;
PRF(printf(" %7u", (unsigned)p->unpackSize));
return LZMA2_IS_UNCOMPRESSED_STATE(p) ? LZMA2_STATE_DATA : LZMA2_STATE_PACK0;
case LZMA2_STATE_PACK0:
p->packSize = (UInt32)b << 8;
return LZMA2_STATE_PACK1;
case LZMA2_STATE_PACK1:
p->packSize |= (UInt32)b;
p->packSize++;
// if (p->packSize < 5) return LZMA2_STATE_ERROR;
PRF(printf(" %5u", (unsigned)p->packSize));
return (p->control & 0x40) ? LZMA2_STATE_PROP : LZMA2_STATE_DATA;
case LZMA2_STATE_PROP:
{
unsigned lc, lp;
if (b >= (9 * 5 * 5))
return LZMA2_STATE_ERROR;
lc = b % 9;
b /= 9;
p->decoder.prop.pb = (Byte)(b / 5);
lp = b % 5;
if (lc + lp > LZMA2_LCLP_MAX)
return LZMA2_STATE_ERROR;
p->decoder.prop.lc = (Byte)lc;
p->decoder.prop.lp = (Byte)lp;
return LZMA2_STATE_DATA;
}
default:
return LZMA2_STATE_ERROR;
}
}
static void LzmaDec_UpdateWithUncompressed(CLzmaDec *p, const Byte *src, SizeT size)
{
memcpy(p->dic + p->dicPos, src, size);
p->dicPos += size;
if (p->checkDicSize == 0 && p->prop.dicSize - p->processedPos <= size)
p->checkDicSize = p->prop.dicSize;
p->processedPos += (UInt32)size;
}
void LzmaDec_InitDicAndState(CLzmaDec *p, BoolInt initDic, BoolInt initState);
SRes Lzma2Dec_DecodeToDic(CLzma2Dec *p, SizeT dicLimit,
const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status)
{
SizeT inSize = *srcLen;
*srcLen = 0;
*status = LZMA_STATUS_NOT_SPECIFIED;
while (p->state != LZMA2_STATE_ERROR)
{
SizeT dicPos;
if (p->state == LZMA2_STATE_FINISHED)
{
*status = LZMA_STATUS_FINISHED_WITH_MARK;
return SZ_OK;
}
dicPos = p->decoder.dicPos;
if (dicPos == dicLimit && finishMode == LZMA_FINISH_ANY)
{
*status = LZMA_STATUS_NOT_FINISHED;
return SZ_OK;
}
if (p->state != LZMA2_STATE_DATA && p->state != LZMA2_STATE_DATA_CONT)
{
if (*srcLen == inSize)
{
*status = LZMA_STATUS_NEEDS_MORE_INPUT;
return SZ_OK;
}
(*srcLen)++;
p->state = Lzma2Dec_UpdateState(p, *src++);
if (dicPos == dicLimit && p->state != LZMA2_STATE_FINISHED)
break;
continue;
}
{
SizeT inCur = inSize - *srcLen;
SizeT outCur = dicLimit - dicPos;
ELzmaFinishMode curFinishMode = LZMA_FINISH_ANY;
if (outCur >= p->unpackSize)
{
outCur = (SizeT)p->unpackSize;
curFinishMode = LZMA_FINISH_END;
}
if (LZMA2_IS_UNCOMPRESSED_STATE(p))
{
if (inCur == 0)
{
*status = LZMA_STATUS_NEEDS_MORE_INPUT;
return SZ_OK;
}
if (p->state == LZMA2_STATE_DATA)
{
BoolInt initDic = (p->control == LZMA2_CONTROL_COPY_RESET_DIC);
LzmaDec_InitDicAndState(&p->decoder, initDic, False);
}
if (inCur > outCur)
inCur = outCur;
if (inCur == 0)
break;
LzmaDec_UpdateWithUncompressed(&p->decoder, src, inCur);
src += inCur;
*srcLen += inCur;
p->unpackSize -= (UInt32)inCur;
p->state = (p->unpackSize == 0) ? LZMA2_STATE_CONTROL : LZMA2_STATE_DATA_CONT;
}
else
{
SRes res;
if (p->state == LZMA2_STATE_DATA)
{
BoolInt initDic = (p->control >= 0xE0);
BoolInt initState = (p->control >= 0xA0);
LzmaDec_InitDicAndState(&p->decoder, initDic, initState);
p->state = LZMA2_STATE_DATA_CONT;
}
if (inCur > p->packSize)
inCur = (SizeT)p->packSize;
res = LzmaDec_DecodeToDic(&p->decoder, dicPos + outCur, src, &inCur, curFinishMode, status);
src += inCur;
*srcLen += inCur;
p->packSize -= (UInt32)inCur;
outCur = p->decoder.dicPos - dicPos;
p->unpackSize -= (UInt32)outCur;
if (res != 0)
break;
if (*status == LZMA_STATUS_NEEDS_MORE_INPUT)
{
if (p->packSize == 0)
break;
return SZ_OK;
}
if (inCur == 0 && outCur == 0)
{
if (*status != LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK
|| p->unpackSize != 0
|| p->packSize != 0)
break;
p->state = LZMA2_STATE_CONTROL;
}
*status = LZMA_STATUS_NOT_SPECIFIED;
}
}
}
*status = LZMA_STATUS_NOT_SPECIFIED;
p->state = LZMA2_STATE_ERROR;
return SZ_ERROR_DATA;
}
ELzma2ParseStatus Lzma2Dec_Parse(CLzma2Dec *p,
SizeT outSize,
const Byte *src, SizeT *srcLen,
int checkFinishBlock)
{
SizeT inSize = *srcLen;
*srcLen = 0;
while (p->state != LZMA2_STATE_ERROR)
{
if (p->state == LZMA2_STATE_FINISHED)
return (ELzma2ParseStatus)LZMA_STATUS_FINISHED_WITH_MARK;
if (outSize == 0 && !checkFinishBlock)
return (ELzma2ParseStatus)LZMA_STATUS_NOT_FINISHED;
if (p->state != LZMA2_STATE_DATA && p->state != LZMA2_STATE_DATA_CONT)
{
if (*srcLen == inSize)
return (ELzma2ParseStatus)LZMA_STATUS_NEEDS_MORE_INPUT;
(*srcLen)++;
p->state = Lzma2Dec_UpdateState(p, *src++);
if (p->state == LZMA2_STATE_UNPACK0)
{
// if (p->decoder.dicPos != 0)
if (p->control == LZMA2_CONTROL_COPY_RESET_DIC || p->control >= 0xE0)
return LZMA2_PARSE_STATUS_NEW_BLOCK;
// if (outSize == 0) return LZMA_STATUS_NOT_FINISHED;
}
// The following code can be commented.
// It's not big problem, if we read additional input bytes.
// It will be stopped later in LZMA2_STATE_DATA / LZMA2_STATE_DATA_CONT state.
if (outSize == 0 && p->state != LZMA2_STATE_FINISHED)
{
// checkFinishBlock is true. So we expect that block must be finished,
// We can return LZMA_STATUS_NOT_SPECIFIED or LZMA_STATUS_NOT_FINISHED here
// break;
return (ELzma2ParseStatus)LZMA_STATUS_NOT_FINISHED;
}
if (p->state == LZMA2_STATE_DATA)
return LZMA2_PARSE_STATUS_NEW_CHUNK;
continue;
}
if (outSize == 0)
return (ELzma2ParseStatus)LZMA_STATUS_NOT_FINISHED;
{
SizeT inCur = inSize - *srcLen;
if (LZMA2_IS_UNCOMPRESSED_STATE(p))
{
if (inCur == 0)
return (ELzma2ParseStatus)LZMA_STATUS_NEEDS_MORE_INPUT;
if (inCur > p->unpackSize)
inCur = p->unpackSize;
if (inCur > outSize)
inCur = outSize;
p->decoder.dicPos += inCur;
src += inCur;
*srcLen += inCur;
outSize -= inCur;
p->unpackSize -= (UInt32)inCur;
p->state = (p->unpackSize == 0) ? LZMA2_STATE_CONTROL : LZMA2_STATE_DATA_CONT;
}
else
{
p->isExtraMode = True;
if (inCur == 0)
{
if (p->packSize != 0)
return (ELzma2ParseStatus)LZMA_STATUS_NEEDS_MORE_INPUT;
}
else if (p->state == LZMA2_STATE_DATA)
{
p->state = LZMA2_STATE_DATA_CONT;
if (*src != 0)
{
// first byte of lzma chunk must be Zero
*srcLen += 1;
p->packSize--;
break;
}
}
if (inCur > p->packSize)
inCur = (SizeT)p->packSize;
src += inCur;
*srcLen += inCur;
p->packSize -= (UInt32)inCur;
if (p->packSize == 0)
{
SizeT rem = outSize;
if (rem > p->unpackSize)
rem = p->unpackSize;
p->decoder.dicPos += rem;
p->unpackSize -= (UInt32)rem;
outSize -= rem;
if (p->unpackSize == 0)
p->state = LZMA2_STATE_CONTROL;
}
}
}
}
p->state = LZMA2_STATE_ERROR;
return (ELzma2ParseStatus)LZMA_STATUS_NOT_SPECIFIED;
}
SRes Lzma2Dec_DecodeToBuf(CLzma2Dec *p, Byte *dest, SizeT *destLen, const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status)
{
SizeT outSize = *destLen, inSize = *srcLen;
*srcLen = *destLen = 0;
for (;;)
{
SizeT inCur = inSize, outCur, dicPos;
ELzmaFinishMode curFinishMode;
SRes res;
if (p->decoder.dicPos == p->decoder.dicBufSize)
p->decoder.dicPos = 0;
dicPos = p->decoder.dicPos;
curFinishMode = LZMA_FINISH_ANY;
outCur = p->decoder.dicBufSize - dicPos;
if (outCur >= outSize)
{
outCur = outSize;
curFinishMode = finishMode;
}
res = Lzma2Dec_DecodeToDic(p, dicPos + outCur, src, &inCur, curFinishMode, status);
src += inCur;
inSize -= inCur;
*srcLen += inCur;
outCur = p->decoder.dicPos - dicPos;
memcpy(dest, p->decoder.dic + dicPos, outCur);
dest += outCur;
outSize -= outCur;
*destLen += outCur;
if (res != 0)
return res;
if (outCur == 0 || outSize == 0)
return SZ_OK;
}
}
SRes Lzma2Decode(Byte *dest, SizeT *destLen, const Byte *src, SizeT *srcLen,
Byte prop, ELzmaFinishMode finishMode, ELzmaStatus *status, ISzAllocPtr alloc)
{
CLzma2Dec p;
SRes res;
SizeT outSize = *destLen, inSize = *srcLen;
*destLen = *srcLen = 0;
*status = LZMA_STATUS_NOT_SPECIFIED;
Lzma2Dec_CONSTRUCT(&p)
RINOK(Lzma2Dec_AllocateProbs(&p, prop, alloc))
p.decoder.dic = dest;
p.decoder.dicBufSize = outSize;
Lzma2Dec_Init(&p);
*srcLen = inSize;
res = Lzma2Dec_DecodeToDic(&p, outSize, src, srcLen, finishMode, status);
*destLen = p.decoder.dicPos;
if (res == SZ_OK && *status == LZMA_STATUS_NEEDS_MORE_INPUT)
res = SZ_ERROR_INPUT_EOF;
Lzma2Dec_FreeProbs(&p, alloc);
return res;
}
#undef PRF
third_party/7z/Lzma2Dec.h
Vendored
+121
View File
@@ -0,0 +1,121 @@
/* Lzma2Dec.h -- LZMA2 Decoder
2023-03-03 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_LZMA2_DEC_H
#define ZIP7_INC_LZMA2_DEC_H
#include "LzmaDec.h"
EXTERN_C_BEGIN
/* ---------- State Interface ---------- */
typedef struct
{
unsigned state;
Byte control;
Byte needInitLevel;
Byte isExtraMode;
Byte _pad_;
UInt32 packSize;
UInt32 unpackSize;
CLzmaDec decoder;
} CLzma2Dec;
#define Lzma2Dec_CONSTRUCT(p) LzmaDec_CONSTRUCT(&(p)->decoder)
#define Lzma2Dec_Construct(p) Lzma2Dec_CONSTRUCT(p)
#define Lzma2Dec_FreeProbs(p, alloc) LzmaDec_FreeProbs(&(p)->decoder, alloc)
#define Lzma2Dec_Free(p, alloc) LzmaDec_Free(&(p)->decoder, alloc)
SRes Lzma2Dec_AllocateProbs(CLzma2Dec *p, Byte prop, ISzAllocPtr alloc);
SRes Lzma2Dec_Allocate(CLzma2Dec *p, Byte prop, ISzAllocPtr alloc);
void Lzma2Dec_Init(CLzma2Dec *p);
/*
finishMode:
It has meaning only if the decoding reaches output limit (*destLen or dicLimit).
LZMA_FINISH_ANY - use smallest number of input bytes
LZMA_FINISH_END - read EndOfStream marker after decoding
Returns:
SZ_OK
status:
LZMA_STATUS_FINISHED_WITH_MARK
LZMA_STATUS_NOT_FINISHED
LZMA_STATUS_NEEDS_MORE_INPUT
SZ_ERROR_DATA - Data error
*/
SRes Lzma2Dec_DecodeToDic(CLzma2Dec *p, SizeT dicLimit,
const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status);
SRes Lzma2Dec_DecodeToBuf(CLzma2Dec *p, Byte *dest, SizeT *destLen,
const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status);
/* ---------- LZMA2 block and chunk parsing ---------- */
/*
Lzma2Dec_Parse() parses compressed data stream up to next independent block or next chunk data.
It can return LZMA_STATUS_* code or LZMA2_PARSE_STATUS_* code:
- LZMA2_PARSE_STATUS_NEW_BLOCK - there is new block, and 1 additional byte (control byte of next block header) was read from input.
- LZMA2_PARSE_STATUS_NEW_CHUNK - there is new chunk, and only lzma2 header of new chunk was read.
CLzma2Dec::unpackSize contains unpack size of that chunk
*/
typedef enum
{
/*
LZMA_STATUS_NOT_SPECIFIED // data error
LZMA_STATUS_FINISHED_WITH_MARK
LZMA_STATUS_NOT_FINISHED //
LZMA_STATUS_NEEDS_MORE_INPUT
LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK // unused
*/
LZMA2_PARSE_STATUS_NEW_BLOCK = LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK + 1,
LZMA2_PARSE_STATUS_NEW_CHUNK
} ELzma2ParseStatus;
ELzma2ParseStatus Lzma2Dec_Parse(CLzma2Dec *p,
SizeT outSize, // output size
const Byte *src, SizeT *srcLen,
int checkFinishBlock // set (checkFinishBlock = 1), if it must read full input data, if decoder.dicPos reaches blockMax position.
);
/*
LZMA2 parser doesn't decode LZMA chunks, so we must read
full input LZMA chunk to decode some part of LZMA chunk.
Lzma2Dec_GetUnpackExtra() returns the value that shows
max possible number of output bytes that can be output by decoder
at current input positon.
*/
#define Lzma2Dec_GetUnpackExtra(p) ((p)->isExtraMode ? (p)->unpackSize : 0)
/* ---------- One Call Interface ---------- */
/*
finishMode:
It has meaning only if the decoding reaches output limit (*destLen).
LZMA_FINISH_ANY - use smallest number of input bytes
LZMA_FINISH_END - read EndOfStream marker after decoding
Returns:
SZ_OK
status:
LZMA_STATUS_FINISHED_WITH_MARK
LZMA_STATUS_NOT_FINISHED
SZ_ERROR_DATA - Data error
SZ_ERROR_MEM - Memory allocation error
SZ_ERROR_UNSUPPORTED - Unsupported properties
SZ_ERROR_INPUT_EOF - It needs more bytes in input buffer (src).
*/
SRes Lzma2Decode(Byte *dest, SizeT *destLen, const Byte *src, SizeT *srcLen,
Byte prop, ELzmaFinishMode finishMode, ELzmaStatus *status, ISzAllocPtr alloc);
EXTERN_C_END
#endif
third_party/7z/LzmaDec.c
Vendored
+1367
View File
File diff suppressed because it is too large. Load diff
third_party/7z/LzmaDec.h
Vendored
+237
View File
@@ -0,0 +1,237 @@
/* LzmaDec.h -- LZMA Decoder
2023-04-02 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_LZMA_DEC_H
#define ZIP7_INC_LZMA_DEC_H
#include "7zTypes.h"
EXTERN_C_BEGIN
/* #define Z7_LZMA_PROB32 */
/* Z7_LZMA_PROB32 can increase the speed on some CPUs,
but memory usage for CLzmaDec::probs will be doubled in that case */
typedef
#ifdef Z7_LZMA_PROB32
UInt32
#else
UInt16
#endif
CLzmaProb;
/* ---------- LZMA Properties ---------- */
#define LZMA_PROPS_SIZE 5
typedef struct
{
Byte lc;
Byte lp;
Byte pb;
Byte _pad_;
UInt32 dicSize;
} CLzmaProps;
/* LzmaProps_Decode - decodes properties
Returns:
SZ_OK
SZ_ERROR_UNSUPPORTED - Unsupported properties
*/
SRes LzmaProps_Decode(CLzmaProps *p, const Byte *data, unsigned size);
/* ---------- LZMA Decoder state ---------- */
/* LZMA_REQUIRED_INPUT_MAX = number of required input bytes for worst case.
Num bits = log2((2^11 / 31) ^ 22) + 26 < 134 + 26 = 160; */
#define LZMA_REQUIRED_INPUT_MAX 20
typedef struct
{
/* Don't change this structure. ASM code can use it. */
CLzmaProps prop;
CLzmaProb *probs;
CLzmaProb *probs_1664;
Byte *dic;
SizeT dicBufSize;
SizeT dicPos;
const Byte *buf;
UInt32 range;
UInt32 code;
UInt32 processedPos;
UInt32 checkDicSize;
UInt32 reps[4];
UInt32 state;
UInt32 remainLen;
UInt32 numProbs;
unsigned tempBufSize;
Byte tempBuf[LZMA_REQUIRED_INPUT_MAX];
} CLzmaDec;
#define LzmaDec_CONSTRUCT(p) { (p)->dic = NULL; (p)->probs = NULL; }
#define LzmaDec_Construct(p) LzmaDec_CONSTRUCT(p)
void LzmaDec_Init(CLzmaDec *p);
/* There are two types of LZMA streams:
- Stream with end mark. That end mark adds about 6 bytes to compressed size.
- Stream without end mark. You must know exact uncompressed size to decompress such stream. */
typedef enum
{
LZMA_FINISH_ANY, /* finish at any point */
LZMA_FINISH_END /* block must be finished at the end */
} ELzmaFinishMode;
/* ELzmaFinishMode has meaning only if the decoding reaches output limit !!!
You must use LZMA_FINISH_END, when you know that current output buffer
covers last bytes of block. In other cases you must use LZMA_FINISH_ANY.
If LZMA decoder sees end marker before reaching output limit, it returns SZ_OK,
and output value of destLen will be less than output buffer size limit.
You can check status result also.
You can use multiple checks to test data integrity after full decompression:
1) Check Result and "status" variable.
2) Check that output(destLen) = uncompressedSize, if you know real uncompressedSize.
3) Check that output(srcLen) = compressedSize, if you know real compressedSize.
You must use correct finish mode in that case. */
typedef enum
{
LZMA_STATUS_NOT_SPECIFIED, /* use main error code instead */
LZMA_STATUS_FINISHED_WITH_MARK, /* stream was finished with end mark. */
LZMA_STATUS_NOT_FINISHED, /* stream was not finished */
LZMA_STATUS_NEEDS_MORE_INPUT, /* you must provide more input bytes */
LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK /* there is probability that stream was finished without end mark */
} ELzmaStatus;
/* ELzmaStatus is used only as output value for function call */
/* ---------- Interfaces ---------- */
/* There are 3 levels of interfaces:
1) Dictionary Interface
2) Buffer Interface
3) One Call Interface
You can select any of these interfaces, but don't mix functions from different
groups for same object. */
/* There are two variants to allocate state for Dictionary Interface:
1) LzmaDec_Allocate / LzmaDec_Free
2) LzmaDec_AllocateProbs / LzmaDec_FreeProbs
You can use variant 2, if you set dictionary buffer manually.
For Buffer Interface you must always use variant 1.
LzmaDec_Allocate* can return:
SZ_OK
SZ_ERROR_MEM - Memory allocation error
SZ_ERROR_UNSUPPORTED - Unsupported properties
*/
SRes LzmaDec_AllocateProbs(CLzmaDec *p, const Byte *props, unsigned propsSize, ISzAllocPtr alloc);
void LzmaDec_FreeProbs(CLzmaDec *p, ISzAllocPtr alloc);
SRes LzmaDec_Allocate(CLzmaDec *p, const Byte *props, unsigned propsSize, ISzAllocPtr alloc);
void LzmaDec_Free(CLzmaDec *p, ISzAllocPtr alloc);
/* ---------- Dictionary Interface ---------- */
/* You can use it, if you want to eliminate the overhead for data copying from
dictionary to some other external buffer.
You must work with CLzmaDec variables directly in this interface.
STEPS:
LzmaDec_Construct()
LzmaDec_Allocate()
for (each new stream)
{
LzmaDec_Init()
while (it needs more decompression)
{
LzmaDec_DecodeToDic()
use data from CLzmaDec::dic and update CLzmaDec::dicPos
}
}
LzmaDec_Free()
*/
/* LzmaDec_DecodeToDic
The decoding to internal dictionary buffer (CLzmaDec::dic).
You must manually update CLzmaDec::dicPos, if it reaches CLzmaDec::dicBufSize !!!
finishMode:
It has meaning only if the decoding reaches output limit (dicLimit).
LZMA_FINISH_ANY - Decode just dicLimit bytes.
LZMA_FINISH_END - Stream must be finished after dicLimit.
Returns:
SZ_OK
status:
LZMA_STATUS_FINISHED_WITH_MARK
LZMA_STATUS_NOT_FINISHED
LZMA_STATUS_NEEDS_MORE_INPUT
LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK
SZ_ERROR_DATA - Data error
SZ_ERROR_FAIL - Some unexpected error: internal error of code, memory corruption or hardware failure
*/
SRes LzmaDec_DecodeToDic(CLzmaDec *p, SizeT dicLimit,
const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status);
/* ---------- Buffer Interface ---------- */
/* It's zlib-like interface.
See LzmaDec_DecodeToDic description for information about STEPS and return results,
but you must use LzmaDec_DecodeToBuf instead of LzmaDec_DecodeToDic and you don't need
to work with CLzmaDec variables manually.
finishMode:
It has meaning only if the decoding reaches output limit (*destLen).
LZMA_FINISH_ANY - Decode just destLen bytes.
LZMA_FINISH_END - Stream must be finished after (*destLen).
*/
SRes LzmaDec_DecodeToBuf(CLzmaDec *p, Byte *dest, SizeT *destLen,
const Byte *src, SizeT *srcLen, ELzmaFinishMode finishMode, ELzmaStatus *status);
/* ---------- One Call Interface ---------- */
/* LzmaDecode
finishMode:
It has meaning only if the decoding reaches output limit (*destLen).
LZMA_FINISH_ANY - Decode just destLen bytes.
LZMA_FINISH_END - Stream must be finished after (*destLen).
Returns:
SZ_OK
status:
LZMA_STATUS_FINISHED_WITH_MARK
LZMA_STATUS_NOT_FINISHED
LZMA_STATUS_MAYBE_FINISHED_WITHOUT_MARK
SZ_ERROR_DATA - Data error
SZ_ERROR_MEM - Memory allocation error
SZ_ERROR_UNSUPPORTED - Unsupported properties
SZ_ERROR_INPUT_EOF - It needs more bytes in input buffer (src).
SZ_ERROR_FAIL - Some unexpected error: internal error of code, memory corruption or hardware failure
*/
SRes LzmaDecode(Byte *dest, SizeT *destLen, const Byte *src, SizeT *srcLen,
const Byte *propData, unsigned propSize, ELzmaFinishMode finishMode,
ELzmaStatus *status, ISzAllocPtr alloc);
EXTERN_C_END
#endif
third_party/7z/Ppmd.h
Vendored
+169
View File
@@ -0,0 +1,169 @@
/* Ppmd.h -- PPMD codec common code
2023-03-05 : Igor Pavlov : Public domain
This code is based on PPMd var.H (2001): Dmitry Shkarin : Public domain */
#ifndef ZIP7_INC_PPMD_H
#define ZIP7_INC_PPMD_H
#include "CpuArch.h"
EXTERN_C_BEGIN
#if defined(MY_CPU_SIZEOF_POINTER) && (MY_CPU_SIZEOF_POINTER == 4)
/*
PPMD code always uses 32-bit internal fields in PPMD structures to store internal references in main block.
if (PPMD_32BIT is defined), the PPMD code stores internal pointers to 32-bit reference fields.
if (PPMD_32BIT is NOT defined), the PPMD code stores internal UInt32 offsets to reference fields.
if (pointer size is 64-bit), then (PPMD_32BIT) mode is not allowed,
if (pointer size is 32-bit), then (PPMD_32BIT) mode is optional,
and it's allowed to disable PPMD_32BIT mode even if pointer is 32-bit.
PPMD code works slightly faster in (PPMD_32BIT) mode.
*/
#define PPMD_32BIT
#endif
#define PPMD_INT_BITS 7
#define PPMD_PERIOD_BITS 7
#define PPMD_BIN_SCALE (1 << (PPMD_INT_BITS + PPMD_PERIOD_BITS))
#define PPMD_GET_MEAN_SPEC(summ, shift, round) (((summ) + (1 << ((shift) - (round)))) >> (shift))
#define PPMD_GET_MEAN(summ) PPMD_GET_MEAN_SPEC((summ), PPMD_PERIOD_BITS, 2)
#define PPMD_UPDATE_PROB_0(prob) ((prob) + (1 << PPMD_INT_BITS) - PPMD_GET_MEAN(prob))
#define PPMD_UPDATE_PROB_1(prob) ((prob) - PPMD_GET_MEAN(prob))
#define PPMD_N1 4
#define PPMD_N2 4
#define PPMD_N3 4
#define PPMD_N4 ((128 + 3 - 1 * PPMD_N1 - 2 * PPMD_N2 - 3 * PPMD_N3) / 4)
#define PPMD_NUM_INDEXES (PPMD_N1 + PPMD_N2 + PPMD_N3 + PPMD_N4)
MY_CPU_pragma_pack_push_1
/* Most compilers works OK here even without #pragma pack(push, 1), but some GCC compilers need it. */
/* SEE-contexts for PPM-contexts with masked symbols */
typedef struct
{
UInt16 Summ; /* Freq */
Byte Shift; /* Speed of Freq change; low Shift is for fast change */
Byte Count; /* Count to next change of Shift */
} CPpmd_See;
#define Ppmd_See_UPDATE(p) \
{ if ((p)->Shift < PPMD_PERIOD_BITS && --(p)->Count == 0) \
{ (p)->Summ = (UInt16)((p)->Summ << 1); \
(p)->Count = (Byte)(3 << (p)->Shift++); }}
typedef struct
{
Byte Symbol;
Byte Freq;
UInt16 Successor_0;
UInt16 Successor_1;
} CPpmd_State;
typedef struct CPpmd_State2_
{
Byte Symbol;
Byte Freq;
} CPpmd_State2;
typedef struct CPpmd_State4_
{
UInt16 Successor_0;
UInt16 Successor_1;
} CPpmd_State4;
MY_CPU_pragma_pop
/*
PPMD code can write full CPpmd_State structure data to CPpmd*_Context
at (byte offset = 2) instead of some fields of original CPpmd*_Context structure.
If we use pointers to different types, but that point to shared
memory space, we can have aliasing problem (strict aliasing).
XLC compiler in -O2 mode can change the order of memory write instructions
in relation to read instructions, if we have use pointers to different types.
To solve that aliasing problem we use combined CPpmd*_Context structure
with unions that contain the fields from both structures:
the original CPpmd*_Context and CPpmd_State.
So we can access the fields from both structures via one pointer,
and the compiler doesn't change the order of write instructions
in relation to read instructions.
If we don't use memory write instructions to shared memory in
some local code, and we use only reading instructions (read only),
then probably it's safe to use pointers to different types for reading.
*/
#ifdef PPMD_32BIT
#define Ppmd_Ref_Type(type) type *
#define Ppmd_GetRef(p, ptr) (ptr)
#define Ppmd_GetPtr(p, ptr) (ptr)
#define Ppmd_GetPtr_Type(p, ptr, note_type) (ptr)
#else
#define Ppmd_Ref_Type(type) UInt32
#define Ppmd_GetRef(p, ptr) ((UInt32)((Byte *)(ptr) - (p)->Base))
#define Ppmd_GetPtr(p, offs) ((void *)((p)->Base + (offs)))
#define Ppmd_GetPtr_Type(p, offs, type) ((type *)Ppmd_GetPtr(p, offs))
#endif // PPMD_32BIT
typedef Ppmd_Ref_Type(CPpmd_State) CPpmd_State_Ref;
typedef Ppmd_Ref_Type(void) CPpmd_Void_Ref;
typedef Ppmd_Ref_Type(Byte) CPpmd_Byte_Ref;
/*
#ifdef MY_CPU_LE_UNALIGN
// the unaligned 32-bit access latency can be too large, if the data is not in L1 cache.
#define Ppmd_GET_SUCCESSOR(p) ((CPpmd_Void_Ref)*(const UInt32 *)(const void *)&(p)->Successor_0)
#define Ppmd_SET_SUCCESSOR(p, v) *(UInt32 *)(void *)(void *)&(p)->Successor_0 = (UInt32)(v)
#else
*/
/*
We can write 16-bit halves to 32-bit (Successor) field in any selected order.
But the native order is more consistent way.
So we use the native order, if LE/BE order can be detected here at compile time.
*/
#ifdef MY_CPU_BE
#define Ppmd_GET_SUCCESSOR(p) \
( (CPpmd_Void_Ref) (((UInt32)(p)->Successor_0 << 16) | (p)->Successor_1) )
#define Ppmd_SET_SUCCESSOR(p, v) { \
(p)->Successor_0 = (UInt16)(((UInt32)(v) >> 16) /* & 0xFFFF */); \
(p)->Successor_1 = (UInt16)((UInt32)(v) /* & 0xFFFF */); }
#else
#define Ppmd_GET_SUCCESSOR(p) \
( (CPpmd_Void_Ref) ((p)->Successor_0 | ((UInt32)(p)->Successor_1 << 16)) )
#define Ppmd_SET_SUCCESSOR(p, v) { \
(p)->Successor_0 = (UInt16)((UInt32)(v) /* & 0xFFFF */); \
(p)->Successor_1 = (UInt16)(((UInt32)(v) >> 16) /* & 0xFFFF */); }
#endif
// #endif
#define PPMD_SetAllBitsIn256Bytes(p) \
{ size_t z; for (z = 0; z < 256 / sizeof(p[0]); z += 8) { \
p[z+7] = p[z+6] = p[z+5] = p[z+4] = p[z+3] = p[z+2] = p[z+1] = p[z+0] = ~(size_t)0; }}
EXTERN_C_END
#endif
third_party/7z/Ppmd7.c
Vendored
+1131
View File
File diff suppressed because it is too large. Load diff
third_party/7z/Ppmd7.h
Vendored
+181
View File
@@ -0,0 +1,181 @@
/* Ppmd7.h -- Ppmd7 (PPMdH) compression codec
2023-04-02 : Igor Pavlov : Public domain
This code is based on:
PPMd var.H (2001): Dmitry Shkarin : Public domain */
#ifndef ZIP7_INC_PPMD7_H
#define ZIP7_INC_PPMD7_H
#include "Ppmd.h"
EXTERN_C_BEGIN
#define PPMD7_MIN_ORDER 2
#define PPMD7_MAX_ORDER 64
#define PPMD7_MIN_MEM_SIZE (1 << 11)
#define PPMD7_MAX_MEM_SIZE (0xFFFFFFFF - 12 * 3)
struct CPpmd7_Context_;
typedef Ppmd_Ref_Type(struct CPpmd7_Context_) CPpmd7_Context_Ref;
// MY_CPU_pragma_pack_push_1
typedef struct CPpmd7_Context_
{
UInt16 NumStats;
union
{
UInt16 SummFreq;
CPpmd_State2 State2;
} Union2;
union
{
CPpmd_State_Ref Stats;
CPpmd_State4 State4;
} Union4;
CPpmd7_Context_Ref Suffix;
} CPpmd7_Context;
// MY_CPU_pragma_pop
#define Ppmd7Context_OneState(p) ((CPpmd_State *)&(p)->Union2)
typedef struct
{
UInt32 Range;
UInt32 Code;
UInt32 Low;
IByteInPtr Stream;
} CPpmd7_RangeDec;
typedef struct
{
UInt32 Range;
Byte Cache;
// Byte _dummy_[3];
UInt64 Low;
UInt64 CacheSize;
IByteOutPtr Stream;
} CPpmd7z_RangeEnc;
typedef struct
{
CPpmd7_Context *MinContext, *MaxContext;
CPpmd_State *FoundState;
unsigned OrderFall, InitEsc, PrevSuccess, MaxOrder, HiBitsFlag;
Int32 RunLength, InitRL; /* must be 32-bit at least */
UInt32 Size;
UInt32 GlueCount;
UInt32 AlignOffset;
Byte *Base, *LoUnit, *HiUnit, *Text, *UnitsStart;
union
{
CPpmd7_RangeDec dec;
CPpmd7z_RangeEnc enc;
} rc;
Byte Indx2Units[PPMD_NUM_INDEXES + 2]; // +2 for alignment
Byte Units2Indx[128];
CPpmd_Void_Ref FreeList[PPMD_NUM_INDEXES];
Byte NS2BSIndx[256], NS2Indx[256];
Byte ExpEscape[16];
CPpmd_See DummySee, See[25][16];
UInt16 BinSumm[128][64];
// int LastSymbol;
} CPpmd7;
void Ppmd7_Construct(CPpmd7 *p);
BoolInt Ppmd7_Alloc(CPpmd7 *p, UInt32 size, ISzAllocPtr alloc);
void Ppmd7_Free(CPpmd7 *p, ISzAllocPtr alloc);
void Ppmd7_Init(CPpmd7 *p, unsigned maxOrder);
#define Ppmd7_WasAllocated(p) ((p)->Base != NULL)
/* ---------- Internal Functions ---------- */
#define Ppmd7_GetPtr(p, ptr) Ppmd_GetPtr(p, ptr)
#define Ppmd7_GetContext(p, ptr) Ppmd_GetPtr_Type(p, ptr, CPpmd7_Context)
#define Ppmd7_GetStats(p, ctx) Ppmd_GetPtr_Type(p, (ctx)->Union4.Stats, CPpmd_State)
void Ppmd7_Update1(CPpmd7 *p);
void Ppmd7_Update1_0(CPpmd7 *p);
void Ppmd7_Update2(CPpmd7 *p);
#define PPMD7_HiBitsFlag_3(sym) ((((unsigned)sym + 0xC0) >> (8 - 3)) & (1 << 3))
#define PPMD7_HiBitsFlag_4(sym) ((((unsigned)sym + 0xC0) >> (8 - 4)) & (1 << 4))
// #define PPMD7_HiBitsFlag_3(sym) ((sym) < 0x40 ? 0 : (1 << 3))
// #define PPMD7_HiBitsFlag_4(sym) ((sym) < 0x40 ? 0 : (1 << 4))
#define Ppmd7_GetBinSumm(p) \
&p->BinSumm[(size_t)(unsigned)Ppmd7Context_OneState(p->MinContext)->Freq - 1] \
[ p->PrevSuccess + ((p->RunLength >> 26) & 0x20) \
+ p->NS2BSIndx[(size_t)Ppmd7_GetContext(p, p->MinContext->Suffix)->NumStats - 1] \
+ PPMD7_HiBitsFlag_4(Ppmd7Context_OneState(p->MinContext)->Symbol) \
+ (p->HiBitsFlag = PPMD7_HiBitsFlag_3(p->FoundState->Symbol)) ]
CPpmd_See *Ppmd7_MakeEscFreq(CPpmd7 *p, unsigned numMasked, UInt32 *scale);
/*
We support two versions of Ppmd7 (PPMdH) methods that use same CPpmd7 structure:
1) Ppmd7a_*: original PPMdH
2) Ppmd7z_*: modified PPMdH with 7z Range Coder
Ppmd7_*: the structures and functions that are common for both versions of PPMd7 (PPMdH)
*/
/* ---------- Decode ---------- */
#define PPMD7_SYM_END (-1)
#define PPMD7_SYM_ERROR (-2)
/*
You must set (CPpmd7::rc.dec.Stream) before Ppmd7*_RangeDec_Init()
Ppmd7*_DecodeSymbol()
out:
>= 0 : decoded byte
-1 : PPMD7_SYM_END : End of payload marker
-2 : PPMD7_SYM_ERROR : Data error
*/
/* Ppmd7a_* : original PPMdH */
BoolInt Ppmd7a_RangeDec_Init(CPpmd7_RangeDec *p);
#define Ppmd7a_RangeDec_IsFinishedOK(p) ((p)->Code == 0)
int Ppmd7a_DecodeSymbol(CPpmd7 *p);
/* Ppmd7z_* : modified PPMdH with 7z Range Coder */
BoolInt Ppmd7z_RangeDec_Init(CPpmd7_RangeDec *p);
#define Ppmd7z_RangeDec_IsFinishedOK(p) ((p)->Code == 0)
int Ppmd7z_DecodeSymbol(CPpmd7 *p);
// Byte *Ppmd7z_DecodeSymbols(CPpmd7 *p, Byte *buf, const Byte *lim);
/* ---------- Encode ---------- */
void Ppmd7z_Init_RangeEnc(CPpmd7 *p);
void Ppmd7z_Flush_RangeEnc(CPpmd7 *p);
// void Ppmd7z_EncodeSymbol(CPpmd7 *p, int symbol);
void Ppmd7z_EncodeSymbols(CPpmd7 *p, const Byte *buf, const Byte *lim);
EXTERN_C_END
#endif
third_party/7z/Ppmd7Dec.c
Vendored
+312
View File
@@ -0,0 +1,312 @@
/* Ppmd7Dec.c -- Ppmd7z (PPMdH with 7z Range Coder) Decoder
2023-09-07 : Igor Pavlov : Public domain
This code is based on:
PPMd var.H (2001): Dmitry Shkarin : Public domain */
#include "Precomp.h"
#include "Ppmd7.h"
#define kTopValue ((UInt32)1 << 24)
#define READ_BYTE(p) IByteIn_Read((p)->Stream)
BoolInt Ppmd7z_RangeDec_Init(CPpmd7_RangeDec *p)
{
unsigned i;
p->Code = 0;
p->Range = 0xFFFFFFFF;
if (READ_BYTE(p) != 0)
return False;
for (i = 0; i < 4; i++)
p->Code = (p->Code << 8) | READ_BYTE(p);
return (p->Code < 0xFFFFFFFF);
}
#define RC_NORM_BASE(p) if ((p)->Range < kTopValue) \
{ (p)->Code = ((p)->Code << 8) | READ_BYTE(p); (p)->Range <<= 8;
#define RC_NORM_1(p) RC_NORM_BASE(p) }
#define RC_NORM(p) RC_NORM_BASE(p) RC_NORM_BASE(p) }}
// we must use only one type of Normalization from two: LOCAL or REMOTE
#define RC_NORM_LOCAL(p) // RC_NORM(p)
#define RC_NORM_REMOTE(p) RC_NORM(p)
#define R (&p->rc.dec)
Z7_FORCE_INLINE
// Z7_NO_INLINE
static void Ppmd7z_RD_Decode(CPpmd7 *p, UInt32 start, UInt32 size)
{
R->Code -= start * R->Range;
R->Range *= size;
RC_NORM_LOCAL(R)
}
#define RC_Decode(start, size) Ppmd7z_RD_Decode(p, start, size);
#define RC_DecodeFinal(start, size) RC_Decode(start, size) RC_NORM_REMOTE(R)
#define RC_GetThreshold(total) (R->Code / (R->Range /= (total)))
#define CTX(ref) ((CPpmd7_Context *)Ppmd7_GetContext(p, ref))
// typedef CPpmd7_Context * CTX_PTR;
#define SUCCESSOR(p) Ppmd_GET_SUCCESSOR(p)
void Ppmd7_UpdateModel(CPpmd7 *p);
#define MASK(sym) ((Byte *)charMask)[sym]
// Z7_FORCE_INLINE
// static
int Ppmd7z_DecodeSymbol(CPpmd7 *p)
{
size_t charMask[256 / sizeof(size_t)];
if (p->MinContext->NumStats != 1)
{
CPpmd_State *s = Ppmd7_GetStats(p, p->MinContext);
unsigned i;
UInt32 count, hiCnt;
const UInt32 summFreq = p->MinContext->Union2.SummFreq;
count = RC_GetThreshold(summFreq);
hiCnt = count;
if ((Int32)(count -= s->Freq) < 0)
{
Byte sym;
RC_DecodeFinal(0, s->Freq)
p->FoundState = s;
sym = s->Symbol;
Ppmd7_Update1_0(p);
return sym;
}
p->PrevSuccess = 0;
i = (unsigned)p->MinContext->NumStats - 1;
do
{
if ((Int32)(count -= (++s)->Freq) < 0)
{
Byte sym;
RC_DecodeFinal((hiCnt - count) - s->Freq, s->Freq)
p->FoundState = s;
sym = s->Symbol;
Ppmd7_Update1(p);
return sym;
}
}
while (--i);
if (hiCnt >= summFreq)
return PPMD7_SYM_ERROR;
hiCnt -= count;
RC_Decode(hiCnt, summFreq - hiCnt)
p->HiBitsFlag = PPMD7_HiBitsFlag_3(p->FoundState->Symbol);
PPMD_SetAllBitsIn256Bytes(charMask)
// i = p->MinContext->NumStats - 1;
// do { MASK((--s)->Symbol) = 0; } while (--i);
{
CPpmd_State *s2 = Ppmd7_GetStats(p, p->MinContext);
MASK(s->Symbol) = 0;
do
{
const unsigned sym0 = s2[0].Symbol;
const unsigned sym1 = s2[1].Symbol;
s2 += 2;
MASK(sym0) = 0;
MASK(sym1) = 0;
}
while (s2 < s);
}
}
else
{
CPpmd_State *s = Ppmd7Context_OneState(p->MinContext);
UInt16 *prob = Ppmd7_GetBinSumm(p);
UInt32 pr = *prob;
UInt32 size0 = (R->Range >> 14) * pr;
pr = PPMD_UPDATE_PROB_1(pr);
if (R->Code < size0)
{
Byte sym;
*prob = (UInt16)(pr + (1 << PPMD_INT_BITS));
// RangeDec_DecodeBit0(size0);
R->Range = size0;
RC_NORM_1(R)
/* we can use single byte normalization here because of
(min(BinSumm[][]) = 95) > (1 << (14 - 8)) */
// sym = (p->FoundState = Ppmd7Context_OneState(p->MinContext))->Symbol;
// Ppmd7_UpdateBin(p);
{
unsigned freq = s->Freq;
CPpmd7_Context *c = CTX(SUCCESSOR(s));
sym = s->Symbol;
p->FoundState = s;
p->PrevSuccess = 1;
p->RunLength++;
s->Freq = (Byte)(freq + (freq < 128));
// NextContext(p);
if (p->OrderFall == 0 && (const Byte *)c > p->Text)
p->MaxContext = p->MinContext = c;
else
Ppmd7_UpdateModel(p);
}
return sym;
}
*prob = (UInt16)pr;
p->InitEsc = p->ExpEscape[pr >> 10];
// RangeDec_DecodeBit1(size0);
R->Code -= size0;
R->Range -= size0;
RC_NORM_LOCAL(R)
PPMD_SetAllBitsIn256Bytes(charMask)
MASK(Ppmd7Context_OneState(p->MinContext)->Symbol) = 0;
p->PrevSuccess = 0;
}
for (;;)
{
CPpmd_State *s, *s2;
UInt32 freqSum, count, hiCnt;
CPpmd_See *see;
CPpmd7_Context *mc;
unsigned numMasked;
RC_NORM_REMOTE(R)
mc = p->MinContext;
numMasked = mc->NumStats;
do
{
p->OrderFall++;
if (!mc->Suffix)
return PPMD7_SYM_END;
mc = Ppmd7_GetContext(p, mc->Suffix);
}
while (mc->NumStats == numMasked);
s = Ppmd7_GetStats(p, mc);
{
unsigned num = mc->NumStats;
unsigned num2 = num / 2;
num &= 1;
hiCnt = (s->Freq & (UInt32)(MASK(s->Symbol))) & (0 - (UInt32)num);
s += num;
p->MinContext = mc;
do
{
const unsigned sym0 = s[0].Symbol;
const unsigned sym1 = s[1].Symbol;
s += 2;
hiCnt += (s[-2].Freq & (UInt32)(MASK(sym0)));
hiCnt += (s[-1].Freq & (UInt32)(MASK(sym1)));
}
while (--num2);
}
see = Ppmd7_MakeEscFreq(p, numMasked, &freqSum);
freqSum += hiCnt;
count = RC_GetThreshold(freqSum);
if (count < hiCnt)
{
Byte sym;
s = Ppmd7_GetStats(p, p->MinContext);
hiCnt = count;
// count -= s->Freq & (UInt32)(MASK(s->Symbol));
// if ((Int32)count >= 0)
{
for (;;)
{
count -= s->Freq & (UInt32)(MASK((s)->Symbol)); s++; if ((Int32)count < 0) break;
// count -= s->Freq & (UInt32)(MASK((s)->Symbol)); s++; if ((Int32)count < 0) break;
}
}
s--;
RC_DecodeFinal((hiCnt - count) - s->Freq, s->Freq)
// new (see->Summ) value can overflow over 16-bits in some rare cases
Ppmd_See_UPDATE(see)
p->FoundState = s;
sym = s->Symbol;
Ppmd7_Update2(p);
return sym;
}
if (count >= freqSum)
return PPMD7_SYM_ERROR;
RC_Decode(hiCnt, freqSum - hiCnt)
// We increase (see->Summ) for sum of Freqs of all non_Masked symbols.
// new (see->Summ) value can overflow over 16-bits in some rare cases
see->Summ = (UInt16)(see->Summ + freqSum);
s = Ppmd7_GetStats(p, p->MinContext);
s2 = s + p->MinContext->NumStats;
do
{
MASK(s->Symbol) = 0;
s++;
}
while (s != s2);
}
}
/*
Byte *Ppmd7z_DecodeSymbols(CPpmd7 *p, Byte *buf, const Byte *lim)
{
int sym = 0;
if (buf != lim)
do
{
sym = Ppmd7z_DecodeSymbol(p);
if (sym < 0)
break;
*buf = (Byte)sym;
}
while (++buf < lim);
p->LastSymbol = sym;
return buf;
}
*/
#undef kTopValue
#undef READ_BYTE
#undef RC_NORM_BASE
#undef RC_NORM_1
#undef RC_NORM
#undef RC_NORM_LOCAL
#undef RC_NORM_REMOTE
#undef R
#undef RC_Decode
#undef RC_DecodeFinal
#undef RC_GetThreshold
#undef CTX
#undef SUCCESSOR
#undef MASK
third_party/7z/Precomp.h
Vendored
+127
View File
@@ -0,0 +1,127 @@
/* Precomp.h -- precompilation file
: Igor Pavlov : Public domain */
#ifndef ZIP7_INC_PRECOMP_H
#define ZIP7_INC_PRECOMP_H
/*
this file must be included before another *.h files and before <windows.h>.
this file is included from the following files:
C\*.c
C\Util\*\Precomp.h <- C\Util\*\*.c
CPP\Common\Common.h <- *\StdAfx.h <- *\*.cpp
this file can set the following macros:
Z7_LARGE_PAGES 1
Z7_LONG_PATH 1
Z7_WIN32_WINNT_MIN 0x0500 (or higher) : we require at least win2000+ for 7-Zip
_WIN32_WINNT 0x0500 (or higher)
WINVER _WIN32_WINNT
UNICODE 1
_UNICODE 1
*/
#include "Compiler.h"
#ifdef _MSC_VER
// #pragma warning(disable : 4206) // nonstandard extension used : translation unit is empty
#if _MSC_VER >= 1912
// #pragma warning(disable : 5039) // pointer or reference to potentially throwing function passed to 'extern "C"' function under - EHc.Undefined behavior may occur if this function throws an exception.
#endif
#endif
/*
// for debug:
#define UNICODE 1
#define _UNICODE 1
#define _WIN32_WINNT 0x0500 // win2000
#ifndef WINVER
#define WINVER _WIN32_WINNT
#endif
*/
#ifndef Z7_LARGE_PAGES
#if !defined(Z7_NO_LARGE_PAGES) && !defined(UNDER_CE)
#define Z7_LARGE_PAGES 1
#endif
#endif
#ifdef _WIN32
/*
this "Precomp.h" file must be included before <windows.h>,
if we want to define _WIN32_WINNT before <windows.h>.
*/
#ifndef Z7_LONG_PATH
#ifndef Z7_NO_LONG_PATH
#define Z7_LONG_PATH 1
#endif
#endif
#ifndef Z7_DEVICE_FILE
#ifndef Z7_NO_DEVICE_FILE
// #define Z7_DEVICE_FILE 1
#endif
#endif
// we don't change macros if included after <windows.h>
#ifndef _WINDOWS_
#ifndef Z7_WIN32_WINNT_MIN
#if defined(_M_ARM64) || defined(__aarch64__)
// #define Z7_WIN32_WINNT_MIN 0x0a00 // win10
#define Z7_WIN32_WINNT_MIN 0x0600 // vista
#elif defined(_M_ARM) && defined(_M_ARMT) && defined(_M_ARM_NT)
// #define Z7_WIN32_WINNT_MIN 0x0602 // win8
#define Z7_WIN32_WINNT_MIN 0x0600 // vista
#elif defined(_M_X64) || defined(_M_AMD64) || defined(__x86_64__) || defined(_M_IA64)
#define Z7_WIN32_WINNT_MIN 0x0503 // win2003
// #elif defined(_M_IX86) || defined(__i386__)
// #define Z7_WIN32_WINNT_MIN 0x0500 // win2000
#else // x86 and another(old) systems
#define Z7_WIN32_WINNT_MIN 0x0500 // win2000
// #define Z7_WIN32_WINNT_MIN 0x0502 // win2003 // for debug
#endif
#endif // Z7_WIN32_WINNT_MIN
#ifndef Z7_DO_NOT_DEFINE_WIN32_WINNT
#ifdef _WIN32_WINNT
// #error Stop_Compiling_Bad_WIN32_WINNT
#else
#ifndef Z7_NO_DEFINE_WIN32_WINNT
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define _WIN32_WINNT Z7_WIN32_WINNT_MIN
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // _WIN32_WINNT
#ifndef WINVER
#define WINVER _WIN32_WINNT
#endif
#endif // Z7_DO_NOT_DEFINE_WIN32_WINNT
#ifndef _MBCS
#ifndef Z7_NO_UNICODE
// UNICODE and _UNICODE are used by <windows.h> and by 7-zip code.
#ifndef UNICODE
#define UNICODE 1
#endif
#ifndef _UNICODE
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define _UNICODE 1
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // Z7_NO_UNICODE
#endif // _MBCS
#endif // _WINDOWS_
// #include "7zWindows.h"
#endif // _WIN32
#endif
third_party/7z/README.md
Vendored
+85
View File
@@ -0,0 +1,85 @@
# 7z decoder sources (LZMA SDK)
Vendored subset of the **LZMA SDK 26.03** (2026-09-03) by Igor Pavlov.
Public domain — see `DOC/lzma-sdk.txt`:
> LZMA SDK is written and placed in the public domain by Igor Pavlov.
Upstream: <https://www.7-zip.org/sdk.html> — release archive
`lzma2603.7z` from <https://github.com/ip7z/7zip/releases>.
These files back the 7z branch of the extraction engine (`src/sevenz_*.c`).
The SDK ships both a C and a C++ implementation; only the **C** one is used
here, because it builds with the plain `prospero-clang` C toolchain on PS5 and
has no dependency on the C++ runtime.
## What is vendored
| Group | Files |
|---|---|
| container | `7z.h`, `7zArcIn.c`, `7zDec.c` |
| streams / infra | `7zTypes.h`, `7zStream.c`, `7zFile.c`, `7zAlloc.c`, `7zBuf.c`, `7zBuf2.c`, `Alloc.c` |
| checksums | `7zCrc.c`, `7zCrcOpt.c`, `SwapBytes.c` |
| CPU / portability | `CpuArch.c`, `Compiler.h`, `Precomp.h`, `RotateDefs.h`, `7zWindows.h` |
| codecs | `LzmaDec.c`, `Lzma2Dec.c`, `Ppmd7.c`, `Ppmd7Dec.c`, `Ppmd.h`, `Bcj2.c`, `Bra.c`, `Bra86.c`, `BraIA64.c`, `Delta.c` |
| crypto | `Aes.c`, `AesOpt.c`, `Sha256.c`, `Sha256Opt.c` |
| misc | `DllSecur.c` (only referenced by the Windows path of `7zFile.c`), `7zVersion.h` |
The encoder half of the SDK (`LzmaEnc.c`, `Lzma2Enc.c`, `Xz*.c`, `Sort.c`,
`Threads.c`, `Mt*.c`, `Ppmd*Enc.c`, …) is deliberately **not** vendored: the
engine only ever decodes.
Build flags that matter:
* `-DZ7_PPMD_SUPPORT` — without it `7zDec.c` drops the PPMd coder entirely.
* `-D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE`
* `-w` — the sources are not expected to be warning clean.
## Known SDK limitations the engine has to work around
Both were confirmed empirically against real 7-Zip 26.03 output, see
`tests/sevenz_e2e.c` and `tests/make_sevenz_fixtures.py`.
### 1. `CSzFolder` caps a folder at 4 coders / 3 bonds
`7z.h` defines `SZ_NUM_CODERS_IN_FOLDER_MAX 4` and
`SZ_NUM_BONDS_IN_FOLDER_MAX 3`, and `SzGetNextFolderItem()` refuses anything
larger. But the chain 7-Zip produces for `-m0=bcj2` is **BCJ2 + 4×LZMA2 = 5
coders / 4 bonds**, so `SzAr_DecodeFolder()` returns `SZ_ERROR_UNSUPPORTED` on
an archive that 7-Zip itself created without complaint.
Note this only affects *decoding*: `SzArEx_Open()` parses folder blobs with a
separate, far more permissive scanner (`k_Scan_NumCoders_MAX 64`, in
`7zArcIn.c`), so the file list and all unpack sizes are still correct — the
failure shows up per-entry, at extract time.
Consequence: the engine parses folder blobs with its own dynamic parser and
drives the codec chain itself instead of calling `SzAr_DecodeFolder()`.
### 2. There is no 7zAES coder in the C decoder
`IS_SUPPORTED_CODER()` in `7zDec.c` accepts only `Copy`, `LZMA`, `LZMA2`,
`PPMd` and the branch/delta/BCJ2 filters. `Aes.c` is present but never wired
up, so both `-p<password>` archives and `-mhe=on` (encrypted header) archives
are rejected.
Consequence: the engine implements the 7zAES coder itself on top of the
vendored `Aes.c` / `Sha256.c`. The reference implementation is in the SDK's
C++ tree at `CPP/7zip/Crypto/7zAes.cpp` (not vendored, kept in the SDK tarball):
* properties: `b0 & 0x3F` = numCyclesPower, `b0 bit7` and `b1>>4` add to the
salt size, `b0 bit6` and `b1 & 0x0F` add to the IV size
* key: `SHA-256(salt || password_utf16le || counter_le64)` iterated
`1 << numCyclesPower` times, unless `numCyclesPower == 0x3F` in which case
the key is `salt || password` truncated/zero-padded to 32 bytes
* then AES-256-CBC with that key and the (zero-padded) IV
## Regenerating this directory
```sh
curl -L -o lzma2603.7z https://github.com/ip7z/7zip/releases/download/26.03/lzma2603.7z
7zr x lzma2603.7z -osdk
# copy the files listed above out of sdk/C/
```
`tests/run-sevenz-tests.sh` builds this subset and runs the fixture matrix.
third_party/7z/RotateDefs.h
Vendored
+50
View File
@@ -0,0 +1,50 @@
/* RotateDefs.h -- Rotate functions
2023-06-18 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_ROTATE_DEFS_H
#define ZIP7_INC_ROTATE_DEFS_H
#ifdef _MSC_VER
#include <stdlib.h>
/* don't use _rotl with old MINGW. It can insert slow call to function. */
/* #if (_MSC_VER >= 1200) */
#pragma intrinsic(_rotl)
#pragma intrinsic(_rotr)
/* #endif */
#define rotlFixed(x, n) _rotl((x), (n))
#define rotrFixed(x, n) _rotr((x), (n))
#if (_MSC_VER >= 1300)
#define Z7_ROTL64(x, n) _rotl64((x), (n))
#define Z7_ROTR64(x, n) _rotr64((x), (n))
#else
#define Z7_ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#define Z7_ROTR64(x, n) (((x) >> (n)) | ((x) << (64 - (n))))
#endif
#else
/* new compilers can translate these macros to fast commands. */
#if defined(__clang__) && (__clang_major__ >= 4) \
|| defined(__GNUC__) && (__GNUC__ >= 5)
/* GCC 4.9.0 and clang 3.5 can recognize more correct version: */
#define rotlFixed(x, n) (((x) << (n)) | ((x) >> (-(n) & 31)))
#define rotrFixed(x, n) (((x) >> (n)) | ((x) << (-(n) & 31)))
#define Z7_ROTL64(x, n) (((x) << (n)) | ((x) >> (-(n) & 63)))
#define Z7_ROTR64(x, n) (((x) >> (n)) | ((x) << (-(n) & 63)))
#else
/* for old GCC / clang: */
#define rotlFixed(x, n) (((x) << (n)) | ((x) >> (32 - (n))))
#define rotrFixed(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#define Z7_ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#define Z7_ROTR64(x, n) (((x) >> (n)) | ((x) << (64 - (n))))
#endif
#endif
#endif
third_party/7z/Sha256.c
Vendored
+492
View File
@@ -0,0 +1,492 @@
/* Sha256.c -- SHA-256 Hash
: Igor Pavlov : Public domain
This code is based on public domain code from Wei Dai's Crypto++ library. */
#include "Precomp.h"
#include <string.h>
#include "Sha256.h"
#include "RotateDefs.h"
#include "CpuArch.h"
#ifdef MY_CPU_X86_OR_AMD64
#if defined(Z7_LLVM_CLANG_VERSION) && (Z7_LLVM_CLANG_VERSION >= 30800) \
|| defined(Z7_APPLE_CLANG_VERSION) && (Z7_APPLE_CLANG_VERSION >= 50100) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40900) \
|| defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1600) \
|| defined(_MSC_VER) && (_MSC_VER >= 1200)
#define Z7_COMPILER_SHA256_SUPPORTED
#endif
#elif defined(MY_CPU_ARM_OR_ARM64) && defined(MY_CPU_LE)
#if defined(__ARM_FEATURE_SHA2) \
|| defined(__ARM_FEATURE_CRYPTO)
#define Z7_COMPILER_SHA256_SUPPORTED
#else
#if defined(MY_CPU_ARM64) \
|| defined(__ARM_ARCH) && (__ARM_ARCH >= 4) \
|| defined(Z7_MSC_VER_ORIGINAL)
#if defined(__ARM_FP) && \
( defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(__GNUC__) && (__GNUC__ >= 6) \
) \
|| defined(Z7_MSC_VER_ORIGINAL) && (_MSC_VER >= 1910)
#if defined(MY_CPU_ARM64) \
|| !defined(Z7_CLANG_VERSION) \
|| defined(__ARM_NEON) && \
(Z7_CLANG_VERSION < 170000 || \
Z7_CLANG_VERSION > 170001)
#define Z7_COMPILER_SHA256_SUPPORTED
#endif
#endif
#endif
#endif
#endif
void Z7_FASTCALL Sha256_UpdateBlocks(UInt32 state[8], const Byte *data, size_t numBlocks);
#ifdef Z7_COMPILER_SHA256_SUPPORTED
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks);
static SHA256_FUNC_UPDATE_BLOCKS g_SHA256_FUNC_UPDATE_BLOCKS = Sha256_UpdateBlocks;
static SHA256_FUNC_UPDATE_BLOCKS g_SHA256_FUNC_UPDATE_BLOCKS_HW;
#define SHA256_UPDATE_BLOCKS(p) p->v.vars.func_UpdateBlocks
#else
#define SHA256_UPDATE_BLOCKS(p) Sha256_UpdateBlocks
#endif
BoolInt Sha256_SetFunction(CSha256 *p, unsigned algo)
{
SHA256_FUNC_UPDATE_BLOCKS func = Sha256_UpdateBlocks;
#ifdef Z7_COMPILER_SHA256_SUPPORTED
if (algo != SHA256_ALGO_SW)
{
if (algo == SHA256_ALGO_DEFAULT)
func = g_SHA256_FUNC_UPDATE_BLOCKS;
else
{
if (algo != SHA256_ALGO_HW)
return False;
func = g_SHA256_FUNC_UPDATE_BLOCKS_HW;
if (!func)
return False;
}
}
#else
if (algo > 1)
return False;
#endif
p->v.vars.func_UpdateBlocks = func;
return True;
}
/* define it for speed optimization */
#ifdef Z7_SFX
#define STEP_PRE 1
#define STEP_MAIN 1
#else
#define STEP_PRE 2
#define STEP_MAIN 4
// #define Z7_SHA256_UNROLL
#endif
#undef Z7_SHA256_BIG_W
#if STEP_MAIN != 16
#define Z7_SHA256_BIG_W
#endif
void Sha256_InitState(CSha256 *p)
{
p->v.vars.count = 0;
p->state[0] = 0x6a09e667;
p->state[1] = 0xbb67ae85;
p->state[2] = 0x3c6ef372;
p->state[3] = 0xa54ff53a;
p->state[4] = 0x510e527f;
p->state[5] = 0x9b05688c;
p->state[6] = 0x1f83d9ab;
p->state[7] = 0x5be0cd19;
}
void Sha256_Init(CSha256 *p)
{
p->v.vars.func_UpdateBlocks =
#ifdef Z7_COMPILER_SHA256_SUPPORTED
g_SHA256_FUNC_UPDATE_BLOCKS;
#else
NULL;
#endif
Sha256_InitState(p);
}
#define S0(x) (rotrFixed(x, 2) ^ rotrFixed(x,13) ^ rotrFixed(x,22))
#define S1(x) (rotrFixed(x, 6) ^ rotrFixed(x,11) ^ rotrFixed(x,25))
#define s0(x) (rotrFixed(x, 7) ^ rotrFixed(x,18) ^ (x >> 3))
#define s1(x) (rotrFixed(x,17) ^ rotrFixed(x,19) ^ (x >>10))
#define Ch(x,y,z) (z^(x&(y^z)))
#define Maj(x,y,z) ((x&y)|(z&(x|y)))
#define W_PRE(i) (W[(i) + (size_t)(j)] = GetBe32(data + ((size_t)(j) + i) * 4))
#define blk2_main(j, i) s1(w(j, (i)-2)) + w(j, (i)-7) + s0(w(j, (i)-15))
#ifdef Z7_SHA256_BIG_W
// we use +i instead of +(i) to change the order to solve CLANG compiler warning for signed/unsigned.
#define w(j, i) W[(size_t)(j) + i]
#define blk2(j, i) (w(j, i) = w(j, (i)-16) + blk2_main(j, i))
#else
#if STEP_MAIN == 16
#define w(j, i) W[(i) & 15]
#else
#define w(j, i) W[((size_t)(j) + (i)) & 15]
#endif
#define blk2(j, i) (w(j, i) += blk2_main(j, i))
#endif
#define W_MAIN(i) blk2(j, i)
#define T1(wx, i) \
tmp = h + S1(e) + Ch(e,f,g) + K[(i)+(size_t)(j)] + wx(i); \
h = g; \
g = f; \
f = e; \
e = d + tmp; \
tmp += S0(a) + Maj(a, b, c); \
d = c; \
c = b; \
b = a; \
a = tmp; \
#define R1_PRE(i) T1( W_PRE, i)
#define R1_MAIN(i) T1( W_MAIN, i)
#if (!defined(Z7_SHA256_UNROLL) || STEP_MAIN < 8) && (STEP_MAIN >= 4)
#define R2_MAIN(i) \
R1_MAIN(i) \
R1_MAIN(i + 1) \
#endif
#if defined(Z7_SHA256_UNROLL) && STEP_MAIN >= 8
#define T4( a,b,c,d,e,f,g,h, wx, i) \
h += S1(e) + Ch(e,f,g) + K[(i)+(size_t)(j)] + wx(i); \
tmp = h; \
h += d; \
d = tmp + S0(a) + Maj(a, b, c); \
#define R4( wx, i) \
T4 ( a,b,c,d,e,f,g,h, wx, (i )); \
T4 ( d,a,b,c,h,e,f,g, wx, (i+1)); \
T4 ( c,d,a,b,g,h,e,f, wx, (i+2)); \
T4 ( b,c,d,a,f,g,h,e, wx, (i+3)); \
#define R4_PRE(i) R4( W_PRE, i)
#define R4_MAIN(i) R4( W_MAIN, i)
#define T8( a,b,c,d,e,f,g,h, wx, i) \
h += S1(e) + Ch(e,f,g) + K[(i)+(size_t)(j)] + wx(i); \
d += h; \
h += S0(a) + Maj(a, b, c); \
#define R8( wx, i) \
T8 ( a,b,c,d,e,f,g,h, wx, i ); \
T8 ( h,a,b,c,d,e,f,g, wx, i+1); \
T8 ( g,h,a,b,c,d,e,f, wx, i+2); \
T8 ( f,g,h,a,b,c,d,e, wx, i+3); \
T8 ( e,f,g,h,a,b,c,d, wx, i+4); \
T8 ( d,e,f,g,h,a,b,c, wx, i+5); \
T8 ( c,d,e,f,g,h,a,b, wx, i+6); \
T8 ( b,c,d,e,f,g,h,a, wx, i+7); \
#define R8_PRE(i) R8( W_PRE, i)
#define R8_MAIN(i) R8( W_MAIN, i)
#endif
extern
MY_ALIGN(64) const UInt32 SHA256_K_ARRAY[64];
MY_ALIGN(64) const UInt32 SHA256_K_ARRAY[64] = {
0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
};
#define K SHA256_K_ARRAY
Z7_NO_INLINE
void Z7_FASTCALL Sha256_UpdateBlocks(UInt32 state[8], const Byte *data, size_t numBlocks)
{
UInt32 W
#ifdef Z7_SHA256_BIG_W
[64];
#else
[16];
#endif
unsigned j;
UInt32 a,b,c,d,e,f,g,h;
#if !defined(Z7_SHA256_UNROLL) || (STEP_MAIN <= 4) || (STEP_PRE <= 4)
UInt32 tmp;
#endif
if (numBlocks == 0) return;
a = state[0];
b = state[1];
c = state[2];
d = state[3];
e = state[4];
f = state[5];
g = state[6];
h = state[7];
do
{
for (j = 0; j < 16; j += STEP_PRE)
{
#if STEP_PRE > 4
#if STEP_PRE < 8
R4_PRE(0);
#else
R8_PRE(0);
#if STEP_PRE == 16
R8_PRE(8);
#endif
#endif
#else
R1_PRE(0)
#if STEP_PRE >= 2
R1_PRE(1)
#if STEP_PRE >= 4
R1_PRE(2)
R1_PRE(3)
#endif
#endif
#endif
}
for (j = 16; j < 64; j += STEP_MAIN)
{
#if defined(Z7_SHA256_UNROLL) && STEP_MAIN >= 8
#if STEP_MAIN < 8
R4_MAIN(0)
#else
R8_MAIN(0)
#if STEP_MAIN == 16
R8_MAIN(8)
#endif
#endif
#else
R1_MAIN(0)
#if STEP_MAIN >= 2
R1_MAIN(1)
#if STEP_MAIN >= 4
R2_MAIN(2)
#if STEP_MAIN >= 8
R2_MAIN(4)
R2_MAIN(6)
#if STEP_MAIN >= 16
R2_MAIN(8)
R2_MAIN(10)
R2_MAIN(12)
R2_MAIN(14)
#endif
#endif
#endif
#endif
#endif
}
a += state[0]; state[0] = a;
b += state[1]; state[1] = b;
c += state[2]; state[2] = c;
d += state[3]; state[3] = d;
e += state[4]; state[4] = e;
f += state[5]; state[5] = f;
g += state[6]; state[6] = g;
h += state[7]; state[7] = h;
data += SHA256_BLOCK_SIZE;
}
while (--numBlocks);
}
#define Sha256_UpdateBlock(p) SHA256_UPDATE_BLOCKS(p)(p->state, p->buffer, 1)
void Sha256_Update(CSha256 *p, const Byte *data, size_t size)
{
if (size == 0)
return;
{
const unsigned pos = (unsigned)p->v.vars.count & (SHA256_BLOCK_SIZE - 1);
const unsigned num = SHA256_BLOCK_SIZE - pos;
p->v.vars.count += size;
if (num > size)
{
memcpy(p->buffer + pos, data, size);
return;
}
if (pos != 0)
{
size -= num;
memcpy(p->buffer + pos, data, num);
data += num;
Sha256_UpdateBlock(p);
}
}
{
const size_t numBlocks = size >> 6;
// if (numBlocks)
SHA256_UPDATE_BLOCKS(p)(p->state, data, numBlocks);
size &= SHA256_BLOCK_SIZE - 1;
if (size == 0)
return;
data += (numBlocks << 6);
memcpy(p->buffer, data, size);
}
}
void Sha256_Final(CSha256 *p, Byte *digest)
{
unsigned pos = (unsigned)p->v.vars.count & (SHA256_BLOCK_SIZE - 1);
p->buffer[pos++] = 0x80;
if (pos > (SHA256_BLOCK_SIZE - 4 * 2))
{
while (pos != SHA256_BLOCK_SIZE) { p->buffer[pos++] = 0; }
// memset(&p->buf.buffer[pos], 0, SHA256_BLOCK_SIZE - pos);
Sha256_UpdateBlock(p);
pos = 0;
}
memset(&p->buffer[pos], 0, (SHA256_BLOCK_SIZE - 4 * 2) - pos);
{
const UInt64 numBits = p->v.vars.count << 3;
SetBe32(p->buffer + SHA256_BLOCK_SIZE - 4 * 2, (UInt32)(numBits >> 32))
SetBe32(p->buffer + SHA256_BLOCK_SIZE - 4 * 1, (UInt32)(numBits))
}
Sha256_UpdateBlock(p);
#if 1 && defined(MY_CPU_BE)
memcpy(digest, p->state, SHA256_DIGEST_SIZE);
#else
{
unsigned i;
for (i = 0; i < 8; i += 2)
{
const UInt32 v0 = p->state[i];
const UInt32 v1 = p->state[(size_t)i + 1];
SetBe32(digest , v0)
SetBe32(digest + 4, v1)
digest += 4 * 2;
}
}
#endif
Sha256_InitState(p);
}
void Sha256Prepare(void)
{
#ifdef Z7_COMPILER_SHA256_SUPPORTED
SHA256_FUNC_UPDATE_BLOCKS f, f_hw;
f = Sha256_UpdateBlocks;
f_hw = NULL;
#ifdef MY_CPU_X86_OR_AMD64
if (CPU_IsSupported_SHA()
&& CPU_IsSupported_SSSE3()
)
#else
if (CPU_IsSupported_SHA2())
#endif
{
// printf("\n========== HW SHA256 ======== \n");
f = f_hw = Sha256_UpdateBlocks_HW;
}
g_SHA256_FUNC_UPDATE_BLOCKS = f;
g_SHA256_FUNC_UPDATE_BLOCKS_HW = f_hw;
#endif
}
#undef U64C
#undef K
#undef S0
#undef S1
#undef s0
#undef s1
#undef Ch
#undef Maj
#undef W_MAIN
#undef W_PRE
#undef w
#undef blk2_main
#undef blk2
#undef T1
#undef T4
#undef T8
#undef R1_PRE
#undef R1_MAIN
#undef R2_MAIN
#undef R4
#undef R4_PRE
#undef R4_MAIN
#undef R8
#undef R8_PRE
#undef R8_MAIN
#undef STEP_PRE
#undef STEP_MAIN
#undef Z7_SHA256_BIG_W
#undef Z7_SHA256_UNROLL
#undef Z7_COMPILER_SHA256_SUPPORTED
third_party/7z/Sha256.h
Vendored
+86
View File
@@ -0,0 +1,86 @@
/* Sha256.h -- SHA-256 Hash
: Igor Pavlov : Public domain */
#ifndef ZIP7_INC_SHA256_H
#define ZIP7_INC_SHA256_H
#include "7zTypes.h"
EXTERN_C_BEGIN
#define SHA256_NUM_BLOCK_WORDS 16
#define SHA256_NUM_DIGEST_WORDS 8
#define SHA256_BLOCK_SIZE (SHA256_NUM_BLOCK_WORDS * 4)
#define SHA256_DIGEST_SIZE (SHA256_NUM_DIGEST_WORDS * 4)
typedef void (Z7_FASTCALL *SHA256_FUNC_UPDATE_BLOCKS)(UInt32 state[8], const Byte *data, size_t numBlocks);
/*
if (the system supports different SHA256 code implementations)
{
(CSha256::func_UpdateBlocks) will be used
(CSha256::func_UpdateBlocks) can be set by
Sha256_Init() - to default (fastest)
Sha256_SetFunction() - to any algo
}
else
{
(CSha256::func_UpdateBlocks) is ignored.
}
*/
typedef struct
{
union
{
struct
{
SHA256_FUNC_UPDATE_BLOCKS func_UpdateBlocks;
UInt64 count;
} vars;
UInt64 _pad_64bit[4];
void *_pad_align_ptr[2];
} v;
UInt32 state[SHA256_NUM_DIGEST_WORDS];
Byte buffer[SHA256_BLOCK_SIZE];
} CSha256;
#define SHA256_ALGO_DEFAULT 0
#define SHA256_ALGO_SW 1
#define SHA256_ALGO_HW 2
/*
Sha256_SetFunction()
return:
0 - (algo) value is not supported, and func_UpdateBlocks was not changed
1 - func_UpdateBlocks was set according (algo) value.
*/
BoolInt Sha256_SetFunction(CSha256 *p, unsigned algo);
void Sha256_InitState(CSha256 *p);
void Sha256_Init(CSha256 *p);
void Sha256_Update(CSha256 *p, const Byte *data, size_t size);
void Sha256_Final(CSha256 *p, Byte *digest);
// void Z7_FASTCALL Sha256_UpdateBlocks(UInt32 state[8], const Byte *data, size_t numBlocks);
/*
call Sha256Prepare() once at program start.
It prepares all supported implementations, and detects the fastest implementation.
*/
void Sha256Prepare(void);
EXTERN_C_END
#endif
third_party/7z/Sha256Opt.c
Vendored
+451
View File
@@ -0,0 +1,451 @@
/* Sha256Opt.c -- SHA-256 optimized code for SHA-256 hardware instructions
: Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Compiler.h"
#include "CpuArch.h"
// #define Z7_USE_HW_SHA_STUB // for debug
#ifdef MY_CPU_X86_OR_AMD64
#if defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1600) // fix that check
#define USE_HW_SHA
#elif defined(Z7_LLVM_CLANG_VERSION) && (Z7_LLVM_CLANG_VERSION >= 30800) \
|| defined(Z7_APPLE_CLANG_VERSION) && (Z7_APPLE_CLANG_VERSION >= 50100) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40900)
#define USE_HW_SHA
#if !defined(__INTEL_COMPILER)
// icc defines __GNUC__, but icc doesn't support __attribute__(__target__)
#if !defined(__SHA__) || !defined(__SSSE3__)
#define ATTRIB_SHA __attribute__((__target__("sha,ssse3")))
#endif
#endif
#elif defined(_MSC_VER)
#if (_MSC_VER >= 1900)
#define USE_HW_SHA
#else
#define Z7_USE_HW_SHA_STUB
#endif
#endif
// #endif // MY_CPU_X86_OR_AMD64
#ifndef USE_HW_SHA
// #define Z7_USE_HW_SHA_STUB // for debug
#endif
#ifdef USE_HW_SHA
// #pragma message("Sha256 HW")
// sse/sse2/ssse3:
#include <tmmintrin.h>
// sha*:
#include <immintrin.h>
#if defined (__clang__) && defined(_MSC_VER)
#if !defined(__SHA__)
#include <shaintrin.h>
#endif
#else
#endif
/*
SHA256 uses:
SSE2:
_mm_loadu_si128
_mm_storeu_si128
_mm_set_epi32
_mm_add_epi32
_mm_shuffle_epi32 / pshufd
SSSE3:
_mm_shuffle_epi8 / pshufb
_mm_alignr_epi8
SHA:
_mm_sha256*
*/
// K array must be aligned for 16-bytes at least.
// The compiler can look align attribute and selects
// movdqu - for code without align attribute
// movdqa - for code with align attribute
extern
MY_ALIGN(64)
const UInt32 SHA256_K_ARRAY[64];
#define K SHA256_K_ARRAY
#define ADD_EPI32(dest, src) dest = _mm_add_epi32(dest, src);
#define SHA256_MSG1(dest, src) dest = _mm_sha256msg1_epu32(dest, src);
#define SHA256_MSG2(dest, src) dest = _mm_sha256msg2_epu32(dest, src);
#define LOAD_SHUFFLE(m, k) \
m = _mm_loadu_si128((const __m128i *)(const void *)(data + (k) * 16)); \
m = _mm_shuffle_epi8(m, mask); \
#define NNN(m0, m1, m2, m3)
#define SM1(m1, m2, m3, m0) \
SHA256_MSG1(m0, m1); \
#define SM2(m2, m3, m0, m1) \
ADD_EPI32(m0, _mm_alignr_epi8(m3, m2, 4)) \
SHA256_MSG2(m0, m3); \
#define RND2(t0, t1) \
t0 = _mm_sha256rnds2_epu32(t0, t1, msg);
#define R4(k, m0, m1, m2, m3, OP0, OP1) \
msg = _mm_add_epi32(m0, *(const __m128i *) (const void *) &K[(k) * 4]); \
RND2(state0, state1); \
msg = _mm_shuffle_epi32(msg, 0x0E); \
OP0(m0, m1, m2, m3) \
RND2(state1, state0); \
OP1(m0, m1, m2, m3) \
#define R16(k, OP0, OP1, OP2, OP3, OP4, OP5, OP6, OP7) \
R4 ( (k)*4+0, m0,m1,m2,m3, OP0, OP1 ) \
R4 ( (k)*4+1, m1,m2,m3,m0, OP2, OP3 ) \
R4 ( (k)*4+2, m2,m3,m0,m1, OP4, OP5 ) \
R4 ( (k)*4+3, m3,m0,m1,m2, OP6, OP7 ) \
#define PREPARE_STATE \
tmp = _mm_shuffle_epi32(state0, 0x1B); /* abcd */ \
state0 = _mm_shuffle_epi32(state1, 0x1B); /* efgh */ \
state1 = state0; \
state0 = _mm_unpacklo_epi64(state0, tmp); /* cdgh */ \
state1 = _mm_unpackhi_epi64(state1, tmp); /* abef */ \
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks);
#ifdef ATTRIB_SHA
ATTRIB_SHA
#endif
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks)
{
const __m128i mask = _mm_set_epi32(0x0c0d0e0f, 0x08090a0b, 0x04050607, 0x00010203);
__m128i tmp, state0, state1;
if (numBlocks == 0)
return;
state0 = _mm_loadu_si128((const __m128i *) (const void *) &state[0]);
state1 = _mm_loadu_si128((const __m128i *) (const void *) &state[4]);
PREPARE_STATE
do
{
__m128i state0_save, state1_save;
__m128i m0, m1, m2, m3;
__m128i msg;
// #define msg tmp
state0_save = state0;
state1_save = state1;
LOAD_SHUFFLE (m0, 0)
LOAD_SHUFFLE (m1, 1)
LOAD_SHUFFLE (m2, 2)
LOAD_SHUFFLE (m3, 3)
R16 ( 0, NNN, NNN, SM1, NNN, SM1, SM2, SM1, SM2 )
R16 ( 1, SM1, SM2, SM1, SM2, SM1, SM2, SM1, SM2 )
R16 ( 2, SM1, SM2, SM1, SM2, SM1, SM2, SM1, SM2 )
R16 ( 3, SM1, SM2, NNN, SM2, NNN, NNN, NNN, NNN )
ADD_EPI32(state0, state0_save)
ADD_EPI32(state1, state1_save)
data += 64;
}
while (--numBlocks);
PREPARE_STATE
_mm_storeu_si128((__m128i *) (void *) &state[0], state0);
_mm_storeu_si128((__m128i *) (void *) &state[4], state1);
}
#endif // USE_HW_SHA
#elif defined(MY_CPU_ARM_OR_ARM64) && defined(MY_CPU_LE)
#if defined(__ARM_FEATURE_SHA2) \
|| defined(__ARM_FEATURE_CRYPTO)
#define USE_HW_SHA
#else
#if defined(MY_CPU_ARM64) \
|| defined(__ARM_ARCH) && (__ARM_ARCH >= 4) \
|| defined(Z7_MSC_VER_ORIGINAL)
#if defined(__ARM_FP) && \
( defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(__GNUC__) && (__GNUC__ >= 6) \
) \
|| defined(Z7_MSC_VER_ORIGINAL) && (_MSC_VER >= 1910)
#if defined(MY_CPU_ARM64) \
|| !defined(Z7_CLANG_VERSION) \
|| defined(__ARM_NEON) && \
(Z7_CLANG_VERSION < 170000 || \
Z7_CLANG_VERSION > 170001)
#define USE_HW_SHA
#endif
#endif
#endif
#endif
#ifdef USE_HW_SHA
// #pragma message("=== Sha256 HW === ")
#if defined(__clang__) || defined(__GNUC__)
#if !defined(__ARM_FEATURE_SHA2) && \
!defined(__ARM_FEATURE_CRYPTO)
#ifdef MY_CPU_ARM64
#if defined(__clang__)
#define ATTRIB_SHA __attribute__((__target__("crypto")))
#else
#define ATTRIB_SHA __attribute__((__target__("+crypto")))
#endif
#else
#if defined(__clang__) && (__clang_major__ >= 1)
#define ATTRIB_SHA __attribute__((__target__("armv8-a,sha2")))
#else
#define ATTRIB_SHA __attribute__((__target__("fpu=crypto-neon-fp-armv8")))
#endif
#endif
#endif
#else
// _MSC_VER
// for arm32
#define _ARM_USE_NEW_NEON_INTRINSICS
#endif
#if defined(Z7_MSC_VER_ORIGINAL) && defined(MY_CPU_ARM64)
#include <arm64_neon.h>
#else
#if defined(__clang__) && __clang_major__ < 16
#if !defined(__ARM_FEATURE_SHA2) && \
!defined(__ARM_FEATURE_CRYPTO)
// #pragma message("=== we set __ARM_FEATURE_CRYPTO 1 === ")
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define Z7_ARM_FEATURE_CRYPTO_WAS_SET 1
// #if defined(__clang__) && __clang_major__ < 13
#define __ARM_FEATURE_CRYPTO 1
// #else
#define __ARM_FEATURE_SHA2 1
// #endif
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // clang
#if defined(__clang__)
#if defined(__ARM_ARCH) && __ARM_ARCH < 8
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
// #pragma message("#define __ARM_ARCH 8")
#undef __ARM_ARCH
#define __ARM_ARCH 8
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // clang
#include <arm_neon.h>
#if defined(Z7_ARM_FEATURE_CRYPTO_WAS_SET) && \
defined(__ARM_FEATURE_CRYPTO) && \
defined(__ARM_FEATURE_SHA2)
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#undef __ARM_FEATURE_CRYPTO
#undef __ARM_FEATURE_SHA2
#undef Z7_ARM_FEATURE_CRYPTO_WAS_SET
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
// #pragma message("=== we undefine __ARM_FEATURE_CRYPTO === ")
#endif
#endif // Z7_MSC_VER_ORIGINAL
typedef uint32x4_t v128;
// typedef __n128 v128; // MSVC
#ifdef MY_CPU_BE
#define MY_rev32_for_LE(x) x
#else
#define MY_rev32_for_LE(x) vrev32q_u8(x)
#endif
#if 1 // 0 for debug
// for arm32: it works slower by some reason than direct code
/*
for arm32 it generates:
MSVC-2022, GCC-9:
vld1.32 {d18,d19}, [r10]
vst1.32 {d4,d5}, [r3]
vld1.8 {d20-d21}, [r4]
there is no align hint (like [r10:128]). So instruction allows unaligned access
*/
#define LOAD_128_32(_p) vld1q_u32(_p)
#define LOAD_128_8(_p) vld1q_u8 (_p)
#define STORE_128_32(_p, _v) vst1q_u32(_p, _v)
#else
/*
for arm32:
MSVC-2022:
vldm r10,{d18,d19}
vstm r3,{d4,d5}
does it require strict alignment?
GCC-9:
vld1.64 {d30-d31}, [r0:64]
vldr d28, [r0, #16]
vldr d29, [r0, #24]
vst1.64 {d30-d31}, [r0:64]
vstr d28, [r0, #16]
vstr d29, [r0, #24]
there is hint [r0:64], so does it requires 64-bit alignment.
*/
#define LOAD_128_32(_p) (*(const v128 *)(const void *)(_p))
#define LOAD_128_8(_p) vreinterpretq_u8_u32(*(const v128 *)(const void *)(_p))
#define STORE_128_32(_p, _v) *(v128 *)(void *)(_p) = (_v)
#endif
#define LOAD_SHUFFLE(m, k) \
m = vreinterpretq_u32_u8( \
MY_rev32_for_LE( \
LOAD_128_8(data + (k) * 16))); \
// K array must be aligned for 16-bytes at least.
extern
MY_ALIGN(64)
const UInt32 SHA256_K_ARRAY[64];
#define K SHA256_K_ARRAY
#define SHA256_SU0(dest, src) dest = vsha256su0q_u32(dest, src);
#define SHA256_SU1(dest, src2, src3) dest = vsha256su1q_u32(dest, src2, src3);
#define SM1(m0, m1, m2, m3) SHA256_SU0(m3, m0)
#define SM2(m0, m1, m2, m3) SHA256_SU1(m2, m0, m1)
#define NNN(m0, m1, m2, m3)
#define R4(k, m0, m1, m2, m3, OP0, OP1) \
msg = vaddq_u32(m0, *(const v128 *) (const void *) &K[(k) * 4]); \
tmp = state0; \
state0 = vsha256hq_u32( state0, state1, msg ); \
state1 = vsha256h2q_u32( state1, tmp, msg ); \
OP0(m0, m1, m2, m3); \
OP1(m0, m1, m2, m3); \
#define R16(k, OP0, OP1, OP2, OP3, OP4, OP5, OP6, OP7) \
R4 ( (k)*4+0, m0, m1, m2, m3, OP0, OP1 ) \
R4 ( (k)*4+1, m1, m2, m3, m0, OP2, OP3 ) \
R4 ( (k)*4+2, m2, m3, m0, m1, OP4, OP5 ) \
R4 ( (k)*4+3, m3, m0, m1, m2, OP6, OP7 ) \
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks);
#ifdef ATTRIB_SHA
ATTRIB_SHA
#endif
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks)
{
v128 state0, state1;
if (numBlocks == 0)
return;
state0 = LOAD_128_32(&state[0]);
state1 = LOAD_128_32(&state[4]);
do
{
v128 state0_save, state1_save;
v128 m0, m1, m2, m3;
v128 msg, tmp;
state0_save = state0;
state1_save = state1;
LOAD_SHUFFLE (m0, 0)
LOAD_SHUFFLE (m1, 1)
LOAD_SHUFFLE (m2, 2)
LOAD_SHUFFLE (m3, 3)
R16 ( 0, NNN, NNN, SM1, NNN, SM1, SM2, SM1, SM2 )
R16 ( 1, SM1, SM2, SM1, SM2, SM1, SM2, SM1, SM2 )
R16 ( 2, SM1, SM2, SM1, SM2, SM1, SM2, SM1, SM2 )
R16 ( 3, SM1, SM2, NNN, SM2, NNN, NNN, NNN, NNN )
state0 = vaddq_u32(state0, state0_save);
state1 = vaddq_u32(state1, state1_save);
data += 64;
}
while (--numBlocks);
STORE_128_32(&state[0], state0);
STORE_128_32(&state[4], state1);
}
#endif // USE_HW_SHA
#endif // MY_CPU_ARM_OR_ARM64
#if !defined(USE_HW_SHA) && defined(Z7_USE_HW_SHA_STUB)
// #error Stop_Compiling_UNSUPPORTED_SHA
// #include <stdlib.h>
// We can compile this file with another C compiler,
// or we can compile asm version.
// So we can generate real code instead of this stub function.
// #include "Sha256.h"
// #if defined(_MSC_VER)
#pragma message("Sha256 HW-SW stub was used")
// #endif
void Z7_FASTCALL Sha256_UpdateBlocks (UInt32 state[8], const Byte *data, size_t numBlocks);
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks);
void Z7_FASTCALL Sha256_UpdateBlocks_HW(UInt32 state[8], const Byte *data, size_t numBlocks)
{
Sha256_UpdateBlocks(state, data, numBlocks);
/*
UNUSED_VAR(state);
UNUSED_VAR(data);
UNUSED_VAR(numBlocks);
exit(1);
return;
*/
}
#endif
#undef K
#undef RND2
#undef MY_rev32_for_LE
#undef NNN
#undef LOAD_128
#undef STORE_128
#undef LOAD_SHUFFLE
#undef SM1
#undef SM2
#undef R4
#undef R16
#undef PREPARE_STATE
#undef USE_HW_SHA
#undef ATTRIB_SHA
#undef USE_VER_MIN
#undef Z7_USE_HW_SHA_STUB
third_party/7z/SwapBytes.c
Vendored
+835
View File
@@ -0,0 +1,835 @@
/* SwapBytes.c -- Byte Swap conversion filter
2024-03-01 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Compiler.h"
#include "CpuArch.h"
#include "RotateDefs.h"
#include "SwapBytes.h"
typedef UInt16 CSwapUInt16;
typedef UInt32 CSwapUInt32;
// #define k_SwapBytes_Mode_BASE 0
#ifdef MY_CPU_X86_OR_AMD64
#define k_SwapBytes_Mode_SSE2 1
#define k_SwapBytes_Mode_SSSE3 2
#define k_SwapBytes_Mode_AVX2 3
// #if defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1900)
#if defined(__clang__) && (__clang_major__ >= 4) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40701)
#define k_SwapBytes_Mode_MAX k_SwapBytes_Mode_AVX2
#define SWAP_ATTRIB_SSE2 __attribute__((__target__("sse2")))
#define SWAP_ATTRIB_SSSE3 __attribute__((__target__("ssse3")))
#define SWAP_ATTRIB_AVX2 __attribute__((__target__("avx2")))
#elif defined(_MSC_VER)
#if (_MSC_VER == 1900)
#pragma warning(disable : 4752) // found Intel(R) Advanced Vector Extensions; consider using /arch:AVX
#endif
#if (_MSC_VER >= 1900)
#define k_SwapBytes_Mode_MAX k_SwapBytes_Mode_AVX2
#elif (_MSC_VER >= 1500) // (VS2008)
#define k_SwapBytes_Mode_MAX k_SwapBytes_Mode_SSSE3
#elif (_MSC_VER >= 1310) // (VS2003)
#define k_SwapBytes_Mode_MAX k_SwapBytes_Mode_SSE2
#endif
#endif // _MSC_VER
/*
// for debug
#ifdef k_SwapBytes_Mode_MAX
#undef k_SwapBytes_Mode_MAX
#endif
*/
#ifndef k_SwapBytes_Mode_MAX
#define k_SwapBytes_Mode_MAX 0
#endif
#if (k_SwapBytes_Mode_MAX != 0) && defined(MY_CPU_AMD64)
#define k_SwapBytes_Mode_MIN k_SwapBytes_Mode_SSE2
#else
#define k_SwapBytes_Mode_MIN 0
#endif
#if (k_SwapBytes_Mode_MAX >= k_SwapBytes_Mode_AVX2)
#define USE_SWAP_AVX2
#endif
#if (k_SwapBytes_Mode_MAX >= k_SwapBytes_Mode_SSSE3)
#define USE_SWAP_SSSE3
#endif
#if (k_SwapBytes_Mode_MAX >= k_SwapBytes_Mode_SSE2)
#define USE_SWAP_128
#endif
#if k_SwapBytes_Mode_MAX <= k_SwapBytes_Mode_MIN || !defined(USE_SWAP_128)
#define FORCE_SWAP_MODE
#endif
#ifdef USE_SWAP_128
/*
<mmintrin.h> MMX
<xmmintrin.h> SSE
<emmintrin.h> SSE2
<pmmintrin.h> SSE3
<tmmintrin.h> SSSE3
<smmintrin.h> SSE4.1
<nmmintrin.h> SSE4.2
<ammintrin.h> SSE4A
<wmmintrin.h> AES
<immintrin.h> AVX, AVX2, FMA
*/
#include <emmintrin.h> // sse2
// typedef __m128i v128;
#define SWAP2_128(i) { \
const __m128i v = *(const __m128i *)(const void *)(items + (i) * 8); \
*( __m128i *)( void *)(items + (i) * 8) = \
_mm_or_si128( \
_mm_slli_epi16(v, 8), \
_mm_srli_epi16(v, 8)); }
// _mm_or_si128() has more ports to execute than _mm_add_epi16().
static
#ifdef SWAP_ATTRIB_SSE2
SWAP_ATTRIB_SSE2
#endif
void
Z7_FASTCALL
SwapBytes2_128(CSwapUInt16 *items, const CSwapUInt16 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP2_128(0) SWAP2_128(1) items += 2 * 8;
SWAP2_128(0) SWAP2_128(1) items += 2 * 8;
}
while (items != lim);
}
/*
// sse2
#define SWAP4_128_pack(i) { \
__m128i v = *(const __m128i *)(const void *)(items + (i) * 4); \
__m128i v0 = _mm_unpacklo_epi8(v, mask); \
__m128i v1 = _mm_unpackhi_epi8(v, mask); \
v0 = _mm_shufflelo_epi16(v0, 0x1b); \
v1 = _mm_shufflelo_epi16(v1, 0x1b); \
v0 = _mm_shufflehi_epi16(v0, 0x1b); \
v1 = _mm_shufflehi_epi16(v1, 0x1b); \
*(__m128i *)(void *)(items + (i) * 4) = _mm_packus_epi16(v0, v1); }
static
#ifdef SWAP_ATTRIB_SSE2
SWAP_ATTRIB_SSE2
#endif
void
Z7_FASTCALL
SwapBytes4_128_pack(CSwapUInt32 *items, const CSwapUInt32 *lim)
{
const __m128i mask = _mm_setzero_si128();
// const __m128i mask = _mm_set_epi16(0, 0, 0, 0, 0, 0, 0, 0);
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP4_128_pack(0); items += 1 * 4;
// SWAP4_128_pack(0); SWAP4_128_pack(1); items += 2 * 4;
}
while (items != lim);
}
// sse2
#define SWAP4_128_shift(i) { \
__m128i v = *(const __m128i *)(const void *)(items + (i) * 4); \
__m128i v2; \
v2 = _mm_or_si128( \
_mm_slli_si128(_mm_and_si128(v, mask), 1), \
_mm_and_si128(_mm_srli_si128(v, 1), mask)); \
v = _mm_or_si128( \
_mm_slli_epi32(v, 24), \
_mm_srli_epi32(v, 24)); \
*(__m128i *)(void *)(items + (i) * 4) = _mm_or_si128(v2, v); }
static
#ifdef SWAP_ATTRIB_SSE2
SWAP_ATTRIB_SSE2
#endif
void
Z7_FASTCALL
SwapBytes4_128_shift(CSwapUInt32 *items, const CSwapUInt32 *lim)
{
#define M1 0xff00
const __m128i mask = _mm_set_epi32(M1, M1, M1, M1);
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
// SWAP4_128_shift(0) SWAP4_128_shift(1) items += 2 * 4;
// SWAP4_128_shift(0) SWAP4_128_shift(1) items += 2 * 4;
SWAP4_128_shift(0); items += 1 * 4;
}
while (items != lim);
}
*/
#if defined(USE_SWAP_SSSE3) || defined(USE_SWAP_AVX2)
#define SWAP_SHUF_REV_SEQ_2_VALS(v) (v)+1, (v)
#define SWAP_SHUF_REV_SEQ_4_VALS(v) (v)+3, (v)+2, (v)+1, (v)
#define SWAP2_SHUF_MASK_16_BYTES \
SWAP_SHUF_REV_SEQ_2_VALS (0 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (1 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (2 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (3 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (4 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (5 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (6 * 2), \
SWAP_SHUF_REV_SEQ_2_VALS (7 * 2)
#define SWAP4_SHUF_MASK_16_BYTES \
SWAP_SHUF_REV_SEQ_4_VALS (0 * 4), \
SWAP_SHUF_REV_SEQ_4_VALS (1 * 4), \
SWAP_SHUF_REV_SEQ_4_VALS (2 * 4), \
SWAP_SHUF_REV_SEQ_4_VALS (3 * 4)
#if defined(USE_SWAP_AVX2)
/* if we use 256_BIT_INIT_MASK, each static array mask will be larger for 16 bytes */
// #define SWAP_USE_256_BIT_INIT_MASK
#endif
#if defined(SWAP_USE_256_BIT_INIT_MASK) && defined(USE_SWAP_AVX2)
#define SWAP_MASK_INIT_SIZE 32
#else
#define SWAP_MASK_INIT_SIZE 16
#endif
MY_ALIGN(SWAP_MASK_INIT_SIZE)
static const Byte k_ShufMask_Swap2[] =
{
SWAP2_SHUF_MASK_16_BYTES
#if SWAP_MASK_INIT_SIZE > 16
, SWAP2_SHUF_MASK_16_BYTES
#endif
};
MY_ALIGN(SWAP_MASK_INIT_SIZE)
static const Byte k_ShufMask_Swap4[] =
{
SWAP4_SHUF_MASK_16_BYTES
#if SWAP_MASK_INIT_SIZE > 16
, SWAP4_SHUF_MASK_16_BYTES
#endif
};
#ifdef USE_SWAP_SSSE3
#include <tmmintrin.h> // ssse3
#define SHUF_128(i) *(items + (i)) = \
_mm_shuffle_epi8(*(items + (i)), mask); // SSSE3
// Z7_NO_INLINE
static
#ifdef SWAP_ATTRIB_SSSE3
SWAP_ATTRIB_SSSE3
#endif
Z7_ATTRIB_NO_VECTORIZE
void
Z7_FASTCALL
ShufBytes_128(void *items8, const void *lim8, const void *mask128_ptr)
{
__m128i *items = (__m128i *)items8;
const __m128i *lim = (const __m128i *)lim8;
// const __m128i mask = _mm_set_epi8(SHUF_SWAP2_MASK_16_VALS);
// const __m128i mask = _mm_set_epi8(SHUF_SWAP4_MASK_16_VALS);
// const __m128i mask = _mm_load_si128((const __m128i *)(const void *)&(k_ShufMask_Swap4[0]));
// const __m128i mask = _mm_load_si128((const __m128i *)(const void *)&(k_ShufMask_Swap4[0]));
// const __m128i mask = *(const __m128i *)(const void *)&(k_ShufMask_Swap4[0]);
const __m128i mask = *(const __m128i *)mask128_ptr;
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SHUF_128(0) SHUF_128(1) items += 2;
SHUF_128(0) SHUF_128(1) items += 2;
}
while (items != lim);
}
#endif // USE_SWAP_SSSE3
#ifdef USE_SWAP_AVX2
#include <immintrin.h> // avx, avx2
#if defined(__clang__)
#include <avxintrin.h>
#include <avx2intrin.h>
#endif
#define SHUF_256(i) *(items + (i)) = \
_mm256_shuffle_epi8(*(items + (i)), mask); // AVX2
// Z7_NO_INLINE
static
#ifdef SWAP_ATTRIB_AVX2
SWAP_ATTRIB_AVX2
#endif
Z7_ATTRIB_NO_VECTORIZE
void
Z7_FASTCALL
ShufBytes_256(void *items8, const void *lim8, const void *mask128_ptr)
{
__m256i *items = (__m256i *)items8;
const __m256i *lim = (const __m256i *)lim8;
/*
UNUSED_VAR(mask128_ptr)
__m256i mask =
for Swap4: _mm256_setr_epi8(SWAP4_SHUF_MASK_16_BYTES, SWAP4_SHUF_MASK_16_BYTES);
for Swap2: _mm256_setr_epi8(SWAP2_SHUF_MASK_16_BYTES, SWAP2_SHUF_MASK_16_BYTES);
*/
const __m256i mask =
#if SWAP_MASK_INIT_SIZE > 16
*(const __m256i *)(const void *)mask128_ptr;
#else
/* msvc: broadcastsi128() version reserves the stack for no reason
msvc 19.29-: _mm256_insertf128_si256() / _mm256_set_m128i)) versions use non-avx movdqu xmm0,XMMWORD PTR [r8]
msvc 19.30+ (VS2022): replaces _mm256_set_m128i(m,m) to vbroadcastf128(m) as we want
*/
// _mm256_broadcastsi128_si256(*mask128_ptr);
#if defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION < 80000)
#define MY_mm256_set_m128i(hi, lo) _mm256_insertf128_si256(_mm256_castsi128_si256(lo), (hi), 1)
#else
#define MY_mm256_set_m128i _mm256_set_m128i
#endif
MY_mm256_set_m128i(
*(const __m128i *)mask128_ptr,
*(const __m128i *)mask128_ptr);
#endif
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SHUF_256(0) SHUF_256(1) items += 2;
SHUF_256(0) SHUF_256(1) items += 2;
}
while (items != lim);
}
#endif // USE_SWAP_AVX2
#endif // USE_SWAP_SSSE3 || USE_SWAP_AVX2
#endif // USE_SWAP_128
// compile message "NEON intrinsics not available with the soft-float ABI"
#elif defined(MY_CPU_ARM_OR_ARM64) \
&& defined(MY_CPU_LE) \
&& !defined(Z7_DISABLE_ARM_NEON)
#if defined(__clang__) && (__clang_major__ >= 8) \
|| defined(__GNUC__) && (__GNUC__ >= 6)
#if defined(__ARM_FP)
#if (defined(__ARM_ARCH) && (__ARM_ARCH >= 4)) \
|| defined(MY_CPU_ARM64)
#if defined(MY_CPU_ARM64) \
|| !defined(Z7_CLANG_VERSION) \
|| defined(__ARM_NEON)
#define USE_SWAP_128
#ifdef MY_CPU_ARM64
// #define SWAP_ATTRIB_NEON __attribute__((__target__("")))
#else
#if defined(Z7_CLANG_VERSION)
// #define SWAP_ATTRIB_NEON __attribute__((__target__("neon")))
#else
// #pragma message("SWAP_ATTRIB_NEON __attribute__((__target__(fpu=neon))")
#define SWAP_ATTRIB_NEON __attribute__((__target__("fpu=neon")))
#endif
#endif // MY_CPU_ARM64
#endif // __ARM_NEON
#endif // __ARM_ARCH
#endif // __ARM_FP
#elif defined(_MSC_VER)
#if (_MSC_VER >= 1910)
#define USE_SWAP_128
#endif
#endif
#ifdef USE_SWAP_128
#if defined(Z7_MSC_VER_ORIGINAL) && defined(MY_CPU_ARM64)
#include <arm64_neon.h>
#else
/*
#if !defined(__ARM_NEON)
#if defined(Z7_GCC_VERSION) && (__GNUC__ < 5) \
|| defined(Z7_GCC_VERSION) && (__GNUC__ == 5) && (Z7_GCC_VERSION < 90201) \
|| defined(Z7_GCC_VERSION) && (__GNUC__ == 5) && (Z7_GCC_VERSION < 100100)
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#pragma message("#define __ARM_NEON 1")
// #define __ARM_NEON 1
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif
*/
#include <arm_neon.h>
#endif
#endif
#ifndef USE_SWAP_128
#define FORCE_SWAP_MODE
#else
#ifdef MY_CPU_ARM64
// for debug : comment it
#define FORCE_SWAP_MODE
#else
#define k_SwapBytes_Mode_NEON 1
#endif
// typedef uint8x16_t v128;
#define SWAP2_128(i) *(uint8x16_t *) (void *)(items + (i) * 8) = \
vrev16q_u8(*(const uint8x16_t *)(const void *)(items + (i) * 8));
#define SWAP4_128(i) *(uint8x16_t *) (void *)(items + (i) * 4) = \
vrev32q_u8(*(const uint8x16_t *)(const void *)(items + (i) * 4));
// Z7_NO_INLINE
static
#ifdef SWAP_ATTRIB_NEON
SWAP_ATTRIB_NEON
#endif
Z7_ATTRIB_NO_VECTORIZE
void
Z7_FASTCALL
SwapBytes2_128(CSwapUInt16 *items, const CSwapUInt16 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP2_128(0) SWAP2_128(1) items += 2 * 8;
SWAP2_128(0) SWAP2_128(1) items += 2 * 8;
}
while (items != lim);
}
// Z7_NO_INLINE
static
#ifdef SWAP_ATTRIB_NEON
SWAP_ATTRIB_NEON
#endif
Z7_ATTRIB_NO_VECTORIZE
void
Z7_FASTCALL
SwapBytes4_128(CSwapUInt32 *items, const CSwapUInt32 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP4_128(0) SWAP4_128(1) items += 2 * 4;
SWAP4_128(0) SWAP4_128(1) items += 2 * 4;
}
while (items != lim);
}
#endif // USE_SWAP_128
#else // MY_CPU_ARM_OR_ARM64
#define FORCE_SWAP_MODE
#endif // MY_CPU_ARM_OR_ARM64
#if defined(Z7_MSC_VER_ORIGINAL) && defined(MY_CPU_X86)
/* _byteswap_ushort() in MSVC x86 32-bit works via slow { mov dh, al; mov dl, ah }
So we use own versions of byteswap function */
#if (_MSC_VER < 1400 ) // old MSVC-X86 without _rotr16() support
#define SWAP2_16(i) { UInt32 v = items[i]; v += (v << 16); v >>= 8; items[i] = (CSwapUInt16)v; }
#else // is new MSVC-X86 with fast _rotr16()
#include <intrin.h>
#define SWAP2_16(i) { items[i] = _rotr16(items[i], 8); }
#endif
#else // is not MSVC-X86
#define SWAP2_16(i) { CSwapUInt16 v = items[i]; items[i] = Z7_BSWAP16(v); }
#endif // MSVC-X86
#if defined(Z7_CPU_FAST_BSWAP_SUPPORTED)
#define SWAP4_32(i) { CSwapUInt32 v = items[i]; items[i] = Z7_BSWAP32(v); }
#else
#define SWAP4_32(i) \
{ UInt32 v = items[i]; \
v = ((v & 0xff00ff) << 8) + ((v >> 8) & 0xff00ff); \
v = rotlFixed(v, 16); \
items[i] = v; }
#endif
#if defined(FORCE_SWAP_MODE) && defined(USE_SWAP_128)
#define DEFAULT_Swap2 SwapBytes2_128
#if !defined(MY_CPU_X86_OR_AMD64)
#define DEFAULT_Swap4 SwapBytes4_128
#endif
#endif
#if !defined(DEFAULT_Swap2) || !defined(DEFAULT_Swap4)
#define SWAP_BASE_FUNCS_PREFIXES \
Z7_FORCE_INLINE \
static \
Z7_ATTRIB_NO_VECTOR \
void Z7_FASTCALL
#if defined(MY_CPU_ARM_OR_ARM64)
#if defined(__clang__)
#pragma GCC diagnostic ignored "-Wlanguage-extension-token"
#endif
#endif
#ifdef MY_CPU_64BIT
#if defined(MY_CPU_ARM64) \
&& defined(__ARM_ARCH) && (__ARM_ARCH >= 8) \
&& ( (defined(__GNUC__) && (__GNUC__ >= 4)) \
|| (defined(__clang__) && (__clang_major__ >= 4)))
#define SWAP2_64_VAR(v) asm ("rev16 %x0,%x0" : "+r" (v));
#define SWAP4_64_VAR(v) asm ("rev32 %x0,%x0" : "+r" (v));
#else // is not ARM64-GNU
#if !defined(MY_CPU_X86_OR_AMD64) || (k_SwapBytes_Mode_MIN == 0) || !defined(USE_SWAP_128)
#define SWAP2_64_VAR(v) \
v = ( 0x00ff00ff00ff00ff & (v >> 8)) \
+ ((0x00ff00ff00ff00ff & v) << 8);
/* plus gives faster code in MSVC */
#endif
#ifdef Z7_CPU_FAST_BSWAP_SUPPORTED
#define SWAP4_64_VAR(v) \
v = Z7_BSWAP64(v); \
v = Z7_ROTL64(v, 32);
#else
#define SWAP4_64_VAR(v) \
v = ( 0x000000ff000000ff & (v >> 24)) \
+ ((0x000000ff000000ff & v) << 24 ) \
+ ( 0x0000ff000000ff00 & (v >> 8)) \
+ ((0x0000ff000000ff00 & v) << 8 ) \
;
#endif
#endif // ARM64-GNU
#ifdef SWAP2_64_VAR
#define SWAP2_64(i) { \
UInt64 v = *(const UInt64 *)(const void *)(items + (i) * 4); \
SWAP2_64_VAR(v) \
*(UInt64 *)(void *)(items + (i) * 4) = v; }
SWAP_BASE_FUNCS_PREFIXES
SwapBytes2_64(CSwapUInt16 *items, const CSwapUInt16 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP2_64(0) SWAP2_64(1) items += 2 * 4;
SWAP2_64(0) SWAP2_64(1) items += 2 * 4;
}
while (items != lim);
}
#define DEFAULT_Swap2 SwapBytes2_64
#if !defined(FORCE_SWAP_MODE)
#define SWAP2_DEFAULT_MODE 0
#endif
#else // !defined(SWAP2_64_VAR)
#define DEFAULT_Swap2 SwapBytes2_128
#if !defined(FORCE_SWAP_MODE)
#define SWAP2_DEFAULT_MODE 1
#endif
#endif // SWAP2_64_VAR
#define SWAP4_64(i) { \
UInt64 v = *(const UInt64 *)(const void *)(items + (i) * 2); \
SWAP4_64_VAR(v) \
*(UInt64 *)(void *)(items + (i) * 2) = v; }
SWAP_BASE_FUNCS_PREFIXES
SwapBytes4_64(CSwapUInt32 *items, const CSwapUInt32 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP4_64(0) SWAP4_64(1) items += 2 * 2;
SWAP4_64(0) SWAP4_64(1) items += 2 * 2;
}
while (items != lim);
}
#define DEFAULT_Swap4 SwapBytes4_64
#else // is not 64BIT
#if defined(MY_CPU_ARM_OR_ARM64) \
&& defined(__ARM_ARCH) && (__ARM_ARCH >= 6) \
&& ( (defined(__GNUC__) && (__GNUC__ >= 4)) \
|| (defined(__clang__) && (__clang_major__ >= 4)))
#ifdef MY_CPU_64BIT
#define SWAP2_32_VAR(v) asm ("rev16 %w0,%w0" : "+r" (v));
#else
#define SWAP2_32_VAR(v) asm ("rev16 %0,%0" : "+r" (v)); // for clang/gcc
// asm ("rev16 %r0,%r0" : "+r" (a)); // for gcc
#endif
#elif defined(_MSC_VER) && (_MSC_VER < 1300) && defined(MY_CPU_X86) \
|| !defined(Z7_CPU_FAST_BSWAP_SUPPORTED) \
|| !defined(Z7_CPU_FAST_ROTATE_SUPPORTED)
// old msvc doesn't support _byteswap_ulong()
#define SWAP2_32_VAR(v) \
v = ((v & 0xff00ff) << 8) + ((v >> 8) & 0xff00ff);
#else // is not ARM and is not old-MSVC-X86 and fast BSWAP/ROTATE are supported
#define SWAP2_32_VAR(v) \
v = Z7_BSWAP32(v); \
v = rotlFixed(v, 16);
#endif // GNU-ARM*
#define SWAP2_32(i) { \
UInt32 v = *(const UInt32 *)(const void *)(items + (i) * 2); \
SWAP2_32_VAR(v); \
*(UInt32 *)(void *)(items + (i) * 2) = v; }
SWAP_BASE_FUNCS_PREFIXES
SwapBytes2_32(CSwapUInt16 *items, const CSwapUInt16 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP2_32(0) SWAP2_32(1) items += 2 * 2;
SWAP2_32(0) SWAP2_32(1) items += 2 * 2;
}
while (items != lim);
}
SWAP_BASE_FUNCS_PREFIXES
SwapBytes4_32(CSwapUInt32 *items, const CSwapUInt32 *lim)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SWAP4_32(0) SWAP4_32(1) items += 2;
SWAP4_32(0) SWAP4_32(1) items += 2;
}
while (items != lim);
}
#define DEFAULT_Swap2 SwapBytes2_32
#define DEFAULT_Swap4 SwapBytes4_32
#if !defined(FORCE_SWAP_MODE)
#define SWAP2_DEFAULT_MODE 0
#endif
#endif // MY_CPU_64BIT
#endif // if !defined(DEFAULT_Swap2) || !defined(DEFAULT_Swap4)
#if !defined(FORCE_SWAP_MODE)
static unsigned g_SwapBytes_Mode;
#endif
/* size of largest unrolled loop iteration: 128 bytes = 4 * 32 bytes (AVX). */
#define SWAP_ITERATION_BLOCK_SIZE_MAX (1 << 7)
// 32 bytes for (AVX) or 2 * 16-bytes for NEON.
#define SWAP_VECTOR_ALIGN_SIZE (1 << 5)
Z7_NO_INLINE
void z7_SwapBytes2(CSwapUInt16 *items, size_t numItems)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
for (; numItems != 0 && ((unsigned)(ptrdiff_t)items & (SWAP_VECTOR_ALIGN_SIZE - 1)) != 0; numItems--)
{
SWAP2_16(0)
items++;
}
{
const size_t k_Align_Mask = SWAP_ITERATION_BLOCK_SIZE_MAX / sizeof(CSwapUInt16) - 1;
size_t numItems2 = numItems;
CSwapUInt16 *lim;
numItems &= k_Align_Mask;
numItems2 &= ~(size_t)k_Align_Mask;
lim = items + numItems2;
if (numItems2 != 0)
{
#if !defined(FORCE_SWAP_MODE)
#ifdef MY_CPU_X86_OR_AMD64
#ifdef USE_SWAP_AVX2
if (g_SwapBytes_Mode > k_SwapBytes_Mode_SSSE3)
ShufBytes_256((__m256i *)(void *)items,
(const __m256i *)(const void *)lim,
(const __m128i *)(const void *)&(k_ShufMask_Swap2[0]));
else
#endif
#ifdef USE_SWAP_SSSE3
if (g_SwapBytes_Mode >= k_SwapBytes_Mode_SSSE3)
ShufBytes_128((__m128i *)(void *)items,
(const __m128i *)(const void *)lim,
(const __m128i *)(const void *)&(k_ShufMask_Swap2[0]));
else
#endif
#endif // MY_CPU_X86_OR_AMD64
#if SWAP2_DEFAULT_MODE == 0
if (g_SwapBytes_Mode != 0)
SwapBytes2_128(items, lim);
else
#endif
#endif // FORCE_SWAP_MODE
DEFAULT_Swap2(items, lim);
}
items = lim;
}
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
for (; numItems != 0; numItems--)
{
SWAP2_16(0)
items++;
}
}
Z7_NO_INLINE
void z7_SwapBytes4(CSwapUInt32 *items, size_t numItems)
{
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
for (; numItems != 0 && ((unsigned)(ptrdiff_t)items & (SWAP_VECTOR_ALIGN_SIZE - 1)) != 0; numItems--)
{
SWAP4_32(0)
items++;
}
{
const size_t k_Align_Mask = SWAP_ITERATION_BLOCK_SIZE_MAX / sizeof(CSwapUInt32) - 1;
size_t numItems2 = numItems;
CSwapUInt32 *lim;
numItems &= k_Align_Mask;
numItems2 &= ~(size_t)k_Align_Mask;
lim = items + numItems2;
if (numItems2 != 0)
{
#if !defined(FORCE_SWAP_MODE)
#ifdef MY_CPU_X86_OR_AMD64
#ifdef USE_SWAP_AVX2
if (g_SwapBytes_Mode > k_SwapBytes_Mode_SSSE3)
ShufBytes_256((__m256i *)(void *)items,
(const __m256i *)(const void *)lim,
(const __m128i *)(const void *)&(k_ShufMask_Swap4[0]));
else
#endif
#ifdef USE_SWAP_SSSE3
if (g_SwapBytes_Mode >= k_SwapBytes_Mode_SSSE3)
ShufBytes_128((__m128i *)(void *)items,
(const __m128i *)(const void *)lim,
(const __m128i *)(const void *)&(k_ShufMask_Swap4[0]));
else
#endif
#else // MY_CPU_X86_OR_AMD64
if (g_SwapBytes_Mode != 0)
SwapBytes4_128(items, lim);
else
#endif // MY_CPU_X86_OR_AMD64
#endif // FORCE_SWAP_MODE
DEFAULT_Swap4(items, lim);
}
items = lim;
}
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
for (; numItems != 0; numItems--)
{
SWAP4_32(0)
items++;
}
}
// #define SHOW_HW_STATUS
#ifdef SHOW_HW_STATUS
#include <stdio.h>
#define PRF(x) x
#else
#define PRF(x)
#endif
void z7_SwapBytesPrepare(void)
{
#ifndef FORCE_SWAP_MODE
unsigned mode = 0; // k_SwapBytes_Mode_BASE;
#ifdef MY_CPU_ARM_OR_ARM64
{
if (CPU_IsSupported_NEON())
{
// #pragma message ("=== SwapBytes NEON")
PRF(printf("\n=== SwapBytes NEON\n");)
mode = k_SwapBytes_Mode_NEON;
}
}
#else // MY_CPU_ARM_OR_ARM64
{
#ifdef USE_SWAP_AVX2
if (CPU_IsSupported_AVX2())
{
// #pragma message ("=== SwapBytes AVX2")
PRF(printf("\n=== SwapBytes AVX2\n");)
mode = k_SwapBytes_Mode_AVX2;
}
else
#endif
#ifdef USE_SWAP_SSSE3
if (CPU_IsSupported_SSSE3())
{
// #pragma message ("=== SwapBytes SSSE3")
PRF(printf("\n=== SwapBytes SSSE3\n");)
mode = k_SwapBytes_Mode_SSSE3;
}
else
#endif
#if !defined(MY_CPU_AMD64)
if (CPU_IsSupported_SSE2())
#endif
{
// #pragma message ("=== SwapBytes SSE2")
PRF(printf("\n=== SwapBytes SSE2\n");)
mode = k_SwapBytes_Mode_SSE2;
}
}
#endif // MY_CPU_ARM_OR_ARM64
g_SwapBytes_Mode = mode;
// g_SwapBytes_Mode = 0; // for debug
#endif // FORCE_SWAP_MODE
PRF(printf("\n=== SwapBytesPrepare\n");)
}
#undef PRF
third_party/7z/SwapBytes.h
Vendored
+17
View File
@@ -0,0 +1,17 @@
/* SwapBytes.h -- Byte Swap conversion filter
2023-04-02 : Igor Pavlov : Public domain */
#ifndef ZIP7_INC_SWAP_BYTES_H
#define ZIP7_INC_SWAP_BYTES_H
#include "7zTypes.h"
EXTERN_C_BEGIN
void z7_SwapBytes2(UInt16 *data, size_t numItems);
void z7_SwapBytes4(UInt32 *data, size_t numItems);
void z7_SwapBytesPrepare(void);
EXTERN_C_END
#endif