Lets us reduce inserts by seeing which bits in the selector mask
indicates a particular source is used more than the other one, and
then just uses that as the base to be inserted into, cutting down
on overall insertion overhead.
In some cases, this can be quite drastic, like with:
vpblendw ymm0, ymm1, ymm2, 0b00000001
being cut down from 98 instructions to 14.
We had a few places that were using Bind, and a few other places
that were using specializations as a means to composing the instruction
tables. Instead, we can just use Bind consistently, which lets us tidy
up a bunch of the implementations (and gets rid of some unnecessary
codegen).
While not a leak in the traditional sense, we were causing pool
allocations to never become free until the thread was closed.
This meant in the case of a game running with >200 threads or so, these
would add up very quickly. So some minor reworking so the IREmitter
doesn't allocate a buffer until first JIT, and making sure to actually
disown the buffer on dispatch error resolved the problems.
Fixes an edge case where Ender Lilies was consuming 409MB with THP
enabled on my desktop, and now it is something like 6MB once idling for
a bit to have the pool allocations do its magic.
The AMD documentation about this instruction is very vague and
misleading in multiple ways. While the Intel documentation is much
cleaner and explains how we need to implement these.
8 of these "new" operations are just inverted signaling versions of the
original 8 SSE versions.
The remaining 16 new operations fill gaps in the original x86 version of
the instructions, exposing the 5 bit truth tables directly, which is why
we also have a "true" and "false" version as well.
Both scalar and vector wide.
Fixes#5326
For the upper-half of the registers it is more efficient to zero the
context with `dc zva` on Ampere1A hardware, while Cortex implements this
as equivalent uops in their store pipeline and aren't affected one way
or the other. ARM C1-Pro and newer with FEAT_MOPS also match `dc zva`
performance with 64B/c, but theoretically slightly fewer instructions.
C1-Nano on the other hand, clearly loses to `dc zva`, where mops can
only do 16B/c, but `dc zva` does 64B/c. So we'll need to benchmark or
not if MOPS is a clear win once hardware is actually shipping.
Most constants don't need to be padded for relocations. So now that
these have all been audited, switch to defaulting to NoPad to reduce
verbosity.
The number of constant that need to be explicitly padded are now marked
and with all the prior changes, this allows bisecting if something has
gone wrong.
In order to support code caching of 32-bit libraries, any library-base
relative relocations on the guest must be transformed into FEX
relocations so e.g. absolute jumps or loads refer to the correct
location when the library is loaded at a different base address.
A few games were generating "Can't handle adddress size".
I implemented 0x67 prefix handling for CMPSOp and SCASOP and improved
the error messages for the remainder. This will implement the address
modifier on 64bit systems, and keep issuing an error on 32bits.
Just a trivial rearrangement, Drops the struct down to 40 bytes.
We can also make some functions take it by const reference so we
aren't churning some copies.
Before:
text data bss dec hex filename
4160559 1471360 4336824 9968743 981c67 Bin/FEX
After:
text data bss dec hex filename
4159927 1471360 4336824 9968111 9819ef Bin/FEX
Avoids actively doing this wonky thing where we're passing
iInvalid all over the place to mean variable alignment depending
on store element size or GPR size.
Makes using the API a little more visibly straightforward and makes
cases where alignment matters more explicit.
Reduces a bunch of noise related to the register classes and hoists them
out so that converting the classes over to enums should be fairly
straightforward.
Instead of having this sort of odd indirection through a struct type,
we can add support for defining custom enums in the IR description.
This lets us both get strong typing (and allow for weak typing, should
any enum in the future need it), without needing a struct for a basic
value type.
Even then, if we do need a struct for anything in the future, then
we still allow strong typing for values themselves while allowing
them to be used in various ways.
This adds a new mode X87StrictReducedPrecision.
The strict reduced precision is like the reduced precision but adds extra checks,
like the currently implemented nan and snan propagations.
Fix for __builtin_issignaling() test of SPEC2017 classify test.
We were paying a large cost per Literal type that we can special case
for the two class of instructions that use a 64-bit literal.
If we packed this would get to a further 62 bytes but probably not worth
it.
This is now cached internally in the X87 pass, and all operations
outside of that that use it are rare so can afford loading/storing
directly from context after flushing x87 regs.
LC