diff --git a/FEXCore/Source/Interface/Core/CPUBackend.cpp b/FEXCore/Source/Interface/Core/CPUBackend.cpp
index 5f7416430..dec22b54f 100644
--- a/FEXCore/Source/Interface/Core/CPUBackend.cpp
+++ b/FEXCore/Source/Interface/Core/CPUBackend.cpp
@@ -186,6 +186,9 @@ CPUBackend::CPUBackend(FEXCore::Core::InternalThreadState *ThreadState, size_t I
     Common.NamedVectorConstantPointers[i] = reinterpret_cast<uint64_t>(NamedVectorConstants[i]);
   }
 
+  // Copy named vector constants.
+  memcpy(Common.NamedVectorConstants, NamedVectorConstants, sizeof(NamedVectorConstants));
+
   // Initialize Indexed named vector constants.
   Common.IndexedNamedVectorConstantPointers[FEXCore::IR::IndexNamedVectorConstant::INDEXED_NAMED_VECTOR_PSHUFLW] = reinterpret_cast<uint64_t>(PSHUFLW_LUT.data());
   Common.IndexedNamedVectorConstantPointers[FEXCore::IR::IndexNamedVectorConstant::INDEXED_NAMED_VECTOR_PSHUFHW] = reinterpret_cast<uint64_t>(PSHUFHW_LUT.data());
diff --git a/FEXCore/Source/Interface/Core/JIT/Arm64/VectorOps.cpp b/FEXCore/Source/Interface/Core/JIT/Arm64/VectorOps.cpp
index a7b68a428..a5423982a 100644
--- a/FEXCore/Source/Interface/Core/JIT/Arm64/VectorOps.cpp
+++ b/FEXCore/Source/Interface/Core/JIT/Arm64/VectorOps.cpp
@@ -101,28 +101,42 @@ DEF_OP(LoadNamedVectorConstant) {
     }
   }
   // Load the pointer.
-  ldr(TMP1, STATE_PTR(CpuStateFrame, Pointers.Common.NamedVectorConstantPointers[Op->Constant]));
+  auto GenerateMemOperand = [this](uint8_t OpSize, uint32_t NamedConstant, FEXCore::ARMEmitter::Register Base) {
+    const auto ConstantOffset = offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.NamedVectorConstants[NamedConstant]);
 
+    if (ConstantOffset <= 255 || // Unscaled 9-bit signed
+        ((ConstantOffset & (OpSize - 1)) == 0 && FEXCore::DividePow2(ConstantOffset, OpSize) <= 4095)) /* 12-bit unsigned scaled */ {
+      return ARMEmitter::ExtendedMemOperand(Base.X(), ARMEmitter::IndexType::OFFSET, ConstantOffset);
+    }
+
+    ldr(TMP1, STATE_PTR(CpuStateFrame, Pointers.Common.NamedVectorConstantPointers[NamedConstant]));
+    return ARMEmitter::ExtendedMemOperand(TMP1, ARMEmitter::IndexType::OFFSET, 0);
+  };
+
+  if (OpSize == 32) {
+    // Handle SVE 32-byte variant upfront.
+    ldr(TMP1, STATE_PTR(CpuStateFrame, Pointers.Common.NamedVectorConstantPointers[Op->Constant]));
+    ld1b<ARMEmitter::SubRegSize::i8Bit>(Dst.Z(), PRED_TMP_32B.Zeroing(), TMP1, 0);
+    return;
+  }
+
+  auto MemOperand = GenerateMemOperand(OpSize, Op->Constant, STATE);
   switch (OpSize) {
     case 1:
-      ldrb(Dst, TMP1, 0);
+      ldrb(Dst, MemOperand);
       break;
     case 2:
-      ldrh(Dst, TMP1, 0);
+      ldrh(Dst, MemOperand);
       break;
     case 4:
-      ldr(Dst.S(), TMP1, 0);
+      ldr(Dst.S(), MemOperand);
       break;
     case 8:
-      ldr(Dst.D(), TMP1, 0);
+      ldr(Dst.D(), MemOperand);
       break;
     case 16:
-      ldr(Dst.Q(), TMP1, 0);
+      ldr(Dst.Q(), MemOperand);
       break;
-    case 32: {
-      ld1b<ARMEmitter::SubRegSize::i8Bit>(Dst.Z(), PRED_TMP_32B.Zeroing(), TMP1, 0);
-      break;
-    }
     default:
       LOGMAN_MSG_A_FMT("Unhandled {} size: {}", __func__, OpSize);
       break;
diff --git a/FEXCore/include/FEXCore/Core/CoreState.h b/FEXCore/include/FEXCore/Core/CoreState.h
index 42d8f1ebd..1eb250245 100644
--- a/FEXCore/include/FEXCore/Core/CoreState.h
+++ b/FEXCore/include/FEXCore/Core/CoreState.h
@@ -259,6 +259,9 @@ namespace FEXCore::Core {
       uint64_t L1Pointer{};
       uint64_t L2Pointer{};
       /**  @} */
+
+      // Copy of process-wide named vector constants data.
+      alignas(16) uint64_t NamedVectorConstants[FEXCore::IR::NamedVectorConstant::NAMED_VECTOR_CONST_POOL_MAX][2];
     } Common;
 
     union {
diff --git a/FEXCore/include/FEXCore/Utils/MathUtils.h b/FEXCore/include/FEXCore/Utils/MathUtils.h
index 76740ea5d..d86aa7fa6 100644
--- a/FEXCore/include/FEXCore/Utils/MathUtils.h
+++ b/FEXCore/include/FEXCore/Utils/MathUtils.h
@@ -25,4 +25,12 @@ requires(std::is_unsigned_v<T>)
   return std::countr_zero(Value);
 }
 
+// Divide a number by a power-of-2 by avoiding integer division.
+// Can be a faster implementation than regular integer divide.
+// Divisor requires to be power-of-2, is enforced in ilog2 helper.
+template<typename T, typename TT>
+requires(std::is_unsigned_v<T> && std::is_unsigned_v<TT>)
+[[nodiscard]] constexpr T DividePow2(T Dividend, TT Divisor) {
+  return Dividend >> ilog2(Divisor);
+}
 } // namespace FEXCore
diff --git a/FEXCore/unittests/APITests/ILog2.cpp b/FEXCore/unittests/APITests/ILog2.cpp
index e4fca68de..5555fd71e 100644
--- a/FEXCore/unittests/APITests/ILog2.cpp
+++ b/FEXCore/unittests/APITests/ILog2.cpp
@@ -5,3 +5,9 @@ TEST_CASE("ILog2") {
   auto i = GENERATE(range(0, 64));
   REQUIRE(FEXCore::ilog2(1ull << i) == i);
 }
+
+TEST_CASE("DividePow2") {
+  auto j = GENERATE(range(0, 64));
+  auto i = GENERATE(range(0, 64));
+  REQUIRE(FEXCore::DividePow2(1ull << j, 1ull << i) == ((1ull << j) / (1ull << i)));
+}
diff --git a/unittests/InstructionCountCI/H0F38.json b/unittests/InstructionCountCI/H0F38.json
index ab50cb0d4..ebf8529d9 100644
--- a/unittests/InstructionCountCI/H0F38.json
+++ b/unittests/InstructionCountCI/H0F38.json
@@ -674,14 +674,13 @@
       ]
     },
     "phminposuw xmm0, xmm1": {
-      "ExpectedInstructionCount": 7,
+      "ExpectedInstructionCount": 6,
       "Optimal": "Yes",
       "Comment": [
         "0x66 0x0f 0x38 0x41"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1544]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1888]",
         "zip1 v3.8h, v2.8h, v17.8h",
         "zip2 v2.8h, v2.8h, v17.8h",
         "umin v2.4s, v3.4s, v2.4s",
diff --git a/unittests/InstructionCountCI/H0F3A.json b/unittests/InstructionCountCI/H0F3A.json
index 9a784780b..9ec45c218 100644
--- a/unittests/InstructionCountCI/H0F3A.json
+++ b/unittests/InstructionCountCI/H0F3A.json
@@ -1550,14 +1550,13 @@
       ]
     },
     "aeskeygenassist xmm0, xmm1, 0": {
-      "ExpectedInstructionCount": 6,
+      "ExpectedInstructionCount": 5,
       "Optimal": "Yes",
       "Comment": [
         "0x66 0x0f 0x3a 0xdf"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1600]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #2000]",
         "movi v3.2d, #0x0",
         "mov v16.16b, v17.16b",
         "unimplemented (Unimplemented)",
@@ -1565,14 +1564,13 @@
       ]
     },
     "aeskeygenassist xmm0, xmm1, 0xFF": {
-      "ExpectedInstructionCount": 9,
+      "ExpectedInstructionCount": 8,
       "Optimal": "No",
       "Comment": [
         "0x66 0x0f 0x3a 0xdf"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1600]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #2000]",
         "movi v3.2d, #0x0",
         "mov v16.16b, v17.16b",
         "unimplemented (Unimplemented)",
diff --git a/unittests/InstructionCountCI/PrimaryGroup.json b/unittests/InstructionCountCI/PrimaryGroup.json
index 0bb417746..3753de59b 100644
--- a/unittests/InstructionCountCI/PrimaryGroup.json
+++ b/unittests/InstructionCountCI/PrimaryGroup.json
@@ -3795,7 +3795,7 @@
         "mov x0, x6",
         "mov x1, x20",
         "mov x2, x7",
-        "ldr x3, [x28, #1912]",
+        "ldr x3, [x28, #2048]",
         "str x30, [sp, #-16]!",
         "blr x3",
         "ldr x30, [sp], #16",
@@ -3806,7 +3806,7 @@
         "mov x0, x6",
         "mov x1, x20",
         "mov x2, x7",
-        "ldr x3, [x28, #1928]",
+        "ldr x3, [x28, #2064]",
         "str x30, [sp, #-16]!",
         "blr x3",
         "ldr x30, [sp], #16",
@@ -3870,7 +3870,7 @@
         "mov x0, x6",
         "mov x1, x20",
         "mov x2, x7",
-        "ldr x3, [x28, #1920]",
+        "ldr x3, [x28, #2056]",
         "str x30, [sp, #-16]!",
         "blr x3",
         "ldr x30, [sp], #16",
@@ -3883,7 +3883,7 @@
         "mov x0, x6",
         "mov x1, x20",
         "mov x2, x7",
-        "ldr x3, [x28, #1936]",
+        "ldr x3, [x28, #2072]",
         "str x30, [sp, #-16]!",
         "blr x3",
         "ldr x30, [sp], #16",
diff --git a/unittests/InstructionCountCI/Secondary.json b/unittests/InstructionCountCI/Secondary.json
index c839791d8..f201f7737 100644
--- a/unittests/InstructionCountCI/Secondary.json
+++ b/unittests/InstructionCountCI/Secondary.json
@@ -833,26 +833,24 @@
       ]
     },
     "movmskps eax, xmm0": {
-      "ExpectedInstructionCount": 6,
+      "ExpectedInstructionCount": 5,
       "Optimal": "Yes",
       "Comment": "0x0f 0x50",
       "ExpectedArm64ASM": [
         "ushr v2.4s, v16.4s, #31",
-        "ldr x0, [x28, #1592]",
-        "ldr q3, [x0]",
+        "ldr q3, [x28, #1984]",
         "ushl v2.4s, v2.4s, v3.4s",
         "addv s2, v2.4s",
         "mov w4, v2.s[0]"
       ]
     },
     "movmskps rax, xmm0": {
-      "ExpectedInstructionCount": 6,
+      "ExpectedInstructionCount": 5,
       "Optimal": "Yes",
       "Comment": "0x0f 0x50",
       "ExpectedArm64ASM": [
         "ushr v2.4s, v16.4s, #31",
-        "ldr x0, [x28, #1592]",
-        "ldr q3, [x0]",
+        "ldr q3, [x28, #1984]",
         "ushl v2.4s, v2.4s, v3.4s",
         "addv s2, v2.4s",
         "mov w4, v2.s[0]"
diff --git a/unittests/InstructionCountCI/Secondary_OpSize.json b/unittests/InstructionCountCI/Secondary_OpSize.json
index fb43aaf8b..bf53ba026 100644
--- a/unittests/InstructionCountCI/Secondary_OpSize.json
+++ b/unittests/InstructionCountCI/Secondary_OpSize.json
@@ -1130,12 +1130,11 @@
       ]
     },
     "addsubpd xmm0, xmm1": {
-      "ExpectedInstructionCount": 4,
+      "ExpectedInstructionCount": 3,
       "Optimal": "Yes",
       "Comment": "0x66 0x0f 0xd0",
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1576]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1952]",
         "eor v2.16b, v17.16b, v2.16b",
         "fadd v16.2d, v16.2d, v2.2d"
       ]
diff --git a/unittests/InstructionCountCI/Secondary_REPNE.json b/unittests/InstructionCountCI/Secondary_REPNE.json
index 1230a4622..16c435991 100644
--- a/unittests/InstructionCountCI/Secondary_REPNE.json
+++ b/unittests/InstructionCountCI/Secondary_REPNE.json
@@ -511,12 +511,11 @@
       ]
     },
     "addsubps xmm0, xmm1": {
-      "ExpectedInstructionCount": 4,
+      "ExpectedInstructionCount": 3,
       "Optimal": "Yes",
       "Comment": "0xf2 0x0f 0xd0",
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1560]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1920]",
         "eor v2.16b, v17.16b, v2.16b",
         "fadd v16.4s, v16.4s, v2.4s"
       ]
diff --git a/unittests/InstructionCountCI/VEX_map1.json b/unittests/InstructionCountCI/VEX_map1.json
index b6df17a2e..51f129f0b 100644
--- a/unittests/InstructionCountCI/VEX_map1.json
+++ b/unittests/InstructionCountCI/VEX_map1.json
@@ -4449,14 +4449,13 @@
       ]
     },
     "vaddsubpd xmm0, xmm1, xmm2": {
-      "ExpectedInstructionCount": 5,
+      "ExpectedInstructionCount": 4,
       "Optimal": "No",
       "Comment": [
         "Map 1 0b01 0xd0 128-bit"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1576]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1952]",
         "eor v2.16b, v18.16b, v2.16b",
         "fadd v2.2d, v17.2d, v2.2d",
         "mov v16.16b, v2.16b"
@@ -4476,14 +4475,13 @@
       ]
     },
     "vaddsubps xmm0, xmm1, xmm2": {
-      "ExpectedInstructionCount": 5,
+      "ExpectedInstructionCount": 4,
       "Optimal": "No",
       "Comment": [
         "Map 1 0b11 0xd0 128-bit"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1560]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1920]",
         "eor v2.16b, v18.16b, v2.16b",
         "fadd v2.4s, v17.4s, v2.4s",
         "mov v16.16b, v2.16b"
diff --git a/unittests/InstructionCountCI/VEX_map2.json b/unittests/InstructionCountCI/VEX_map2.json
index 202210415..cd5081c81 100644
--- a/unittests/InstructionCountCI/VEX_map2.json
+++ b/unittests/InstructionCountCI/VEX_map2.json
@@ -1579,14 +1579,13 @@
       ]
     },
     "vphminposuw xmm0, xmm1": {
-      "ExpectedInstructionCount": 8,
+      "ExpectedInstructionCount": 7,
       "Optimal": "No",
       "Comment": [
         "Map 2 0b01 0x41 256-bit"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1544]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #1888]",
         "zip1 v3.8h, v2.8h, v17.8h",
         "zip2 v2.8h, v2.8h, v17.8h",
         "umin v2.4s, v3.4s, v2.4s",
diff --git a/unittests/InstructionCountCI/VEX_map3.json b/unittests/InstructionCountCI/VEX_map3.json
index 65fa07bb3..d8a39ee2e 100644
--- a/unittests/InstructionCountCI/VEX_map3.json
+++ b/unittests/InstructionCountCI/VEX_map3.json
@@ -4713,14 +4713,13 @@
       ]
     },
     "vaeskeygenassist xmm0, xmm1, 0": {
-      "ExpectedInstructionCount": 7,
+      "ExpectedInstructionCount": 6,
       "Optimal": "No",
       "Comment": [
         "Map 3 0b01 0xdf 128-bit"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1600]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #2000]",
         "movi v3.2d, #0x0",
         "mov v2.16b, v17.16b",
         "unimplemented (Unimplemented)",
@@ -4729,14 +4728,13 @@
       ]
     },
     "vaeskeygenassist xmm0, xmm1, 0xFF": {
-      "ExpectedInstructionCount": 10,
+      "ExpectedInstructionCount": 9,
       "Optimal": "No",
       "Comment": [
         "Map 3 0b01 0xdf 128-bit"
       ],
       "ExpectedArm64ASM": [
-        "ldr x0, [x28, #1600]",
-        "ldr q2, [x0]",
+        "ldr q2, [x28, #2000]",
         "movi v3.2d, #0x0",
         "mov v2.16b, v17.16b",
         "unimplemented (Unimplemented)",