Add AvxIfma + AvxIfma.V512 hardware intrinsics by jamesburton · Pull Request #130079 · dotnet/runtime

jamesburton · 2026-07-01T12:46:56Z

Summary

Adds the managed AvxIfma surface for the AVX-IFMA (VEX) and AVX-512-IFMA (EVEX) ISAs, exposing VPMADD52LUQ / VPMADD52HUQ on Vector128<ulong> / Vector256<ulong> / Vector512<ulong>.

Closes #98833 (AVX-IFMA — VEX, V128/V256)
Closes #96476 (AVX-512-IFMA — EVEX, V512)

Both handled together per Tanner's direction on #98833:

"@jamesburton, feel free to pick this up. #96476 should be done simultaneously and should have the same fixup we've had approved and done for the other ISAs, where its a nested V512 class."

Shape

Mirrors the AvxVnni / AvxVnni.V512 pattern that landed in #128365, and the shape @saucecontrol pre-proposed for IFMA on 2024-11-06:

public abstract class AvxIfma : Avx2
{
    public static new bool IsSupported { get; }

    public static Vector128<ulong> MultiplyAdd52Low (Vector128<ulong> addend, Vector128<ulong> left, Vector128<ulong> right);
    public static Vector128<ulong> MultiplyAdd52High(Vector128<ulong> addend, Vector128<ulong> left, Vector128<ulong> right);
    public static Vector256<ulong> MultiplyAdd52Low (Vector256<ulong> addend, Vector256<ulong> left, Vector256<ulong> right);
    public static Vector256<ulong> MultiplyAdd52High(Vector256<ulong> addend, Vector256<ulong> left, Vector256<ulong> right);

    public new abstract class X64 : Avx2.X64 { public static new bool IsSupported { get; } }

    public abstract class V512
    {
        public static bool IsSupported { get; }
        public static Vector512<ulong> MultiplyAdd52Low (Vector512<ulong> addend, Vector512<ulong> left, Vector512<ulong> right);
        public static Vector512<ulong> MultiplyAdd52High(Vector512<ulong> addend, Vector512<ulong> left, Vector512<ulong> right);
    }
}

AvxIfma.V512 folds under the existing AVX512v2 JIT instruction set (same relationship as AvxVnni.V512 → AVX512v3 in #128365). The class-name dispatch in lookupInstructionSet chooses AVXIFMA when that CPUID bit is present, otherwise AVX512v2 — so on a hypothetical machine with AVX-512-IFMA but not VEX AVX-IFMA, AvxIfma.V512.IsSupported (and AvxIfma.IsSupported) still report true.

What was already wired

Significant pre-existing plumbing meant this PR is small:

INS_vpmadd52luq / INS_vpmadd52huq already in instrsxarch.h (FIRST_AVXIFMA_INSTRUCTION..LAST_AVXIFMA_INSTRUCTION)
emitxarch.cpp already gates VEX form on InstructionSet_AVXIFMA (line 353) and EVEX form on InstructionSet_AVX512v2 (line 430)
Is3OpRmwInstruction already returns true via the AVXIFMA range check
InstructionSet_AVXIFMA / AVXIFMA_X64 enums, R2R IDs (AvxIfma=66, Avx512Ifma=76), EnableAVXIFMA config, and codeman.cpp CPUID gate — all present

What this PR adds

JIT wire-up:

hwintrinsiclistxarch.h: new AVXIFMA range (MultiplyAdd52Low/MultiplyAdd52High at ULONG slot with HW_Flag_NoEvexSemantics) + AVX512v2 V512 entries
hwintrinsicxarch.cpp: V512VersionOfIsa maps AVXIFMA/AVX512v2 → AVX512v2; lookupInstructionSet dispatches "Ifma" class between AVXIFMA and AVX512v2 by compSupportsHWIntrinsic
hwintrinsiccodegenxarch.cpp: dispatch cases for the 4 new NIs alongside the existing VNNI/AVX512v3 group
lsraxarch.cpp: RMW handling for the 4 new NIs
lowerxarch.cpp: default-branch assert extended for the new NIs + AVXIFMA range

R2R + AOT metadata:

CorInfoInstructionSet.cs: nested-type lookup for AvxIfma (X64/V512/V512_X64) on X64 and X86 + LookupPlatformIntrinsicTypes for the AVX512v2 case + new InstructionSetInfo("avx512v2", "AvxIfma_V512", ...) entries
InstructionSetDesc.txt: adds AvxIfma_V512 alias to the Avx512Ifma row
InstructionSetHelpers.cs: adds avxifma_v512 to the optimistic set

Managed surface:

AvxIfma.cs + AvxIfma.PlatformNotSupported.cs (new files)
.Shared.projitems: includes both
System.Runtime.Intrinsics.cs (ref assembly): full API surface

Tests:

X86_Avx/AvxIfma/ — V128/V256 VEX form sample test
X86_Avx/AvxIfma_V512/ — V512 EVEX form sample test
NativeAOT smoke: uncomments AvxIfma / AvxIfma.X64 checks (were pre-approved stubs) and adds AvxIfma.V512 check

Validation

build.cmd clr.jit+clr.runtime+libs -c Release -arch x64 — 0 errors / 0 warnings
Local runtime verification via ifmaprobe / probe DLL swap deferred to CI: locally the packs pipeline is currently broken by a WiX installer failure, so the standard RestoreAdditionalProjectSources → Shipping approach doesn't refresh. Runtime verification will land via the CI test legs.

Compatibility note

The R2R ID Avx512Ifma = 76 retains its numeric value; existing R2R images continue to work. The InstructionSetDesc.txt row now records AvxIfma_V512 as the nested-type alias, matching the pattern PR #128365 used for Avx512Vnni → AvxVnni_V512.

🤖 Generated with Claude Code

Adds the managed AvxIfma class (V128/V256 VEX form) and nested AvxIfma.V512 (EVEX form) for VPMADD52LUQ / VPMADD52HUQ. Structural changes only (no tests yet, no build validation yet): - hwintrinsiclistxarch.h: new AVXIFMA range + AVX512v2 IFMA entries - hwintrinsicxarch.cpp: V512VersionOfIsa AVXIFMA/AVX512v2 case, lookupInstructionSet Ifma dispatch - hwintrinsiccodegenxarch.cpp: dispatch cases for MultiplyAdd52Low/High - lsraxarch.cpp: RMW cases - lowerxarch.cpp: default-branch assert extension - CorInfoInstructionSet.cs: nested type lookup + LookupPlatformIntrinsicTypes - InstructionSetDesc.txt: nested type alias AvxIfma_V512 - InstructionSetHelpers.cs: optimistic support - projitems: include new files - AvxIfma.cs + AvxIfma.PlatformNotSupported.cs: managed surface - System.Runtime.Intrinsics.cs: ref assembly Closes dotnet#98833, closes dotnet#96476.

Adds sample tests mirroring the AvxVnni.V512 test shape: - X86_Avx/AvxIfma/ — V128/V256 VEX form - X86_Avx/AvxIfma_V512/ — V512 EVEX form NativeAOT smoke test: uncomments AvxIfma / AvxIfma.X64 checks (were pre-approved but not yet wired) and adds AvxIfma.V512 check. AvxIfma.V512 folds under AVX512v2 (Avx512Vbmi is a representative sibling).

dotnet-policy-service · 2026-07-01T12:48:33Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Copilot

Pull request overview

This PR adds the managed System.Runtime.Intrinsics.X86.AvxIfma surface (VEX AVX-IFMA for Vector128/256<ulong>) plus a nested AvxIfma.V512 surface (EVEX AVX-512-IFMA for Vector512<ulong>), and wires the corresponding JIT/AOT instruction-set plumbing and sample tests.

Changes:

Add AvxIfma / AvxIfma.V512 public API surface (ref + CoreLib implementation + PNSE stubs) and include it in CoreLib build items.
Extend JIT HWIntrinsic tables and codegen/LSRA/lowering handling for VPMADD52LUQ/HUQ (VEX + EVEX forms).
Add JIT and NativeAOT smoke tests covering the new intrinsics.

Reviewed changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/X86/AvxIfma.cs	Adds managed intrinsic declarations for AVX-IFMA + nested V512 surface.
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/X86/AvxIfma.PlatformNotSupported.cs	Adds PNSE stubs for unsupported platforms/hardware.
src/libraries/System.Private.CoreLib/src/System.Private.CoreLib.Shared.projitems	Wires new CoreLib source files into the build.
src/libraries/System.Runtime.Intrinsics/ref/System.Runtime.Intrinsics.cs	Adds public ref-assembly API surface for `AvxIfma` / `AvxIfma.V512`.
src/coreclr/jit/hwintrinsicxarch.cpp	Updates ISA lookup / V512 ISA mapping for the new intrinsics.
src/coreclr/jit/hwintrinsiclistxarch.h	Adds AVXIFMA and AVX512v2 intrinsic table entries for `MultiplyAdd52*`.
src/coreclr/jit/hwintrinsiccodegenxarch.cpp	Adds codegen dispatch coverage for the new intrinsic IDs.
src/coreclr/jit/lsraxarch.cpp	Adds LSRA handling for the new RMW-style 3-operand intrinsics.
src/coreclr/jit/lowerxarch.cpp	Extends containment asserts/logic to include the new intrinsic IDs.
src/coreclr/tools/Common/JitInterface/CorInfoInstructionSet.cs	Adds type/ISA mapping entries for nested `AvxIfma` forms and AVX512v2 aliasing.
src/coreclr/tools/Common/JitInterface/ThunkGenerator/InstructionSetDesc.txt	Adds `AvxIfma_V512` alias mapping for the Avx512Ifma row.
src/coreclr/tools/Common/InstructionSetHelpers.cs	Adds `avxifma_v512` to the optimistic instruction-set set.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma/Program.AvxIfma.cs	Adds test program stub/namespace for AVXIFMA tests.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma/AvxIfmaSampleTest.cs	Adds V128/V256 semantic sample tests for `MultiplyAdd52*`.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma/*.csproj	Adds project files to run the AVXIFMA JIT tests in r/ro variants.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma_V512/Program.AvxIfma_V512.cs	Adds test program stub/namespace for AVXIFMA V512 tests.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma_V512/AvxIfma_V512SampleTest.cs	Adds V512 semantic sample tests for `MultiplyAdd52*`.
src/tests/JIT/HardwareIntrinsics/X86_Avx/AvxIfma_V512/*.csproj	Adds project files to run the AVXIFMA V512 JIT tests in r/ro variants.
src/tests/nativeaot/SmokeTests/HardwareIntrinsics/Program.cs	Enables NativeAOT smoke checks for `AvxIfma` and adds `AvxIfma.V512`.

jamesburton · 2026-07-01T14:27:55Z

+HARDWARE_INTRINSIC(AVX512v2,        MultiplyAdd52Low,                                                64,              3,     {INS_invalid,           INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_vpmadd52luq,        INS_invalid,            INS_invalid},            5,        -1,         HW_Category_SimpleSIMD,             HW_Flag_NoFlag)
+HARDWARE_INTRINSIC(AVX512v2,        MultiplyAdd52High,                                               64,              3,     {INS_invalid,           INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_vpmadd52huq,        INS_invalid,            INS_invalid},            5,        -1,         HW_Category_SimpleSIMD,             HW_Flag_NoFlag)


Good catch — but the fix was actually the ordering of the HARDWARE_INTRINSIC entries, not the simdSize itself. binarySearchId uses strcmp (case-sensitive ASCII) to locate intrinsics within an ISA range, so entries must be alphabetically sorted. My initial ordering had MultiplyAdd52Low before MultiplyAdd52High, but 'H' (0x48) < 'L' (0x4C), so the search couldn't find High and ILC failed with "Code generation failed" on the NativeAOT builds.

Fixed in add8927e21d: swapped so High precedes Low in both the AVXIFMA range and the AVX512v2 range, and moved the pair to the correct alphabetical position between MultiShift and PermuteVar16x8.

On the fixed-simdSize=64 concern for AVX512v2 specifically: this mirrors the AvxVnni.V512 pattern in #128365 exactly (NI_AVX512v3_MultiplyWideningAndAdd is also simdSize=64). The class-name dispatch in lookupInstructionSet chooses AVXIFMA vs AVX512v2 based on compSupportsHWIntrinsic — for a call site using V128/V256 args, the class-name lookup returns AVXIFMA (V128/V256, simdSize=-1) whenever the AVXIFMA CPUID bit is available (which is true across all AOT targets in optimisticInstructionSetSupport).

Update: the real root cause is different — the simdSize=64 mismatch is a theoretical issue but wasn't what fired here. The actual failure was a KeyNotFoundException in HardwareIntrinsicILProvider.GetMethodIL.

InstructionSetDesc.txt had AvxIfma in the R2R-name column (col 4) instead of the managed-name column (col 3), so the generated ArchitectureToValidInstructionSets emitted the avxifma entry with an empty ManagedName. HardwareIntrinsicILProvider skips empty ManagedNames when building its class-name→InstructionSet dictionary — so ILC's _instructionSetMap["AvxIfma"] threw:

System.Collections.Generic.KeyNotFoundException: The given key 'AvxIfma' was not present in the dictionary. at ILCompiler.HardwareIntrinsicILProvider.GetMethodIL(...) HardwareIntrinsicILProvider.cs:line 44

Fixed in ff40b908575 — moved AvxIfma from col 4 to col 3 in InstructionSetDesc.txt and updated the already-generated CorInfoInstructionSet.cs accordingly.

jamesburton · 2026-07-01T14:28:02Z

+    public abstract class AvxIfma : System.Runtime.Intrinsics.X86.Avx2
+    {
+        internal AvxIfma() { }
+        public static new bool IsSupported { get { throw null; } }
+        public static System.Runtime.Intrinsics.Vector128<ulong> MultiplyAdd52Low(System.Runtime.Intrinsics.Vector128<ulong> addend, System.Runtime.Intrinsics.Vector128<ulong> left, System.Runtime.Intrinsics.Vector128<ulong> right) { throw null; }
+        public static System.Runtime.Intrinsics.Vector128<ulong> MultiplyAdd52High(System.Runtime.Intrinsics.Vector128<ulong> addend, System.Runtime.Intrinsics.Vector128<ulong> left, System.Runtime.Intrinsics.Vector128<ulong> right) { throw null; }
+        public static System.Runtime.Intrinsics.Vector256<ulong> MultiplyAdd52Low(System.Runtime.Intrinsics.Vector256<ulong> addend, System.Runtime.Intrinsics.Vector256<ulong> left, System.Runtime.Intrinsics.Vector256<ulong> right) { throw null; }
+        public static System.Runtime.Intrinsics.Vector256<ulong> MultiplyAdd52High(System.Runtime.Intrinsics.Vector256<ulong> addend, System.Runtime.Intrinsics.Vector256<ulong> left, System.Runtime.Intrinsics.Vector256<ulong> right) { throw null; }
+        public new abstract partial class X64 : System.Runtime.Intrinsics.X86.Avx2.X64
+        {
+            internal X64() { }
+            public static new bool IsSupported { get { throw null; } }
+        }
+        public abstract partial class V512
+        {
+            internal V512() { }
+            public static bool IsSupported { get { throw null; } }
+            public static System.Runtime.Intrinsics.Vector512<ulong> MultiplyAdd52Low(System.Runtime.Intrinsics.Vector512<ulong> addend, System.Runtime.Intrinsics.Vector512<ulong> left, System.Runtime.Intrinsics.Vector512<ulong> right) { throw null; }
+            public static System.Runtime.Intrinsics.Vector512<ulong> MultiplyAdd52High(System.Runtime.Intrinsics.Vector512<ulong> addend, System.Runtime.Intrinsics.Vector512<ulong> left, System.Runtime.Intrinsics.Vector512<ulong> right) { throw null; }
+        }


Tanner explicitly authorised the AvxIfma.V512 nested-class shape (as opposed to a separate Avx512Ifma type) in his 2026-06-30 reply on #98833:

"@jamesburton, feel free to pick this up. #96476 should be done simultaneously and should have the same fixup we've had approved and done for the other ISAs, where its a nested V512 class."

The nested-V512 pattern mirrors the AvxVnni.V512 restructure that landed in #128365 (which itself replaced an original Avx512Vnni proposal on the same grounds). @saucecontrol also pre-proposed this exact shape on #96476 on 2024-11-06. So the shape carries prior consensus even though the individual comment approving the original Avx512Ifma still exists in the issue history.

…tical binarySearchId uses strcmp (case-sensitive ASCII order) to find intrinsics within an ISA range. My initial ordering put MultiplyAdd52Low before MultiplyAdd52High, but 'H' (0x48) < 'L' (0x4C), so the binary search would fail on High. Same issue for both AVXIFMA and AVX512v2 ranges. Also reorders MultiplyAdd52*/PermuteVar* so Multiply* precedes PermuteVar* (M < P). This surfaced as ILC 'Code generation failed' on the AvxIfma_handwritten_r NativeAOT tests.

The LookupPlatformIntrinsicTypes switch was missing cases for InstructionSet.X64_AVXIFMA/AVXIFMA_X64 and X86_AVXIFMA. When the AOT compiler tried to resolve AvxIfma against an AVXIFMA-only target (e.g. NativeAOT smoke tests X64Baseline / X64Avx2), the type wasn't found and ILC failed with 'Code generation failed for AvxIfma.get_IsSupported()'. Mirrors the existing X64_AVXVNNI/X86_AVXVNNI cases.

Copilot

Pull request overview

Copilot reviewed 25 out of 25 changed files in this pull request and generated 1 comment.

+HARDWARE_INTRINSIC(AVX512v2,        MultiplyAdd52High,                                               64,              3,     {INS_invalid,           INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_vpmadd52huq,        INS_invalid,            INS_invalid},            5,        -1,         HW_Category_SimpleSIMD,             HW_Flag_NoFlag)
+HARDWARE_INTRINSIC(AVX512v2,        MultiplyAdd52Low,                                                64,              3,     {INS_invalid,           INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_invalid,            INS_vpmadd52luq,        INS_invalid,            INS_invalid},            5,        -1,         HW_Category_SimpleSIMD,             HW_Flag_NoFlag)


InstructionSetDesc.txt had AvxIfma in the R2R-name column (col 4) instead of the managed-name column (col 3), so the generated ArchitectureToValidInstructionSets emitted 'avxifma' with an empty ManagedName. HardwareIntrinsicILProvider builds its class-name -> InstructionSet dictionary from ArchitectureToValidInstructionSets, skipping entries with empty ManagedName, so lookups for 'AvxIfma' threw KeyNotFoundException at ILC time. Fixed the .txt source-of-truth (so any regeneration is consistent) and the already-generated CorInfoInstructionSet.cs to set ManagedName='AvxIfma'.

jamesburton added 2 commits July 1, 2026 13:27

jamesburton requested a review from MichalStrehovsky as a code owner July 1, 2026 12:46

Copilot AI review requested due to automatic review settings July 1, 2026 12:46

github-actions Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Jul 1, 2026

Copilot started reviewing on behalf of jamesburton July 1, 2026 12:47 View session

dotnet-policy-service Bot added the community-contribution Indicates that the PR has been added by a community member label Jul 1, 2026

Copilot AI reviewed Jul 1, 2026

View reviewed changes

jamesburton added 2 commits July 1, 2026 15:27

Copilot AI review requested due to automatic review settings July 1, 2026 15:38

Copilot started reviewing on behalf of jamesburton July 1, 2026 15:39 View session

Copilot AI reviewed Jul 1, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add AvxIfma + AvxIfma.V512 hardware intrinsics#130079

Add AvxIfma + AvxIfma.V512 hardware intrinsics#130079
jamesburton wants to merge 5 commits into
dotnet:mainfrom
jamesburton:feature/avxifma

jamesburton commented Jul 1, 2026

Uh oh!

dotnet-policy-service Bot commented Jul 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

jamesburton Jul 1, 2026

Uh oh!

jamesburton Jul 1, 2026

Uh oh!

jamesburton Jul 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		HARDWARE_INTRINSIC(AVX512v2, MultiplyAdd52Low, 64, 3, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vpmadd52luq, INS_invalid, INS_invalid}, 5, -1, HW_Category_SimpleSIMD, HW_Flag_NoFlag)
		HARDWARE_INTRINSIC(AVX512v2, MultiplyAdd52High, 64, 3, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vpmadd52huq, INS_invalid, INS_invalid}, 5, -1, HW_Category_SimpleSIMD, HW_Flag_NoFlag)

Uh oh!

Conversation

jamesburton commented Jul 1, 2026

Summary

Shape

What was already wired

What this PR adds

Validation

Compatibility note

Uh oh!

dotnet-policy-service Bot commented Jul 1, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

jamesburton Jul 1, 2026

Choose a reason for hiding this comment

Uh oh!

jamesburton Jul 1, 2026

Choose a reason for hiding this comment

Uh oh!

jamesburton Jul 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants