[llvm] [AMDGPU]: Fix type signatures for wmma intrinsics, NFC (PR #80087)
Changpeng Fang via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 30 16:56:02 PST 2024
https://github.com/changpeng created https://github.com/llvm/llvm-project/pull/80087
Make the wmma intrinsic type signatures to be canonical. We need
a type signature as long as the type is not fixed. However, when an argument's type matches a previous argument's type, we do not need the signature for this argument.
This patch fixes three general cases:
1. add missing signatures
2. remove signatures for matching arguments
3. reorer the signatures -- return type signature should always appear first
>From cc62e973209d9259b46d868e26bbbb8194ae8bf0 Mon Sep 17 00:00:00 2001
From: Changpeng Fang <changpeng.fang at amd.com>
Date: Tue, 30 Jan 2024 16:35:25 -0800
Subject: [PATCH] [AMDGPU]: Fix type signatures for wmma intrinsics, NFC
Make the wmma intrinsic type signatures to be canonical. We need
a type signature as long as the type is not fixed. However, when an
argument's type matches a previous argument's type, we do not need the
signature for this argument.
This patch fixes three general cases:
1. add missing signatures
2. remove signatures for matching arguments
3. reorer the signatures -- return type signature should always
appear first
---
.../UniformityAnalysis/AMDGPU/intrinsics.ll | 36 +++----
.../CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll | 76 ++++++-------
...wmma-gfx12-w32-f16-f32-matrix-modifiers.ll | 58 +++++-----
.../test/CodeGen/AMDGPU/wmma-gfx12-w32-imm.ll | 77 ++++++-------
.../AMDGPU/wmma-gfx12-w32-iu-modifiers.ll | 48 ++++-----
.../AMDGPU/wmma-gfx12-w32-swmmac-index_key.ll | 60 +++++------
llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32.ll | 88 +++++++--------
...wmma-gfx12-w64-f16-f32-matrix-modifiers.ll | 58 +++++-----
.../test/CodeGen/AMDGPU/wmma-gfx12-w64-imm.ll | 66 ++++++------
.../AMDGPU/wmma-gfx12-w64-iu-modifiers.ll | 53 +++++----
.../AMDGPU/wmma-gfx12-w64-swmmac-index_key.ll | 102 +++++++++---------
llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64.ll | 88 +++++++--------
llvm/test/CodeGen/AMDGPU/wmma_modifiers.ll | 4 +-
llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll | 100 ++++++++---------
llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll | 100 ++++++++---------
15 files changed, 499 insertions(+), 515 deletions(-)
diff --git a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
index 4b6f6159ff23f..333ec6dcaf488 100644
--- a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
+++ b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
@@ -112,7 +112,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
define amdgpu_kernel void @swmmac_f32_16x16x32_f16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
+ %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -120,7 +120,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
define amdgpu_kernel void @swmmac_f32_16x16x32_bf16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
+ %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -128,7 +128,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
define amdgpu_kernel void @swmmac_f16_16x16x32_f16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
+ %tmp0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
store <8 x half> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -136,7 +136,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
define amdgpu_kernel void @swmmac_bf16_16x16x32_bf16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
+ %tmp0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
store <8 x i16> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -144,7 +144,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
define amdgpu_kernel void @swmmac_i32_16x16x32_iu8(<2 x i32> %A, <4 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
+ %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -152,7 +152,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 false, i32 %A, i1 false, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
define amdgpu_kernel void @swmmac_i32_16x16x32_iu4(i32 %A, <2 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4(i1 false, i32 %A, i1 false, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
+ %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 false, i32 %A, i1 false, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -160,7 +160,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
define amdgpu_kernel void @swmmac_i32_16x16x64_iu4(<2 x i32> %A, <4 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
+ %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -168,7 +168,7 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
define amdgpu_kernel void @swmmac_f32_16x16x32_fp8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
@@ -176,13 +176,13 @@ bb:
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
define amdgpu_kernel void @swmmac_f32_16x16x32_fp8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
- %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
ret void
}
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
-define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
+define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
%tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
@@ -190,7 +190,7 @@ bb:
}
; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
-define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
+define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
bb:
%tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
@@ -284,13 +284,13 @@ declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half>
declare <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16>, <16 x i16> , <16 x i16>, i1 immarg) #1
declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 immarg, <4 x i32>, i1 immarg, <4 x i32> , <8 x i32>, i1 immarg) #1
declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32> , <8 x i32>, i1 immarg) #1
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16(<8 x half>, <16 x half>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16(<8 x i16>, <16 x i16>, <8 x float>, i16)
-declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16(<8 x half>, <16 x half>, <8 x half>, i16)
-declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16, i1)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
+declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
+declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16, i1)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8(<2 x i32>, <4 x i32>, <8 x float>, i16)
declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
index 60db306609139..368d637f8ac1b 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
@@ -1,14 +1,14 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs < %s | FileCheck %s --check-prefix=W64
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half>, <16 x half>, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16>, <16 x i16>, <4 x float>)
-declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
-declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
-declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
-declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 immarg, <4 x i32>, i1 immarg, <4 x i32>, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half>, <16 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v16i16(<16 x i16>, <16 x i16>, <4 x float>)
+declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.b8f16.v16f16(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
+declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied.v8f16.v16f16(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
+declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
+declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied.v8i16.v16i16(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 immarg, <4 x i32>, i1 immarg, <4 x i32>, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <4 x i32>, i1 immarg)
; @llvm.amdgcn.wmma.f32.16x16x16.f16
@@ -21,7 +21,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<16 x half> %A, <16 x half> %B
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> %A, <16 x half> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half> %A, <16 x half> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -37,7 +37,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<16 x i16> %A, <16 x i16> %B,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v16i16(<16 x i16> %A, <16 x i16> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -53,7 +53,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<16 x half> %A, <16 x half>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 0)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 0)
store <8 x half> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -67,7 +67,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<16 x half> %A, <16 x half>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 1)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 1)
store <8 x half> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -83,8 +83,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_untied(<16 x half> %A.0, <16 x
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res.0 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A.0, <16 x half> %B.0, <8 x half> %C, i1 0)
- %res.1 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A.1, <16 x half> %B.1, <8 x half> %C, i1 0)
+ %res.0 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A.0, <16 x half> %B.0, <8 x half> %C, i1 0)
+ %res.1 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A.1, <16 x half> %B.1, <8 x half> %C, i1 0)
store <8 x half> %res.0, ptr addrspace(1) %out.0, align 32
store <8 x half> %res.1, ptr addrspace(1) %out.1, align 32
ret void
@@ -106,8 +106,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_tied(<16 x half> %A.0, <16 x h
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res.0 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied(<16 x half> %A.0, <16 x half> %B.0, <8 x half> %C, i1 0)
- %res.1 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied(<16 x half> %A.1, <16 x half> %B.1, <8 x half> %C, i1 0)
+ %res.0 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied.v8f16.v16f16(<16 x half> %A.0, <16 x half> %B.0, <8 x half> %C, i1 0)
+ %res.1 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.tied.v8f16.v16f16(<16 x half> %A.1, <16 x half> %B.1, <8 x half> %C, i1 0)
store <8 x half> %res.0, ptr addrspace(1) %out.0, align 32
store <8 x half> %res.1, ptr addrspace(1) %out.1, align 32
ret void
@@ -124,7 +124,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<16 x i16> %A, <16 x i16>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 0)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 0)
store <8 x i16> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -138,7 +138,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<16 x i16> %A, <16 x i16>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 1)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 1)
store <8 x i16> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -154,8 +154,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_untied(<16 x i16> %A.0, <16
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res.0 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A.0, <16 x i16> %B.0, <8 x i16> %C, i1 0)
- %res.1 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A.1, <16 x i16> %B.1, <8 x i16> %C, i1 0)
+ %res.0 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A.0, <16 x i16> %B.0, <8 x i16> %C, i1 0)
+ %res.1 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A.1, <16 x i16> %B.1, <8 x i16> %C, i1 0)
store <8 x i16> %res.0, ptr addrspace(1) %out.0, align 32
store <8 x i16> %res.1, ptr addrspace(1) %out.1, align 32
ret void
@@ -177,8 +177,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_tied(<16 x i16> %A.0, <16 x
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res.0 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied(<16 x i16> %A.0, <16 x i16> %B.0, <8 x i16> %C, i1 0)
- %res.1 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied(<16 x i16> %A.1, <16 x i16> %B.1, <8 x i16> %C, i1 0)
+ %res.0 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied.v8i16.v16i16(<16 x i16> %A.0, <16 x i16> %B.0, <8 x i16> %C, i1 0)
+ %res.1 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.tied.v8i16.v16i16(<16 x i16> %A.1, <16 x i16> %B.1, <8 x i16> %C, i1 0)
store <8 x i16> %res.0, ptr addrspace(1) %out.0, align 32
store <8 x i16> %res.1, ptr addrspace(1) %out.1, align 32
ret void
@@ -195,7 +195,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -210,7 +210,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -224,7 +224,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -238,7 +238,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -252,7 +252,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -266,7 +266,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -280,7 +280,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -294,7 +294,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -310,7 +310,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -324,7 +324,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -338,7 +338,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -352,7 +352,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -366,7 +366,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -380,7 +380,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -394,7 +394,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
@@ -408,7 +408,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
ret void
}
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-f16-f32-matrix-modifiers.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-f16-f32-matrix-modifiers.ll
index fb2af36839b5d..bbb97902905c7 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-f16-f32-matrix-modifiers.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-f16-f32-matrix-modifiers.ll
@@ -13,7 +13,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negA(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <8 x half> %A
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %fneg.A, <8 x half> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %fneg.A, <8 x half> %B, <8 x float> %C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -30,7 +30,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negB(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <8 x half> %B
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %A, <8 x half> %fneg.B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %fneg.B, <8 x float> %C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -47,7 +47,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negC(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -64,7 +64,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_absC(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -81,7 +81,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_negC(<8 x i16> %A, <8 x i16>
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8i16.v8f32(<8 x i16> %A, <8 x i16> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -98,7 +98,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_absC(<8 x i16> %A, <8 x i16>
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8i16.v8f32(<8 x i16> %A, <8 x i16> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -175,7 +175,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_negC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -192,7 +192,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_absC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -209,7 +209,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_negC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -226,7 +226,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_absC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -243,7 +243,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_negC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -260,7 +260,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_absC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -277,7 +277,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_negC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <8 x float> %C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fneg.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -294,7 +294,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_absC(<2 x i32> %A, <2 x i3
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -311,7 +311,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_negA(<8 x half> %A, <16 x ha
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <8 x half> %A
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f16.v16f16.v8f32.i16(<8 x half> %fneg.A, <16 x half> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %fneg.A, <16 x half> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -328,7 +328,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_negB(<8 x half> %A, <16 x ha
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <16 x half> %B
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f16.v16f16.v8f32.i16(<8 x half> %A, <16 x half> %fneg.B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %fneg.B, <8 x float> %C, i16 %Index)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -343,7 +343,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_negA(<8 x half> %A, <16 x ha
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <8 x half> %A
- %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v16f16.v8f16.i16(<8 x half> %fneg.A, <16 x half> %B, <8 x half> %C, i16 %Index)
+ %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %fneg.A, <16 x half> %B, <8 x half> %C, i16 %Index)
store <8 x half> %res, <8 x half> addrspace(1)* %out
ret void
}
@@ -358,7 +358,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_negB(<8 x half> %A, <16 x ha
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <16 x half> %B
- %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v16f16.v8f16.i16(<8 x half> %A, <16 x half> %fneg.B, <8 x half> %C, i16 %Index)
+ %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %fneg.B, <8 x half> %C, i16 %Index)
store <8 x half> %res, <8 x half> addrspace(1)* %out
ret void
}
@@ -378,7 +378,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negabsC(<8 x half> %A, <8 x ha
bb:
%fabs.C = call <8 x float> @llvm.fabs.v8f32(<8 x float> %C)
%fneg.fabs.C = fneg <8 x float> %fabs.C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -416,7 +416,7 @@ bb:
%el3.fabs = call float @llvm.fabs.f32(float %el3)
%partial.fabs.C = insertelement <8 x float> %C, float %el3.fabs, i32 3
%fneg.partial.fabs.C = fneg <8 x float> %partial.fabs.C
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.partial.fabs.C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> %fneg.partial.fabs.C)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -435,7 +435,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negA_constantC(<8 x half> %A,
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <8 x half> %A
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half> %fneg.A, <8 x half> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %fneg.A, <8 x half> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, <8 x float> addrspace(1)* %out
ret void
}
@@ -488,12 +488,12 @@ declare <8 x half> @llvm.fabs.v8f16(<8 x half>)
declare <8 x float> @llvm.fabs.v8f32(<8 x float>)
declare float @llvm.fabs.f32(float)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f16.v8f32(<8 x half>, <8 x half>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8i16.v8f32(<8 x i16>, <8 x i16>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half>, <8 x half>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16>, <8 x i16>, <8 x float>)
declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half>, <8 x half>, <8 x half>, i1 immarg)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f16.v16f16.v8f32.i16(<8 x half>, <16 x half>, <8 x float>, i16)
-declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v16f16.v8f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
+declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-imm.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-imm.ll
index 51f93b57f38e9..c80e5e0e3506c 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-imm.ll
@@ -12,7 +12,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -34,7 +34,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm_non_inlineable(<8 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -50,7 +50,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm(<8 x i16> %A, <8 x i16> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16.v8i16.v8f32(<8 x i16> %A, <8 x i16> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -72,7 +72,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm_non_inlineable(<8 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16.v8i16.v8f32(<8 x i16> %A, <8 x i16> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -86,7 +86,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm(<8 x half> %A, <8 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> <half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0>, i1 0)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> <half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0, half 1.0>, i1 0)
store <8 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -104,7 +104,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm_non_inlineable(<8 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> <half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0>, i1 0)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> <half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0, half 3.0>, i1 0)
store <8 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -118,7 +118,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm(<8 x i16> %A, <8 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> <i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256>, i1 0)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> <i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256, i16 16256>, i1 0)
store <8 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -136,7 +136,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm_non_inlineable(<8 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> <i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320>, i1 0)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> <i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320, i16 16320>, i1 0)
store <8 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -152,7 +152,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -174,7 +174,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm_non_inlineable(<2 x i32> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -190,7 +190,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm(i32 %A, i32 %B, ptr addrsp
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -212,7 +212,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm_non_inlineable(i32 %A, i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -228,7 +228,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm(<2 x i32> %A, <2 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -250,7 +250,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable(<2 x i3
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -266,7 +266,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm(<2 x i32> %A, <2 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -288,7 +288,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable(<2 x i3
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -304,7 +304,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm(<2 x i32> %A, <2 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -326,7 +326,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable(<2 x i3
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -342,7 +342,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm(<2 x i32> %A, <2 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -364,7 +364,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable(<2 x i3
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> <float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0, float 3.0>)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -380,7 +380,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -402,30 +402,19 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm_non_inlineable(<2 x i32> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> <i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16.v8f16.v8f32(<8 x half>, <8 x half>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16.v8i16.v8f32(<8 x i16>, <8 x i16>, <8 x float>)
-declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16.v8f16.v8f16(<8 x half>, <8 x half>, <8 x half>, i1 immarg)
-declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16.v8i16.v8i16(<8 x i16>, <8 x i16>, <8 x i16>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half>, <16 x half>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
-declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
-declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i32 %Index, i1 immarg)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half>, <8 x half>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16>, <8 x i16>, <8 x float>)
+declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half>, <8 x half>, <8 x half>, i1 immarg)
+declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16(<8 x i16>, <8 x i16>, <8 x i16>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-iu-modifiers.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-iu-modifiers.ll
index 48297932aa598..5426458e6b1df 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-iu-modifiers.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-iu-modifiers.ll
@@ -12,7 +12,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_zext_src0(<2 x i32> %A, <2 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -28,7 +28,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_zext_src1(<2 x i32> %A, <2 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -44,7 +44,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_clamp(<2 x i32> %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -62,7 +62,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_zext_src0(i32 %A, i32 %B, <8 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 1, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 1, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -78,7 +78,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_zext_src1(i32 %A, i32 %B, <8 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 0, i32 %A, i1 1, i32 %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 0, i32 %A, i1 1, i32 %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -94,7 +94,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_clamp(i32 %A, i32 %B, <8 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -112,7 +112,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_zext_src0(<2 x i32> %A, <2 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -128,7 +128,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_zext_src1(<2 x i32> %A, <2 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -144,7 +144,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_clamp(<2 x i32> %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -164,7 +164,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_zext_src0(<2 x i32> %A, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 1, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 1, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -180,7 +180,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_zext_src1(<2 x i32> %A, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 0, <2 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 0, <2 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -196,7 +196,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_clamp(<2 x i32> %A, <4 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -214,7 +214,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_zext_src0(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 1, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 1, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -230,7 +230,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_zext_src1(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 0, i32 %A, i1 1, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 0, i32 %A, i1 1, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -246,7 +246,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_clamp(i32 %A, <2 x i32> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -264,7 +264,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_zext_src0(<2 x i32> %A, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 1, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 1, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -280,7 +280,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_zext_src1(<2 x i32> %A, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 0, <2 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 0, <2 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -296,14 +296,14 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_clamp(<2 x i32> %A, <4 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i32 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i32 %Index, i1 immarg)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-swmmac-index_key.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-swmmac-index_key.ll
index 43538768f00fd..b0213abba9048 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-swmmac-index_key.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32-swmmac-index_key.ll
@@ -25,10 +25,10 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_index_key(<8 x half> %A, <16
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
@@ -57,10 +57,10 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf16_index_key(<8 x i16> %A, <16
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
@@ -83,10 +83,10 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_index_key(<8 x half> %A, <16
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index0)
+ %res0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index0)
store <8 x half> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index1)
+ %res1 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index1)
store <8 x half> %res1, ptr addrspace(1) %out1
ret void
}
@@ -109,10 +109,10 @@ define amdgpu_ps void @test_swmmac_bf16_16x16x32_bf16_index_key(<8 x i16> %A, <1
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index0)
+ %res0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index0)
store <8 x i16> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index1)
+ %res1 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index1)
store <8 x i16> %res1, ptr addrspace(1) %out1
ret void
}
@@ -141,10 +141,10 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_index_key(<2 x i32> %A, <4 x
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index0, i1 0)
+ %res0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index0, i1 0)
store <8 x i32> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index1, i1 0)
+ %res1 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index1, i1 0)
store <8 x i32> %res1, ptr addrspace(1) %out1
ret void
}
@@ -173,10 +173,10 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_index_key(i32 %A, <2 x i32>
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index0, i1 0)
+ %res0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index0, i1 0)
store <8 x i32> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index1, i1 0)
+ %res1 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index1, i1 0)
store <8 x i32> %res1, ptr addrspace(1) %out1
ret void
}
@@ -205,10 +205,10 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_fp8_index_key(<2 x i32> %A,
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
@@ -237,10 +237,10 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_bf8_index_key(<2 x i32> %A,
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
@@ -269,10 +269,10 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_fp8_index_key(<2 x i32> %A,
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
@@ -301,21 +301,21 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_bf8_index_key(<2 x i32> %A,
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
+ %res0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index0)
store <8 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
+ %res1 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index1)
store <8 x float> %res1, ptr addrspace(1) %out1
ret void
}
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half>, <16 x half>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
-declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
-declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
+declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
+declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32.ll
index 2db3b07de54d0..a66747567dd3d 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w32.ll
@@ -12,7 +12,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x half> %A, <8 x half> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16.v8f16.v8f32(<8 x half> %A, <8 x half> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half> %A, <8 x half> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -28,7 +28,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i16> %A, <8 x i16> %B, <
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16.v8i16.v8f32(<8 x i16> %A, <8 x i16> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -42,7 +42,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16(<8 x half> %A, <8 x half> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> %C, i1 0)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half> %A, <8 x half> %B, <8 x half> %C, i1 0)
store <8 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -56,7 +56,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16(<8 x i16> %A, <8 x i16> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, i1 0)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, i1 0)
store <8 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -72,7 +72,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8(<2 x i32> %A, <2 x i32> %B, <8
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -88,7 +88,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4(i32 %A, i32 %B, <8 x i32> %C,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 0, i32 %A, i1 0, i32 %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -104,7 +104,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -120,7 +120,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -136,7 +136,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -152,7 +152,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8(<2 x i32> %A, <2 x i32> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32> %A, <2 x i32> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -168,7 +168,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4(<2 x i32> %A, <2 x i32> %B, <8
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -185,7 +185,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16(<8 x half> %A, <16 x half> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -201,7 +201,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf16(<8 x i16> %A, <16 x i16> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -215,7 +215,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16(<8 x half> %A, <16 x half> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
+ %res = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
store <8 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -229,7 +229,7 @@ define amdgpu_ps void @test_swmmac_bf16_16x16x32_bf16(<8 x i16> %A, <16 x i16> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
+ %res = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
store <8 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -245,7 +245,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8(<2 x i32> %A, <4 x i32> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -261,7 +261,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4(i32 %A, <2 x i32> %B, <8 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -277,7 +277,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4(<2 x i32> %A, <4 x i32> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 0, <2 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i32 %Index, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -293,7 +293,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_fp8(<2 x i32> %A, <4 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -309,7 +309,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_bf8(<2 x i32> %A, <4 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -325,7 +325,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_fp8(<2 x i32> %A, <4 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -341,30 +341,30 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_bf8(<2 x i32> %A, <4 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
+ %res = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
store <8 x float> %res, ptr addrspace(1) %out
ret void
}
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16.v8f16.v8f32(<8 x half>, <8 x half>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16.v8i16.v8f32(<8 x i16>, <8 x i16>, <8 x float>)
-declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16.v8f16.v8f16(<8 x half>, <8 x half>, <8 x half>, i1 immarg)
-declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16.v8i16.v8i16(<8 x i16>, <8 x i16>, <8 x i16>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32.i32.v8i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32.v2i32.v8f32(<2 x i32>, <2 x i32>, <8 x float>)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32.v2i32.v8i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.v8f32.i16(<8 x half>, <16 x half>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.v8f32.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
-declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.v8f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
-declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.v8i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.v8i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.v8i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.v8i32.i32(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i32 %Index, i1 immarg)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
-declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.v8f32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v8f16(<8 x half>, <8 x half>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v8i16(<8 x i16>, <8 x i16>, <8 x float>)
+declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<8 x half>, <8 x half>, <8 x half>, i1 immarg)
+declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v8i16(<8 x i16>, <8 x i16>, <8 x i16>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v2i32.(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.i32(i1 immarg, i32, i1 immarg, i32, <8 x i32>, i1 immarg)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v8f32.v2i32(<2 x i32>, <2 x i32>, <8 x float>)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <8 x i32>, i1 immarg)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
+declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16..i16(<8 x half>, <16 x half>, <8 x half>, i16)
+declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16 %Index, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i32(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i32 %Index, i1 immarg)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
+declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32>, <4 x i32>, <8 x float>, i16)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-f16-f32-matrix-modifiers.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-f16-f32-matrix-modifiers.ll
index 1f2ffaf762712..a62c3c0643362 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-f16-f32-matrix-modifiers.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-f16-f32-matrix-modifiers.ll
@@ -11,7 +11,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negA(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <4 x half> %A
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %fneg.A, <4 x half> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %fneg.A, <4 x half> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -26,7 +26,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negB(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <4 x half> %B
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %A, <4 x half> %fneg.B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %fneg.B, <4 x float> %C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -41,7 +41,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negC(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -56,7 +56,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_absC(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -71,7 +71,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_negC(<4 x i16> %A, <4 x i16>
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4i16.v4f32(<4 x i16> %A, <4 x i16> %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -86,7 +86,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_absC(<4 x i16> %A, <4 x i16>
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4i16.v4f32(<4 x i16> %A, <4 x i16> %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -161,7 +161,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_negC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -176,7 +176,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_absC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -191,7 +191,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_negC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -206,7 +206,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_absC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -221,7 +221,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_negC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -236,7 +236,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_absC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -251,7 +251,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_negC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fneg.C = fneg <4 x float> %C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fneg.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fneg.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -266,7 +266,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_absC(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_endpgm
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.i32.v4f32(i32 %A, i32 %B, <4 x float> %fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -281,7 +281,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_negA(<4 x half> %A, <8 x hal
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <4 x half> %A
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f16.v8f16.v4f32.i16(<4 x half> %fneg.A, <8 x half> %B, <4 x float> %C, i16 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i16(<4 x half> %fneg.A, <8 x half> %B, <4 x float> %C, i16 %Index)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -296,7 +296,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_negB(<4 x half> %A, <8 x hal
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <8 x half> %B
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f16.v8f16.v4f32.i16(<4 x half> %A, <8 x half> %fneg.B, <4 x float> %C, i16 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i16(<4 x half> %A, <8 x half> %fneg.B, <4 x float> %C, i16 %Index)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -311,7 +311,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_negA(<4 x half> %A, <8 x hal
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <4 x half> %A
- %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v8f16.v4f16.i16(<4 x half> %fneg.A, <8 x half> %B, <4 x half> %C, i16 %Index)
+ %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i16(<4 x half> %fneg.A, <8 x half> %B, <4 x half> %C, i16 %Index)
store <4 x half> %res, <4 x half> addrspace(1)* %out
ret void
}
@@ -326,7 +326,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_negB(<4 x half> %A, <8 x hal
; GFX12-NEXT: s_endpgm
bb:
%fneg.B = fneg <8 x half> %B
- %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v8f16.v4f16.i16(<4 x half> %A, <8 x half> %fneg.B, <4 x half> %C, i16 %Index)
+ %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i16(<4 x half> %A, <8 x half> %fneg.B, <4 x half> %C, i16 %Index)
store <4 x half> %res, <4 x half> addrspace(1)* %out
ret void
}
@@ -344,7 +344,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negabsC(<4 x half> %A, <4 x ha
bb:
%fabs.C = call <4 x float> @llvm.fabs.v4f32(<4 x float> %C)
%fneg.fabs.C = fneg <4 x float> %fabs.C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -380,7 +380,7 @@ bb:
%el3.fabs = call float @llvm.fabs.f32(float %el3)
%partial.fabs.C = insertelement <4 x float> %C, float %el3.fabs, i32 3
%fneg.partial.fabs.C = fneg <4 x float> %partial.fabs.C
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.partial.fabs.C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> %fneg.partial.fabs.C)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -397,7 +397,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_negA_constantC(<4 x half> %A,
; GFX12-NEXT: s_endpgm
bb:
%fneg.A = fneg <4 x half> %A
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half> %fneg.A, <4 x half> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %fneg.A, <4 x half> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, <4 x float> addrspace(1)* %out
ret void
}
@@ -445,12 +445,12 @@ declare <4 x half> @llvm.fabs.v4f16(<4 x half>)
declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
declare float @llvm.fabs.f32(float)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f16.v4f32(<4 x half>, <4 x half>, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4i16.v4f32(<4 x i16>, <4 x i16>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half>, <4 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16>, <4 x i16>, <4 x float>)
declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v8f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f16.v8f16.v4f32.i16(<4 x half>, <8 x half>, <4 x float>, i16)
-declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v8f16.v4f16.i16(<4 x half>, <8 x half>, <4 x half>, i16)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i16(<4 x half>, <8 x half>, <4 x float>, i16)
+declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i16(<4 x half>, <8 x half>, <4 x half>, i16)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-imm.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-imm.ll
index fa0a7c98cea32..19b0e697183f4 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-imm.ll
@@ -10,7 +10,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -29,7 +29,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm_non_inlineable(<4 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -43,7 +43,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm(<4 x i16> %A, <4 x i16> %
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16.v4i16.v4f32(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -62,7 +62,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm_non_inlineable(<4 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16.v4i16.v4f32(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -76,7 +76,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm(<4 x half> %A, <4 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 1.0, half 1.0, half 1.0, half 1.0>, i1 0)
+ %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 1.0, half 1.0, half 1.0, half 1.0>, i1 0)
store <4 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -93,7 +93,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm_non_inlineable(<4 x half>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 3.0, half 3.0, half 3.0, half 3.0>, i1 0)
+ %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 3.0, half 3.0, half 3.0, half 3.0>, i1 0)
store <4 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -107,7 +107,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm(<4 x i16> %A, <4 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16256, i16 16256, i16 16256, i16 16256>, i1 0)
+ %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16256, i16 16256, i16 16256, i16 16256>, i1 0)
store <4 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -124,7 +124,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm_non_inlineable(<4 x i16>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16320, i16 16320, i16 16320, i16 16320>, i1 0)
+ %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16320, i16 16320, i16 16320, i16 16320>, i1 0)
store <4 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -138,7 +138,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm(i32 %A, i32 %B, ptr addrsp
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -157,7 +157,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm_non_inlineable(i32 %A, i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -171,7 +171,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm(i32 %A, i32 %B, ptr addrsp
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -190,7 +190,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm_non_inlineable(i32 %A, i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -204,7 +204,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm(i32 %A, i32 %B, <4 x f
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -223,7 +223,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable(i32 %A,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -237,7 +237,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm(i32 %A, i32 %B, <4 x f
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -256,7 +256,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable(i32 %A,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -270,7 +270,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm(i32 %A, i32 %B, <4 x f
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -289,7 +289,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable(i32 %A,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -303,7 +303,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm(i32 %A, i32 %B, <4 x f
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -322,7 +322,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable(i32 %A,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -336,7 +336,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm(i32 %A, i32 %B, <4 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -355,19 +355,19 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm_non_inlineable(i32 %A, i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16.v4f16.v4f32(<4 x half>, <4 x half>, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16.v4i16.v4f32(<4 x i16>, <4 x i16>, <4 x float>)
-declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16.v4f16.v4f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
-declare <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16.v4i16.v4i16(<4 x i16>, <4 x i16>, <4 x i16>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half>, <4 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16>, <4 x i16>, <4 x float>)
+declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
+declare <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16>, <4 x i16>, <4 x i16>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-iu-modifiers.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-iu-modifiers.ll
index 803453e45ced0..fa5eb3605e67a 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-iu-modifiers.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-iu-modifiers.ll
@@ -10,7 +10,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_zext_src0(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -24,7 +24,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_zext_src1(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -38,7 +38,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_clamp(i32 %A, i32 %B, <4 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -54,7 +54,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_zext_src0(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -68,7 +68,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_zext_src1(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -82,7 +82,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_clamp(i32 %A, i32 %B, <4 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -98,7 +98,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_zext_src0(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -112,7 +112,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_zext_src1(i32 %A, i32 %B, <4 x
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -126,16 +126,11 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_clamp(i32 %A, i32 %B, <4 x i32
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
-
-
-
-
-
define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_zext_src0(i32 %A, <2 x i32> %B, <4 x i32> %C, i8 %Index, ptr addrspace(1) %out) {
; GFX12-LABEL: test_swmmac_i32_16x16x32_iu8_zext_src0:
; GFX12: ; %bb.0: ; %bb
@@ -145,7 +140,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_zext_src0(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 1, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 1, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -159,7 +154,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_zext_src1(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 1, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 1, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -173,7 +168,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_clamp(i32 %A, <2 x i32> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -189,7 +184,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_zext_src0(i32 %A, i32 %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 1, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -203,7 +198,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_zext_src1(i32 %A, i32 %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 0, i32 %A, i1 1, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -217,7 +212,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_clamp(i32 %A, i32 %B, <4 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -233,7 +228,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_zext_src0(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 1, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 1, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -247,7 +242,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_zext_src1(i32 %A, <2 x i32>
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 0, i32 %A, i1 1, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 0, i32 %A, i1 1, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -261,14 +256,14 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_clamp(i32 %A, <2 x i32> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-swmmac-index_key.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-swmmac-index_key.ll
index 0c5b19a8416fb..861eb1aaa333a 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-swmmac-index_key.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64-swmmac-index_key.ll
@@ -33,16 +33,16 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16_index_key(<4 x half> %A, <8
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
@@ -79,16 +79,16 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf16_index_key(<4 x i16> %A, <8
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
@@ -119,16 +119,16 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16_index_key(<4 x half> %A, <8
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index0)
+ %res0 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index0)
store <4 x half> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index1)
+ %res1 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index1)
store <4 x half> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index2)
+ %res2 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index2)
store <4 x half> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index3)
+ %res3 = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index3)
store <4 x half> %res3, ptr addrspace(1) %out3
ret void
}
@@ -159,16 +159,16 @@ define amdgpu_ps void @test_swmmac_bf16_16x16x32_bf16_index_key(<4 x i16> %A, <8
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index0)
+ %res0 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index0)
store <4 x i16> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index1)
+ %res1 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index1)
store <4 x i16> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index2)
+ %res2 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index2)
store <4 x i16> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index3)
+ %res3 = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index3)
store <4 x i16> %res3, ptr addrspace(1) %out3
ret void
}
@@ -205,16 +205,16 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8_index_key(i32 %A, <2 x i32>
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index0, i1 0)
+ %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index0, i1 0)
store <4 x i32> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index1, i1 0)
+ %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index1, i1 0)
store <4 x i32> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index2, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index2, i1 0)
store <4 x i32> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index3, i1 0)
+ %res3 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index3, i1 0)
store <4 x i32> %res3, ptr addrspace(1) %out3
ret void
}
@@ -239,10 +239,10 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4_index_key(i32 %A, i32 %B, <4
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index0, i1 0)
+ %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index0, i1 0)
store <4 x i32> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index1, i1 0)
+ %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index1, i1 0)
store <4 x i32> %res1, ptr addrspace(1) %out1
ret void
}
@@ -267,10 +267,10 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4_index_key(i32 %A, <2 x i32>
bb:
%IndexVec = load <2 x i16>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <2 x i16> %IndexVec, i32 0
- %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index0, i1 0)
+ %res0 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index0, i1 0)
store <4 x i32> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <2 x i16> %IndexVec, i32 1
- %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index1, i1 0)
+ %res1 = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index1, i1 0)
store <4 x i32> %res1, ptr addrspace(1) %out1
ret void
}
@@ -307,16 +307,16 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_fp8_index_key(i32 %A, <2 x i
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
@@ -353,16 +353,16 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_bf8_index_key(i32 %A, <2 x i
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
@@ -399,16 +399,16 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_fp8_index_key(i32 %A, <2 x i
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
@@ -445,28 +445,28 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_bf8_index_key(i32 %A, <2 x i
bb:
%IndexVec = load <4 x i8>, ptr addrspace(1) %IndexVecPtr, align 4
%Index0 = extractelement <4 x i8> %IndexVec, i32 0
- %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
+ %res0 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index0)
store <4 x float> %res0, ptr addrspace(1) %out0
%Index1 = extractelement <4 x i8> %IndexVec, i32 1
- %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
+ %res1 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index1)
store <4 x float> %res1, ptr addrspace(1) %out1
%Index2 = extractelement <4 x i8> %IndexVec, i32 2
- %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
+ %res2 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index2)
store <4 x float> %res2, ptr addrspace(1) %out2
%Index3 = extractelement <4 x i8> %IndexVec, i32 3
- %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
+ %res3 = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index3)
store <4 x float> %res3, ptr addrspace(1) %out3
ret void
}
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half>, <8 x half>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16>, <8 x i16>, <4 x float>, i8)
-declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half>, <8 x half>, <4 x half>, i8)
-declare <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16>, <8 x i16>, <4 x i16>, i8)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half>, <8 x half>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16>, <8 x i16>, <4 x float>, i8)
+declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half>, <8 x half>, <4 x half>, i8)
+declare <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16>, <8 x i16>, <4 x i16>, i8)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64.ll b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64.ll
index b25f278513331..a05a8f4117ece 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-gfx12-w64.ll
@@ -10,7 +10,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<4 x half> %A, <4 x half> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16.v4f16.v4f32(<4 x half> %A, <4 x half> %B, <4 x float> %C)
+ %res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -24,7 +24,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<4 x i16> %A, <4 x i16> %B, <
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16.v4i16.v4f32(<4 x i16> %A, <4 x i16> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -38,7 +38,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16(<4 x half> %A, <4 x half> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> %C, i1 0)
+ %res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> %C, i1 0)
store <4 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -52,7 +52,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16(<4 x i16> %A, <4 x i16> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, i1 0)
+ %res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, i1 0)
store <4 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -66,7 +66,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu8(i32 %A, i32 %B, <4 x i32> %C,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -80,7 +80,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_iu4(i32 %A, i32 %B, <4 x i32> %C,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -94,7 +94,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8(i32 %A, i32 %B, <4 x float
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -108,7 +108,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8(i32 %A, i32 %B, <4 x float
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -122,7 +122,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8(i32 %A, i32 %B, <4 x float
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -136,7 +136,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8(i32 %A, i32 %B, <4 x float
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32.i32.v4f32(i32 %A, i32 %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -150,7 +150,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x32_iu4(i32 %A, i32 %B, <4 x i32> %C,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -164,7 +164,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_f16(<4 x half> %A, <8 x half> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -178,7 +178,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf16(<4 x i16> %A, <8 x i16> %B,
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -192,7 +192,7 @@ define amdgpu_ps void @test_swmmac_f16_16x16x32_f16(<4 x half> %A, <8 x half> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index)
+ %res = call <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half> %A, <8 x half> %B, <4 x half> %C, i8 %Index)
store <4 x half> %res, ptr addrspace(1) %out
ret void
}
@@ -206,7 +206,7 @@ define amdgpu_ps void @test_swmmac_bf16_16x16x32_bf16(<4 x i16> %A, <8 x i16> %B
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index)
+ %res = call <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16> %A, <8 x i16> %B, <4 x i16> %C, i8 %Index)
store <4 x i16> %res, ptr addrspace(1) %out
ret void
}
@@ -220,7 +220,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu8(i32 %A, <2 x i32> %B, <4 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i8 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -234,7 +234,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x32_iu4(i32 %A, i32 %B, <4 x i32> %C
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -248,7 +248,7 @@ define amdgpu_ps void @test_swmmac_i32_16x16x64_iu4(i32 %A, <2 x i32> %B, <4 x i
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 0, i32 %A, i1 0, <2 x i32> %B, <4 x i32> %C, i16 %Index, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out
ret void
}
@@ -262,7 +262,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_fp8(i32 %A, <2 x i32> %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -276,7 +276,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_fp8_bf8(i32 %A, <2 x i32> %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -290,7 +290,7 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_fp8(i32 %A, <2 x i32> %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
@@ -304,30 +304,30 @@ define amdgpu_ps void @test_swmmac_f32_16x16x32_bf8_bf8(i32 %A, <2 x i32> %B, <4
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
+ %res = call <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32 %A, <2 x i32> %B, <4 x float> %C, i8 %Index)
store <4 x float> %res, ptr addrspace(1) %out
ret void
}
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16.v4f16.v4f32(<4 x half>, <4 x half>, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16.v4i16.v4f32(<4 x i16>, <4 x i16>, <4 x float>)
-declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16.v4f16.v4f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
-declare <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16.v4i16.v4i16(<4 x i16>, <4 x i16>, <4 x i16>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32.i32.v4f32(i32, i32, <4 x float>)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32.i32.v4i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.v4f32.i8(<4 x half>, <8 x half>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.v4f32.i8(<4 x i16>, <8 x i16>, <4 x float>, i8)
-declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.v4f16.i8(<4 x half>, <8 x half>, <4 x half>, i8)
-declare <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.v4i16.i8(<4 x i16>, <8 x i16>, <4 x i16>, i8)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.v4i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.v4i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.v4i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
-declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.v4f32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half>, <4 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16>, <4 x i16>, <4 x float>)
+declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
+declare <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16>, <4 x i16>, <4 x i16>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32.(i32, i32, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32, i32, <4 x float>)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v4f32.v4f16.v8f16.i8(<4 x half>, <8 x half>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v4f32.v4i16.v8i16.i8(<4 x i16>, <8 x i16>, <4 x float>, i8)
+declare <4 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v4f16.v4f16.v8f16.i8(<4 x half>, <8 x half>, <4 x half>, i8)
+declare <4 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v4i16.v4i16.v8i16.i8(<4 x i16>, <8 x i16>, <4 x i16>, i8)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v4i32.i32.v2i32.i8(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i8 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v4i32.i32.i32.i16(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v4i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <4 x i32>, i16 %Index, i1 immarg)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
+declare <4 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v4f32.i32.v2i32.i8(i32, <2 x i32>, <4 x float>, i8)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma_modifiers.ll b/llvm/test/CodeGen/AMDGPU/wmma_modifiers.ll
index 0ed2f7029c5c0..801bb06aacb63 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma_modifiers.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma_modifiers.ll
@@ -1,6 +1,6 @@
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs < %s | FileCheck %s
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half>, <16 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half>, <16 x half>, <4 x float>)
; Make sure we don't crash when trying to select modifiers in SelectVOP3PMods.
@@ -13,7 +13,7 @@ define amdgpu_cs void @xyz () {
loop:
%ld = load <8 x float>, ptr addrspace(5) null, align 32
%in_shuffle = shufflevector <8 x float> %ld, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %wmma = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> undef, <16 x half> undef, <4 x float> %in_shuffle)
+ %wmma = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half> undef, <16 x half> undef, <4 x float> %in_shuffle)
%out_shuffle = shufflevector <4 x float> %wmma, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
store <8 x float> %out_shuffle, ptr addrspace(5) null, align 32
br i1 false, label %.exit, label %loop
diff --git a/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll b/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
index 364ef76b8b246..12b65986cf1c7 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
@@ -1,12 +1,12 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs < %s | FileCheck %s --check-prefix=W32
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half>, <16 x half> , <8 x float>)
-declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16>, <16 x i16> , <8 x float>)
-declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half>, <16 x half> , <16 x half>, i1 immarg)
-declare <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16>, <16 x i16> , <16 x i16>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 immarg, <4 x i32>, i1 immarg, <4 x i32> , <8 x i32>, i1 immarg)
-declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 immarg, <2 x i32>, i1 immarg, <2 x i32> , <8 x i32>, i1 immarg)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half>, <16 x half> , <8 x float>)
+declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16>, <16 x i16> , <8 x float>)
+declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half>, <16 x half> , <16 x half>, i1 immarg)
+declare <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16>, <16 x i16> , <16 x i16>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 immarg, <4 x i32>, i1 immarg, <4 x i32> , <8 x i32>, i1 immarg)
+declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32> , <8 x i32>, i1 immarg)
; The tests demonstrate that the following WMMA register constraints are satisfied.
;
@@ -35,8 +35,8 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<16 x half> %A, <16 x half> %B
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> %A, <16 x half> %B, <8 x float> %C)
- %res2 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> %B, <16 x half> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half> %A, <16 x half> %B, <8 x float> %C)
+ %res2 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half> %B, <16 x half> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out, align 32
store <8 x float> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -59,8 +59,8 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<16 x i16> %A, <16 x i16> %B,
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <8 x float> %C)
- %res2 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <8 x float> %C)
+ %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x float> %C)
+ %res2 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16> %B, <16 x i16> %B, <8 x float> %C)
store <8 x float> %res, ptr addrspace(1) %out, align 32
store <8 x float> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -83,8 +83,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<16 x half> %A, <16 x half>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 0)
- %res2 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %B, <16 x half> %B, <16 x half> %C, i1 0)
+ %res = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 0)
+ %res2 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %B, <16 x half> %B, <16 x half> %C, i1 0)
store <16 x half> %res, ptr addrspace(1) %out, align 32
store <16 x half> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -105,8 +105,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<16 x half> %A, <16 x half>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 1)
- %res2 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %B, <16 x half> %B, <16 x half> %C, i1 1)
+ %res = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 1)
+ %res2 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %B, <16 x half> %B, <16 x half> %C, i1 1)
store <16 x half> %res, ptr addrspace(1) %out, align 32
store <16 x half> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -129,8 +129,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<16 x i16> %A, <16 x i16>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 0)
- %res2 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <16 x i16> %C, i1 0)
+ %res = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 0)
+ %res2 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %B, <16 x i16> %B, <16 x i16> %C, i1 0)
store <16 x i16> %res, ptr addrspace(1) %out, align 32
store <16 x i16> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -151,8 +151,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<16 x i16> %A, <16 x i16>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 1)
- %res2 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <16 x i16> %C, i1 1)
+ %res = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 1)
+ %res2 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %B, <16 x i16> %B, <16 x i16> %C, i1 1)
store <16 x i16> %res, ptr addrspace(1) %out, align 32
store <16 x i16> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -175,8 +175,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -197,8 +197,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -219,8 +219,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -241,8 +241,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -263,8 +263,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -285,8 +285,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -307,8 +307,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -329,8 +329,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -353,8 +353,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -375,8 +375,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -397,8 +397,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -419,8 +419,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -442,8 +442,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -464,8 +464,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -486,8 +486,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
@@ -508,8 +508,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
; W32-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W32-NEXT: s_endpgm
bb:
- %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
- %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
+ %res2 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
store <8 x i32> %res, ptr addrspace(1) %out, align 32
store <8 x i32> %res2, ptr addrspace(1) %out2, align 32
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll b/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
index f7b45212cc8b4..b43969dd4f5c0 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
@@ -1,12 +1,12 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs < %s | FileCheck %s --check-prefix=W64
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half>, <16 x half>, <4 x float>)
-declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16>, <16 x i16>, <4 x float>)
-declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
-declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 immarg, <4 x i32>, i1 immarg, <4 x i32>, <4 x i32>, i1 immarg)
-declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <4 x i32>, i1 immarg)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half>, <16 x half>, <4 x float>)
+declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v16i16(<16 x i16>, <16 x i16>, <4 x float>)
+declare <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half>, <16 x half>, <8 x half>, i1 immarg)
+declare <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16>, <16 x i16>, <8 x i16>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 immarg, <4 x i32>, i1 immarg, <4 x i32>, <4 x i32>, i1 immarg)
+declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32>, <4 x i32>, i1 immarg)
; The tests demonstrate that the following WMMA register constraints are satisfied.
;
@@ -31,8 +31,8 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<16 x half> %A, <16 x half> %B
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> %A, <16 x half> %B, <4 x float> %C)
- %res2 = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<16 x half> %B, <16 x half> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half> %A, <16 x half> %B, <4 x float> %C)
+ %res2 = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v16f16(<16 x half> %B, <16 x half> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out, align 16
store <4 x float> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -51,8 +51,8 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<16 x i16> %A, <16 x i16> %B,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <4 x float> %C)
- %res2 = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <4 x float> %C)
+ %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v16i16(<16 x i16> %A, <16 x i16> %B, <4 x float> %C)
+ %res2 = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v16i16(<16 x i16> %B, <16 x i16> %B, <4 x float> %C)
store <4 x float> %res, ptr addrspace(1) %out, align 16
store <4 x float> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -71,8 +71,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<16 x half> %A, <16 x half>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 0)
- %res2 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %B, <16 x half> %B, <8 x half> %C, i1 0)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 0)
+ %res2 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %B, <16 x half> %B, <8 x half> %C, i1 0)
store <8 x half> %res, ptr addrspace(1) %out, align 16
store <8 x half> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -89,8 +89,8 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<16 x half> %A, <16 x half>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 1)
- %res2 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16(<16 x half> %B, <16 x half> %B, <8 x half> %C, i1 1)
+ %res = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %A, <16 x half> %B, <8 x half> %C, i1 1)
+ %res2 = call <8 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v8f16.v16f16(<16 x half> %B, <16 x half> %B, <8 x half> %C, i1 1)
store <8 x half> %res, ptr addrspace(1) %out, align 16
store <8 x half> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -109,8 +109,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<16 x i16> %A, <16 x i16>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 0)
- %res2 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <8 x i16> %C, i1 0)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 0)
+ %res2 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %B, <16 x i16> %B, <8 x i16> %C, i1 0)
store <8 x i16> %res, ptr addrspace(1) %out, align 16
store <8 x i16> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -127,8 +127,8 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<16 x i16> %A, <16 x i16>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 1)
- %res2 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<16 x i16> %B, <16 x i16> %B, <8 x i16> %C, i1 1)
+ %res = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x i16> %C, i1 1)
+ %res2 = call <8 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v8i16.v16i16(<16 x i16> %B, <16 x i16> %B, <8 x i16> %C, i1 1)
store <8 x i16> %res, ptr addrspace(1) %out, align 16
store <8 x i16> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -147,8 +147,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -166,8 +166,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -184,8 +184,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -202,8 +202,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -220,8 +220,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -238,8 +238,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 0, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -256,8 +256,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %B, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -274,8 +274,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.v4i32(i1 1, <4 x i32> %B, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -294,8 +294,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -312,8 +312,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -330,8 +330,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -348,8 +348,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -366,8 +366,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -384,8 +384,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 0, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -402,8 +402,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %B, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
@@ -420,8 +420,8 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
; W64-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; W64-NEXT: s_endpgm
bb:
- %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
- %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
+ %res2 = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.v2i32(i1 1, <2 x i32> %B, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
store <4 x i32> %res, ptr addrspace(1) %out, align 16
store <4 x i32> %res2, ptr addrspace(1) %out2, align 16
ret void
More information about the llvm-commits
mailing list