; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=instcombine < %s | FileCheck %s

; ------------------------------------------------------------------------------------
; Incorrect signature for format cases (IR vector too large) wmma.f32.16x16x128.f8f6f4
; ------------------------------------------------------------------------------------

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 2, <12 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 2, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 2, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 3, <12 x i32> [[TMP0]], i32 1, <16 x i32> [[B]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 1, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 3, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 3, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v8i32.v12i32(i32 4, <8 x i32> [[TMP0]], i32 2, <12 x i32> [[TMP1]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v12i32.v8i32(i32 3, <12 x i32> [[TMP0]], i32 4, <8 x i32> [[TMP1]], i16 0, <8 x float> [[C]])
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

; ------------------------------------------------------------------------------------------
; Incorrect signature for format cases (IR vector too large) wmma.scale.f32.16x16x128.f8f6f4
; ------------------------------------------------------------------------------------------

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 2, <12 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 2, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 2, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 3, <12 x i32> [[TMP0]], i32 1, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 1, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 3, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 3, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v8i32.v12i32(i32 4, <8 x i32> [[TMP0]], i32 2, <12 x i32> [[TMP1]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v12i32.v8i32(i32 3, <12 x i32> [[TMP0]], i32 4, <8 x i32> [[TMP1]], i16 0, <8 x float> [[C]], i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

; --------------------------------------------------------------------------------------------
; Incorrect signature for format cases (IR vector too large) wmma.scale16.f32.16x16x128.f8f6f4
; --------------------------------------------------------------------------------------------

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp6___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 2, <12 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 2, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_fp6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 2, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf6___v16i32_bf8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v12i32.v16i32(i32 3, <12 x i32> [[TMP0]], i32 1, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 1, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf8___v16i32_bf6(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v12i32(i32 1, <16 x i32> [[A]], i32 3, <12 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 1, <16 x i32> %A, i32 3, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp4___v16i32_fp8(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <16 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(<16 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp8___v16i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp8(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v8i32.v16i32(i32 4, <8 x i32> [[TMP0]], i32 0, <16 x i32> [[B]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 0, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_fp8___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v8i32(i32 0, <16 x i32> [[A]], i32 4, <8 x i32> [[TMP0]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 0, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(<12 x i32> %A, <16 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v12i32_fp4___v16i32_fp6(
; CHECK-SAME: <12 x i32> [[A:%.*]], <16 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <12 x i32> [[A]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <16 x i32> [[B]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v8i32.v12i32(i32 4, <8 x i32> [[TMP0]], i32 2, <12 x i32> [[TMP1]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 4, <12 x i32> %A, i32 2, <16 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(<16 x i32> %A, <12 x i32> %B, <8 x float> %C, ptr addrspace(1) %out) {
; CHECK-LABEL: define amdgpu_ps void @test_wmma_scale16_f32_16x16x128_f8f6f4___v16i32_bf6___v12i32_fp4(
; CHECK-SAME: <16 x i32> [[A:%.*]], <12 x i32> [[B:%.*]], <8 x float> [[C:%.*]], ptr addrspace(1) [[OUT:%.*]]) {
; CHECK-NEXT:  [[BB:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <16 x i32> [[A]], <16 x i32> poison, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <12 x i32> [[B]], <12 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[RES:%.*]] = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v12i32.v8i32(i32 3, <12 x i32> [[TMP0]], i32 4, <8 x i32> [[TMP1]], i16 0, <8 x float> [[C]], i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
; CHECK-NEXT:    store <8 x float> [[RES]], ptr addrspace(1) [[OUT]], align 32
; CHECK-NEXT:    ret void
;
bb:
  %res = call <8 x float> @llvm.amdgcn.wmma.scale16.f32.16x16x128.f8f6f4.v8f32.v16i32.v16i32(i32 3, <16 x i32> %A, i32 4, <12 x i32> %B, i16 0, <8 x float> %C, i32 0, i32 0, i64 0, i32 0, i32 0, i64 0, i1 false, i1 false)
  store <8 x float> %res, ptr addrspace(1) %out
  ret void
}