AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950 #117232

arsenm · 2024-11-21T20:33:22Z

No description provided.

arsenm · 2024-11-21T20:33:47Z

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmbot · 2024-11-21T20:34:15Z

@llvm/pr-subscribers-llvm-analysis

@llvm/pr-subscribers-mc

Author: Matt Arsenault (arsenm)

Changes

Patch is 26.30 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/117232.diff

13 Files Affected:

(modified) clang/include/clang/Basic/BuiltinsAMDGPU.def (+1)
(modified) clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl (+7)
(modified) clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl (+6)
(modified) clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl (+1)
(modified) llvm/include/llvm/IR/IntrinsicsAMDGPU.td (+1)
(modified) llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp (+4)
(modified) llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp (+2-1)
(modified) llvm/lib/Target/AMDGPU/SIInstrInfo.td (+1)
(modified) llvm/lib/Target/AMDGPU/VOP3PInstructions.td (+5)
(modified) llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll (+8)
(modified) llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll (+215)
(modified) llvm/test/MC/AMDGPU/mai-gfx950.s (+36)
(modified) llvm/test/MC/Disassembler/AMDGPU/gfx950_mai.txt (+22)

diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def
index 306f9d2a5deac1..5029e1a6153f9a 100644
--- a/clang/include/clang/Basic/BuiltinsAMDGPU.def
+++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def
@@ -450,6 +450,7 @@ TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_16x16x64_bf16, "V4fV8yV16yV4fiIiIi",
 TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_32x32x32_bf16, "V16fV8yV16yV16fiIiIi", "nc", "gfx950-insts")
 TARGET_BUILTIN(__builtin_amdgcn_smfmac_i32_16x16x128_i8, "V4iV4iV8iV4iiIiIi", "nc", "gfx950-insts")
 TARGET_BUILTIN(__builtin_amdgcn_smfmac_i32_32x32x64_i8, "V16iV4iV8iV16iiIiIi", "nc", "gfx950-insts")
+TARGET_BUILTIN(__builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8, "V4fV4iV8iV4fiIiIi", "nc", "gfx950-insts")
 
 //===----------------------------------------------------------------------===//
 // GFX12+ only builtins.
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl
index fc1924ec187b1c..f07bec5ec91eb3 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-mfma.cl
@@ -510,4 +510,11 @@ void test_smfmac_i32_32x32x64_i8(global v16i* out, v4i a, v8i b, v16i c, int idx
   *out = __builtin_amdgcn_smfmac_i32_32x32x64_i8(a, b, c, idx, 0, 0);
 }
 
+// CHECK-GFX950-LABEL: @test_smfmac_f32_16x16x128_bf8_bf8
+// CHECK-GFX950: call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %a, <8 x i32> %b, <4 x float> %c, i32 %idx, i32 0, i32 0)
+void test_smfmac_f32_16x16x128_bf8_bf8(global v4f* out, v4i a, v8i b, v4f c, int idx)
+{
+  *out = __builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8(a, b, c, idx, 0, 0);
+}
+
 #endif
diff --git a/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl b/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl
index 1812c6fd6203a5..e2235cec42d84b 100644
--- a/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl
+++ b/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950-param.cl
@@ -100,3 +100,9 @@ void test_smfmac_i32_32x32x64_i8(global int16* out, int4 a, int8 b, int16 c, int
   *out = __builtin_amdgcn_smfmac_i32_32x32x64_i8(a, b, c, idx, d, 0); // expected-error{{argument to '__builtin_amdgcn_smfmac_i32_32x32x64_i8' must be a constant integer}}
   *out = __builtin_amdgcn_smfmac_i32_32x32x64_i8(a, b, c, idx, 0, d); // expected-error{{argument to '__builtin_amdgcn_smfmac_i32_32x32x64_i8' must be a constant integer}}
 }
+
+void test_smfmac_f32_16x16x128_bf8_bf8(global float4* out, int4 a, int8 b, float4 c, int idx, int d)
+{
+  *out = __builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8(a, b, c, idx, d, 0); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8' must be a constant integer}}
+  *out = __builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8(a, b, c, idx, 0, d); // expected-error{{argument to '__builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8' must be a constant integer}}
+}
diff --git a/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl b/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl
index 37a7f2a502e435..f65abeb2b58c4d 100644
--- a/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl
+++ b/clang/test/SemaOpenCL/builtins-amdgcn-error-gfx950.cl
@@ -40,6 +40,7 @@ void test(__global float4* out0, half8 a0, half8 b0, float4 c0,
   *out9 = __builtin_amdgcn_smfmac_f32_32x32x32_bf16(a9, b9, c9, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_32x32x32_bf16' needs target feature gfx950-insts}}
   *out10 = __builtin_amdgcn_smfmac_i32_16x16x128_i8(a10, b10, c10, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_i32_16x16x128_i8' needs target feature gfx950-insts}}
   *out11 = __builtin_amdgcn_smfmac_i32_32x32x64_i8(a11, b11, c11, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_i32_32x32x64_i8' needs target feature gfx950-insts}}
+  *out12 = __builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8(a12, b12, c12, 0, 0, 0); // expected-error{{'__builtin_amdgcn_smfmac_f32_16x16x128_bf8_bf8' needs target feature gfx950-insts}}
   *out14 = __builtin_amdgcn_mfma_scale_f32_16x16x128_f8f6f4(a14, b14, c14, 0, 0, 0, d14, 0, e14); // expected-error{{'__builtin_amdgcn_mfma_scale_f32_16x16x128_f8f6f4' needs target feature gfx950-insts}}
   *out15 = __builtin_amdgcn_mfma_scale_f32_32x32x64_f8f6f4(a15, b15, c15, 0, 0, 0, d15, 0, e15); // expected-error{{'__builtin_amdgcn_mfma_scale_f32_32x32x64_f8f6f4' needs target feature gfx950-insts}}
 }
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index b0c4ed6ce40da5..7ad23afb7010de 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -3158,6 +3158,7 @@ def int_amdgcn_smfmac_f32_16x16x64_bf16 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_ty,
 def int_amdgcn_smfmac_f32_32x32x32_bf16 : AMDGPUMSmfmacIntrinsic<llvm_v16f32_ty, llvm_v8bf16_ty, llvm_v16bf16_ty>;
 def int_amdgcn_smfmac_i32_16x16x128_i8 : AMDGPUMSmfmacIntrinsic<llvm_v4i32_ty,  llvm_v4i32_ty, llvm_v8i32_ty>;
 def int_amdgcn_smfmac_i32_32x32x64_i8 : AMDGPUMSmfmacIntrinsic<llvm_v16i32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
+def int_amdgcn_smfmac_f32_16x16x128_bf8_bf8 : AMDGPUMSmfmacIntrinsic<llvm_v4f32_ty, llvm_v4i32_ty, llvm_v8i32_ty>;
 }
 
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
index 10c7825dc3e402..909fc349eca302 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -1096,6 +1096,7 @@ bool AMDGPUInstructionSelector::selectG_INTRINSIC(MachineInstr &I) const {
   case Intrinsic::amdgcn_smfmac_f32_32x32x32_bf16:
   case Intrinsic::amdgcn_smfmac_i32_16x16x128_i8:
   case Intrinsic::amdgcn_smfmac_i32_32x32x64_i8:
+  case Intrinsic::amdgcn_smfmac_f32_16x16x128_bf8_bf8:
     return selectSMFMACIntrin(I);
   default:
     return selectImpl(I, *CoverageInfo);
@@ -3506,6 +3507,9 @@ bool AMDGPUInstructionSelector::selectSMFMACIntrin(MachineInstr &MI) const {
   case Intrinsic::amdgcn_smfmac_i32_32x32x64_i8:
     Opc = AMDGPU::V_SMFMAC_I32_32X32X64_I8_e64;
     break;
+  case Intrinsic::amdgcn_smfmac_f32_16x16x128_bf8_bf8:
+    Opc = AMDGPU::V_SMFMAC_F32_16X16X128_BF8_BF8_e64;
+    break;
   default:
     llvm_unreachable("unhandled smfmac intrinsic");
   }
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
index 596e30ce8b935d..7cf0dc17a25e7a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
@@ -4810,7 +4810,8 @@ AMDGPURegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
     case Intrinsic::amdgcn_smfmac_f32_16x16x64_bf16:
     case Intrinsic::amdgcn_smfmac_f32_32x32x32_bf16:
     case Intrinsic::amdgcn_smfmac_i32_16x16x128_i8:
-    case Intrinsic::amdgcn_smfmac_i32_32x32x64_i8: {
+    case Intrinsic::amdgcn_smfmac_i32_32x32x64_i8:
+    case Intrinsic::amdgcn_smfmac_f32_16x16x128_bf8_bf8: {
       // vdst, srcA, srcB, srcC, idx
       OpdsMapping[0] = getAGPROpMapping(MI.getOperand(0).getReg(), MRI, *TRI);
       OpdsMapping[2] = getVGPROpMapping(MI.getOperand(2).getReg(), MRI, *TRI);
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index e155a9579b6958..afb0b2cede045c 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -2882,6 +2882,7 @@ def VOP_V4F32_V2I32_V4I32_I32     : VOPProfile <[v4f32,  v2i32, v4i32, i32]>;
 def VOP_V16F32_V2I32_V4I32_I32    : VOPProfile <[v16f32, v2i32, v4i32, i32]>;
 def VOP_V4I32_V4I32_V8I32_I32     : VOPProfile <[v4i32,  v4i32, v8i32, i32]>;
 def VOP_V16I32_V4I32_V8I32_I32    : VOPProfile <[v16i32, v4i32, v8i32, i32]>;
+def VOP_V4F32_V4I32_V8I32_I32     : VOPProfile <[v4f32, v4i32, v8i32, i32]>;
 
 def VOP_V4F32_V8F16_V8F16_V4F32   : VOPProfile <[v4f32,  v8f16, v8f16, v4f32]>;
 def VOP_V16F32_V8F16_V8F16_V16F32 : VOPProfile <[v16f32, v8f16, v8f16, v16f32]>;
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index b62ac8c7cfb7da..8be84c250166ba 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -643,6 +643,8 @@ def VOPProfileSMFMAC_F32_32X32X32_F8  : VOPProfileSMFMAC<VOP_V16F32_V2I32_V4I32_
 def VOPProfileSMFMAC_I32_16X16X128_I8 : VOPProfileSMFMAC<VOP_V4I32_V4I32_V8I32_I32, AVDst_128, AVSrc_128, AVSrc_256>;
 def VOPProfileSMFMAC_I32_32X32X64_I8 : VOPProfileSMFMAC<VOP_V16I32_V4I32_V8I32_I32, AVDst_512, AVSrc_128, AVSrc_256>;
 
+def VOPProfileSMFMAC_F32_16X16X128_F8  : VOPProfileSMFMAC<VOP_V4F32_V4I32_V8I32_I32,  AVDst_128, AVSrc_128, AVSrc_256>;
+
 def VOPProfileMAI_F32_V8F16_X32     : VOPProfileMAI<VOP_V4F32_V8F16_V8F16_V4F32,       AISrc_128_f32,  ADst_128, AVSrc_128>;
 def VOPProfileMAI_F32_V8F16_X32_VCD : VOPProfileMAI<VOP_V4F32_V8F16_V8F16_V4F32,       VISrc_128_f32,  VDst_128, AVSrc_128>;
 def VOPProfileMAI_F32_V8F16_X16     : VOPProfileMAI<VOP_V16F32_V8F16_V8F16_V16F32,     AISrc_512_f32,  ADst_512, AVSrc_128>;
@@ -1055,6 +1057,7 @@ defm V_SMFMAC_F32_16X16X64_BF16    : SMFMACInst<"v_smfmac_f32_16x16x64_bf16",
 defm V_SMFMAC_F32_32X32X32_BF16    : SMFMACInst<"v_smfmac_f32_32x32x32_bf16",    "F32_32X32X32_BF16", int_amdgcn_smfmac_f32_32x32x32_bf16>;
 defm V_SMFMAC_I32_16X16X128_I8     : SMFMACInst<"v_smfmac_i32_16x16x128_i8",     "I32_16X16X128_I8", int_amdgcn_smfmac_i32_16x16x128_i8>;
 defm V_SMFMAC_I32_32X32X64_I8      : SMFMACInst<"v_smfmac_i32_32x32x64_i8",      "I32_32X32X64_I8", int_amdgcn_smfmac_i32_32x32x64_i8>;
+defm V_SMFMAC_F32_16X16X128_BF8_BF8 : SMFMACInst<"v_smfmac_f32_16x16x128_bf8_bf8", "F32_16X16X128_F8", int_amdgcn_smfmac_f32_16x16x128_bf8_bf8>;
 }
 
 def MAIInstInfoTable : GenericTable {
@@ -2153,6 +2156,8 @@ defm V_SMFMAC_F32_32X32X32_BF16    : VOP3P_Real_SMFMAC <0x46, "v_smfmac_f32_32x3
 defm V_SMFMAC_I32_16X16X128_I8     : VOP3P_Real_SMFMAC <0x3a, "v_smfmac_i32_16x16x128i8">;
 defm V_SMFMAC_I32_32X32X64_I8      : VOP3P_Real_SMFMAC <0x47, "v_smfmac_i32_32x32x64i8">;
 
+defm V_SMFMAC_F32_16X16X128_BF8_BF8 : VOP3P_Real_SMFMAC <0x3b, "v_smfmac_f32_16x16x128bf8bf8">;
+
 defm V_PK_FMA_F32 : VOP3P_Real_vi <0x30>;
 defm V_PK_MUL_F32 : VOP3P_Real_vi <0x31>;
 defm V_PK_ADD_F32 : VOP3P_Real_vi <0x32>;
diff --git a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
index 74b832b91043fe..bea81706b652a0 100644
--- a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
+++ b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
@@ -368,6 +368,14 @@ define amdgpu_kernel void @smfmac_i32_16x16x128_i8(<4 x i32> %arg0, <8 x i32> %a
   ret void
 }
 
+declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
+
+; CHECK: DIVERGENT:  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
+define amdgpu_kernel void @smfmac_f32_16x16x128_bf8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
+  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
+  store <4 x float> %result, ptr addrspace(1) %out
+  ret void
+}
 
 declare i32 @llvm.amdgcn.ds.swizzle(i32, i32) #1
 declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1) #1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
index 16d6e0dd3dd588..98b65803887129 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
@@ -2110,6 +2110,221 @@ define <16 x i32> @test_smfmac_i32_32x32x64_i8__sgpr(<4 x i32> inreg %arg0, <8 x
   ret <16 x i32> %result
 }
 
+; --------------------------------------------------------------------
+; llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8
+; --------------------------------------------------------------------
+
+declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32 immarg, i32 immarg)
+
+define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_bf8__vgpr(ptr addrspace(1) %arg, <4 x i32> %a, <8 x i32> %b, i32 %idx) #0 {
+; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__vgpr:
+; SDAG:       ; %bb.0: ; %bb
+; SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
+; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
+; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
+; SDAG-NEXT:    v_mov_b32_e32 v12, s8
+; SDAG-NEXT:    v_mov_b32_e32 v13, s9
+; SDAG-NEXT:    v_mov_b32_e32 v14, s10
+; SDAG-NEXT:    v_mov_b32_e32 v15, s11
+; SDAG-NEXT:    v_mov_b32_e32 v0, s12
+; SDAG-NEXT:    v_mov_b32_e32 v1, s13
+; SDAG-NEXT:    v_mov_b32_e32 v2, s14
+; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; SDAG-NEXT:    v_mov_b32_e32 v4, s0
+; SDAG-NEXT:    v_mov_b32_e32 v5, s1
+; SDAG-NEXT:    v_mov_b32_e32 v6, s2
+; SDAG-NEXT:    v_mov_b32_e32 v7, s3
+; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    s_waitcnt vmcnt(0)
+; SDAG-NEXT:    s_nop 0
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    s_nop 6
+; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    s_endpgm
+;
+; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__vgpr:
+; GISEL:       ; %bb.0: ; %bb
+; GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
+; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
+; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
+; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
+; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
+; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
+; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_nop 0
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GISEL-NEXT:    s_nop 5
+; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    s_endpgm
+bb:
+  %id = call i32 @llvm.amdgcn.workitem.id.x()
+  %gep = getelementptr <4 x float>, ptr addrspace(1) %arg, i32 %id
+  %in.1 = load <4 x float>, ptr addrspace(1) %gep
+  %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %a, <8 x i32> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
+  store <4 x float> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
+define <4 x float> @test_smfmac_f32_16x16x128_bf8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3) {
+; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8:
+; SDAG:       ; %bb.0:
+; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v12
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v13
+; SDAG-NEXT:    v_accvgpr_write_b32 a2, v14
+; SDAG-NEXT:    v_accvgpr_write_b32 a3, v15
+; SDAG-NEXT:    s_nop 1
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 a[0:3], v[0:3], v[4:11], v16
+; SDAG-NEXT:    s_nop 6
+; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
+; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
+; SDAG-NEXT:    v_accvgpr_read_b32 v2, a2
+; SDAG-NEXT:    v_accvgpr_read_b32 v3, a3
+; SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[12:15], v[0:3], v[4:11], v16
+; GISEL-NEXT:    s_nop 6
+; GISEL-NEXT:    v_mov_b32_e32 v0, v12
+; GISEL-NEXT:    v_mov_b32_e32 v1, v13
+; GISEL-NEXT:    v_mov_b32_e32 v2, v14
+; GISEL-NEXT:    v_mov_b32_e32 v3, v15
+; GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
+  ret <4 x float> %result
+}
+
+define <4 x float> @test_smfmac_f32_16x16x128_bf8_bf8__flags0(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3) {
+; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__flags0:
+; SDAG:       ; %bb.0:
+; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v12
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v13
+; SDAG-NEXT:    v_accvgpr_write_b32 a2, v14
+; SDAG-NEXT:    v_accvgpr_write_b32 a3, v15
+; SDAG-NEXT:    s_nop 1
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 a[0:3], v[0:3], v[4:11], v16 cbsz:1 abid:3
+; SDAG-NEXT:    s_nop 6
+; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
+; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
+; SDAG-NEXT:    v_accvgpr_read_b32 v2, a2
+; SDAG-NEXT:    v_accvgpr_read_b32 v3, a3
+; SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__flags0:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[12:15], v[0:3], v[4:11], v16 cbsz:1 abid:3
+; GISEL-NEXT:    s_nop 6
+; GISEL-NEXT:    v_mov_b32_e32 v0, v12
+; GISEL-NEXT:    v_mov_b32_e32 v1, v13
+; GISEL-NEXT:    v_mov_b32_e32 v2, v14
+; GISEL-NEXT:    v_mov_b32_e32 v3, v15
+; GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 1, i32 immarg 3)
+  ret <4 x float> %result
+}
+
+define <4 x float> @test_smfmac_f32_16x16x128_bf8_bf8__flags1(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3) {
+; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__flags1:
+; SDAG:       ; %bb.0:
+; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v12
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v13
+; SDAG-NEXT:    v_accvgpr_write_b32 a2, v14
+; SDAG-NEXT:    v_accvgpr_write_b32 a3, v15
+; SDAG-NEXT:    s_nop 1
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 a[0:3], v[0:3], v[4:11], v16 cbsz:3 abid:1
+; SDAG-NEXT:    s_nop 6
+; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
+; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
+; SDAG-NEXT:    v_accvgpr_read_b32 v2, a2
+; SDAG-NEXT:    v_accvgpr_read_b32 v3, a3
+; SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__flags1:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[12:15], v[0:3], v[4:11], v16 cbsz:3 abid:1
+; GISEL-NEXT:    s_nop 6
+; GISEL-NEXT:    v_mov_b32_e32 v0, v12
+; GISEL-NEXT:    v_mov_b32_e32 v1, v13
+; GISEL-NEXT:    v_mov_b32_e32 v2, v14
+; GISEL-NEXT:    v_mov_b32_e32 v3, v15
+; GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 3, i32 immarg 1)
+  ret <4 x float> %result
+}
+
+define <4 x float> @test_smfmac_f32_16x16x128_bf8_bf8__sgpr(<4 x i32> inreg %arg0, <8 x i32> inreg %arg1, <4 x float> inreg %arg2, i32 inreg %arg3) {
+; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__sgpr:
+; SDAG:       ; %bb.0:
+; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SDAG-NEXT:    v_mov_b32_e32 v8, s0
+; SDAG-NEXT:    v_mov_b32_e32 v9, s1
+; SDAG-NEXT:    v_mov_b32_e32 v10, s2
+; SDAG-NEXT:    v_mov_b32_e32 v11, s3
+; SDAG-NEXT:    v_mov_b32_e32 v0, s16
+; SDAG-NEXT:    v_mov_b32_e32 v1, s17
+; SDAG-NEXT:    v_mov_b32_e32 v2, s18
+; SDAG-NEXT:    v_mov_b32_e32 v3, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s20
+; SDAG-NEXT:    v_mov_b32_e32 v5, s21
+; SDAG-NEXT:    v_mov_b32_e32 v6, s22
+; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, s...
[truncated]

arsenm · 2024-11-22T00:53:28Z

Merge activity

Nov 21, 7:53 PM EST: A user started a stack merge that includes this pull request via Graphite.
Nov 21, 7:55 PM EST: Graphite rebased this pull request as part of a merge.
Nov 21, 7:57 PM EST: A user merged this pull request with Graphite.

arsenm added backend:AMDGPU clang Clang issues not falling into any other category clang:frontend Language frontend issues, e.g. anything involving "Sema" llvm:analysis llvm:ir mc Machine (object) code labels Nov 21, 2024 — with Graphite App

arsenm requested review from jayfoad, pravinjagtap, rampitec, shiltian, Sisyph and srpande November 21, 2024 20:33

arsenm mentioned this pull request Nov 21, 2024

AMDGPU: Add v_smfmac_f32_16x16x128_fp8_fp8 for gfx950 #117235

Merged

arsenm marked this pull request as ready for review November 21, 2024 20:34

shiltian approved these changes Nov 21, 2024

View reviewed changes

arsenm force-pushed the users/arsenm/gfx950/v_smfmac_i32_32x32x64_i8 branch from f10f99c to 7f20442 Compare November 21, 2024 22:58

Base automatically changed from users/arsenm/gfx950/v_smfmac_i32_32x32x64_i8 to main November 21, 2024 23:01

arsenm force-pushed the users/arsenm/gfx950/v_smfmac_f32_16x16x128_bf8_bf8 branch from 629e1aa to a8894cf Compare November 21, 2024 23:03

arsenm force-pushed the users/arsenm/gfx950/v_smfmac_f32_16x16x128_bf8_bf8 branch from a8894cf to 801c0d3 Compare November 21, 2024 23:34

AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950

e85d020

arsenm force-pushed the users/arsenm/gfx950/v_smfmac_f32_16x16x128_bf8_bf8 branch from 801c0d3 to e85d020 Compare November 22, 2024 00:54

arsenm merged commit 7baadb2 into main Nov 22, 2024
5 of 7 checks passed

arsenm deleted the users/arsenm/gfx950/v_smfmac_f32_16x16x128_bf8_bf8 branch November 22, 2024 00:57

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950 #117232

AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950 #117232

arsenm commented Nov 21, 2024

arsenm commented Nov 21, 2024 •

edited

Loading

llvmbot commented Nov 21, 2024 •

edited

Loading

arsenm commented Nov 22, 2024 •

edited

Loading

AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950 #117232

AMDGPU: Add v_smfmac_f32_16x16x128_bf8_bf8 for gfx950 #117232

Conversation

arsenm commented Nov 21, 2024

arsenm commented Nov 21, 2024 • edited Loading

llvmbot commented Nov 21, 2024 • edited Loading

arsenm commented Nov 22, 2024 • edited Loading

Merge activity

arsenm commented Nov 21, 2024 •

edited

Loading

llvmbot commented Nov 21, 2024 •

edited

Loading

arsenm commented Nov 22, 2024 •

edited

Loading