Skip to content

Commit 6360d09

Browse files
gryppgrypp
authored
[MLIR][NVVM] Fix the register number of predicate (#65970)
The register number of predicate is calculated incorrectly. This PR fixes that.
1 parent d2c2fa9 commit 6360d09

File tree

2 files changed

+15
-15
lines changed

2 files changed

+15
-15
lines changed

‎mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -903,7 +903,7 @@ std::string NVVM::WgmmaMmaAsyncOp::getPtx() {
903903
ss << "{\n"
904904
".reg .pred p;\n"
905905
"setp.ne.b32 p, $"
906-
<< (expectedOutputRegisters + 2)
906+
<< ((expectedOutputRegisters * 2) + 2)
907907
<< ", 0;\n"
908908
"wgmma.mma_async.sync.aligned.m"
909909
<< m << "n" << n << "k" << k << "." << outputTypeName << "."

‎mlir/test/Conversion/NVVMToLLVM/nvvm-to-llvm.mlir

Lines changed: 14 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -124,15 +124,15 @@ func.func @wgmma_f32_f16_f16(%descA : i64, %descB : i64) -> !mat64f32{
124124
// CHECK: %[[V4:.*]] = llvm.extractvalue %[[RES]][4] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
125125
// CHECK: %[[V11:.*]] = llvm.extractvalue %[[RES]][11] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
126126
// CHECK: %[[V13:.*]] = llvm.extractvalue %[[RES]][13] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
127-
// CHECK: %[[RES1:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $18, 0;\0Awgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15}, $32, $33, p, $35, $36, $37, $38;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,l,l,n,n,n,n,n" %[[V0]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11]], %{{.*}}, %[[V13]], %{{.*}}, %{{.*}}, %[[ARG0]], %[[ARG1]], %[[A1]], %[[A2]], %[[A3]], %[[A4]], %[[A5]] : (f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, i64, i64, i32, i32, i32, i32, i32) -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
127+
// CHECK: %[[RES1:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15}, $32, $33, p, $35, $36, $37, $38;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,l,l,n,n,n,n,n" %[[V0]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11]], %{{.*}}, %[[V13]], %{{.*}}, %{{.*}}, %[[ARG0]], %[[ARG1]], %[[A1]], %[[A2]], %[[A3]], %[[A4]], %[[A5]] : (f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, i64, i64, i32, i32, i32, i32, i32) -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
128128
// CHECK: %[[C2:.*]] = llvm.mlir.constant(2 : i64) : i64
129129
// CHECK: %[[DESCa:.+]] = llvm.add %[[ARG0]], %[[C2]] : i64
130130
// CHECK: %[[DESCb:.+]] = llvm.add %[[ARG1]], %[[C2]] : i64
131131
// CHECK: %[[V0_2:.*]] = llvm.extractvalue %[[RES1]][0] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
132132
// CHECK: %[[V4_2:.*]] = llvm.extractvalue %[[RES1]][4] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
133133
// CHECK: %[[V11_2:.*]] = llvm.extractvalue %[[RES1]][11] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
134134
// CHECK: %[[V13_2:.*]] = llvm.extractvalue %[[RES1]][13] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
135-
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $18, 0;\0Awgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15}, $32, $33, p, $35, $36, $37, $38;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,l,l,n,n,n,n,n" %[[V0_2]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4_2]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11_2]], %{{.*}}, %[[V13_2]], %{{.*}}, %{{.*}}, %[[DESCa]], %[[DESCb]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}} : (f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, i64, i64, i32, i32, i32, i32, i32) -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
135+
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15}, $32, $33, p, $35, $36, $37, $38;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,l,l,n,n,n,n,n" %[[V0_2]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4_2]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11_2]], %{{.*}}, %[[V13_2]], %{{.*}}, %{{.*}}, %[[DESCa]], %[[DESCb]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}} : (f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, i64, i64, i32, i32, i32, i32, i32) -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
136136
%result = llvm.mlir.undef : !mat64f32
137137
%result1 = nvvm.wgmma.mma_async
138138
%descA, %descB,
@@ -168,17 +168,17 @@ func.func @wgmma_s32_s8_s8_satfinite(%descA : i64, %descB : i64) -> !mat16i32{
168168
// CHECK: %[[V1:.*]] = llvm.extractvalue %[[RES]][1]
169169
// CHECK: %[[V2:.*]] = llvm.extractvalue %[[RES]][2]
170170
// CHECK: %[[V3:.*]] = llvm.extractvalue %[[RES]][3]
171-
// CHECK: %[[RES_2:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : (i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
171+
// CHECK: %[[RES_2:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : (i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
172172
// CHECK: %[[V0_2:.*]] = llvm.extractvalue %[[RES_2]][0]
173173
// CHECK: %[[V1_2:.*]] = llvm.extractvalue %[[RES_2]][1]
174174
// CHECK: %[[V2_2:.*]] = llvm.extractvalue %[[RES_2]][2]
175175
// CHECK: %[[V3_2:.*]] = llvm.extractvalue %[[RES_2]][3]
176-
// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
176+
// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
177177
// CHECK: %[[V0_3:.*]] = llvm.extractvalue %[[RES_3]][0]
178178
// CHECK: %[[V1_3:.*]] = llvm.extractvalue %[[RES_3]][1]
179179
// CHECK: %[[V2_3:.*]] = llvm.extractvalue %[[RES_3]][2]
180180
// CHECK: %[[V3_3:.*]] = llvm.extractvalue %[[RES_3]][3]
181-
// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}}
181+
// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}}
182182
%result1 = nvvm.wgmma.mma_async %descA, %descB,
183183
#nvvm.shape<m = 64, n = 8, k = 32>,
184184
D [%result, #nvvm.wgmma_scale_out<one>, <satfinite>],
@@ -209,17 +209,17 @@ func.func @wgmma_s32_u8_u8(%descA : i64, %descB : i64) -> !mat16i32 {
209209
// CHECK: %[[V1:.*]] = llvm.extractvalue %[[RES]][1]
210210
// CHECK: %[[V2:.*]] = llvm.extractvalue %[[RES]][2]
211211
// CHECK: %[[V3:.*]] = llvm.extractvalue %[[RES]][3]
212-
// CHECK: %[[RES_2:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : (i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
212+
// CHECK: %[[RES_2:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : (i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
213213
// CHECK: %[[V0_2:.*]] = llvm.extractvalue %[[RES_2]][0]
214214
// CHECK: %[[V1_2:.*]] = llvm.extractvalue %[[RES_2]][1]
215215
// CHECK: %[[V2_2:.*]] = llvm.extractvalue %[[RES_2]][2]
216216
// CHECK: %[[V3_2:.*]] = llvm.extractvalue %[[RES_2]][3]
217-
// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
217+
// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
218218
// CHECK: %[[V0_3:.*]] = llvm.extractvalue %[[RES_3]][0]
219219
// CHECK: %[[V1_3:.*]] = llvm.extractvalue %[[RES_3]][1]
220220
// CHECK: %[[V2_3:.*]] = llvm.extractvalue %[[RES_3]][2]
221221
// CHECK: %[[V3_3:.*]] = llvm.extractvalue %[[RES_3]][3]
222-
// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $6, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}}
222+
// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $10, 0;\0Awgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}}
223223
%result = llvm.mlir.undef : !mat16i32
224224
%result1 = nvvm.wgmma.mma_async %descA, %descB,
225225
#nvvm.shape<m = 64, n = 8, k = 32>,
@@ -252,8 +252,8 @@ func.func @wgmma_s32_u8_u8(%descA : i64, %descB : i64) -> !mat16i32 {
252252

253253
// CHECK-LABEL: @wgmma_f32_tf32_tf32
254254
func.func @wgmma_f32_tf32_tf32(%descA : i64, %descB : i64) -> !mat32f32 {
255-
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k8.f32.tf32.tf32 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
256-
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k8.f32.tf32.tf32 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
255+
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k8.f32.tf32.tf32 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
256+
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k8.f32.tf32.tf32 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
257257
%result = llvm.mlir.undef : !mat32f32
258258
%result1 = nvvm.wgmma.mma_async %descA, %descB,
259259
#nvvm.shape<m = 64, n = 64, k = 8>,
@@ -281,8 +281,8 @@ func.func @wgmma_f32_tf32_tf32(%descA : i64, %descB : i64) -> !mat32f32 {
281281

282282
// CHECK-LABEL: @wgmma_f32_e4m3_e4m3
283283
func.func @wgmma_f32_e4m3_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {
284-
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e4m3.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
285-
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e4m3.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
284+
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e4m3.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
285+
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e4m3.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
286286
%result = llvm.mlir.undef : !mat32f32
287287
%result1 = nvvm.wgmma.mma_async %descA, %descB,
288288
#nvvm.shape<m = 64, n = 64, k = 32>,
@@ -309,8 +309,8 @@ func.func @wgmma_f32_e4m3_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {
309309

310310
// CHECK-LABEL: @wgmma_f32_e5m2_e4m3
311311
func.func @wgmma_f32_e5m2_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {
312-
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e5m2.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
313-
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $34, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e5m2.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
312+
// CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e5m2.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
313+
// CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;\0Awgmma.mma_async.sync.aligned.m64n64k32.f32.e5m2.e4m3 {$0, $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16, $17, $18, $19, $20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31}, $64, $65, p, $67, $68;\0A}\0A", "=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,=f,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,l,l,n,n,n"
314314
%result = llvm.mlir.undef : !mat32f32
315315
%result1 = nvvm.wgmma.mma_async %descA, %descB,
316316
#nvvm.shape<m = 64, n = 64, k = 32>,

0 commit comments

Comments
 (0)