[Mlir-commits] [mlir] d205624 - [nvvm] use check-next in nvvm-to-llvm test (nfc) (#68326)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Fri Oct 6 01:37:26 PDT 2023


Author: Guray Ozen
Date: 2023-10-06T10:37:23+02:00
New Revision: d2056247b56ec5801afe0c94d458bcd2e3aa90fc

URL: https://github.com/llvm/llvm-project/commit/d2056247b56ec5801afe0c94d458bcd2e3aa90fc
DIFF: https://github.com/llvm/llvm-project/commit/d2056247b56ec5801afe0c94d458bcd2e3aa90fc.diff

LOG: [nvvm] use check-next in nvvm-to-llvm test (nfc) (#68326)

Added: 
    

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

Removed: 
    


################################################################################
diff  --git a/mlir/test/Conversion/NVVMToLLVM/nvvm-to-llvm.mlir b/mlir/test/Conversion/NVVMToLLVM/nvvm-to-llvm.mlir
index 7ffe1ad2bb2b111..0092dca3506403c 100644
--- a/mlir/test/Conversion/NVVMToLLVM/nvvm-to-llvm.mlir
+++ b/mlir/test/Conversion/NVVMToLLVM/nvvm-to-llvm.mlir
@@ -6,7 +6,7 @@
 
 // CHECK-LABEL: @init_mbarrier_arrive_expect_tx
 llvm.func @init_mbarrier_arrive_expect_tx(%barrier : !llvm.ptr<3>, %txcount : i32) {
-  //CHECK: llvm.inline_asm has_side_effects asm_dialect = att "mbarrier.arrive.expect_tx.shared.b64 _, [$0], $1;", "r,r"
+  //CHECK: llvm.inline_asm has_side_effects asm_dialect = att "mbarrier.arrive.expect_tx.shared.b64 _, [$0], $1;", "r,r"  
   nvvm.mbarrier.arrive.expect_tx.shared %barrier, %txcount : !llvm.ptr<3>, i32
   llvm.return
 }
@@ -20,14 +20,32 @@ llvm.func @init_mbarrier_arrive_expect_tx_generic(%barrier : !llvm.ptr, %txcount
 
 // CHECK-LABEL: @init_mbarrier_try_wait_shared
 llvm.func @init_mbarrier_try_wait_shared(%barrier : !llvm.ptr<3>, %ticks : i32, %phase : i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "{\0A\09.reg .pred       P1; \0A\09LAB_WAIT: \0A\09mbarrier.try_wait.parity.shared.b64 P1, [$0], $1, $2; \0A\09 at P1 bra.uni DONE; \0A\09bra.uni     LAB_WAIT; \0A\09DONE: \0A\09}", "r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{
+  // CHECK-SAME: .reg .pred       P1;
+  // CHECK-SAME: LAB_WAIT: 
+  // CHECK-SAME: mbarrier.try_wait.parity.shared.b64 P1, [$0], $1, $2;
+  // CHECK-SAME: @P1 bra.uni DONE;
+  // CHECK-SAME: bra.uni     LAB_WAIT;
+  // CHECK-SAME: DONE:
+  // CHECK-SAME: }",
+  // CHECK-SAME: "r,r,r"
    nvvm.mbarrier.try_wait.parity.shared %barrier, %phase, %ticks : !llvm.ptr<3>, i32, i32
   llvm.return
 }
 
 // CHECK-LABEL: @init_mbarrier_try_wait
 llvm.func @init_mbarrier_try_wait(%barrier : !llvm.ptr, %ticks : i32, %phase : i32){
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "{\0A\09.reg .pred       P1; \0A\09LAB_WAIT: \0A\09mbarrier.try_wait.parity.b64 P1, [$0], $1, $2; \0A\09 at P1 bra.uni DONE; \0A\09bra.uni     LAB_WAIT; \0A\09DONE: \0A\09}", "l,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att
+  // CHECK-SAME: "{
+  // CHECK-SAME: .reg .pred       P1;
+  // CHECK-SAME: LAB_WAIT: 
+  // CHECK-SAME: mbarrier.try_wait.parity.b64 P1, [$0], $1, $2;
+  // CHECK-SAME: @P1 bra.uni DONE;
+  // CHECK-SAME: bra.uni     LAB_WAIT;
+  // CHECK-SAME: DONE:
+  // CHECK-SAME: }",
+  // CHECK-SAME: "l,r,r"
   nvvm.mbarrier.try_wait.parity %barrier, %phase, %ticks : !llvm.ptr, i32, i32
   llvm.return
 }
@@ -43,79 +61,93 @@ func.func @async_cp(%dst: !llvm.ptr<3>, %src: !llvm.ptr<1>) {
 
 // CHECK-LABEL: @async_cp_zfill
 func.func @async_cp_zfill(%dst: !llvm.ptr<3>, %src: !llvm.ptr<1>, %cpSize: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.cg.shared.global [$0], [$1], $2, $3;\0A", "r,l,n,r" %{{.*}}, %{{.*}}, %{{.*}} : (!llvm.ptr<3>, !llvm.ptr<1>, i32, i32) -> ()
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.cg.shared.global [$0], [$1], $2, $3;\0A", 
+  // CHECK-SAME: "r,l,n,r" %{{.*}}, %{{.*}}, %{{.*}} : (!llvm.ptr<3>, !llvm.ptr<1>, i32, i32) -> ()
   nvvm.cp.async.shared.global %dst, %src, 16, cache =  cg, %cpSize : !llvm.ptr<3>, !llvm.ptr<1>, i32
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.ca.shared.global [$0], [$1], $2, $3;\0A", "r,l,n,r" %{{.*}}, %{{.*}}, %{{.*}} : (!llvm.ptr<3>, !llvm.ptr<1>, i32, i32) -> ()
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.ca.shared.global [$0], [$1], $2, $3;\0A", 
+  // CHECK-SAME: "r,l,n,r" %{{.*}}, %{{.*}}, %{{.*}} : (!llvm.ptr<3>, !llvm.ptr<1>, i32, i32) -> ()
   nvvm.cp.async.shared.global %dst, %src, 4, cache =  ca, %cpSize : !llvm.ptr<3>, !llvm.ptr<1>, i32
   return
 }
 
 // CHECK-LABEL: @tma_load_1d
 func.func @tma_load_1d(%tmaDescriptor: !llvm.ptr, %dest : !llvm.ptr<3>, %barrier: !llvm.ptr<3>, %crd0: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.1d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3} ], [$2];", "r,l,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.1d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3} ], [$2];", "r,l,r,r"
   nvvm.cp.async.bulk.tensor.shared.cluster.global %dest, %tmaDescriptor,  %barrier, box[%crd0] : !llvm.ptr<3>, !llvm.ptr, !llvm.ptr<3>, i32
   return
 }
 
 // CHECK-LABEL: @tma_load_2d
 func.func @tma_load_2d(%tmaDescriptor: !llvm.ptr, %dest : !llvm.ptr<3>, %barrier: !llvm.ptr<3>, %crd0: i32, %crd1: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.2d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4} ], [$2];", "r,l,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.2d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4} ], [$2];", "r,l,r,r,r"
   nvvm.cp.async.bulk.tensor.shared.cluster.global %dest, %tmaDescriptor,  %barrier, box[%crd0,%crd1] : !llvm.ptr<3>, !llvm.ptr, !llvm.ptr<3>, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_load_3d
 func.func @tma_load_3d(%tmaDescriptor: !llvm.ptr, %dest : !llvm.ptr<3>, %barrier: !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.3d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5} ], [$2];", "r,l,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.3d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5} ], [$2];", "r,l,r,r,r,r"
   nvvm.cp.async.bulk.tensor.shared.cluster.global %dest, %tmaDescriptor,  %barrier, box[%crd0,%crd1,%crd2] : !llvm.ptr<3>, !llvm.ptr, !llvm.ptr<3>, i32, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_load_4d
 func.func @tma_load_4d(%tmaDescriptor: !llvm.ptr, %dest : !llvm.ptr<3>, %barrier: !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32, %crd3: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.4d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5, $6} ], [$2];", "r,l,r,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.4d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5, $6} ], [$2];", "r,l,r,r,r,r,r"
   nvvm.cp.async.bulk.tensor.shared.cluster.global %dest, %tmaDescriptor,  %barrier, box[%crd0,%crd1,%crd2,%crd3] : !llvm.ptr<3>, !llvm.ptr, !llvm.ptr<3>, i32, i32, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_load_5d
 func.func @tma_load_5d(%tmaDescriptor: !llvm.ptr, %dest : !llvm.ptr<3>, %barrier: !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32, %crd3: i32, %crd4: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.5d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5, $6, $7} ], [$2];", "r,l,r,r,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.5d.shared::cluster.global.mbarrier::complete_tx::bytes [$0], [$1, {$3, $4, $5, $6, $7} ], [$2];", "r,l,r,r,r,r,r,r"
   nvvm.cp.async.bulk.tensor.shared.cluster.global %dest, %tmaDescriptor,  %barrier, box[%crd0,%crd1,%crd2,%crd3,%crd4] : !llvm.ptr<3>, !llvm.ptr, !llvm.ptr<3>, i32, i32, i32, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_store_1d
 func.func @tma_store_1d(%tmaDescriptor: !llvm.ptr, %src : !llvm.ptr<3>, %crd0: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.1d.global.shared::cta.bulk_group [$0, {$2} ], [$1];", "l,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.1d.global.shared::cta.bulk_group [$0, {$2} ], [$1];", "l,r,r"
   nvvm.cp.async.bulk.tensor.global.shared.cta %tmaDescriptor, %src, box[%crd0] : !llvm.ptr, !llvm.ptr<3>, i32
   return
 }
 
 // CHECK-LABEL: @tma_store_2d
 func.func @tma_store_2d(%tmaDescriptor: !llvm.ptr, %src : !llvm.ptr<3>, %crd0: i32, %crd1: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.2d.global.shared::cta.bulk_group [$0, {$2, $3} ], [$1];", "l,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.2d.global.shared::cta.bulk_group [$0, {$2, $3} ], [$1];", "l,r,r,r"
   nvvm.cp.async.bulk.tensor.global.shared.cta %tmaDescriptor, %src, box[%crd0,%crd1] : !llvm.ptr, !llvm.ptr<3>, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_store_3d
 func.func @tma_store_3d(%tmaDescriptor: !llvm.ptr, %src : !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.3d.global.shared::cta.bulk_group [$0, {$2, $3, $4} ], [$1];", "l,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.3d.global.shared::cta.bulk_group [$0, {$2, $3, $4} ], [$1];", "l,r,r,r,r"
   nvvm.cp.async.bulk.tensor.global.shared.cta %tmaDescriptor, %src, box[%crd0,%crd1,%crd2] : !llvm.ptr, !llvm.ptr<3>, i32, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_store_4d
 func.func @tma_store_4d(%tmaDescriptor: !llvm.ptr, %src : !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32, %crd3: i32) {
-  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.4d.global.shared::cta.bulk_group [$0, {$2, $3, $4, $5} ], [$1];", "l,r,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.4d.global.shared::cta.bulk_group [$0, {$2, $3, $4, $5} ], [$1];", "l,r,r,r,r,r"
   nvvm.cp.async.bulk.tensor.global.shared.cta %tmaDescriptor, %src, box[%crd0,%crd1,%crd2,%crd3] : !llvm.ptr, !llvm.ptr<3>, i32, i32, i32, i32
   return
 }
 
 // CHECK-LABEL: @tma_store_5d
 func.func @tma_store_5d(%tmaDescriptor: !llvm.ptr, %src : !llvm.ptr<3>, %crd0: i32, %crd1: i32, %crd2: i32, %crd3: i32, %crd4: i32) {
-  // CHECK-NEXT: llvm.inline_asm has_side_effects asm_dialect = att "cp.async.bulk.tensor.5d.global.shared::cta.bulk_group [$0, {$2, $3, $4, $5, $6} ], [$1];", "l,r,r,r,r,r,r"
+  // CHECK: llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "cp.async.bulk.tensor.5d.global.shared::cta.bulk_group [$0, {$2, $3, $4, $5, $6} ], [$1];", "l,r,r,r,r,r,r"
   nvvm.cp.async.bulk.tensor.global.shared.cta %tmaDescriptor, %src, box[%crd0,%crd1,%crd2,%crd3,%crd4] : !llvm.ptr, !llvm.ptr<3>, i32, i32, i32, i32, i32
   return
 }
@@ -161,7 +193,16 @@ func.func @wgmma_f32_f16_f16(%descA : i64, %descB : i64) -> !mat64f32{
   // CHECK: %[[V4:.*]] = llvm.extractvalue %[[RES]][4] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)> 
   // CHECK: %[[V11:.*]] = llvm.extractvalue %[[RES]][11] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>  
   // CHECK: %[[V13:.*]] = llvm.extractvalue %[[RES]][13] : !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)> 
-  // 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)>
+  // CHECK: %[[RES1:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{
+  // CHECK-SAME: reg .pred p;
+  // CHECK-SAME: setp.ne.b32 p, $34, 0;
+  // CHECK-SAME: wgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 
+  // CHECK-SAME: {$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", 
+  // CHECK-SAME: "=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" 
+  // CHECK-SAME: %[[V0]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11]], %{{.*}}, %[[V13]], %{{.*}}, %{{.*}}, %[[ARG0]], %[[ARG1]], %[[A1]], %[[A2]], %[[A3]], %[[A4]], %[[A5]] 
+  // CHECK-SAME: : (f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, i64, i64, i32, i32, i32, i32, i32) 
+  // CHECK-SAME: -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32)>
   // CHECK: %[[C2:.*]] = llvm.mlir.constant(2 : i64) : i64
   // CHECK: %[[DESCa:.+]] = llvm.add %[[ARG0]], %[[C2]] : i64
   // CHECK: %[[DESCb:.+]] = llvm.add %[[ARG1]], %[[C2]] : i64
@@ -169,7 +210,14 @@ func.func @wgmma_f32_f16_f16(%descA : i64, %descB : i64) -> !mat64f32{
   // 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)> 
   // 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)>  
   // 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)> 
-  // 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)>
+  // CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{
+    // CHECK-SAME: .reg .pred p;
+    // CHECK-SAME: setp.ne.b32 p, $34, 0;
+    // CHECK-SAME: wgmma.mma_async.sync.aligned.m64n32k16.f32.f16.f16 
+    // CHECK-SAME: {$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", 
+    // CHECK-SAME: "=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" 
+    // CHECK-SAME: %[[V0_2]], %{{.*}}, %{{.*}}, %{{.*}}, %[[V4_2]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %[[V11_2]], %{{.*}}, %[[V13_2]], %{{.*}}, %{{.*}}, %[[DESCa]], %[[DESCb]], %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}} 
   %result = llvm.mlir.undef : !mat64f32
   %result1 = nvvm.wgmma.mma_async 
       %descA, %descB, 
@@ -205,17 +253,37 @@ func.func @wgmma_s32_s8_s8_satfinite(%descA : i64, %descB : i64) -> !mat16i32{
 // CHECK: %[[V1:.*]] = llvm.extractvalue %[[RES]][1]
 // CHECK: %[[V2:.*]] = llvm.extractvalue %[[RES]][2]
 // CHECK: %[[V3:.*]] = llvm.extractvalue %[[RES]][3]
-// 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)>
+// CHECK: %[[RES_2:.*]] =  llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME: "{
+// CHECK-SAME: .reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite 
+// CHECK-SAME: {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" 
+// CHECK-SAME: %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : 
+// CHECK-SAME: (i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
 // CHECK: %[[V0_2:.*]] = llvm.extractvalue %[[RES_2]][0]
 // CHECK: %[[V1_2:.*]] = llvm.extractvalue %[[RES_2]][1]
 // CHECK: %[[V2_2:.*]] = llvm.extractvalue %[[RES_2]][2]
 // CHECK: %[[V3_2:.*]] = llvm.extractvalue %[[RES_2]][3]
-// 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]], %{{.*}}
+// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME: "{
+// CHECK-SAME: .reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite 
+// CHECK-SAME: {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", 
+// CHECK-SAME: "=r,=r,=r,=r,0,1,2,3,l,l,n" 
+// CHECK-SAME: %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
 // CHECK: %[[V0_3:.*]] = llvm.extractvalue %[[RES_3]][0]
 // CHECK: %[[V1_3:.*]] = llvm.extractvalue %[[RES_3]][1]
 // CHECK: %[[V2_3:.*]] = llvm.extractvalue %[[RES_3]][2]
 // CHECK: %[[V3_3:.*]] = llvm.extractvalue %[[RES_3]][3]
-// 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]], %{{.*}} 
+// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME:"{
+// CHECK-SAME:.reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.s8.s8.satfinite
+// CHECK-SAME: {$0, $1, $2, $3}, $8, $9, p;\0A}\0A", "=r,=r,=r,=r,0,1,2,3,l,l,n" 
+// CHECK-SAME: %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}} 
   %result1 = nvvm.wgmma.mma_async %descA, %descB, 
       #nvvm.shape<m = 64, n = 8, k = 32>, 
       D [%result, #nvvm.wgmma_scale_out<one>, <satfinite>],
@@ -246,17 +314,36 @@ func.func @wgmma_s32_u8_u8(%descA : i64, %descB : i64) -> !mat16i32 {
 // CHECK: %[[V1:.*]] = llvm.extractvalue %[[RES]][1]
 // CHECK: %[[V2:.*]] = llvm.extractvalue %[[RES]][2]
 // CHECK: %[[V3:.*]] = llvm.extractvalue %[[RES]][3]
-// 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)>
+// CHECK: %[[RES_2:.*]] =  llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME: "{
+// CHECK-SAME: .reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;
+// CHECK-SAME: }\0A",
+// CHECK-SAME: "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0]], %[[V1]], %[[V2]], %[[V3]], %[[ARG0]], %[[ARG1]], %[[A1]] : 
+// CHECK-SAME:(i32, i32, i32, i32, i64, i64, i32) -> !llvm.struct<(i32, i32, i32, i32)>
 // CHECK: %[[V0_2:.*]] = llvm.extractvalue %[[RES_2]][0]
 // CHECK: %[[V1_2:.*]] = llvm.extractvalue %[[RES_2]][1]
 // CHECK: %[[V2_2:.*]] = llvm.extractvalue %[[RES_2]][2]
 // CHECK: %[[V3_2:.*]] = llvm.extractvalue %[[RES_2]][3]
-// 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]], %{{.*}}
+// CHECK: %[[RES_3:.*]] = llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME:"{
+// CHECK-SAME: .reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;
+// CHECK-SAME: }\0A",
+// CHECK-SAME: "=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_2]], %[[V1_2]], %[[V2_2]], %[[V3_2]], %[[ARG0]], %[[ARG1]], %{{.*}}
 // CHECK: %[[V0_3:.*]] = llvm.extractvalue %[[RES_3]][0]
 // CHECK: %[[V1_3:.*]] = llvm.extractvalue %[[RES_3]][1]
 // CHECK: %[[V2_3:.*]] = llvm.extractvalue %[[RES_3]][2]
 // CHECK: %[[V3_3:.*]] = llvm.extractvalue %[[RES_3]][3]
-// 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]], %{{.*}} 
+// CHECK: %[[RES1:.*]] = llvm.inline_asm has_side_effects asm_dialect = att 
+// CHECK-SAME:"{
+// CHECK-SAME: .reg .pred p;
+// CHECK-SAME: setp.ne.b32 p, $10, 0;
+// CHECK-SAME: wgmma.mma_async.sync.aligned.m64n8k32.s32.u8.u8 {$0, $1, $2, $3}, $8, $9, p;
+// CHECK-SAME:}\0A", 
+// CHECK-SAME:"=r,=r,=r,=r,0,1,2,3,l,l,n" %[[V0_3]], %[[V1_3]], %[[V2_3]], %[[V3_3]], %[[ARG0]], %[[ARG1]], %{{.*}} 
   %result = llvm.mlir.undef : !mat16i32
   %result1 = nvvm.wgmma.mma_async %descA, %descB, 
       #nvvm.shape<m = 64, n = 8, k = 32>, 
@@ -289,8 +376,16 @@ func.func @wgmma_s32_u8_u8(%descA : i64, %descB : i64) -> !mat16i32 {
 
 // CHECK-LABEL: @wgmma_f32_tf32_tf32
 func.func @wgmma_f32_tf32_tf32(%descA : i64, %descB : i64) -> !mat32f32 {  
-  // 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"
-  // 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"
+  // CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME:"{
+  // CHECK-SAME: .reg .pred p;
+  // CHECK-SAME: setp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
+  // CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{
+  // CHECK-SAME: .reg .pred p;
+  // CHECK-SAME: setp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
   %result = llvm.mlir.undef : !mat32f32
   %result1 = nvvm.wgmma.mma_async %descA, %descB, 
       #nvvm.shape<m = 64, n = 64, k = 8>, 
@@ -318,8 +413,12 @@ func.func @wgmma_f32_tf32_tf32(%descA : i64, %descB : i64) -> !mat32f32 {
 
 // CHECK-LABEL: @wgmma_f32_e4m3_e4m3
 func.func @wgmma_f32_e4m3_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {  
-  // 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"
-  // 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"
+  // CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
+  // CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
   %result = llvm.mlir.undef : !mat32f32
   %result1 = nvvm.wgmma.mma_async %descA, %descB, 
       #nvvm.shape<m = 64, n = 64, k = 32>, 
@@ -346,8 +445,12 @@ func.func @wgmma_f32_e4m3_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {
 
 // CHECK-LABEL: @wgmma_f32_e5m2_e4m3
 func.func @wgmma_f32_e5m2_e4m3(%descA : i64, %descB : i64) -> !mat32f32 {  
-  // 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"
-  // 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"
+  // CHECK: %[[RES:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
+  // CHECK: %[[RES_2:.+]] = llvm.inline_asm has_side_effects asm_dialect = att 
+  // CHECK-SAME: "{\0A.reg .pred p;\0Asetp.ne.b32 p, $66, 0;
+  // CHECK-SAME: wgmma.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"
   %result = llvm.mlir.undef : !mat32f32
   %result1 = nvvm.wgmma.mma_async %descA, %descB, 
       #nvvm.shape<m = 64, n = 64, k = 32>, 


        


More information about the Mlir-commits mailing list