[PATCH] D146131: [AMDGPU] Avoid constant bus limitation on V_BFE GISel pattern

[Pierre van Houtryve via Phabricator via llvm-commits]

Pierre-vh updated this revision to Diff 505452.
Pierre-vh added a comment.

comment


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D146131/new/

https://reviews.llvm.org/D146131

Files:
  llvm/lib/Target/AMDGPU/VOP3Instructions.td
  llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sext.mir
  llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-zext.mir
  llvm/test/CodeGen/AMDGPU/v_bfe_i32.ll


Index: llvm/test/CodeGen/AMDGPU/v_bfe_i32.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/AMDGPU/v_bfe_i32.ll
@@ -0,0 +1,25 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -verify-machineinstrs < %s | FileCheck --check-prefix=PREGFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -mcpu=hawaii -verify-machineinstrs < %s | FileCheck --check-prefix=PREGFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -mcpu=fiji -verify-machineinstrs < %s | FileCheck --check-prefix=PREGFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck --check-prefix=PREGFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -mcpu=gfx1030 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10PLUS %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa --global-isel -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10PLUS %s
+
+define i32 @check_v_bfe(i16 %a) {
+; PREGFX9-LABEL: check_v_bfe:
+; PREGFX9:       ; %bb.0: ; %entry
+; PREGFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; PREGFX9-NEXT:    v_bfe_i32 v0, v0, 0, 16
+; PREGFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10PLUS-LABEL: check_v_bfe:
+; GFX10PLUS:       ; %bb.0: ; %entry
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10PLUS-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10PLUS-NEXT:    v_bfe_i32 v0, v0, 0, 16
+; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
+entry:
+  %res = sext i16 %a to i32
+  ret i32 %res
+}
Index: llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-zext.mir
===================================================================
--- llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-zext.mir
+++ llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-zext.mir
@@ -165,7 +165,9 @@
     ; GCN-NEXT: {{  $}}
     ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GCN-NEXT: [[V_AND_B32_e32_:%[0-9]+]]:vgpr_32 = V_AND_B32_e32 1, [[COPY]], implicit $exec
-    ; GCN-NEXT: [[V_BFE_I32_e64_:%[0-9]+]]:vgpr_32 = V_BFE_I32_e64 [[V_AND_B32_e32_]], 0, 16, implicit $exec
+    ; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 16, implicit $exec
+    ; GCN-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GCN-NEXT: [[V_BFE_I32_e64_:%[0-9]+]]:vgpr_32 = V_BFE_I32_e64 [[V_AND_B32_e32_]], [[V_MOV_B32_e32_1]], [[V_MOV_B32_e32_]], implicit $exec
     ; GCN-NEXT: $vgpr0 = COPY [[V_BFE_I32_e64_]]
     %0:vgpr(s32) = COPY $vgpr0
     %1:vgpr(s1) = G_TRUNC %0
Index: llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sext.mir
===================================================================
--- llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sext.mir
+++ llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sext.mir
@@ -208,7 +208,9 @@
     ; GCN: liveins: $vgpr0
     ; GCN-NEXT: {{  $}}
     ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
-    ; GCN-NEXT: [[V_BFE_I32_e64_:%[0-9]+]]:vgpr_32 = V_BFE_I32_e64 [[COPY]], 0, 16, implicit $exec
+    ; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 16, implicit $exec
+    ; GCN-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    ; GCN-NEXT: [[V_BFE_I32_e64_:%[0-9]+]]:vgpr_32 = V_BFE_I32_e64 [[COPY]], [[V_MOV_B32_e32_1]], [[V_MOV_B32_e32_]], implicit $exec
     ; GCN-NEXT: $vgpr0 = COPY [[V_BFE_I32_e64_]]
     %0:vgpr(s32) = COPY $vgpr0
     %1:vgpr(s16) = G_TRUNC %0
Index: llvm/lib/Target/AMDGPU/VOP3Instructions.td
===================================================================
--- llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -263,7 +263,7 @@
 
 def : GCNPat<
   (i32 (DivergentUnaryFrag<sext> i16:$src)),
-  (i32 (V_BFE_I32_e64 $src, (S_MOV_B32 (i32 0)), (S_MOV_B32 (i32 0x10))))
+  (i32 (V_BFE_I32_e64 i16:$src, (V_MOV_B32_e32 (i32 0)), (V_MOV_B32_e32 (i32 0x10))))
 >;
 
 let isReMaterializable = 1 in {


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D146131.505452.patch
Type: text/x-patch
Size: 4011 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230315/e5efbb01/attachment.bin>