[llvm] 7e681ef - [AMDGPU] Add GFX11 codegen for llvm.amdgcn.mov.dpp8

[Jay Foad via llvm-commits]

Author: Jay Foad
Date: 2022-06-16T19:44:28+01:00
New Revision: 7e681ef35e09abc5b4c5294a0ea5df62fc0345ae

URL: https://github.com/llvm/llvm-project/commit/7e681ef35e09abc5b4c5294a0ea5df62fc0345ae
DIFF: https://github.com/llvm/llvm-project/commit/7e681ef35e09abc5b4c5294a0ea5df62fc0345ae.diff

LOG: [AMDGPU] Add GFX11 codegen for llvm.amdgcn.mov.dpp8

Differential Revision: https://reviews.llvm.org/D127980

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/VOP1Instructions.td
    llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp8.ll

Removed: 
    llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp8.ll


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/VOP1Instructions.td b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
index d735110e2e17..f307155df6fe 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -1111,3 +1111,15 @@ def : GCNPat <
                         (as_i32timm $dpp8), (i32 DPP8Mode.FI_0))
 >;
 } // End OtherPredicates = [isGFX10Only]
+
+//===----------------------------------------------------------------------===//
+// GFX11
+//===----------------------------------------------------------------------===//
+
+let OtherPredicates = [isGFX11Only] in {
+def : GCNPat <
+  (i32 (int_amdgcn_mov_dpp8 i32:$src, timm:$dpp8)),
+  (V_MOV_B32_dpp8_gfx11 VGPR_32:$src, VGPR_32:$src,
+                        (as_i32timm $dpp8), (i32 DPP8Mode.FI_0))
+>;
+} // End OtherPredicates = [isGFX11Only]

diff  --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp8.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp8.ll
deleted file mode 100644
index 48ffc9a2e99d..000000000000
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp8.ll
+++ /dev/null
@@ -1 +0,0 @@
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-flat-for-global -verify-machineinstrs < %S/../llvm.amdgcn.mov.dpp8.ll | FileCheck -check-prefix=GFX10 %S/../llvm.amdgcn.mov.dpp8.ll

diff  --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp8.ll
index d77a7a1b38df..dbc0721a6ca4 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp8.ll
@@ -1,19 +1,22 @@
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
 
-; GFX10-LABEL: {{^}}dpp8_test:
-; GFX10: v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
-; GFX10: v_mov_b32_dpp [[SRC]], [[SRC]]  dpp8:[1,0,0,0,0,0,0,0]{{$}}
+; GFX10PLUS-LABEL: {{^}}dpp8_test:
+; GFX10PLUS: v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
+; GFX10PLUS: v_mov_b32_dpp [[SRC]], [[SRC]]  dpp8:[1,0,0,0,0,0,0,0]{{$}}
 define amdgpu_kernel void @dpp8_test(i32 addrspace(1)* %out, i32 %in) {
   %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
   store i32 %tmp0, i32 addrspace(1)* %out
   ret void
 }
 
-; GFX10-LABEL: {{^}}dpp8_wait_states:
-; GFX10-NOOPT: v_mov_b32_e32 [[VGPR1:v[0-9]+]], s{{[0-9]+}}
-; GFX10: v_mov_b32_e32 [[VGPR0:v[0-9]+]], s{{[0-9]+}}
-; GFX10: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[1,0,0,0,0,0,0,0]{{$}}
-; GFX10: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[5,0,0,0,0,0,0,0]{{$}}
+; GFX10PLUS-LABEL: {{^}}dpp8_wait_states:
+; GFX10PLUS-NOOPT: v_mov_b32_e32 [[VGPR1:v[0-9]+]], s{{[0-9]+}}
+; GFX10PLUS: v_mov_b32_e32 [[VGPR0:v[0-9]+]], s{{[0-9]+}}
+; GFX10PLUS: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[1,0,0,0,0,0,0,0]{{$}}
+; GFX10PLUS: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[5,0,0,0,0,0,0,0]{{$}}
 define amdgpu_kernel void @dpp8_wait_states(i32 addrspace(1)* %out, i32 %in) {
   %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
   %tmp1 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %tmp0, i32 5) #0