squiid/llvm
1
0
Fork 0
Code Issues Pull requests Activity

[AMDGPU] Updated some tests to run on gfx940. NFC.

Browse source
This commit is contained in:
Stanislav Mekhanoshin 2022-03-10 12:32:38 -08:00
parent 895653d904
commit c7f25b6fd4
5 changed files with 18 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.ll
View file

@ -1,11 +1,12 @@
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX906
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9,GFX906
; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9,GFX940
; RUN: llc -march=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
; RUN: llc -march=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
declare float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float %c, i1 %clamp)
; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_clamp
; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX9: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX10: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_fdot2_clamp(
float addrspace(1)* %r,
@ -23,6 +24,7 @@ entry:
; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_no_clamp
; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX10: v_dot2c_f32_f16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_fdot2_no_clamp(
float addrspace(1)* %r,
@ -38,8 +40,9 @@ entry:
ret void
}
; GFX906-LABEL: {{^}}fdot2_inline_literal
; GFX9-LABEL: {{^}}fdot2_inline_literal
; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 1.0
; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
define float @fdot2_inline_literal(<2 x half> %a, <2 x half> %b) {
%ret = tail call float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float 1.0, i1 false)
ret float %ret

7
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.udot4.ll
View file

@ -1,11 +1,12 @@
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX906
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
; RUN: llc -march=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
; RUN: llc -march=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
declare i32 @llvm.amdgcn.udot4(i32 %a, i32 %b, i32 %c, i1 %clamp)
; GCN-LABEL: {{^}}test_llvm_amdgcn_udot4_clamp
; GFX906: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX9: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX10: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_udot4_clamp(
i32 addrspace(1)* %r,
@ -24,7 +25,7 @@ entry:
}
; GCN-LABEL: {{^}}test_llvm_amdgcn_udot4_no_clamp
; GFX906: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX9: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX10: v_dot4_u32_u8 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_udot4_no_clamp(
i32 addrspace(1)* %r,

7
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.udot8.ll
View file

@ -1,11 +1,12 @@
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX906
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
; RUN: llc -march=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
; RUN: llc -march=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
declare i32 @llvm.amdgcn.udot8(i32 %a, i32 %b, i32 %c, i1 %clamp)
; GCN-LABEL: {{^}}test_llvm_amdgcn_udot8_clamp
; GFX906: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX9: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
; GFX10: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_udot8_clamp(
i32 addrspace(1)* %r,
@ -24,7 +25,7 @@ entry:
}
; GCN-LABEL: {{^}}test_llvm_amdgcn_udot8_no_clamp
; GFX906: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX9: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
; GFX10: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @test_llvm_amdgcn_udot8_no_clamp(
i32 addrspace(1)* %r,

2
llvm/test/CodeGen/AMDGPU/mfma-cd-select.ll
View file

@ -1,6 +1,8 @@
; RUN: llc -march=amdgcn -mcpu=gfx908 -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX908 %s
; RUN: llc -march=amdgcn -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float, float, <32 x float>, i32, i32, i32)

2
llvm/test/CodeGen/AMDGPU/mfma-vgpr-cd-select.ll
View file

@ -1,5 +1,7 @@
; RUN: llc -march=amdgcn -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN %s
; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck --enable-var-scope --check-prefixes=GCN %s
declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float, float, <32 x float>, i32, i32, i32)
declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float, float, <16 x float>, i32, i32, i32)