130 lines
3.5 KiB
Plaintext
130 lines
3.5 KiB
Plaintext
|
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx2 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX2
|
||
|
|
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f,+avx512vl -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL
|
||
|
|
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f,+avx512vl,+avx512bw -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512BWVL
|
||
|
|
|
||
|
|
--- |
|
||
|
|
define <32 x i8> @test_sub_v32i8(<32 x i8> %arg1, <32 x i8> %arg2) {
|
||
|
|
%ret = sub <32 x i8> %arg1, %arg2
|
||
|
|
ret <32 x i8> %ret
|
||
|
|
}
|
||
|
|
|
||
|
|
define <16 x i16> @test_sub_v16i16(<16 x i16> %arg1, <16 x i16> %arg2) {
|
||
|
|
%ret = sub <16 x i16> %arg1, %arg2
|
||
|
|
ret <16 x i16> %ret
|
||
|
|
}
|
||
|
|
|
||
|
|
define <8 x i32> @test_sub_v8i32(<8 x i32> %arg1, <8 x i32> %arg2) {
|
||
|
|
%ret = sub <8 x i32> %arg1, %arg2
|
||
|
|
ret <8 x i32> %ret
|
||
|
|
}
|
||
|
|
|
||
|
|
define <4 x i64> @test_sub_v4i64(<4 x i64> %arg1, <4 x i64> %arg2) {
|
||
|
|
%ret = sub <4 x i64> %arg1, %arg2
|
||
|
|
ret <4 x i64> %ret
|
||
|
|
}
|
||
|
|
...
|
||
|
|
---
|
||
|
|
name: test_sub_v32i8
|
||
|
|
# ALL-LABEL: name: test_sub_v32i8
|
||
|
|
alignment: 16
|
||
|
|
legalized: true
|
||
|
|
regBankSelected: true
|
||
|
|
registers:
|
||
|
|
- { id: 0, class: vecr }
|
||
|
|
- { id: 1, class: vecr }
|
||
|
|
- { id: 2, class: vecr }
|
||
|
|
# AVX2: %2:vr256 = VPSUBBYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512VL: %2:vr256 = VPSUBBYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512BWVL: %2:vr256x = VPSUBBZ256rr %0, %1
|
||
|
|
body: |
|
||
|
|
bb.1 (%ir-block.0):
|
||
|
|
liveins: $ymm0, $ymm1
|
||
|
|
|
||
|
|
%0(<32 x s8>) = COPY $ymm0
|
||
|
|
%1(<32 x s8>) = COPY $ymm1
|
||
|
|
%2(<32 x s8>) = G_SUB %0, %1
|
||
|
|
$ymm0 = COPY %2(<32 x s8>)
|
||
|
|
RET 0, implicit $ymm0
|
||
|
|
|
||
|
|
...
|
||
|
|
---
|
||
|
|
name: test_sub_v16i16
|
||
|
|
# ALL-LABEL: name: test_sub_v16i16
|
||
|
|
alignment: 16
|
||
|
|
legalized: true
|
||
|
|
regBankSelected: true
|
||
|
|
registers:
|
||
|
|
- { id: 0, class: vecr }
|
||
|
|
- { id: 1, class: vecr }
|
||
|
|
- { id: 2, class: vecr }
|
||
|
|
# AVX2: %2:vr256 = VPSUBWYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512VL: %2:vr256 = VPSUBWYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512BWVL: %2:vr256x = VPSUBWZ256rr %0, %1
|
||
|
|
body: |
|
||
|
|
bb.1 (%ir-block.0):
|
||
|
|
liveins: $ymm0, $ymm1
|
||
|
|
|
||
|
|
%0(<16 x s16>) = COPY $ymm0
|
||
|
|
%1(<16 x s16>) = COPY $ymm1
|
||
|
|
%2(<16 x s16>) = G_SUB %0, %1
|
||
|
|
$ymm0 = COPY %2(<16 x s16>)
|
||
|
|
RET 0, implicit $ymm0
|
||
|
|
|
||
|
|
...
|
||
|
|
---
|
||
|
|
name: test_sub_v8i32
|
||
|
|
# ALL-LABEL: name: test_sub_v8i32
|
||
|
|
alignment: 16
|
||
|
|
legalized: true
|
||
|
|
regBankSelected: true
|
||
|
|
registers:
|
||
|
|
- { id: 0, class: vecr }
|
||
|
|
- { id: 1, class: vecr }
|
||
|
|
- { id: 2, class: vecr }
|
||
|
|
# AVX2: %2:vr256 = VPSUBDYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512VL: %2:vr256x = VPSUBDZ256rr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512BWVL: %2:vr256x = VPSUBDZ256rr %0, %1
|
||
|
|
body: |
|
||
|
|
bb.1 (%ir-block.0):
|
||
|
|
liveins: $ymm0, $ymm1
|
||
|
|
|
||
|
|
%0(<8 x s32>) = COPY $ymm0
|
||
|
|
%1(<8 x s32>) = COPY $ymm1
|
||
|
|
%2(<8 x s32>) = G_SUB %0, %1
|
||
|
|
$ymm0 = COPY %2(<8 x s32>)
|
||
|
|
RET 0, implicit $ymm0
|
||
|
|
|
||
|
|
...
|
||
|
|
---
|
||
|
|
name: test_sub_v4i64
|
||
|
|
# ALL-LABEL: name: test_sub_v4i64
|
||
|
|
alignment: 16
|
||
|
|
legalized: true
|
||
|
|
regBankSelected: true
|
||
|
|
registers:
|
||
|
|
- { id: 0, class: vecr }
|
||
|
|
- { id: 1, class: vecr }
|
||
|
|
- { id: 2, class: vecr }
|
||
|
|
# AVX2: %2:vr256 = VPSUBQYrr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512VL: %2:vr256x = VPSUBQZ256rr %0, %1
|
||
|
|
#
|
||
|
|
# AVX512BWVL: %2:vr256x = VPSUBQZ256rr %0, %1
|
||
|
|
body: |
|
||
|
|
bb.1 (%ir-block.0):
|
||
|
|
liveins: $ymm0, $ymm1
|
||
|
|
|
||
|
|
%0(<4 x s64>) = COPY $ymm0
|
||
|
|
%1(<4 x s64>) = COPY $ymm1
|
||
|
|
%2(<4 x s64>) = G_SUB %0, %1
|
||
|
|
$ymm0 = COPY %2(<4 x s64>)
|
||
|
|
RET 0, implicit $ymm0
|
||
|
|
|
||
|
|
...
|