maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/avx512vbmi-intrinsics-fast-...

163 lines
6.4 KiB
LLVM
Raw Normal View History

[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+avx512f,+avx512vbmi | FileCheck %s --check-prefixes=CHECK,X86
; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vbmi | FileCheck %s --check-prefixes=CHECK,X64
[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/avx512vbmi-builtins.c
define <8 x i64> @test_mm512_mask2_permutex2var_epi8(<8 x i64> %__A, <8 x i64> %__I, i64 %__U, <8 x i64> %__B) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_mm512_mask2_permutex2var_epi8:
; X86: # %bb.0: # %entry
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1
; X86-NEXT: kunpckdq %k1, %k0, %k1
; X86-NEXT: vpermi2b %zmm2, %zmm0, %zmm1 {%k1}
; X86-NEXT: vmovdqa64 %zmm1, %zmm0
; X86-NEXT: retl
[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
;
; X64-LABEL: test_mm512_mask2_permutex2var_epi8:
; X64: # %bb.0: # %entry
; X64-NEXT: kmovq %rdi, %k1
; X64-NEXT: vpermi2b %zmm2, %zmm0, %zmm1 {%k1}
; X64-NEXT: vmovdqa64 %zmm1, %zmm0
; X64-NEXT: retq
entry:
%0 = bitcast <8 x i64> %__A to <64 x i8>
%1 = bitcast <8 x i64> %__I to <64 x i8>
%2 = bitcast <8 x i64> %__B to <64 x i8>
%3 = tail call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %0, <64 x i8> %1, <64 x i8> %2)
%4 = bitcast i64 %__U to <64 x i1>
%5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %1
%6 = bitcast <64 x i8> %5 to <8 x i64>
ret <8 x i64> %6
}
define <8 x i64> @test_mm512_permutex2var_epi8(<8 x i64> %__A, <8 x i64> %__I, <8 x i64> %__B) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; CHECK-LABEL: test_mm512_permutex2var_epi8:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vpermt2b %zmm2, %zmm1, %zmm0
; CHECK-NEXT: ret{{[l|q]}}
[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
entry:
%0 = bitcast <8 x i64> %__A to <64 x i8>
%1 = bitcast <8 x i64> %__I to <64 x i8>
%2 = bitcast <8 x i64> %__B to <64 x i8>
%3 = tail call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %0, <64 x i8> %1, <64 x i8> %2)
%4 = bitcast <64 x i8> %3 to <8 x i64>
ret <8 x i64> %4
}
define <8 x i64> @test_mm512_mask_permutex2var_epi8(<8 x i64> %__A, i64 %__U, <8 x i64> %__I, <8 x i64> %__B) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_mm512_mask_permutex2var_epi8:
; X86: # %bb.0: # %entry
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1
; X86-NEXT: kunpckdq %k1, %k0, %k1
; X86-NEXT: vpermt2b %zmm2, %zmm1, %zmm0 {%k1}
; X86-NEXT: retl
[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
;
; X64-LABEL: test_mm512_mask_permutex2var_epi8:
; X64: # %bb.0: # %entry
; X64-NEXT: kmovq %rdi, %k1
; X64-NEXT: vpermt2b %zmm2, %zmm1, %zmm0 {%k1}
; X64-NEXT: retq
entry:
%0 = bitcast <8 x i64> %__A to <64 x i8>
%1 = bitcast <8 x i64> %__I to <64 x i8>
%2 = bitcast <8 x i64> %__B to <64 x i8>
%3 = tail call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %0, <64 x i8> %1, <64 x i8> %2)
%4 = bitcast i64 %__U to <64 x i1>
%5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %0
%6 = bitcast <64 x i8> %5 to <8 x i64>
ret <8 x i64> %6
}
define <8 x i64> @test_mm512_maskz_permutex2var_epi8(i64 %__U, <8 x i64> %__A, <8 x i64> %__I, <8 x i64> %__B) {
[X86][AVX512] Cleanup intrinsics tests Ensure we test on 32-bit and 64-bit targets, and strip -mcpu usage. Part of ongoing work to ensure we test all intrinsic style tests on 32 and 64 bit targets where possible. llvm-svn: 333843
2018-06-03 22:56:04 +08:00
; X86-LABEL: test_mm512_maskz_permutex2var_epi8:
; X86: # %bb.0: # %entry
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1
; X86-NEXT: kunpckdq %k1, %k0, %k1
; X86-NEXT: vpermt2b %zmm2, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT: retl
[X86] Add unmasked vermi2var intrinsics so we can use explicit select instructions for masking in clang. This will allow us to remove the 3 different flavors of masked intrinsics. I'm leaving the actual intrinsic removal for another patch. llvm-svn: 333386
2018-05-29 11:26:30 +08:00
;
; X64-LABEL: test_mm512_maskz_permutex2var_epi8:
; X64: # %bb.0: # %entry
; X64-NEXT: kmovq %rdi, %k1
; X64-NEXT: vpermt2b %zmm2, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT: retq
entry:
%0 = bitcast <8 x i64> %__A to <64 x i8>
%1 = bitcast <8 x i64> %__I to <64 x i8>
%2 = bitcast <8 x i64> %__B to <64 x i8>
%3 = tail call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %0, <64 x i8> %1, <64 x i8> %2)
%4 = bitcast i64 %__U to <64 x i1>
%5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
%6 = bitcast <64 x i8> %5 to <8 x i64>
ret <8 x i64> %6
}
declare <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8>, <64 x i8>, <64 x i8>)
[X86] Remove mask parameter from avx512 pmultishiftqb intrinsics. Use select in IR instead. Fixes PR40259 llvm-svn: 351035
2019-01-14 16:46:45 +08:00
define <8 x i64> @test_mm512_mask_multishift_epi64_epi8(<8 x i64> %__W, i64 %__M, <8 x i64> %__X, <8 x i64> %__Y) {
; X86-LABEL: test_mm512_mask_multishift_epi64_epi8:
; X86: # %bb.0: # %entry
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1
; X86-NEXT: kunpckdq %k1, %k0, %k1
; X86-NEXT: vpmultishiftqb %zmm2, %zmm1, %zmm0 {%k1}
; X86-NEXT: retl
;
; X64-LABEL: test_mm512_mask_multishift_epi64_epi8:
; X64: # %bb.0: # %entry
; X64-NEXT: kmovq %rdi, %k1
; X64-NEXT: vpmultishiftqb %zmm2, %zmm1, %zmm0 {%k1}
; X64-NEXT: retq
entry:
%0 = bitcast <8 x i64> %__X to <64 x i8>
%1 = bitcast <8 x i64> %__Y to <64 x i8>
%2 = tail call <64 x i8> @llvm.x86.avx512.pmultishift.qb.512(<64 x i8> %0, <64 x i8> %1)
%3 = bitcast <8 x i64> %__W to <64 x i8>
%4 = bitcast i64 %__M to <64 x i1>
%5 = select <64 x i1> %4, <64 x i8> %2, <64 x i8> %3
%6 = bitcast <64 x i8> %5 to <8 x i64>
ret <8 x i64> %6
}
define <8 x i64> @test_mm512_maskz_multishift_epi64_epi8(i64 %__M, <8 x i64> %__X, <8 x i64> %__Y) {
; X86-LABEL: test_mm512_maskz_multishift_epi64_epi8:
; X86: # %bb.0: # %entry
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0
; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1
; X86-NEXT: kunpckdq %k1, %k0, %k1
; X86-NEXT: vpmultishiftqb %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT: retl
;
; X64-LABEL: test_mm512_maskz_multishift_epi64_epi8:
; X64: # %bb.0: # %entry
; X64-NEXT: kmovq %rdi, %k1
; X64-NEXT: vpmultishiftqb %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT: retq
entry:
%0 = bitcast <8 x i64> %__X to <64 x i8>
%1 = bitcast <8 x i64> %__Y to <64 x i8>
%2 = tail call <64 x i8> @llvm.x86.avx512.pmultishift.qb.512(<64 x i8> %0, <64 x i8> %1)
%3 = bitcast i64 %__M to <64 x i1>
%4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
%5 = bitcast <64 x i8> %4 to <8 x i64>
ret <8 x i64> %5
}
define <8 x i64> @test_mm512_multishift_epi64_epi8(<8 x i64> %__X, <8 x i64> %__Y) {
; CHECK-LABEL: test_mm512_multishift_epi64_epi8:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: vpmultishiftqb %zmm1, %zmm0, %zmm0
; CHECK-NEXT: ret{{[l|q]}}
entry:
%0 = bitcast <8 x i64> %__X to <64 x i8>
%1 = bitcast <8 x i64> %__Y to <64 x i8>
%2 = tail call <64 x i8> @llvm.x86.avx512.pmultishift.qb.512(<64 x i8> %0, <64 x i8> %1)
%3 = bitcast <64 x i8> %2 to <8 x i64>
ret <8 x i64> %3
}
declare <64 x i8> @llvm.x86.avx512.pmultishift.qb.512(<64 x i8>, <64 x i8>)