OSchip
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
13
0
Fork 6
Code Issues Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll

650 lines
23 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE,SSE42
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX,AVX2-FAST
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512,AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi,+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL
; PR31551
; Pairs of shufflevector:trunc functions with functional equivalence.
; Ideally, the shuffles should be lowered to code with the same quality as the truncates.
define void @shuffle_v16i8_to_v8i8(<16 x i8>* %L, <8 x i8>* %S) nounwind {
; SSE2-LABEL: shuffle_v16i8_to_v8i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movq %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: shuffle_v16i8_to_v8i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movq %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: shuffle_v16i8_to_v8i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovq %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: shuffle_v16i8_to_v8i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i8_to_v8i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v16i8_to_v8i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: shuffle_v16i8_to_v8i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
store <8 x i8> %strided.vec, <8 x i8>* %S
ret void
}
define void @trunc_v8i16_to_v8i8(<16 x i8>* %L, <8 x i8>* %S) nounwind {
; SSE2-LABEL: trunc_v8i16_to_v8i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movq %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: trunc_v8i16_to_v8i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movq %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: trunc_v8i16_to_v8i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovq %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: trunc_v8i16_to_v8i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v8i16_to_v8i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512VL-NEXT: vmovq %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v8i16_to_v8i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v8i16_to_v8i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%bc = bitcast <16 x i8> %vec to <8 x i16>
%strided.vec = trunc <8 x i16> %bc to <8 x i8>
store <8 x i8> %strided.vec, <8 x i8>* %S
ret void
}
define void @shuffle_v8i16_to_v4i16(<8 x i16>* %L, <4 x i16>* %S) nounwind {
; SSE2-LABEL: shuffle_v8i16_to_v4i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = mem[0,2,2,3,4,5,6,7]
; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE2-NEXT: movq %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: shuffle_v8i16_to_v4i16:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movq %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: shuffle_v8i16_to_v4i16:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovq %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: shuffle_v8i16_to_v4i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v8i16_to_v4i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovdw %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v8i16_to_v4i16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: shuffle_v8i16_to_v4i16:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovdw %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <8 x i16>, <8 x i16>* %L
%strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
store <4 x i16> %strided.vec, <4 x i16>* %S
ret void
}
define void @trunc_v4i32_to_v4i16(<8 x i16>* %L, <4 x i16>* %S) nounwind {
; SSE2-LABEL: trunc_v4i32_to_v4i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = mem[0,2,2,3,4,5,6,7]
; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE2-NEXT: movq %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: trunc_v4i32_to_v4i16:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movq %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: trunc_v4i32_to_v4i16:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovq %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: trunc_v4i32_to_v4i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovq %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v4i32_to_v4i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovdw %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v4i32_to_v4i16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovq %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v4i32_to_v4i16:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovdw %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <8 x i16>, <8 x i16>* %L
%bc = bitcast <8 x i16> %vec to <4 x i32>
%strided.vec = trunc <4 x i32> %bc to <4 x i16>
store <4 x i16> %strided.vec, <4 x i16>* %S
ret void
}
define void @shuffle_v4i32_to_v2i32(<4 x i32>* %L, <2 x i32>* %S) nounwind {
; SSE-LABEL: shuffle_v4i32_to_v2i32:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; SSE-NEXT: movq %xmm0, (%rsi)
; SSE-NEXT: retq
;
; AVX-LABEL: shuffle_v4i32_to_v2i32:
; AVX: # %bb.0:
; AVX-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX-NEXT: vmovlps %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512-LABEL: shuffle_v4i32_to_v2i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX512-NEXT: vmovlps %xmm0, (%rsi)
; AVX512-NEXT: retq
%vec = load <4 x i32>, <4 x i32>* %L
%strided.vec = shufflevector <4 x i32> %vec, <4 x i32> undef, <2 x i32> <i32 0, i32 2>
store <2 x i32> %strided.vec, <2 x i32>* %S
ret void
}
define void @trunc_v2i64_to_v2i32(<4 x i32>* %L, <2 x i32>* %S) nounwind {
; SSE-LABEL: trunc_v2i64_to_v2i32:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; SSE-NEXT: movq %xmm0, (%rsi)
; SSE-NEXT: retq
;
; AVX-LABEL: trunc_v2i64_to_v2i32:
; AVX: # %bb.0:
; AVX-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX-NEXT: vmovlps %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: trunc_v2i64_to_v2i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX512F-NEXT: vmovlps %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v2i64_to_v2i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovqd %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v2i64_to_v2i32:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX512BW-NEXT: vmovlps %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v2i64_to_v2i32:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovqd %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <4 x i32>, <4 x i32>* %L
%bc = bitcast <4 x i32> %vec to <2 x i64>
%strided.vec = trunc <2 x i64> %bc to <2 x i32>
store <2 x i32> %strided.vec, <2 x i32>* %S
ret void
}
define void @shuffle_v16i8_to_v4i8(<16 x i8>* %L, <4 x i8>* %S) nounwind {
; SSE2-LABEL: shuffle_v16i8_to_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: shuffle_v16i8_to_v4i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movd %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: shuffle_v16i8_to_v4i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovd %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: shuffle_v16i8_to_v4i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i8_to_v4i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v16i8_to_v4i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: shuffle_v16i8_to_v4i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
store <4 x i8> %strided.vec, <4 x i8>* %S
ret void
}
define void @trunc_v4i32_to_v4i8(<16 x i8>* %L, <4 x i8>* %S) nounwind {
; SSE2-LABEL: trunc_v4i32_to_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: trunc_v4i32_to_v4i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; SSE42-NEXT: movd %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: trunc_v4i32_to_v4i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT: vmovd %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: trunc_v4i32_to_v4i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v4i32_to_v4i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v4i32_to_v4i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v4i32_to_v4i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%bc = bitcast <16 x i8> %vec to <4 x i32>
%strided.vec = trunc <4 x i32> %bc to <4 x i8>
store <4 x i8> %strided.vec, <4 x i8>* %S
ret void
}
define void @shuffle_v8i16_to_v2i16(<8 x i16>* %L, <2 x i16>* %S) nounwind {
; SSE-LABEL: shuffle_v8i16_to_v2i16:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: movd %xmm0, (%rsi)
; SSE-NEXT: retq
;
; AVX1-LABEL: shuffle_v8i16_to_v2i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX1-NEXT: vmovd %xmm0, (%rsi)
; AVX1-NEXT: retq
;
; AVX2-SLOW-LABEL: shuffle_v8i16_to_v2i16:
; AVX2-SLOW: # %bb.0:
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovd %xmm0, (%rsi)
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: shuffle_v8i16_to_v2i16:
; AVX2-FAST: # %bb.0:
; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT: vmovd %xmm0, (%rsi)
; AVX2-FAST-NEXT: retq
;
; AVX512F-LABEL: shuffle_v8i16_to_v2i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v8i16_to_v2i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v8i16_to_v2i16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <8 x i16>, <8 x i16>* %L
%strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 0, i32 4>
store <2 x i16> %strided.vec, <2 x i16>* %S
ret void
}
define void @trunc_v2i64_to_v2i16(<8 x i16>* %L, <2 x i16>* %S) nounwind {
; SSE-LABEL: trunc_v2i64_to_v2i16:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: movd %xmm0, (%rsi)
; SSE-NEXT: retq
;
; AVX1-LABEL: trunc_v2i64_to_v2i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX1-NEXT: vmovd %xmm0, (%rsi)
; AVX1-NEXT: retq
;
; AVX2-SLOW-LABEL: trunc_v2i64_to_v2i16:
; AVX2-SLOW: # %bb.0:
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovd %xmm0, (%rsi)
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: trunc_v2i64_to_v2i16:
; AVX2-FAST: # %bb.0:
; AVX2-FAST-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT: vmovd %xmm0, (%rsi)
; AVX2-FAST-NEXT: retq
;
; AVX512F-LABEL: trunc_v2i64_to_v2i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpshufd {{.*#+}} xmm0 = mem[0,2,2,3]
; AVX512F-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX512F-NEXT: vmovd %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v2i64_to_v2i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v2i64_to_v2i16:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vmovd %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v2i64_to_v2i16:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <8 x i16>, <8 x i16>* %L
%bc = bitcast <8 x i16> %vec to <2 x i64>
%strided.vec = trunc <2 x i64> %bc to <2 x i16>
store <2 x i16> %strided.vec, <2 x i16>* %S
ret void
}
define void @shuffle_v16i8_to_v2i8(<16 x i8>* %L, <2 x i8>* %S) nounwind {
; SSE2-LABEL: shuffle_v16i8_to_v2i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: shuffle_v16i8_to_v2i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: shuffle_v16i8_to_v2i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: shuffle_v16i8_to_v2i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i8_to_v2i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: shuffle_v16i8_to_v2i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 0, i32 8>
store <2 x i8> %strided.vec, <2 x i8>* %S
ret void
}
define void @trunc_v2i64_to_v2i8(<16 x i8>* %L, <2 x i8>* %S) nounwind {
; SSE2-LABEL: trunc_v2i64_to_v2i8:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa (%rdi), %xmm0
; SSE2-NEXT: pand {{.*}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, (%rsi)
; SSE2-NEXT: retq
;
; SSE42-LABEL: trunc_v2i64_to_v2i8:
; SSE42: # %bb.0:
; SSE42-NEXT: movdqa (%rdi), %xmm0
; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; SSE42-NEXT: pextrw $0, %xmm0, (%rsi)
; SSE42-NEXT: retq
;
; AVX-LABEL: trunc_v2i64_to_v2i8:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX-NEXT: retq
;
; AVX512F-LABEL: trunc_v2i64_to_v2i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: trunc_v2i64_to_v2i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi)
; AVX512VL-NEXT: retq
;
; AVX512BW-LABEL: trunc_v2i64_to_v2i8:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi)
; AVX512BW-NEXT: retq
;
; AVX512BWVL-LABEL: trunc_v2i64_to_v2i8:
; AVX512BWVL: # %bb.0:
; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi)
; AVX512BWVL-NEXT: retq
%vec = load <16 x i8>, <16 x i8>* %L
%bc = bitcast <16 x i8> %vec to <2 x i64>
%strided.vec = trunc <2 x i64> %bc to <2 x i8>
store <2 x i8> %strided.vec, <2 x i8>* %S
ret void
}
Reference in New Issue View Git Blame Copy Permalink