diff options
author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2018-03-11 11:19:19 +0000 |
---|---|---|
committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2018-03-11 11:19:19 +0000 |
commit | 2565bd421e20e34373041ac4a281947e70748c49 (patch) | |
tree | 3cc20f42e1476715c3f2a64edb627d6b920848bd /llvm/test | |
parent | 1c861c582fbd121ee1db6d9858e443b551a8efc1 (diff) | |
download | bcm5719-llvm-2565bd421e20e34373041ac4a281947e70748c49.tar.gz bcm5719-llvm-2565bd421e20e34373041ac4a281947e70748c49.zip |
[X86][AVX512] createVariablePermute - Non-VLX targets can widen v4i64/v8f64 variable permutes to v8i64/v8f64
Permutes in the upper elements will be undefined, but they will be discarded anyway.
llvm-svn: 327238
Diffstat (limited to 'llvm/test')
-rw-r--r-- | llvm/test/CodeGen/X86/var-permute-256.ll | 100 |
1 files changed, 28 insertions, 72 deletions
diff --git a/llvm/test/CodeGen/X86/var-permute-256.ll b/llvm/test/CodeGen/X86/var-permute-256.ll index 8538ed71c0a..87cd0918f75 100644 --- a/llvm/test/CodeGen/X86/var-permute-256.ll +++ b/llvm/test/CodeGen/X86/var-permute-256.ll @@ -66,28 +66,18 @@ define <4 x i64> @var_shuffle_v4i64(<4 x i64> %v, <4 x i64> %indices) nounwind { ; ; AVX512F-LABEL: var_shuffle_v4i64: ; AVX512F: # %bb.0: -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,2,2,2] -; AVX512F-NEXT: vpmuludq %ymm2, %ymm1, %ymm2 -; AVX512F-NEXT: vpsrlq $32, %ymm1, %ymm3 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm4 = [8589934594,8589934594,8589934594,8589934594] -; AVX512F-NEXT: vpmuludq %ymm4, %ymm3, %ymm3 -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpsllq $32, %ymm2, %ymm2 -; AVX512F-NEXT: vpmuludq %ymm4, %ymm1, %ymm1 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm3 = [4294967296,4294967296,4294967296,4294967296] -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512F-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0 +; AVX512F-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512F-NEXT: retq ; ; AVX512DQ-LABEL: var_shuffle_v4i64: ; AVX512DQ: # %bb.0: ; AVX512DQ-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [8589934594,8589934594,8589934594,8589934594] -; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4294967296,4294967296,4294967296,4294967296] -; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512DQ-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0 +; AVX512DQ-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512DQ-NEXT: retq ; ; AVX512VL-LABEL: var_shuffle_v4i64: @@ -1377,28 +1367,18 @@ define <4 x double> @var_shuffle_v4f64(<4 x double> %v, <4 x i64> %indices) noun ; ; AVX512F-LABEL: var_shuffle_v4f64: ; AVX512F: # %bb.0: -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,2,2,2] -; AVX512F-NEXT: vpmuludq %ymm2, %ymm1, %ymm2 -; AVX512F-NEXT: vpsrlq $32, %ymm1, %ymm3 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm4 = [8589934594,8589934594,8589934594,8589934594] -; AVX512F-NEXT: vpmuludq %ymm4, %ymm3, %ymm3 -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpsllq $32, %ymm2, %ymm2 -; AVX512F-NEXT: vpmuludq %ymm4, %ymm1, %ymm1 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm3 = [4294967296,4294967296,4294967296,4294967296] -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512F-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0 +; AVX512F-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512F-NEXT: retq ; ; AVX512DQ-LABEL: var_shuffle_v4f64: ; AVX512DQ: # %bb.0: ; AVX512DQ-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [8589934594,8589934594,8589934594,8589934594] -; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4294967296,4294967296,4294967296,4294967296] -; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512DQ-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0 +; AVX512DQ-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512DQ-NEXT: retq ; ; AVX512VL-LABEL: var_shuffle_v4f64: @@ -1558,30 +1538,18 @@ define <4 x i64> @var_shuffle_v4i64_from_v2i64(<2 x i64> %v, <4 x i64> %indices) ; ; AVX512F-LABEL: var_shuffle_v4i64_from_v2i64: ; AVX512F: # %bb.0: -; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,2,2,2] -; AVX512F-NEXT: vpmuludq %ymm2, %ymm1, %ymm2 -; AVX512F-NEXT: vpsrlq $32, %ymm1, %ymm3 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm4 = [8589934594,8589934594,8589934594,8589934594] -; AVX512F-NEXT: vpmuludq %ymm4, %ymm3, %ymm3 -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpsllq $32, %ymm2, %ymm2 -; AVX512F-NEXT: vpmuludq %ymm4, %ymm1, %ymm1 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm3 = [4294967296,4294967296,4294967296,4294967296] -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512F-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 +; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 +; AVX512F-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512F-NEXT: retq ; ; AVX512DQ-LABEL: var_shuffle_v4i64_from_v2i64: ; AVX512DQ: # %bb.0: ; AVX512DQ-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 -; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [8589934594,8589934594,8589934594,8589934594] -; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4294967296,4294967296,4294967296,4294967296] -; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512DQ-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 +; AVX512DQ-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512DQ-NEXT: retq ; ; AVX512VL-LABEL: var_shuffle_v4i64_from_v2i64: @@ -2806,30 +2774,18 @@ define <4 x double> @var_shuffle_v4f64_from_v2f64(<2 x double> %v, <4 x i64> %in ; ; AVX512F-LABEL: var_shuffle_v4f64_from_v2f64: ; AVX512F: # %bb.0: -; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,2,2,2] -; AVX512F-NEXT: vpmuludq %ymm2, %ymm1, %ymm2 -; AVX512F-NEXT: vpsrlq $32, %ymm1, %ymm3 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm4 = [8589934594,8589934594,8589934594,8589934594] -; AVX512F-NEXT: vpmuludq %ymm4, %ymm3, %ymm3 -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpsllq $32, %ymm2, %ymm2 -; AVX512F-NEXT: vpmuludq %ymm4, %ymm1, %ymm1 -; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm3 = [4294967296,4294967296,4294967296,4294967296] -; AVX512F-NEXT: vpaddq %ymm3, %ymm2, %ymm2 -; AVX512F-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512F-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 +; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 +; AVX512F-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512F-NEXT: retq ; ; AVX512DQ-LABEL: var_shuffle_v4f64_from_v2f64: ; AVX512DQ: # %bb.0: ; AVX512DQ-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1 -; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [8589934594,8589934594,8589934594,8589934594] -; AVX512DQ-NEXT: vpmullq %zmm2, %zmm1, %zmm1 -; AVX512DQ-NEXT: vpbroadcastq {{.*#+}} ymm2 = [4294967296,4294967296,4294967296,4294967296] -; AVX512DQ-NEXT: vpaddq %ymm2, %ymm1, %ymm1 -; AVX512DQ-NEXT: vpermd %ymm0, %ymm1, %ymm0 +; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0 +; AVX512DQ-NEXT: vpermpd %zmm0, %zmm1, %zmm0 +; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 ; AVX512DQ-NEXT: retq ; ; AVX512VL-LABEL: var_shuffle_v4f64_from_v2f64: |