diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2016-08-04 10:14:39 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2016-08-04 10:14:39 +0000 |
| commit | 8ae6dad49b4440641073a2ef4001875ebf406352 (patch) | |
| tree | 983d87fce96ccad8f54a741f3f53dd487d635dc9 /llvm/test/CodeGen/X86/vec_ctbits.ll | |
| parent | 87e6d99487ddd7ba4bdc582600f0dfc2ef271ae7 (diff) | |
| download | bcm5719-llvm-8ae6dad49b4440641073a2ef4001875ebf406352.tar.gz bcm5719-llvm-8ae6dad49b4440641073a2ef4001875ebf406352.zip | |
[X86][SSE] Don't decide when to scalarize CTTZ/CTLZ for performance at lowering - this is what cost models are for
Improved CTTZ/CTLZ costings will be added shortly
llvm-svn: 277713
Diffstat (limited to 'llvm/test/CodeGen/X86/vec_ctbits.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vec_ctbits.ll | 61 |
1 files changed, 40 insertions, 21 deletions
diff --git a/llvm/test/CodeGen/X86/vec_ctbits.ll b/llvm/test/CodeGen/X86/vec_ctbits.ll index e151317c658..23f70db586f 100644 --- a/llvm/test/CodeGen/X86/vec_ctbits.ll +++ b/llvm/test/CodeGen/X86/vec_ctbits.ll @@ -8,15 +8,26 @@ declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>) define <2 x i64> @footz(<2 x i64> %a) nounwind { ; CHECK-LABEL: footz: ; CHECK: # BB#0: -; CHECK-NEXT: movd %xmm0, %rax -; CHECK-NEXT: bsfq %rax, %rax -; CHECK-NEXT: movd %rax, %xmm1 -; CHECK-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; CHECK-NEXT: movd %xmm0, %rax -; CHECK-NEXT: bsfq %rax, %rax -; CHECK-NEXT: movd %rax, %xmm0 -; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0] -; CHECK-NEXT: movdqa %xmm1, %xmm0 +; CHECK-NEXT: pxor %xmm1, %xmm1 +; CHECK-NEXT: pxor %xmm2, %xmm2 +; CHECK-NEXT: psubq %xmm0, %xmm2 +; CHECK-NEXT: pand %xmm0, %xmm2 +; CHECK-NEXT: psubq {{.*}}(%rip), %xmm2 +; CHECK-NEXT: movdqa %xmm2, %xmm0 +; CHECK-NEXT: psrlq $1, %xmm0 +; CHECK-NEXT: pand {{.*}}(%rip), %xmm0 +; CHECK-NEXT: psubq %xmm0, %xmm2 +; CHECK-NEXT: movdqa {{.*#+}} xmm0 = [3689348814741910323,3689348814741910323] +; CHECK-NEXT: movdqa %xmm2, %xmm3 +; CHECK-NEXT: pand %xmm0, %xmm3 +; CHECK-NEXT: psrlq $2, %xmm2 +; CHECK-NEXT: pand %xmm0, %xmm2 +; CHECK-NEXT: paddq %xmm3, %xmm2 +; CHECK-NEXT: movdqa %xmm2, %xmm0 +; CHECK-NEXT: psrlq $4, %xmm0 +; CHECK-NEXT: paddq %xmm2, %xmm0 +; CHECK-NEXT: pand {{.*}}(%rip), %xmm0 +; CHECK-NEXT: psadbw %xmm1, %xmm0 ; CHECK-NEXT: retq %c = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 true) ret <2 x i64> %c @@ -75,18 +86,26 @@ define <2 x i32> @promtz(<2 x i32> %a) nounwind { ; CHECK-LABEL: promtz: ; CHECK: # BB#0: ; CHECK-NEXT: por {{.*}}(%rip), %xmm0 -; CHECK-NEXT: movd %xmm0, %rax -; CHECK-NEXT: bsfq %rax, %rax -; CHECK-NEXT: movl $64, %ecx -; CHECK-NEXT: cmoveq %rcx, %rax -; CHECK-NEXT: movd %rax, %xmm1 -; CHECK-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] -; CHECK-NEXT: movd %xmm0, %rax -; CHECK-NEXT: bsfq %rax, %rax -; CHECK-NEXT: cmoveq %rcx, %rax -; CHECK-NEXT: movd %rax, %xmm0 -; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0] -; CHECK-NEXT: movdqa %xmm1, %xmm0 +; CHECK-NEXT: pxor %xmm1, %xmm1 +; CHECK-NEXT: pxor %xmm2, %xmm2 +; CHECK-NEXT: psubq %xmm0, %xmm2 +; CHECK-NEXT: pand %xmm0, %xmm2 +; CHECK-NEXT: psubq {{.*}}(%rip), %xmm2 +; CHECK-NEXT: movdqa %xmm2, %xmm0 +; CHECK-NEXT: psrlq $1, %xmm0 +; CHECK-NEXT: pand {{.*}}(%rip), %xmm0 +; CHECK-NEXT: psubq %xmm0, %xmm2 +; CHECK-NEXT: movdqa {{.*#+}} xmm0 = [3689348814741910323,3689348814741910323] +; CHECK-NEXT: movdqa %xmm2, %xmm3 +; CHECK-NEXT: pand %xmm0, %xmm3 +; CHECK-NEXT: psrlq $2, %xmm2 +; CHECK-NEXT: pand %xmm0, %xmm2 +; CHECK-NEXT: paddq %xmm3, %xmm2 +; CHECK-NEXT: movdqa %xmm2, %xmm0 +; CHECK-NEXT: psrlq $4, %xmm0 +; CHECK-NEXT: paddq %xmm2, %xmm0 +; CHECK-NEXT: pand {{.*}}(%rip), %xmm0 +; CHECK-NEXT: psadbw %xmm1, %xmm0 ; CHECK-NEXT: retq %c = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %a, i1 false) ret <2 x i32> %c |
