1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=sse2 < %s | FileCheck --check-prefix=SSE %s 3; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=sse4.1 < %s | FileCheck --check-prefix=SSE %s 4; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=avx < %s | FileCheck --check-prefix=AVX %s 5 6; PR21507 - https://llvm.org/bugs/show_bug.cgi?id=21507 7; Each function should be a single math op; no extra moves. 8 9 10define <4 x float> @recip(<4 x float> %x) { 11; SSE-LABEL: recip: 12; SSE: # BB#0: 13; SSE-NEXT: rcpss %xmm0, %xmm0 14; SSE-NEXT: retq 15; 16; AVX-LABEL: recip: 17; AVX: # BB#0: 18; AVX-NEXT: vrcpss %xmm0, %xmm0, %xmm0 19; AVX-NEXT: retq 20 %y = tail call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %x) 21 %shuf = shufflevector <4 x float> %y, <4 x float> %x, <4 x i32> <i32 0, i32 5, i32 6, i32 7> 22 ret <4 x float> %shuf 23} 24 25define <4 x float> @recip_square_root(<4 x float> %x) { 26; SSE-LABEL: recip_square_root: 27; SSE: # BB#0: 28; SSE-NEXT: rsqrtss %xmm0, %xmm0 29; SSE-NEXT: retq 30; 31; AVX-LABEL: recip_square_root: 32; AVX: # BB#0: 33; AVX-NEXT: vrsqrtss %xmm0, %xmm0, %xmm0 34; AVX-NEXT: retq 35 %y = tail call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %x) 36 %shuf = shufflevector <4 x float> %y, <4 x float> %x, <4 x i32> <i32 0, i32 5, i32 6, i32 7> 37 ret <4 x float> %shuf 38} 39 40define <4 x float> @square_root(<4 x float> %x) { 41; SSE-LABEL: square_root: 42; SSE: # BB#0: 43; SSE-NEXT: sqrtss %xmm0, %xmm0 44; SSE-NEXT: retq 45; 46; AVX-LABEL: square_root: 47; AVX: # BB#0: 48; AVX-NEXT: vsqrtss %xmm0, %xmm0, %xmm0 49; AVX-NEXT: retq 50 %y = tail call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %x) 51 %shuf = shufflevector <4 x float> %y, <4 x float> %x, <4 x i32> <i32 0, i32 5, i32 6, i32 7> 52 ret <4 x float> %shuf 53} 54 55define <2 x double> @square_root_double(<2 x double> %x) { 56; SSE-LABEL: square_root_double: 57; SSE: # BB#0: 58; SSE-NEXT: sqrtsd %xmm0, %xmm0 59; SSE-NEXT: retq 60; 61; AVX-LABEL: square_root_double: 62; AVX: # BB#0: 63; AVX-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0 64; AVX-NEXT: retq 65 %y = tail call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %x) 66 %shuf = shufflevector <2 x double> %y, <2 x double> %x, <2 x i32> <i32 0, i32 3> 67 ret <2 x double> %shuf 68} 69 70declare <4 x float> @llvm.x86.sse.rcp.ss(<4 x float>) 71declare <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float>) 72declare <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float>) 73declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) 74 75