From: Simon Pilgrim Date: Wed, 8 Feb 2023 21:09:41 +0000 (+0000) Subject: [X86] midpoint-int-vec - cleanup common check prefixes X-Git-Tag: upstream/17.0.6~18159 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=e887627c00134d7180c3aef4e26bd49dbbec9e24;p=platform%2Fupstream%2Fllvm.git [X86] midpoint-int-vec - cleanup common check prefixes --- diff --git a/llvm/test/CodeGen/X86/midpoint-int-vec-128.ll b/llvm/test/CodeGen/X86/midpoint-int-vec-128.ll index bfcb9ce87d18..e1ecfae9b4cb 100644 --- a/llvm/test/CodeGen/X86/midpoint-int-vec-128.ll +++ b/llvm/test/CodeGen/X86/midpoint-int-vec-128.ll @@ -1,12 +1,12 @@ ; 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=+sse4.1 | FileCheck %s --check-prefixes=SSE,SSE41 -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX1-FALLBACK -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2-FALLBACK +; 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 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop | FileCheck %s --check-prefixes=XOP,XOP-FALLBACK ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefixes=XOP,XOPAVX1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=XOP,XOPAVX2 -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512F +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512F ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl | FileCheck %s --check-prefixes=AVX512VL,AVX512VL-FALLBACK ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX512BW-FALLBACK ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX512VL,AVX512VLBW @@ -65,30 +65,30 @@ define <4 x i32> @vec128_i32_signed_reg_reg(<4 x i32> %a1, <4 x i32> %a2) nounwi ; SSE41-NEXT: paddd %xmm2, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i32_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i32_signed_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] -; AVX2-FALLBACK-NEXT: vpor %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i32_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i32_signed_reg_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] +; AVX2-NEXT: vpor %xmm3, %xmm2, %xmm2 +; AVX2-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX2-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i32_signed_reg_reg: ; XOP-FALLBACK: # %bb.0: @@ -228,34 +228,34 @@ define <4 x i32> @vec128_i32_unsigned_reg_reg(<4 x i32> %a1, <4 x i32> %a2) noun ; SSE41-NEXT: paddd %xmm4, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i32_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpminud %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i32_unsigned_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpminud %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX2-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpbroadcastd {{.*#+}} xmm4 = [1,1,1,1] -; AVX2-FALLBACK-NEXT: vpor %xmm4, %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubd %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmulld %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i32_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpminud %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm3 +; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 +; AVX1-NEXT: vpxor %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 +; AVX1-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubd %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i32_unsigned_reg_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vpminud %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm3 +; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 +; AVX2-NEXT: vpxor %xmm4, %xmm3, %xmm3 +; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm4 = [1,1,1,1] +; AVX2-NEXT: vpor %xmm4, %xmm3, %xmm3 +; AVX2-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubd %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX2-NEXT: vpmulld %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i32_unsigned_reg_reg: ; XOP-FALLBACK: # %bb.0: @@ -394,32 +394,32 @@ define <4 x i32> @vec128_i32_signed_mem_reg(ptr %a1_addr, <4 x i32> %a2) nounwin ; SSE41-NEXT: paddd %xmm1, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i32_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm0, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddd %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i32_signed_mem_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] -; AVX2-FALLBACK-NEXT: vpor %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsd %xmm0, %xmm1, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vpsubd %xmm3, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpmulld %xmm2, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddd %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i32_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsd %xmm0, %xmm1, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vpsubd %xmm3, %xmm0, %xmm0 +; AVX1-NEXT: vpsrld $1, %xmm0, %xmm0 +; AVX1-NEXT: vpmulld %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vpaddd %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i32_signed_mem_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm1 +; AVX2-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm2 +; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] +; AVX2-NEXT: vpor %xmm3, %xmm2, %xmm2 +; AVX2-NEXT: vpminsd %xmm0, %xmm1, %xmm3 +; AVX2-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vpsubd %xmm3, %xmm0, %xmm0 +; AVX2-NEXT: vpsrld $1, %xmm0, %xmm0 +; AVX2-NEXT: vpmulld %xmm2, %xmm0, %xmm0 +; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0 +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i32_signed_mem_reg: ; XOP-FALLBACK: # %bb.0: @@ -561,32 +561,32 @@ define <4 x i32> @vec128_i32_signed_reg_mem(<4 x i32> %a1, ptr %a2_addr) nounwin ; SSE41-NEXT: paddd %xmm1, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i32_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i32_signed_reg_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] -; AVX2-FALLBACK-NEXT: vpor %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i32_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i32_signed_reg_mem: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm1 +; AVX2-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] +; AVX2-NEXT: vpor %xmm3, %xmm2, %xmm2 +; AVX2-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX2-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i32_signed_reg_mem: ; XOP-FALLBACK: # %bb.0: @@ -730,34 +730,34 @@ define <4 x i32> @vec128_i32_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; SSE41-NEXT: paddd %xmm1, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i32_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i32_signed_mem_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX2-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] -; AVX2-FALLBACK-NEXT: vpor %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmulld %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i32_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm0 +; AVX1-NEXT: vmovdqa (%rsi), %xmm1 +; AVX1-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i32_signed_mem_mem: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm0 +; AVX2-NEXT: vmovdqa (%rsi), %xmm1 +; AVX2-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [1,1,1,1] +; AVX2-NEXT: vpor %xmm3, %xmm2, %xmm2 +; AVX2-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX2-NEXT: vpmulld %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i32_signed_mem_mem: ; XOP-FALLBACK: # %bb.0: @@ -947,43 +947,24 @@ define <2 x i64> @vec128_i64_signed_reg_reg(<2 x i64> %a1, <2 x i64> %a2) nounwi ; SSE41-NEXT: paddq %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i64_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i64_signed_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX2-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX2-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX2-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i64_signed_reg_reg: +; AVX: # %bb.0: +; AVX-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 +; AVX-NEXT: vpsubq %xmm1, %xmm0, %xmm4 +; AVX-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $1, %xmm1, %xmm2 +; AVX-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $32, %xmm3, %xmm4 +; AVX-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 +; AVX-NEXT: vpaddq %xmm1, %xmm4, %xmm1 +; AVX-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i64_signed_reg_reg: ; XOP: # %bb.0: @@ -1156,49 +1137,27 @@ define <2 x i64> @vec128_i64_unsigned_reg_reg(<2 x i64> %a1, <2 x i64> %a2) noun ; SSE41-NEXT: paddq %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i64_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808] -; AVX1-FALLBACK-NEXT: vpxor %xmm2, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpxor %xmm2, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i64_unsigned_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808] -; AVX2-FALLBACK-NEXT: vpxor %xmm2, %xmm1, %xmm3 -; AVX2-FALLBACK-NEXT: vpxor %xmm2, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpcmpgtq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX2-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX2-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX2-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i64_unsigned_reg_reg: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808] +; AVX-NEXT: vpxor %xmm2, %xmm1, %xmm3 +; AVX-NEXT: vpxor %xmm2, %xmm0, %xmm2 +; AVX-NEXT: vpcmpgtq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 +; AVX-NEXT: vpsubq %xmm1, %xmm0, %xmm4 +; AVX-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $1, %xmm1, %xmm2 +; AVX-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $32, %xmm3, %xmm4 +; AVX-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 +; AVX-NEXT: vpaddq %xmm1, %xmm4, %xmm1 +; AVX-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i64_unsigned_reg_reg: ; XOP: # %bb.0: @@ -1375,45 +1334,25 @@ define <2 x i64> @vec128_i64_signed_mem_reg(ptr %a1_addr, <2 x i64> %a2) nounwin ; SSE41-NEXT: paddq %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i64_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm4, %xmm0 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm2, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i64_signed_mem_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX2-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm4 -; AVX2-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm4, %xmm0 -; AVX2-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm2, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i64_signed_mem_reg: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm1 +; AVX-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 +; AVX-NEXT: vpsubq %xmm0, %xmm1, %xmm4 +; AVX-NEXT: vpsubq %xmm1, %xmm0, %xmm0 +; AVX-NEXT: vblendvpd %xmm2, %xmm4, %xmm0, %xmm0 +; AVX-NEXT: vpsrlq $1, %xmm0, %xmm2 +; AVX-NEXT: vpsrlq $33, %xmm0, %xmm0 +; AVX-NEXT: vpmuludq %xmm3, %xmm0, %xmm0 +; AVX-NEXT: vpsrlq $32, %xmm3, %xmm4 +; AVX-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 +; AVX-NEXT: vpaddq %xmm0, %xmm4, %xmm0 +; AVX-NEXT: vpsllq $32, %xmm0, %xmm0 +; AVX-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpaddq %xmm1, %xmm2, %xmm1 +; AVX-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i64_signed_mem_reg: ; XOP: # %bb.0: @@ -1591,45 +1530,25 @@ define <2 x i64> @vec128_i64_signed_reg_mem(<2 x i64> %a1, ptr %a2_addr) nounwin ; SSE41-NEXT: paddq %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i64_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i64_signed_reg_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX2-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX2-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX2-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i64_signed_reg_mem: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm1 +; AVX-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 +; AVX-NEXT: vpsubq %xmm1, %xmm0, %xmm4 +; AVX-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $1, %xmm1, %xmm2 +; AVX-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $32, %xmm3, %xmm4 +; AVX-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 +; AVX-NEXT: vpaddq %xmm1, %xmm4, %xmm1 +; AVX-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i64_signed_reg_mem: ; XOP: # %bb.0: @@ -1808,47 +1727,26 @@ define <2 x i64> @vec128_i64_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; SSE41-NEXT: paddq %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i64_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i64_signed_mem_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX2-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 -; AVX2-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm4 -; AVX2-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlq $32, %xmm3, %xmm4 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm4, %xmm1 -; AVX2-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i64_signed_mem_mem: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm0 +; AVX-NEXT: vmovdqa (%rsi), %xmm1 +; AVX-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3 +; AVX-NEXT: vpsubq %xmm1, %xmm0, %xmm4 +; AVX-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX-NEXT: vblendvpd %xmm2, %xmm4, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $1, %xmm1, %xmm2 +; AVX-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlq $32, %xmm3, %xmm4 +; AVX-NEXT: vpmuludq %xmm4, %xmm2, %xmm4 +; AVX-NEXT: vpaddq %xmm1, %xmm4, %xmm1 +; AVX-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX-NEXT: vpmuludq %xmm3, %xmm2, %xmm2 +; AVX-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i64_signed_mem_mem: ; XOP: # %bb.0: @@ -1966,29 +1864,17 @@ define <8 x i16> @vec128_i16_signed_reg_reg(<8 x i16> %a1, <8 x i16> %a2) nounwi ; SSE-NEXT: paddw %xmm2, %xmm0 ; SSE-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i16_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i16_signed_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i16_signed_reg_reg: +; AVX: # %bb.0: +; AVX-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX-NEXT: vpminsw %xmm1, %xmm0, %xmm3 +; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; AVX-NEXT: vpsubw %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i16_signed_reg_reg: ; XOP: # %bb.0: @@ -2100,33 +1986,19 @@ define <8 x i16> @vec128_i16_unsigned_reg_reg(<8 x i16> %a1, <8 x i16> %a2) noun ; SSE41-NEXT: paddw %xmm4, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i16_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpminuw %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmullw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i16_unsigned_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpminuw %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX2-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubw %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmullw %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i16_unsigned_reg_reg: +; AVX: # %bb.0: +; AVX-NEXT: vpminuw %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm3 +; AVX-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 +; AVX-NEXT: vpxor %xmm4, %xmm3, %xmm3 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 +; AVX-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 +; AVX-NEXT: vpsubw %xmm2, %xmm1, %xmm1 +; AVX-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX-NEXT: vpmullw %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i16_unsigned_reg_reg: ; XOP: # %bb.0: @@ -2223,31 +2095,18 @@ define <8 x i16> @vec128_i16_signed_mem_reg(ptr %a1_addr, <8 x i16> %a2) nounwin ; SSE-NEXT: paddw %xmm1, %xmm0 ; SSE-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i16_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm0, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsw %xmm0, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubw %xmm3, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddw %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i16_signed_mem_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtw %xmm0, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsw %xmm0, %xmm1, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vpsubw %xmm3, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpmullw %xmm2, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddw %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i16_signed_mem_reg: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm1 +; AVX-NEXT: vpcmpgtw %xmm0, %xmm1, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX-NEXT: vpminsw %xmm0, %xmm1, %xmm3 +; AVX-NEXT: vpmaxsw %xmm0, %xmm1, %xmm0 +; AVX-NEXT: vpsubw %xmm3, %xmm0, %xmm0 +; AVX-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX-NEXT: vpmullw %xmm2, %xmm0, %xmm0 +; AVX-NEXT: vpaddw %xmm1, %xmm0, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i16_signed_mem_reg: ; XOP: # %bb.0: @@ -2344,31 +2203,18 @@ define <8 x i16> @vec128_i16_signed_reg_mem(<8 x i16> %a1, ptr %a2_addr) nounwin ; SSE-NEXT: paddw %xmm1, %xmm0 ; SSE-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i16_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i16_signed_reg_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i16_signed_reg_mem: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm1 +; AVX-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX-NEXT: vpminsw %xmm1, %xmm0, %xmm3 +; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; AVX-NEXT: vpsubw %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i16_signed_reg_mem: ; XOP: # %bb.0: @@ -2466,33 +2312,19 @@ define <8 x i16> @vec128_i16_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; SSE-NEXT: paddw %xmm1, %xmm0 ; SSE-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i16_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i16_signed_mem_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX2-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubw %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: retq +; AVX-LABEL: vec128_i16_signed_mem_mem: +; AVX: # %bb.0: +; AVX-NEXT: vmovdqa (%rdi), %xmm0 +; AVX-NEXT: vmovdqa (%rsi), %xmm1 +; AVX-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm2 +; AVX-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX-NEXT: vpminsw %xmm1, %xmm0, %xmm3 +; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; AVX-NEXT: vpsubw %xmm3, %xmm1, %xmm1 +; AVX-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX-NEXT: retq ; ; XOP-LABEL: vec128_i16_signed_mem_mem: ; XOP: # %bb.0: @@ -2643,46 +2475,46 @@ define <16 x i8> @vec128_i8_signed_reg_reg(<16 x i8> %a1, <16 x i8> %a2) nounwin ; SSE41-NEXT: paddb %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i8_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i8_signed_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero -; AVX2-FALLBACK-NEXT: vpmullw %ymm2, %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vextracti128 $1, %ymm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vzeroupper -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i8_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i8_signed_reg_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX2-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX2-NEXT: vpmullw %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i8_signed_reg_reg: ; XOP-FALLBACK: # %bb.0: @@ -2870,50 +2702,50 @@ define <16 x i8> @vec128_i8_unsigned_reg_reg(<16 x i8> %a1, <16 x i8> %a2) nounw ; SSE41-NEXT: paddb %xmm2, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i8_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpminub %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i8_unsigned_reg_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vpminub %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 -; AVX2-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubb %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero -; AVX2-FALLBACK-NEXT: vpmullw %ymm2, %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vextracti128 $1, %ymm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vzeroupper -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i8_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpminub %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm3 +; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 +; AVX1-NEXT: vpxor %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 +; AVX1-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmullw %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i8_unsigned_reg_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vpminub %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm3 +; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4 +; AVX2-NEXT: vpxor %xmm4, %xmm3, %xmm3 +; AVX2-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 +; AVX2-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero +; AVX2-NEXT: vpmullw %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i8_unsigned_reg_reg: ; XOP-FALLBACK: # %bb.0: @@ -3113,48 +2945,48 @@ define <16 x i8> @vec128_i8_signed_mem_reg(ptr %a1_addr, <16 x i8> %a2) nounwind ; SSE41-NEXT: movdqa %xmm1, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i8_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm0, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubb %xmm3, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm3, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddb %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i8_signed_mem_reg: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsb %xmm0, %xmm1, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vpsubb %xmm3, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero -; AVX2-FALLBACK-NEXT: vpmullw %ymm2, %ymm0, %ymm0 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 -; AVX2-FALLBACK-NEXT: vextracti128 $1, %ymm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpackuswb %xmm2, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vpaddb %xmm1, %xmm0, %xmm0 -; AVX2-FALLBACK-NEXT: vzeroupper -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i8_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm0, %xmm1, %xmm3 +; AVX1-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vpsubb %xmm3, %xmm0, %xmm0 +; AVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vpand %xmm4, %xmm0, %xmm0 +; AVX1-NEXT: vpackuswb %xmm3, %xmm0, %xmm0 +; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i8_signed_mem_reg: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm1 +; AVX2-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm2 +; AVX2-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX2-NEXT: vpminsb %xmm0, %xmm1, %xmm3 +; AVX2-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vpsubb %xmm3, %xmm0, %xmm0 +; AVX2-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX2-NEXT: vpmullw %ymm2, %ymm0, %ymm0 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 +; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm0, %xmm0 +; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i8_signed_mem_reg: ; XOP-FALLBACK: # %bb.0: @@ -3355,48 +3187,48 @@ define <16 x i8> @vec128_i8_signed_reg_mem(<16 x i8> %a1, ptr %a2_addr) nounwind ; SSE41-NEXT: paddb %xmm3, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i8_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i8_signed_reg_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero -; AVX2-FALLBACK-NEXT: vpmullw %ymm2, %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vextracti128 $1, %ymm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vzeroupper -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i8_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i8_signed_reg_mem: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm1 +; AVX2-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX2-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX2-NEXT: vpmullw %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i8_signed_reg_mem: ; XOP-FALLBACK: # %bb.0: @@ -3599,50 +3431,50 @@ define <16 x i8> @vec128_i8_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; SSE41-NEXT: paddb %xmm1, %xmm0 ; SSE41-NEXT: retq ; -; AVX1-FALLBACK-LABEL: vec128_i8_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm2, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: retq -; -; AVX2-FALLBACK-LABEL: vec128_i8_signed_mem_mem: -; AVX2-FALLBACK: # %bb.0: -; AVX2-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX2-FALLBACK-NEXT: vmovdqa (%rsi), %xmm1 -; AVX2-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 -; AVX2-FALLBACK-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 -; AVX2-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm3 -; AVX2-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX2-FALLBACK-NEXT: vpsubb %xmm3, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero -; AVX2-FALLBACK-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero -; AVX2-FALLBACK-NEXT: vpmullw %ymm2, %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 -; AVX2-FALLBACK-NEXT: vextracti128 $1, %ymm1, %xmm2 -; AVX2-FALLBACK-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 -; AVX2-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX2-FALLBACK-NEXT: vzeroupper -; AVX2-FALLBACK-NEXT: retq +; AVX1-LABEL: vec128_i8_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm0 +; AVX1-NEXT: vmovdqa (%rsi), %xmm1 +; AVX1-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm3 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmullw %xmm2, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpackuswb %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: retq +; +; AVX2-LABEL: vec128_i8_signed_mem_mem: +; AVX2: # %bb.0: +; AVX2-NEXT: vmovdqa (%rdi), %xmm0 +; AVX2-NEXT: vmovdqa (%rsi), %xmm1 +; AVX2-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm2 +; AVX2-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2 +; AVX2-NEXT: vpminsb %xmm1, %xmm0, %xmm3 +; AVX2-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX2-NEXT: vpsubb %xmm3, %xmm1, %xmm1 +; AVX2-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero +; AVX2-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero +; AVX2-NEXT: vpmullw %ymm2, %ymm1, %ymm1 +; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1 +; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2 +; AVX2-NEXT: vpackuswb %xmm2, %xmm1, %xmm1 +; AVX2-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX2-NEXT: vzeroupper +; AVX2-NEXT: retq ; ; XOP-FALLBACK-LABEL: vec128_i8_signed_mem_mem: ; XOP-FALLBACK: # %bb.0: diff --git a/llvm/test/CodeGen/X86/midpoint-int-vec-256.ll b/llvm/test/CodeGen/X86/midpoint-int-vec-256.ll index da255045018c..424274361450 100644 --- a/llvm/test/CodeGen/X86/midpoint-int-vec-256.ll +++ b/llvm/test/CodeGen/X86/midpoint-int-vec-256.ll @@ -1,9 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX1-FALLBACK -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2 -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop | FileCheck %s --check-prefix=XOP-FALLBACK -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefix=XOPAVX1 -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefix=AVX2 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX1 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop | FileCheck %s --check-prefixes=XOP +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx | FileCheck %s --check-prefixes=XOP +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+xop,+avx2 | FileCheck %s --check-prefixes=AVX2 ; 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=+avx512f,+avx512vl | FileCheck %s --check-prefixes=AVX512,AVX512VL,AVX512VL-FALLBACK ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512,AVX512BW-FALLBACK @@ -21,24 +21,24 @@ ; Values come from regs define <8 x i32> @vec256_i32_signed_reg_reg(<8 x i32> %a1, <8 x i32> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i32_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubd %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm1 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmulld %xmm1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i32_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpsubd %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpminsd %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm2, %xmm2 +; AVX1-NEXT: vpmulld %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vpmulld %xmm1, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i32_signed_reg_reg: ; AVX2: # %bb.0: @@ -50,39 +50,22 @@ define <8 x i32> @vec256_i32_signed_reg_reg(<8 x i32> %a1, <8 x i32> %a2) nounwi ; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i32_signed_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpminsd %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i32_signed_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpminsd %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i32_signed_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpminsd %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubd %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpminsd %xmm1, %xmm0, %xmm4 +; XOP-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512-LABEL: vec256_i32_signed_reg_reg: ; AVX512: # %bb.0: @@ -105,24 +88,24 @@ define <8 x i32> @vec256_i32_signed_reg_reg(<8 x i32> %a1, <8 x i32> %a2) nounwi } define <8 x i32> @vec256_i32_unsigned_reg_reg(<8 x i32> %a1, <8 x i32> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i32_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vpminud %xmm1, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpmaxud %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubd %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm1 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpminud %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxud %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmulld %xmm1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i32_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vpminud %xmm1, %xmm0, %xmm2 +; AVX1-NEXT: vpmaxud %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpsubd %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpminud %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpmaxud %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm2, %xmm2 +; AVX1-NEXT: vpmulld %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vpmulld %xmm1, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i32_unsigned_reg_reg: ; AVX2: # %bb.0: @@ -134,39 +117,22 @@ define <8 x i32> @vec256_i32_unsigned_reg_reg(<8 x i32> %a1, <8 x i32> %a2) noun ; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i32_unsigned_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpminud %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxud %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminud %xmm1, %xmm0, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i32_unsigned_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpminud %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpmaxud %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminud %xmm1, %xmm0, %xmm4 -; XOPAVX1-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i32_unsigned_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpminud %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpmaxud %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubd %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpminud %xmm1, %xmm0, %xmm4 +; XOP-NEXT: vpmaxud %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512-LABEL: vec256_i32_unsigned_reg_reg: ; AVX512: # %bb.0: @@ -191,25 +157,25 @@ define <8 x i32> @vec256_i32_unsigned_reg_reg(<8 x i32> %a1, <8 x i32> %a2) noun ; Values are loaded. Only check signed case. define <8 x i32> @vec256_i32_signed_mem_reg(ptr %a1_addr, <8 x i32> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i32_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm0, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm1, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm4, %xmm3 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmulld %xmm3, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmulld %xmm0, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddd %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddd %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i32_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2 +; AVX1-NEXT: vpminsd %xmm0, %xmm1, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm0, %xmm1, %xmm4 +; AVX1-NEXT: vpsubd %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0 +; AVX1-NEXT: vpminsd %xmm0, %xmm2, %xmm4 +; AVX1-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vpsubd %xmm4, %xmm0, %xmm0 +; AVX1-NEXT: vpsrld $1, %xmm0, %xmm0 +; AVX1-NEXT: vpsrld $1, %xmm3, %xmm3 +; AVX1-NEXT: vpmulld %xmm3, %xmm3, %xmm3 +; AVX1-NEXT: vpmulld %xmm0, %xmm0, %xmm0 +; AVX1-NEXT: vpaddd %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vpaddd %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i32_signed_mem_reg: ; AVX2: # %bb.0: @@ -222,41 +188,23 @@ define <8 x i32> @vec256_i32_signed_mem_reg(ptr %a1_addr, <8 x i32> %a2) nounwin ; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i32_signed_mem_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i32_signed_mem_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrld $1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i32_signed_mem_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpminsd %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpminsd %xmm0, %xmm2, %xmm4 +; XOP-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vpsubd %xmm4, %xmm0, %xmm0 +; XOP-NEXT: vpsrld $1, %xmm0, %xmm0 +; XOP-NEXT: vpsrld $1, %xmm1, %xmm1 +; XOP-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 +; XOP-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512-LABEL: vec256_i32_signed_mem_reg: ; AVX512: # %bb.0: @@ -281,25 +229,25 @@ define <8 x i32> @vec256_i32_signed_mem_reg(ptr %a1_addr, <8 x i32> %a2) nounwin } define <8 x i32> @vec256_i32_signed_reg_mem(<8 x i32> %a1, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i32_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpminsd %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmulld %xmm2, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddd %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i32_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2 +; AVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm3 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpminsd %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpsubd %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpsrld $1, %xmm2, %xmm2 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm1, %xmm1, %xmm1 +; AVX1-NEXT: vpmulld %xmm2, %xmm2, %xmm2 +; AVX1-NEXT: vpaddd %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i32_signed_reg_mem: ; AVX2: # %bb.0: @@ -312,41 +260,23 @@ define <8 x i32> @vec256_i32_signed_reg_mem(<8 x i32> %a1, ptr %a2_addr) nounwin ; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i32_signed_reg_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpminsd %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminsd %xmm1, %xmm0, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i32_signed_reg_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm1 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpminsd %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminsd %xmm1, %xmm0, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrld $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i32_signed_reg_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rdi), %xmm1 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpminsd %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpmaxsd %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubd %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpminsd %xmm1, %xmm0, %xmm4 +; XOP-NEXT: vpmaxsd %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrld $1, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm3, %xmm2, %xmm2, %xmm2 +; XOP-NEXT: vpmacsdd %xmm0, %xmm1, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512-LABEL: vec256_i32_signed_reg_mem: ; AVX512: # %bb.0: @@ -371,26 +301,26 @@ define <8 x i32> @vec256_i32_signed_reg_mem(<8 x i32> %a1, ptr %a2_addr) nounwin } define <8 x i32> @vec256_i32_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i32_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmulld %xmm0, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmulld %xmm1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddd %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i32_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rsi), %xmm0 +; AVX1-NEXT: vmovdqa 16(%rsi), %xmm1 +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vpminsd %xmm0, %xmm2, %xmm4 +; AVX1-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vpsubd %xmm4, %xmm0, %xmm0 +; AVX1-NEXT: vpminsd %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm1, %xmm1 +; AVX1-NEXT: vpsrld $1, %xmm0, %xmm0 +; AVX1-NEXT: vpmulld %xmm0, %xmm0, %xmm0 +; AVX1-NEXT: vpmulld %xmm1, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddd %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i32_signed_mem_mem: ; AVX2: # %bb.0: @@ -404,43 +334,24 @@ define <8 x i32> @vec256_i32_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i32_signed_mem_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; XOP-FALLBACK-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i32_signed_mem_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rsi), %xmm0 -; XOPAVX1-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpminsd %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsd %xmm0, %xmm2, %xmm4 -; XOPAVX1-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vpsubd %xmm4, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrld $1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrld $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i32_signed_mem_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rsi), %xmm0 +; XOP-NEXT: vmovdqa 16(%rsi), %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpminsd %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpmaxsd %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpsubd %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpminsd %xmm0, %xmm2, %xmm4 +; XOP-NEXT: vpmaxsd %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vpsubd %xmm4, %xmm0, %xmm0 +; XOP-NEXT: vpsrld $1, %xmm0, %xmm0 +; XOP-NEXT: vpsrld $1, %xmm1, %xmm1 +; XOP-NEXT: vpmacsdd %xmm3, %xmm1, %xmm1, %xmm1 +; XOP-NEXT: vpmacsdd %xmm2, %xmm0, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512-LABEL: vec256_i32_signed_mem_mem: ; AVX512: # %bb.0: @@ -473,43 +384,43 @@ define <8 x i32> @vec256_i32_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; Values come from regs define <4 x i64> @vec256_i64_signed_reg_reg(<4 x i64> %a1, <4 x i64> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i64_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i64_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm5 +; AVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm6 +; AVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlq $1, %xmm2, %xmm6 +; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 +; AVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; AVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 +; AVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; AVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; AVX1-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; AVX1-NEXT: vpsrlq $33, %xmm2, %xmm2 +; AVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 +; AVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; AVX1-NEXT: vpaddq %xmm2, %xmm7, %xmm2 +; AVX1-NEXT: vpsllq $32, %xmm2, %xmm2 +; AVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; AVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpaddq %xmm0, %xmm5, %xmm0 +; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i64_signed_reg_reg: ; AVX2: # %bb.0: @@ -531,81 +442,43 @@ define <4 x i64> @vec256_i64_signed_reg_reg(<4 x i64> %a1, <4 x i64> %a2) nounwi ; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i64_signed_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOP-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOP-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i64_signed_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOPAVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOPAVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOPAVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOPAVX1-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOPAVX1-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOPAVX1-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOPAVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i64_signed_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; XOP-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsubq %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpsubq %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $1, %xmm2, %xmm6 +; XOP-NEXT: vpsrlq $1, %xmm1, %xmm7 +; XOP-NEXT: vpsrlq $33, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; XOP-NEXT: vpor %xmm5, %xmm8, %xmm5 +; XOP-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $32, %xmm5, %xmm9 +; XOP-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; XOP-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; XOP-NEXT: vpsllq $32, %xmm1, %xmm1 +; XOP-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; XOP-NEXT: vpsrlq $33, %xmm2, %xmm2 +; XOP-NEXT: vpor %xmm4, %xmm8, %xmm4 +; XOP-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $32, %xmm4, %xmm7 +; XOP-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; XOP-NEXT: vpaddq %xmm2, %xmm7, %xmm2 +; XOP-NEXT: vpsllq $32, %xmm2, %xmm2 +; XOP-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; XOP-NEXT: vpaddq %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpaddq %xmm0, %xmm5, %xmm0 +; XOP-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i64_signed_reg_reg: ; AVX512F: # %bb.0: @@ -676,48 +549,48 @@ define <4 x i64> @vec256_i64_signed_reg_reg(<4 x i64> %a1, <4 x i64> %a2) nounwi } define <4 x i64> @vec256_i64_unsigned_reg_reg(<4 x i64> %a1, <4 x i64> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i64_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808] -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm3, %xmm5 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpxor %xmm4, %xmm0, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm9 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm7, %xmm4 -; AVX1-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm7 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm7, %xmm3 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm6, %xmm5 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm4, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i64_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm4, %xmm3, %xmm5 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 +; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm6 +; AVX1-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 +; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm6 +; AVX1-NEXT: vpxor %xmm4, %xmm0, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm6 +; AVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm3 +; AVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlq $1, %xmm3, %xmm6 +; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; AVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm9 +; AVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; AVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; AVX1-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm4, %xmm7, %xmm4 +; AVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpsrlq $33, %xmm3, %xmm3 +; AVX1-NEXT: vpmuludq %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlq $32, %xmm5, %xmm7 +; AVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; AVX1-NEXT: vpaddq %xmm3, %xmm7, %xmm3 +; AVX1-NEXT: vpsllq $32, %xmm3, %xmm3 +; AVX1-NEXT: vpmuludq %xmm5, %xmm6, %xmm5 +; AVX1-NEXT: vpaddq %xmm2, %xmm5, %xmm2 +; AVX1-NEXT: vpaddq %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpaddq %xmm0, %xmm4, %xmm0 +; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i64_unsigned_reg_reg: ; AVX2: # %bb.0: @@ -742,81 +615,43 @@ define <4 x i64> @vec256_i64_unsigned_reg_reg(<4 x i64> %a1, <4 x i64> %a2) noun ; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i64_unsigned_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtuq %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtuq %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOP-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOP-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i64_unsigned_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtuq %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtuq %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOPAVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOPAVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOPAVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOPAVX1-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOPAVX1-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOPAVX1-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOPAVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i64_unsigned_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtuq %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtuq %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; XOP-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsubq %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpsubq %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $1, %xmm2, %xmm6 +; XOP-NEXT: vpsrlq $1, %xmm1, %xmm7 +; XOP-NEXT: vpsrlq $33, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; XOP-NEXT: vpor %xmm5, %xmm8, %xmm5 +; XOP-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $32, %xmm5, %xmm9 +; XOP-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; XOP-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; XOP-NEXT: vpsllq $32, %xmm1, %xmm1 +; XOP-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; XOP-NEXT: vpsrlq $33, %xmm2, %xmm2 +; XOP-NEXT: vpor %xmm4, %xmm8, %xmm4 +; XOP-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $32, %xmm4, %xmm7 +; XOP-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; XOP-NEXT: vpaddq %xmm2, %xmm7, %xmm2 +; XOP-NEXT: vpsllq $32, %xmm2, %xmm2 +; XOP-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; XOP-NEXT: vpaddq %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpaddq %xmm0, %xmm5, %xmm0 +; XOP-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i64_unsigned_reg_reg: ; AVX512F: # %bb.0: @@ -889,44 +724,44 @@ define <4 x i64> @vec256_i64_unsigned_reg_reg(<4 x i64> %a1, <4 x i64> %a2) noun ; Values are loaded. Only check signed case. define <4 x i64> @vec256_i64_signed_mem_reg(ptr %a1_addr, <4 x i64> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i64_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm0, %xmm2, %xmm5 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm7 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i64_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm0, %xmm2, %xmm5 +; AVX1-NEXT: vpsubq %xmm0, %xmm2, %xmm6 +; AVX1-NEXT: vpsubq %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpsubq %xmm1, %xmm3, %xmm6 +; AVX1-NEXT: vpsubq %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm6 +; AVX1-NEXT: vpsrlq $1, %xmm0, %xmm7 +; AVX1-NEXT: vpsrlq $33, %xmm0, %xmm0 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; AVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 +; AVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; AVX1-NEXT: vpaddq %xmm0, %xmm9, %xmm0 +; AVX1-NEXT: vpsllq $32, %xmm0, %xmm0 +; AVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; AVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 +; AVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; AVX1-NEXT: vpaddq %xmm1, %xmm7, %xmm1 +; AVX1-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; AVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpaddq %xmm2, %xmm5, %xmm2 +; AVX1-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i64_signed_mem_reg: ; AVX2: # %bb.0: @@ -949,83 +784,44 @@ define <4 x i64> @vec256_i64_signed_mem_reg(ptr %a1_addr, <4 x i64> %a2) nounwin ; AVX2-NEXT: vpaddq %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i64_signed_mem_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 -; XOP-FALLBACK-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm6 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm7 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOP-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOP-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i64_signed_mem_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 -; XOPAVX1-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; XOPAVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; XOPAVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm6 -; XOPAVX1-NEXT: vpsrlq $1, %xmm0, %xmm7 -; XOPAVX1-NEXT: vpsrlq $33, %xmm0, %xmm0 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOPAVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; XOPAVX1-NEXT: vpsllq $32, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOPAVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; XOPAVX1-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOPAVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i64_signed_mem_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 +; XOP-NEXT: vpsubq %xmm0, %xmm2, %xmm6 +; XOP-NEXT: vpsubq %xmm2, %xmm0, %xmm0 +; XOP-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpsubq %xmm1, %xmm3, %xmm6 +; XOP-NEXT: vpsubq %xmm3, %xmm1, %xmm1 +; XOP-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $1, %xmm1, %xmm6 +; XOP-NEXT: vpsrlq $1, %xmm0, %xmm7 +; XOP-NEXT: vpsrlq $33, %xmm0, %xmm0 +; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; XOP-NEXT: vpor %xmm5, %xmm8, %xmm5 +; XOP-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vpsrlq $32, %xmm5, %xmm9 +; XOP-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; XOP-NEXT: vpaddq %xmm0, %xmm9, %xmm0 +; XOP-NEXT: vpsllq $32, %xmm0, %xmm0 +; XOP-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; XOP-NEXT: vpsrlq $33, %xmm1, %xmm1 +; XOP-NEXT: vpor %xmm4, %xmm8, %xmm4 +; XOP-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $32, %xmm4, %xmm7 +; XOP-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; XOP-NEXT: vpaddq %xmm1, %xmm7, %xmm1 +; XOP-NEXT: vpsllq $32, %xmm1, %xmm1 +; XOP-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; XOP-NEXT: vpaddq %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpaddq %xmm2, %xmm5, %xmm2 +; XOP-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i64_signed_mem_reg: ; AVX512F: # %bb.0: @@ -1098,44 +894,44 @@ define <4 x i64> @vec256_i64_signed_mem_reg(ptr %a1_addr, <4 x i64> %a2) nounwin } define <4 x i64> @vec256_i64_signed_reg_mem(<4 x i64> %a1, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i64_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i64_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm5 +; AVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; AVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm6 +; AVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlq $1, %xmm2, %xmm6 +; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 +; AVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; AVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 +; AVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; AVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; AVX1-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; AVX1-NEXT: vpsrlq $33, %xmm2, %xmm2 +; AVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 +; AVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; AVX1-NEXT: vpaddq %xmm2, %xmm7, %xmm2 +; AVX1-NEXT: vpsllq $32, %xmm2, %xmm2 +; AVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; AVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpaddq %xmm0, %xmm5, %xmm0 +; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i64_signed_reg_mem: ; AVX2: # %bb.0: @@ -1158,83 +954,44 @@ define <4 x i64> @vec256_i64_signed_reg_mem(<4 x i64> %a1, ptr %a2_addr) nounwin ; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i64_signed_reg_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOP-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOP-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i64_signed_reg_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm1 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpsubq %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm0, %xmm1, %xmm1 -; XOPAVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsubq %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $1, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm7 -; XOPAVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOPAVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm9, %xmm1 -; XOPAVX1-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOPAVX1-NEXT: vpsrlq $33, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm7, %xmm2 -; XOPAVX1-NEXT: vpsllq $32, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOPAVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm5, %xmm0 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i64_signed_reg_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rdi), %xmm1 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtq %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtq %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpsubq %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpsubq %xmm0, %xmm1, %xmm1 +; XOP-NEXT: vblendvpd %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsubq %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpsubq %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vblendvpd %xmm4, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $1, %xmm2, %xmm6 +; XOP-NEXT: vpsrlq $1, %xmm1, %xmm7 +; XOP-NEXT: vpsrlq $33, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; XOP-NEXT: vpor %xmm5, %xmm8, %xmm5 +; XOP-NEXT: vpmuludq %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $32, %xmm5, %xmm9 +; XOP-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; XOP-NEXT: vpaddq %xmm1, %xmm9, %xmm1 +; XOP-NEXT: vpsllq $32, %xmm1, %xmm1 +; XOP-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; XOP-NEXT: vpsrlq $33, %xmm2, %xmm2 +; XOP-NEXT: vpor %xmm4, %xmm8, %xmm4 +; XOP-NEXT: vpmuludq %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpsrlq $32, %xmm4, %xmm7 +; XOP-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; XOP-NEXT: vpaddq %xmm2, %xmm7, %xmm2 +; XOP-NEXT: vpsllq $32, %xmm2, %xmm2 +; XOP-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; XOP-NEXT: vpaddq %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpaddq %xmm0, %xmm5, %xmm0 +; XOP-NEXT: vpaddq %xmm1, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i64_signed_reg_mem: ; AVX512F: # %bb.0: @@ -1307,45 +1064,45 @@ define <4 x i64> @vec256_i64_signed_reg_mem(<4 x i64> %a1, ptr %a2_addr) nounwin } define <4 x i64> @vec256_i64_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i64_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtq %xmm0, %xmm2, %xmm5 -; AVX1-FALLBACK-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm7 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; AVX1-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; AVX1-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; AVX1-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i64_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rsi), %xmm0 +; AVX1-NEXT: vmovdqa 16(%rsi), %xmm1 +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtq %xmm0, %xmm2, %xmm5 +; AVX1-NEXT: vpsubq %xmm0, %xmm2, %xmm6 +; AVX1-NEXT: vpsubq %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpsubq %xmm1, %xmm3, %xmm6 +; AVX1-NEXT: vpsubq %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $1, %xmm1, %xmm6 +; AVX1-NEXT: vpsrlq $1, %xmm0, %xmm7 +; AVX1-NEXT: vpsrlq $33, %xmm0, %xmm0 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; AVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 +; AVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; AVX1-NEXT: vpaddq %xmm0, %xmm9, %xmm0 +; AVX1-NEXT: vpsllq $32, %xmm0, %xmm0 +; AVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; AVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 +; AVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; AVX1-NEXT: vpaddq %xmm1, %xmm7, %xmm1 +; AVX1-NEXT: vpsllq $32, %xmm1, %xmm1 +; AVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; AVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpaddq %xmm2, %xmm5, %xmm2 +; AVX1-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i64_signed_mem_mem: ; AVX2: # %bb.0: @@ -1369,85 +1126,45 @@ define <4 x i64> @vec256_i64_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i64_signed_mem_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 -; XOP-FALLBACK-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm1, %xmm6 -; XOP-FALLBACK-NEXT: vpsrlq $1, %xmm0, %xmm7 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOP-FALLBACK-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOP-FALLBACK-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOP-FALLBACK-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; XOP-FALLBACK-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOP-FALLBACK-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOP-FALLBACK-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; XOP-FALLBACK-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i64_signed_mem_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rsi), %xmm0 -; XOPAVX1-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 -; XOPAVX1-NEXT: vpsubq %xmm0, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm2, %xmm0, %xmm0 -; XOPAVX1-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsubq %xmm1, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpsubq %xmm3, %xmm1, %xmm1 -; XOPAVX1-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $1, %xmm1, %xmm6 -; XOPAVX1-NEXT: vpsrlq $1, %xmm0, %xmm7 -; XOPAVX1-NEXT: vpsrlq $33, %xmm0, %xmm0 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] -; XOPAVX1-NEXT: vpor %xmm5, %xmm8, %xmm5 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlq $32, %xmm5, %xmm9 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm9, %xmm0 -; XOPAVX1-NEXT: vpsllq $32, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 -; XOPAVX1-NEXT: vpsrlq $33, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpor %xmm4, %xmm8, %xmm4 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlq $32, %xmm4, %xmm7 -; XOPAVX1-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm7, %xmm1 -; XOPAVX1-NEXT: vpsllq $32, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 -; XOPAVX1-NEXT: vpaddq %xmm3, %xmm4, %xmm3 -; XOPAVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpaddq %xmm2, %xmm5, %xmm2 -; XOPAVX1-NEXT: vpaddq %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i64_signed_mem_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rsi), %xmm0 +; XOP-NEXT: vmovdqa 16(%rsi), %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtq %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtq %xmm0, %xmm2, %xmm5 +; XOP-NEXT: vpsubq %xmm0, %xmm2, %xmm6 +; XOP-NEXT: vpsubq %xmm2, %xmm0, %xmm0 +; XOP-NEXT: vblendvpd %xmm5, %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpsubq %xmm1, %xmm3, %xmm6 +; XOP-NEXT: vpsubq %xmm3, %xmm1, %xmm1 +; XOP-NEXT: vblendvpd %xmm4, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $1, %xmm1, %xmm6 +; XOP-NEXT: vpsrlq $1, %xmm0, %xmm7 +; XOP-NEXT: vpsrlq $33, %xmm0, %xmm0 +; XOP-NEXT: vmovdqa {{.*#+}} xmm8 = [1,1] +; XOP-NEXT: vpor %xmm5, %xmm8, %xmm5 +; XOP-NEXT: vpmuludq %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vpsrlq $32, %xmm5, %xmm9 +; XOP-NEXT: vpmuludq %xmm7, %xmm9, %xmm9 +; XOP-NEXT: vpaddq %xmm0, %xmm9, %xmm0 +; XOP-NEXT: vpsllq $32, %xmm0, %xmm0 +; XOP-NEXT: vpmuludq %xmm5, %xmm7, %xmm5 +; XOP-NEXT: vpsrlq $33, %xmm1, %xmm1 +; XOP-NEXT: vpor %xmm4, %xmm8, %xmm4 +; XOP-NEXT: vpmuludq %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpsrlq $32, %xmm4, %xmm7 +; XOP-NEXT: vpmuludq %xmm7, %xmm6, %xmm7 +; XOP-NEXT: vpaddq %xmm1, %xmm7, %xmm1 +; XOP-NEXT: vpsllq $32, %xmm1, %xmm1 +; XOP-NEXT: vpmuludq %xmm4, %xmm6, %xmm4 +; XOP-NEXT: vpaddq %xmm3, %xmm4, %xmm3 +; XOP-NEXT: vpaddq %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpaddq %xmm2, %xmm5, %xmm2 +; XOP-NEXT: vpaddq %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i64_signed_mem_mem: ; AVX512F: # %bb.0: @@ -1528,29 +1245,29 @@ define <4 x i64> @vec256_i64_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; Values come from regs define <16 x i16> @vec256_i16_signed_reg_reg(<16 x i16> %a1, <16 x i16> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i16_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i16_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm5 +; AVX1-NEXT: vpminsw %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpminsw %xmm2, %xmm3, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpsubw %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i16_signed_reg_reg: ; AVX2: # %bb.0: @@ -1564,49 +1281,27 @@ define <16 x i16> @vec256_i16_signed_reg_reg(<16 x i16> %a1, <16 x i16> %a2) nou ; AVX2-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i16_signed_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i16_signed_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i16_signed_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminsw %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubw %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpminsw %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm2, %xmm2 +; XOP-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm6, %xmm5, %xmm5 +; XOP-NEXT: vpor %xmm6, %xmm4, %xmm4 +; XOP-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i16_signed_reg_reg: ; AVX512F: # %bb.0: @@ -1671,32 +1366,32 @@ define <16 x i16> @vec256_i16_signed_reg_reg(<16 x i16> %a1, <16 x i16> %a2) nou } define <16 x i16> @vec256_i16_unsigned_reg_reg(<16 x i16> %a1, <16 x i16> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i16_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpminuw %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpeqw %xmm4, %xmm3, %xmm5 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; AVX1-FALLBACK-NEXT: vpxor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpminuw %xmm1, %xmm0, %xmm7 -; AVX1-FALLBACK-NEXT: vpcmpeqw %xmm7, %xmm0, %xmm8 -; AVX1-FALLBACK-NEXT: vpxor %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm7, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmaxuw %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubw %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm6, %xmm6 -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpor %xmm4, %xmm5, %xmm4 -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i16_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpminuw %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpeqw %xmm4, %xmm3, %xmm5 +; AVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; AVX1-NEXT: vpxor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpminuw %xmm1, %xmm0, %xmm7 +; AVX1-NEXT: vpcmpeqw %xmm7, %xmm0, %xmm8 +; AVX1-NEXT: vpxor %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubw %xmm7, %xmm1, %xmm1 +; AVX1-NEXT: vpmaxuw %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpsubw %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm4, %xmm6, %xmm6 +; AVX1-NEXT: vpmullw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm4, %xmm5, %xmm4 +; AVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i16_unsigned_reg_reg: ; AVX2: # %bb.0: @@ -1712,49 +1407,27 @@ define <16 x i16> @vec256_i16_unsigned_reg_reg(<16 x i16> %a1, <16 x i16> %a2) n ; AVX2-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i16_unsigned_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtuw %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtuw %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminuw %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxuw %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminuw %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i16_unsigned_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtuw %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtuw %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminuw %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxuw %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminuw %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i16_unsigned_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtuw %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtuw %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminuw %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxuw %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubw %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpminuw %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxuw %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm2, %xmm2 +; XOP-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm6, %xmm5, %xmm5 +; XOP-NEXT: vpor %xmm6, %xmm4, %xmm4 +; XOP-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i16_unsigned_reg_reg: ; AVX512F: # %bb.0: @@ -1823,30 +1496,30 @@ define <16 x i16> @vec256_i16_unsigned_reg_reg(<16 x i16> %a1, <16 x i16> %a2) n ; Values are loaded. Only check signed case. define <16 x i16> @vec256_i16_signed_mem_reg(ptr %a1_addr, <16 x i16> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i16_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm0, %xmm2, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i16_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vpcmpgtw %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtw %xmm0, %xmm2, %xmm5 +; AVX1-NEXT: vpminsw %xmm0, %xmm2, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vpsubw %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpminsw %xmm1, %xmm3, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpmullw %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpmullw %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpaddw %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddw %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i16_signed_mem_reg: ; AVX2: # %bb.0: @@ -1861,51 +1534,28 @@ define <16 x i16> @vec256_i16_signed_mem_reg(ptr %a1_addr, <16 x i16> %a2) nounw ; AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i16_signed_mem_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 -; XOP-FALLBACK-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i16_signed_mem_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 -; XOPAVX1-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i16_signed_mem_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 +; XOP-NEXT: vpminsw %xmm1, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpminsw %xmm0, %xmm2, %xmm6 +; XOP-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vpsubw %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpsrlw $1, %xmm0, %xmm0 +; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm6, %xmm5, %xmm5 +; XOP-NEXT: vpor %xmm6, %xmm4, %xmm4 +; XOP-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i16_signed_mem_reg: ; AVX512F: # %bb.0: @@ -1974,30 +1624,30 @@ define <16 x i16> @vec256_i16_signed_mem_reg(ptr %a1_addr, <16 x i16> %a2) nounw } define <16 x i16> @vec256_i16_signed_reg_mem(<16 x i16> %a1, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i16_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm3, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i16_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm5 +; AVX1-NEXT: vpminsw %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpminsw %xmm2, %xmm3, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpsubw %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; AVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm3, %xmm2, %xmm2 +; AVX1-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i16_signed_reg_mem: ; AVX2: # %bb.0: @@ -2012,51 +1662,28 @@ define <16 x i16> @vec256_i16_signed_reg_mem(<16 x i16> %a1, ptr %a2_addr) nounw ; AVX2-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i16_signed_reg_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i16_signed_reg_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm1 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminsw %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpminsw %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i16_signed_reg_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rdi), %xmm1 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtw %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtw %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminsw %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsw %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubw %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpminsw %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxsw %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1 +; XOP-NEXT: vpsrlw $1, %xmm2, %xmm2 +; XOP-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm6, %xmm5, %xmm5 +; XOP-NEXT: vpor %xmm6, %xmm4, %xmm4 +; XOP-NEXT: vpmacsww %xmm3, %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpmacsww %xmm0, %xmm5, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i16_signed_reg_mem: ; AVX512F: # %bb.0: @@ -2125,31 +1752,31 @@ define <16 x i16> @vec256_i16_signed_reg_mem(<16 x i16> %a1, ptr %a2_addr) nounw } define <16 x i16> @vec256_i16_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i16_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm1, %xmm3, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtw %xmm0, %xmm2, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm3, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddw %xmm2, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i16_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rsi), %xmm0 +; AVX1-NEXT: vmovdqa 16(%rsi), %xmm1 +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vpcmpgtw %xmm1, %xmm3, %xmm4 +; AVX1-NEXT: vpcmpgtw %xmm0, %xmm2, %xmm5 +; AVX1-NEXT: vpminsw %xmm0, %xmm2, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vpsubw %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpminsw %xmm1, %xmm3, %xmm6 +; AVX1-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpmullw %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpmullw %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpaddw %xmm3, %xmm1, %xmm1 +; AVX1-NEXT: vpaddw %xmm2, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i16_signed_mem_mem: ; AVX2: # %bb.0: @@ -2165,53 +1792,29 @@ define <16 x i16> @vec256_i16_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwin ; AVX2-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i16_signed_mem_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 -; XOP-FALLBACK-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i16_signed_mem_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rsi), %xmm0 -; XOPAVX1-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 -; XOPAVX1-NEXT: vpminsw %xmm1, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsw %xmm0, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vpsubw %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm6, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpor %xmm6, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i16_signed_mem_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rsi), %xmm0 +; XOP-NEXT: vmovdqa 16(%rsi), %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtw %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtw %xmm0, %xmm2, %xmm5 +; XOP-NEXT: vpminsw %xmm1, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsw %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpsubw %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpminsw %xmm0, %xmm2, %xmm6 +; XOP-NEXT: vpmaxsw %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vpsubw %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpsrlw $1, %xmm0, %xmm0 +; XOP-NEXT: vpsrlw $1, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm6 = [1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm6, %xmm5, %xmm5 +; XOP-NEXT: vpor %xmm6, %xmm4, %xmm4 +; XOP-NEXT: vpmacsww %xmm3, %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpmacsww %xmm2, %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i16_signed_mem_mem: ; AVX512F: # %bb.0: @@ -2290,49 +1893,49 @@ define <16 x i16> @vec256_i16_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwin ; Values come from regs define <32 x i8> @vec256_i8_signed_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i8_signed_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm3 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm3, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpminsb %xmm3, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm3, %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm1, %xmm8, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpand %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpand %xmm3, %xmm8, %xmm3 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddb %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i8_signed_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 +; AVX1-NEXT: vpcmpgtb %xmm3, %xmm2, %xmm4 +; AVX1-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm5 +; AVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpminsb %xmm3, %xmm2, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm3, %xmm2, %xmm3 +; AVX1-NEXT: vpsubb %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] +; AVX1-NEXT: vpand %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; AVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm1, %xmm8, %xmm1 +; AVX1-NEXT: vpackuswb %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpand %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpand %xmm3, %xmm8, %xmm3 +; AVX1-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i8_signed_reg_reg: ; AVX2: # %bb.0: @@ -2356,81 +1959,43 @@ define <32 x i8> @vec256_i8_signed_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounwin ; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i8_signed_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i8_signed_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOPAVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOPAVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i8_signed_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminsb %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpminsb %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; XOP-NEXT: vpshlb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpshlb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm7, %xmm5, %xmm5 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; XOP-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpor %xmm7, %xmm4, %xmm4 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm7, %xmm6, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; XOP-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i8_signed_reg_reg: ; AVX512F: # %bb.0: @@ -2510,52 +2075,52 @@ define <32 x i8> @vec256_i8_signed_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounwin } define <32 x i8> @vec256_i8_unsigned_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i8_unsigned_reg_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm3 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpminub %xmm3, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpeqb %xmm4, %xmm2, %xmm5 -; AVX1-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; AVX1-FALLBACK-NEXT: vpxor %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpminub %xmm1, %xmm0, %xmm7 -; AVX1-FALLBACK-NEXT: vpcmpeqb %xmm7, %xmm0, %xmm8 -; AVX1-FALLBACK-NEXT: vpxor %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpsubb %xmm7, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpmaxub %xmm3, %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubb %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm6, %xmm6 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm6 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpand %xmm1, %xmm8, %xmm1 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm4, %xmm1, %xmm1 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpand %xmm4, %xmm8, %xmm4 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpand %xmm3, %xmm8, %xmm3 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddb %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i8_unsigned_reg_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 +; AVX1-NEXT: vpminub %xmm3, %xmm2, %xmm4 +; AVX1-NEXT: vpcmpeqb %xmm4, %xmm2, %xmm5 +; AVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; AVX1-NEXT: vpxor %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpminub %xmm1, %xmm0, %xmm7 +; AVX1-NEXT: vpcmpeqb %xmm7, %xmm0, %xmm8 +; AVX1-NEXT: vpxor %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 +; AVX1-NEXT: vpsubb %xmm7, %xmm1, %xmm1 +; AVX1-NEXT: vpmaxub %xmm3, %xmm2, %xmm3 +; AVX1-NEXT: vpsubb %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] +; AVX1-NEXT: vpand %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm7, %xmm6, %xmm6 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm6 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero +; AVX1-NEXT: vpmullw %xmm6, %xmm1, %xmm1 +; AVX1-NEXT: vpand %xmm1, %xmm8, %xmm1 +; AVX1-NEXT: vpackuswb %xmm4, %xmm1, %xmm1 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm4, %xmm4 +; AVX1-NEXT: vpand %xmm4, %xmm8, %xmm4 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; AVX1-NEXT: vpmullw %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpand %xmm3, %xmm8, %xmm3 +; AVX1-NEXT: vpackuswb %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i8_unsigned_reg_reg: ; AVX2: # %bb.0: @@ -2581,81 +2146,43 @@ define <32 x i8> @vec256_i8_unsigned_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounw ; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i8_unsigned_reg_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtub %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtub %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminub %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminub %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxub %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i8_unsigned_reg_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtub %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtub %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminub %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminub %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxub %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOPAVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOPAVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i8_unsigned_reg_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm1, %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtub %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtub %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminub %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxub %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpminub %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxub %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; XOP-NEXT: vpshlb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpshlb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm7, %xmm5, %xmm5 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; XOP-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpor %xmm7, %xmm4, %xmm4 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm7, %xmm6, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; XOP-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i8_unsigned_reg_reg: ; AVX512F: # %bb.0: @@ -2739,50 +2266,50 @@ define <32 x i8> @vec256_i8_unsigned_reg_reg(<32 x i8> %a1, <32 x i8> %a2) nounw ; Values are loaded. Only check signed case. define <32 x i8> @vec256_i8_signed_mem_reg(ptr %a1_addr, <32 x i8> %a2) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i8_signed_mem_reg: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm3, %xmm2, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsb %xmm0, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpminsb %xmm3, %xmm2, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm3, %xmm2, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpand %xmm0, %xmm8, %xmm0 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm6, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpand %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpand %xmm3, %xmm8, %xmm3 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddb %xmm2, %xmm3, %xmm2 -; AVX1-FALLBACK-NEXT: vpaddb %xmm1, %xmm0, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i8_signed_mem_reg: +; AVX1: # %bb.0: +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vmovdqa (%rdi), %xmm1 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2 +; AVX1-NEXT: vpcmpgtb %xmm3, %xmm2, %xmm4 +; AVX1-NEXT: vpcmpgtb %xmm0, %xmm1, %xmm5 +; AVX1-NEXT: vpminsb %xmm0, %xmm1, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 +; AVX1-NEXT: vpsubb %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpminsb %xmm3, %xmm2, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm3, %xmm2, %xmm3 +; AVX1-NEXT: vpsubb %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] +; AVX1-NEXT: vpand %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm0, %xmm0 +; AVX1-NEXT: vpand %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; AVX1-NEXT: vpmullw %xmm5, %xmm0, %xmm0 +; AVX1-NEXT: vpand %xmm0, %xmm8, %xmm0 +; AVX1-NEXT: vpackuswb %xmm6, %xmm0, %xmm0 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpand %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpand %xmm3, %xmm8, %xmm3 +; AVX1-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2 +; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i8_signed_mem_reg: ; AVX2: # %bb.0: @@ -2807,83 +2334,44 @@ define <32 x i8> @vec256_i8_signed_mem_reg(ptr %a1_addr, <32 x i8> %a2) nounwind ; AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i8_signed_mem_reg: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm0, %xmm1, %xmm5 -; XOP-FALLBACK-NEXT: vpminsb %xmm0, %xmm1, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm1, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i8_signed_mem_reg: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm1 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtb %xmm0, %xmm1, %xmm5 -; XOPAVX1-NEXT: vpminsb %xmm0, %xmm1, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOPAVX1-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOPAVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i8_signed_mem_reg: +; XOP: # %bb.0: +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm2 +; XOP-NEXT: vmovdqa (%rdi), %xmm1 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtb %xmm0, %xmm1, %xmm5 +; XOP-NEXT: vpminsb %xmm0, %xmm1, %xmm6 +; XOP-NEXT: vpmaxsb %xmm0, %xmm1, %xmm0 +; XOP-NEXT: vpsubb %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpminsb %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; XOP-NEXT: vpshlb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpshlb %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm7, %xmm5, %xmm5 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; XOP-NEXT: vpmullw %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpor %xmm7, %xmm4, %xmm4 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm7, %xmm6, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; XOP-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm1, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i8_signed_mem_reg: ; AVX512F: # %bb.0: @@ -2967,50 +2455,50 @@ define <32 x i8> @vec256_i8_signed_mem_reg(ptr %a1_addr, <32 x i8> %a2) nounwind } define <32 x i8> @vec256_i8_signed_reg_mem(<32 x i8> %a1, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i8_signed_reg_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; AVX1-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm3, %xmm1, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm2, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsb %xmm2, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm2, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm3, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm3, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpand %xmm2, %xmm8, %xmm2 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpand %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpand %xmm3, %xmm8, %xmm3 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddb %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i8_signed_reg_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rdi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm3 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 +; AVX1-NEXT: vpcmpgtb %xmm3, %xmm1, %xmm4 +; AVX1-NEXT: vpcmpgtb %xmm2, %xmm0, %xmm5 +; AVX1-NEXT: vpminsb %xmm2, %xmm0, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm2, %xmm0, %xmm2 +; AVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm3, %xmm1, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm3, %xmm1, %xmm3 +; AVX1-NEXT: vpsubb %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] +; AVX1-NEXT: vpand %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; AVX1-NEXT: vpmullw %xmm5, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm2, %xmm8, %xmm2 +; AVX1-NEXT: vpackuswb %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpand %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpand %xmm3, %xmm8, %xmm3 +; AVX1-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i8_signed_reg_mem: ; AVX2: # %bb.0: @@ -3035,83 +2523,44 @@ define <32 x i8> @vec256_i8_signed_reg_mem(<32 x i8> %a1, ptr %a2_addr) nounwind ; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i8_signed_reg_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 -; XOP-FALLBACK-NEXT: vpminsb %xmm1, %xmm0, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOP-FALLBACK-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i8_signed_reg_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm1 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm2 -; XOPAVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; XOPAVX1-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 -; XOPAVX1-NEXT: vpminsb %xmm1, %xmm0, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpminsb %xmm2, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOPAVX1-NEXT: vpmullw %xmm5, %xmm1, %xmm1 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOPAVX1-NEXT: vpmullw %xmm4, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm3, %xmm2, %xmm2 -; XOPAVX1-NEXT: vpaddb %xmm0, %xmm1, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i8_signed_reg_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rdi), %xmm1 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm2 +; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 +; XOP-NEXT: vpcomgtb %xmm2, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtb %xmm1, %xmm0, %xmm5 +; XOP-NEXT: vpminsb %xmm1, %xmm0, %xmm6 +; XOP-NEXT: vpmaxsb %xmm1, %xmm0, %xmm1 +; XOP-NEXT: vpsubb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpminsb %xmm2, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsb %xmm2, %xmm3, %xmm2 +; XOP-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; XOP-NEXT: vpshlb %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpshlb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm7, %xmm5, %xmm5 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; XOP-NEXT: vpmullw %xmm5, %xmm1, %xmm1 +; XOP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpor %xmm7, %xmm4, %xmm4 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm7, %xmm6, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; XOP-NEXT: vpmullw %xmm4, %xmm2, %xmm2 +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm3, %xmm2, %xmm2 +; XOP-NEXT: vpaddb %xmm0, %xmm1, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i8_signed_reg_mem: ; AVX512F: # %bb.0: @@ -3195,51 +2644,51 @@ define <32 x i8> @vec256_i8_signed_reg_mem(<32 x i8> %a1, ptr %a2_addr) nounwind } define <32 x i8> @vec256_i8_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind { -; AVX1-FALLBACK-LABEL: vec256_i8_signed_mem_mem: -; AVX1-FALLBACK: # %bb.0: -; AVX1-FALLBACK-NEXT: vmovdqa (%rsi), %xmm2 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa (%rdi), %xmm0 -; AVX1-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm1 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm3, %xmm1, %xmm4 -; AVX1-FALLBACK-NEXT: vpcmpgtb %xmm2, %xmm0, %xmm5 -; AVX1-FALLBACK-NEXT: vpminsb %xmm2, %xmm0, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm2, %xmm0, %xmm2 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpminsb %xmm3, %xmm1, %xmm6 -; AVX1-FALLBACK-NEXT: vpmaxsb %xmm3, %xmm1, %xmm3 -; AVX1-FALLBACK-NEXT: vpsubb %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpsrlw $1, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] -; AVX1-FALLBACK-NEXT: vpand %xmm6, %xmm8, %xmm6 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm5, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpand %xmm2, %xmm8, %xmm2 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm6, %xmm2, %xmm2 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; AVX1-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; AVX1-FALLBACK-NEXT: vpmullw %xmm6, %xmm5, %xmm5 -; AVX1-FALLBACK-NEXT: vpand %xmm5, %xmm8, %xmm5 -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero -; AVX1-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; AVX1-FALLBACK-NEXT: vpmullw %xmm4, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpand %xmm3, %xmm8, %xmm3 -; AVX1-FALLBACK-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 -; AVX1-FALLBACK-NEXT: vpaddb %xmm1, %xmm3, %xmm1 -; AVX1-FALLBACK-NEXT: vpaddb %xmm0, %xmm2, %xmm0 -; AVX1-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-FALLBACK-NEXT: retq +; AVX1-LABEL: vec256_i8_signed_mem_mem: +; AVX1: # %bb.0: +; AVX1-NEXT: vmovdqa (%rsi), %xmm2 +; AVX1-NEXT: vmovdqa 16(%rsi), %xmm3 +; AVX1-NEXT: vmovdqa (%rdi), %xmm0 +; AVX1-NEXT: vmovdqa 16(%rdi), %xmm1 +; AVX1-NEXT: vpcmpgtb %xmm3, %xmm1, %xmm4 +; AVX1-NEXT: vpcmpgtb %xmm2, %xmm0, %xmm5 +; AVX1-NEXT: vpminsb %xmm2, %xmm0, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm2, %xmm0, %xmm2 +; AVX1-NEXT: vpsubb %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpminsb %xmm3, %xmm1, %xmm6 +; AVX1-NEXT: vpmaxsb %xmm3, %xmm1, %xmm3 +; AVX1-NEXT: vpsubb %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm3, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127] +; AVX1-NEXT: vpand %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vpsrlw $1, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; AVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255] +; AVX1-NEXT: vpand %xmm6, %xmm8, %xmm6 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; AVX1-NEXT: vpmullw %xmm5, %xmm2, %xmm2 +; AVX1-NEXT: vpand %xmm2, %xmm8, %xmm2 +; AVX1-NEXT: vpackuswb %xmm6, %xmm2, %xmm2 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 +; AVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; AVX1-NEXT: vpmullw %xmm6, %xmm5, %xmm5 +; AVX1-NEXT: vpand %xmm5, %xmm8, %xmm5 +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero +; AVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; AVX1-NEXT: vpmullw %xmm4, %xmm3, %xmm3 +; AVX1-NEXT: vpand %xmm3, %xmm8, %xmm3 +; AVX1-NEXT: vpackuswb %xmm5, %xmm3, %xmm3 +; AVX1-NEXT: vpaddb %xmm1, %xmm3, %xmm1 +; AVX1-NEXT: vpaddb %xmm0, %xmm2, %xmm0 +; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; AVX1-NEXT: retq ; ; AVX2-LABEL: vec256_i8_signed_mem_mem: ; AVX2: # %bb.0: @@ -3265,85 +2714,45 @@ define <32 x i8> @vec256_i8_signed_mem_mem(ptr %a1_addr, ptr %a2_addr) nounwind ; AVX2-NEXT: vpaddb %ymm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; XOP-FALLBACK-LABEL: vec256_i8_signed_mem_mem: -; XOP-FALLBACK: # %bb.0: -; XOP-FALLBACK-NEXT: vmovdqa (%rsi), %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOP-FALLBACK-NEXT: vmovdqa (%rdi), %xmm2 -; XOP-FALLBACK-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm1, %xmm3, %xmm4 -; XOP-FALLBACK-NEXT: vpcomgtb %xmm0, %xmm2, %xmm5 -; XOP-FALLBACK-NEXT: vpminsb %xmm0, %xmm2, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm0, %xmm2, %xmm0 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpminsb %xmm1, %xmm3, %xmm6 -; XOP-FALLBACK-NEXT: vpmaxsb %xmm1, %xmm3, %xmm1 -; XOP-FALLBACK-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpshlb %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOP-FALLBACK-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOP-FALLBACK-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOP-FALLBACK-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOP-FALLBACK-NEXT: vpmullw %xmm4, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpaddb %xmm3, %xmm1, %xmm1 -; XOP-FALLBACK-NEXT: vpaddb %xmm2, %xmm0, %xmm0 -; XOP-FALLBACK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOP-FALLBACK-NEXT: retq -; -; XOPAVX1-LABEL: vec256_i8_signed_mem_mem: -; XOPAVX1: # %bb.0: -; XOPAVX1-NEXT: vmovdqa (%rsi), %xmm0 -; XOPAVX1-NEXT: vmovdqa 16(%rsi), %xmm1 -; XOPAVX1-NEXT: vmovdqa (%rdi), %xmm2 -; XOPAVX1-NEXT: vmovdqa 16(%rdi), %xmm3 -; XOPAVX1-NEXT: vpcomgtb %xmm1, %xmm3, %xmm4 -; XOPAVX1-NEXT: vpcomgtb %xmm0, %xmm2, %xmm5 -; XOPAVX1-NEXT: vpminsb %xmm0, %xmm2, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm0, %xmm2, %xmm0 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpminsb %xmm1, %xmm3, %xmm6 -; XOPAVX1-NEXT: vpmaxsb %xmm1, %xmm3, %xmm1 -; XOPAVX1-NEXT: vpsubb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpshlb %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] -; XOPAVX1-NEXT: vpor %xmm7, %xmm5, %xmm5 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm6, %xmm8, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero -; XOPAVX1-NEXT: vpmullw %xmm5, %xmm0, %xmm0 -; XOPAVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpor %xmm7, %xmm4, %xmm4 -; XOPAVX1-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] -; XOPAVX1-NEXT: vpmullw %xmm7, %xmm6, %xmm6 -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero -; XOPAVX1-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero -; XOPAVX1-NEXT: vpmullw %xmm4, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1 -; XOPAVX1-NEXT: vpaddb %xmm2, %xmm0, %xmm0 -; XOPAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; XOPAVX1-NEXT: retq +; XOP-LABEL: vec256_i8_signed_mem_mem: +; XOP: # %bb.0: +; XOP-NEXT: vmovdqa (%rsi), %xmm0 +; XOP-NEXT: vmovdqa 16(%rsi), %xmm1 +; XOP-NEXT: vmovdqa (%rdi), %xmm2 +; XOP-NEXT: vmovdqa 16(%rdi), %xmm3 +; XOP-NEXT: vpcomgtb %xmm1, %xmm3, %xmm4 +; XOP-NEXT: vpcomgtb %xmm0, %xmm2, %xmm5 +; XOP-NEXT: vpminsb %xmm0, %xmm2, %xmm6 +; XOP-NEXT: vpmaxsb %xmm0, %xmm2, %xmm0 +; XOP-NEXT: vpsubb %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpminsb %xmm1, %xmm3, %xmm6 +; XOP-NEXT: vpmaxsb %xmm1, %xmm3, %xmm1 +; XOP-NEXT: vpsubb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpcmpeqd %xmm6, %xmm6, %xmm6 +; XOP-NEXT: vpshlb %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpshlb %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vmovdqa {{.*#+}} xmm7 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] +; XOP-NEXT: vpor %xmm7, %xmm5, %xmm5 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm8 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm6, %xmm8, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero +; XOP-NEXT: vpmullw %xmm5, %xmm0, %xmm0 +; XOP-NEXT: vmovdqa {{.*#+}} xmm5 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30] +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm0, %xmm0 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpor %xmm7, %xmm4, %xmm4 +; XOP-NEXT: vpunpckhbw {{.*#+}} xmm7 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15] +; XOP-NEXT: vpmullw %xmm7, %xmm6, %xmm6 +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero +; XOP-NEXT: vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero +; XOP-NEXT: vpmullw %xmm4, %xmm1, %xmm1 +; XOP-NEXT: vpperm %xmm5, %xmm6, %xmm1, %xmm1 +; XOP-NEXT: vpaddb %xmm3, %xmm1, %xmm1 +; XOP-NEXT: vpaddb %xmm2, %xmm0, %xmm0 +; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 +; XOP-NEXT: retq ; ; AVX512F-LABEL: vec256_i8_signed_mem_mem: ; AVX512F: # %bb.0: