From 19b3ce0f0729d964458f59012cffdd18012a1fd8 Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Sat, 20 Feb 2016 17:11:32 +0000 Subject: [PATCH] [X86][SSE41] Added fast-isel intrinsics tests As discussed on PR24580, this patch adds some (more to come) initial fast-isel codegen tests to match the IR generated in clang/test/CodeGen/sse41-builtins.c llvm-svn: 261438 --- .../test/CodeGen/X86/sse41-intrinsics-fast-isel.ll | 380 +++++++++++++++++++++ 1 file changed, 380 insertions(+) create mode 100644 llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll diff --git a/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll new file mode 100644 index 0000000..87a91e2 --- /dev/null +++ b/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll @@ -0,0 +1,380 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=X32 +; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=X64 + +; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse41-builtins.c + +define <2 x i64> @test_mm_blend_epi16(<2 x i64> %a0, <2 x i64> %a1) { +; X32-LABEL: test_mm_blend_epi16: +; X32: # BB#0: +; X32-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6,7] +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blend_epi16: +; X64: # BB#0: +; X64-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6,7] +; X64-NEXT: retq + %arg0 = bitcast <2 x i64> %a0 to <8 x i16> + %arg1 = bitcast <2 x i64> %a1 to <8 x i16> + %shuf = shufflevector <8 x i16> %arg0, <8 x i16> %arg1, <8 x i32> + %res = bitcast <8 x i16> %shuf to <2 x i64> + ret <2 x i64> %res +} + +define <2 x double> @test_mm_blend_pd(<2 x double> %a0, <2 x double> %a1) { +; X32-LABEL: test_mm_blend_pd: +; X32: # BB#0: +; X32-NEXT: blendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blend_pd: +; X64: # BB#0: +; X64-NEXT: blendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] +; X64-NEXT: retq + %res = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> + ret <2 x double> %res +} + +define <4 x float> @test_mm_blend_ps(<4 x float> %a0, <4 x float> %a1) { +; X32-LABEL: test_mm_blend_ps: +; X32: # BB#0: +; X32-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blend_ps: +; X64: # BB#0: +; X64-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] +; X64-NEXT: retq + %res = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> + ret <4 x float> %res +} + +define <2 x i64> @test_mm_blendv_epi8(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> %a2) { +; X32-LABEL: test_mm_blendv_epi8: +; X32: # BB#0: +; X32-NEXT: movdqa %xmm0, %xmm3 +; X32-NEXT: movaps %xmm2, %xmm0 +; X32-NEXT: pblendvb %xmm1, %xmm3 +; X32-NEXT: movdqa %xmm3, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blendv_epi8: +; X64: # BB#0: +; X64-NEXT: movdqa %xmm0, %xmm3 +; X64-NEXT: movaps %xmm2, %xmm0 +; X64-NEXT: pblendvb %xmm1, %xmm3 +; X64-NEXT: movdqa %xmm3, %xmm0 +; X64-NEXT: retq + %arg0 = bitcast <2 x i64> %a0 to <16 x i8> + %arg1 = bitcast <2 x i64> %a1 to <16 x i8> + %arg2 = bitcast <2 x i64> %a2 to <16 x i8> + %call = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %arg0, <16 x i8> %arg1, <16 x i8> %arg2) + %res = bitcast <16 x i8> %call to <2 x i64> + ret <2 x i64> %res +} +declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone + +define <2 x double> @test_mm_blendv_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) { +; X32-LABEL: test_mm_blendv_pd: +; X32: # BB#0: +; X32-NEXT: movapd %xmm0, %xmm3 +; X32-NEXT: movaps %xmm2, %xmm0 +; X32-NEXT: blendvpd %xmm1, %xmm3 +; X32-NEXT: movapd %xmm3, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blendv_pd: +; X64: # BB#0: +; X64-NEXT: movapd %xmm0, %xmm3 +; X64-NEXT: movaps %xmm2, %xmm0 +; X64-NEXT: blendvpd %xmm1, %xmm3 +; X64-NEXT: movapd %xmm3, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) + ret <2 x double> %res +} +declare <2 x double> @llvm.x86.sse41.blendvpd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone + +define <4 x float> @test_mm_blendv_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) { +; X32-LABEL: test_mm_blendv_ps: +; X32: # BB#0: +; X32-NEXT: movaps %xmm0, %xmm3 +; X32-NEXT: movaps %xmm2, %xmm0 +; X32-NEXT: blendvps %xmm1, %xmm3 +; X32-NEXT: movaps %xmm3, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_blendv_ps: +; X64: # BB#0: +; X64-NEXT: movaps %xmm0, %xmm3 +; X64-NEXT: movaps %xmm2, %xmm0 +; X64-NEXT: blendvps %xmm1, %xmm3 +; X64-NEXT: movaps %xmm3, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) + ret <4 x float> %res +} +declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone + +define <2 x double> @test_mm_ceil_pd(<2 x double> %a0) { +; X32-LABEL: test_mm_ceil_pd: +; X32: # BB#0: +; X32-NEXT: roundpd $10, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_ceil_pd: +; X64: # BB#0: +; X64-NEXT: roundpd $10, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 10) + ret <2 x double> %res +} +declare <2 x double> @llvm.x86.sse41.round.pd(<2 x double>, i32) nounwind readnone + +define <4 x float> @test_mm_ceil_ps(<4 x float> %a0) { +; X32-LABEL: test_mm_ceil_ps: +; X32: # BB#0: +; X32-NEXT: roundps $10, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_ceil_ps: +; X64: # BB#0: +; X64-NEXT: roundps $10, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 10) + ret <4 x float> %res +} +declare <4 x float> @llvm.x86.sse41.round.ps(<4 x float>, i32) nounwind readnone + +define <2 x double> @test_mm_ceil_sd(<2 x double> %a0, <2 x double> %a1) { +; X32-LABEL: test_mm_ceil_sd: +; X32: # BB#0: +; X32-NEXT: roundsd $10, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_ceil_sd: +; X64: # BB#0: +; X64-NEXT: roundsd $10, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 10) + ret <2 x double> %res +} +declare <2 x double> @llvm.x86.sse41.round.sd(<2 x double>, <2 x double>, i32) nounwind readnone + +define <4 x float> @test_mm_ceil_ss(<4 x float> %a0, <4 x float> %a1) { +; X32-LABEL: test_mm_ceil_ss: +; X32: # BB#0: +; X32-NEXT: roundss $10, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_ceil_ss: +; X64: # BB#0: +; X64-NEXT: roundss $10, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 10) + ret <4 x float> %res +} +declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) nounwind readnone + +define <2 x i64> @test_mm_cmpeq_epi64(<2 x i64> %a0, <2 x i64> %a1) { +; X32-LABEL: test_mm_cmpeq_epi64: +; X32: # BB#0: +; X32-NEXT: pcmpeqq %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_cmpeq_epi64: +; X64: # BB#0: +; X64-NEXT: pcmpeqq %xmm1, %xmm0 +; X64-NEXT: retq + %cmp = icmp eq <2 x i64> %a0, %a1 + %res = sext <2 x i1> %cmp to <2 x i64> + ret <2 x i64> %res +} + +; TODO test_mm_cvtepi8_epi16 +; TODO test_mm_cvtepi8_epi32 +; TODO test_mm_cvtepi8_epi64 +; TODO test_mm_cvtepi16_epi32 +; TODO test_mm_cvtepi16_epi64 +; TODO test_mm_cvtepi32_epi64 + +; TODO test_mm_cvtepu8_epi16 +; TODO test_mm_cvtepu8_epi32 +; TODO test_mm_cvtepu8_epi64 +; TODO test_mm_cvtepu16_epi32 +; TODO test_mm_cvtepu16_epi64 +; TODO test_mm_cvtepu32_epi64 + +define <2 x double> @test_mm_dp_pd(<2 x double> %a0, <2 x double> %a1) { +; X32-LABEL: test_mm_dp_pd: +; X32: # BB#0: +; X32-NEXT: dppd $7, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_dp_pd: +; X64: # BB#0: +; X64-NEXT: dppd $7, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i8 7) + ret <2 x double> %res +} +declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i8) nounwind readnone + +define <4 x float> @test_mm_dp_ps(<4 x float> %a0, <4 x float> %a1) { +; X32-LABEL: test_mm_dp_ps: +; X32: # BB#0: +; X32-NEXT: dpps $7, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_dp_ps: +; X64: # BB#0: +; X64-NEXT: dpps $7, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i8 7) + ret <4 x float> %res +} +declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i8) nounwind readnone + +; TODO test_mm_extract_epi8 +; TODO test_mm_extract_epi32 +; TODO test_mm_extract_epi64 +; TODO test_mm_extract_ps + +define <2 x double> @test_mm_floor_pd(<2 x double> %a0) { +; X32-LABEL: test_mm_floor_pd: +; X32: # BB#0: +; X32-NEXT: roundpd $9, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_floor_pd: +; X64: # BB#0: +; X64-NEXT: roundpd $9, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 9) + ret <2 x double> %res +} + +define <4 x float> @test_mm_floor_ps(<4 x float> %a0) { +; X32-LABEL: test_mm_floor_ps: +; X32: # BB#0: +; X32-NEXT: roundps $9, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_floor_ps: +; X64: # BB#0: +; X64-NEXT: roundps $9, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 9) + ret <4 x float> %res +} + +define <2 x double> @test_mm_floor_sd(<2 x double> %a0, <2 x double> %a1) { +; X32-LABEL: test_mm_floor_sd: +; X32: # BB#0: +; X32-NEXT: roundsd $9, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_floor_sd: +; X64: # BB#0: +; X64-NEXT: roundsd $9, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 9) + ret <2 x double> %res +} + +define <4 x float> @test_mm_floor_ss(<4 x float> %a0, <4 x float> %a1) { +; X32-LABEL: test_mm_floor_ss: +; X32: # BB#0: +; X32-NEXT: roundss $9, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_floor_ss: +; X64: # BB#0: +; X64-NEXT: roundss $9, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 9) + ret <4 x float> %res +} + +; TODO test_mm_insert_epi8 +; TODO test_mm_insert_epi32 +; TODO test_mm_insert_epi64 +; TODO test_mm_insert_ps + +; TODO test_mm_max_epi8 +; TODO test_mm_max_epu16 +; TODO test_mm_max_epi32 +; TODO test_mm_max_epu32 +; TODO test_mm_min_epi8 +; TODO test_mm_min_epu16 +; TODO test_mm_min_epi32 +; TODO test_mm_min_epu32 +; TODO test_mm_minpos_epu16 +; TODO test_mm_mpsadbw_epu8 +; TODO test_mm_mul_epi32 +; TODO test_mm_mullo_epi32 +; TODO test_mm_packus_epi32 + +define <2 x double> @test_mm_round_pd(<2 x double> %a0) { +; X32-LABEL: test_mm_round_pd: +; X32: # BB#0: +; X32-NEXT: roundpd $2, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_round_pd: +; X64: # BB#0: +; X64-NEXT: roundpd $2, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 2) + ret <2 x double> %res +} + +define <4 x float> @test_mm_round_ps(<4 x float> %a0) { +; X32-LABEL: test_mm_round_ps: +; X32: # BB#0: +; X32-NEXT: roundps $2, %xmm0, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_round_ps: +; X64: # BB#0: +; X64-NEXT: roundps $2, %xmm0, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 2) + ret <4 x float> %res +} + +define <2 x double> @test_mm_round_sd(<2 x double> %a0, <2 x double> %a1) { +; X32-LABEL: test_mm_round_sd: +; X32: # BB#0: +; X32-NEXT: roundsd $2, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_round_sd: +; X64: # BB#0: +; X64-NEXT: roundsd $2, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 2) + ret <2 x double> %res +} + +define <4 x float> @test_mm_round_ss(<4 x float> %a0, <4 x float> %a1) { +; X32-LABEL: test_mm_round_ss: +; X32: # BB#0: +; X32-NEXT: roundss $2, %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: test_mm_round_ss: +; X64: # BB#0: +; X64-NEXT: roundss $2, %xmm1, %xmm0 +; X64-NEXT: retq + %res = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 2) + ret <4 x float> %res +} + +; TODO test_mm_stream_load_si128 +; TODO test_mm_test_all_ones +; TODO test_mm_test_all_zeros +; TODO test_mm_test_mix_ones_zeros +; TODO test_mm_testc_si128 +; TODO test_mm_testnzc_si128 +; TODO test_mm_testz_si128 -- 2.7.4