From: Simon Pilgrim Date: Thu, 28 Mar 2019 19:03:53 +0000 (+0000) Subject: [SLP][X86] Add tests showing failure to commute icmp/fcmp operands X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=66b5e322fcc337b65c580d733d04ab41572d643d;p=platform%2Fupstream%2Fllvm.git [SLP][X86] Add tests showing failure to commute icmp/fcmp operands Some predicates are fully commutative - we should be able to easily commute their operands to improve vectorization llvm-svn: 357202 --- diff --git a/llvm/test/Transforms/SLPVectorizer/X86/cmp_commute.ll b/llvm/test/Transforms/SLPVectorizer/X86/cmp_commute.ll new file mode 100644 index 0000000..b51834e --- /dev/null +++ b/llvm/test/Transforms/SLPVectorizer/X86/cmp_commute.ll @@ -0,0 +1,199 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64--- -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE +; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64--- -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX + +; +; Check that we can commute operands based on the predicate. +; + +define <4 x i32> @icmp_eq_v4i32(<4 x i32> %a, i32* %b) { +; CHECK-LABEL: @icmp_eq_v4i32( +; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x i32> [[A:%.*]], i32 0 +; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x i32> [[A]], i32 1 +; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x i32> [[A]], i32 2 +; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x i32> [[A]], i32 3 +; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 1 +; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2 +; CHECK-NEXT: [[P3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3 +; CHECK-NEXT: [[B0:%.*]] = load i32, i32* [[B]], align 4 +; CHECK-NEXT: [[B1:%.*]] = load i32, i32* [[P1]], align 4 +; CHECK-NEXT: [[B2:%.*]] = load i32, i32* [[P2]], align 4 +; CHECK-NEXT: [[B3:%.*]] = load i32, i32* [[P3]], align 4 +; CHECK-NEXT: [[C0:%.*]] = icmp eq i32 [[A0]], [[B0]] +; CHECK-NEXT: [[C1:%.*]] = icmp eq i32 [[B1]], [[A1]] +; CHECK-NEXT: [[C2:%.*]] = icmp eq i32 [[B2]], [[A2]] +; CHECK-NEXT: [[C3:%.*]] = icmp eq i32 [[A3]], [[B3]] +; CHECK-NEXT: [[D0:%.*]] = insertelement <4 x i1> undef, i1 [[C0]], i32 0 +; CHECK-NEXT: [[D1:%.*]] = insertelement <4 x i1> [[D0]], i1 [[C1]], i32 1 +; CHECK-NEXT: [[D2:%.*]] = insertelement <4 x i1> [[D1]], i1 [[C2]], i32 2 +; CHECK-NEXT: [[D3:%.*]] = insertelement <4 x i1> [[D2]], i1 [[C3]], i32 3 +; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[D3]] to <4 x i32> +; CHECK-NEXT: ret <4 x i32> [[R]] +; + %a0 = extractelement <4 x i32> %a, i32 0 + %a1 = extractelement <4 x i32> %a, i32 1 + %a2 = extractelement <4 x i32> %a, i32 2 + %a3 = extractelement <4 x i32> %a, i32 3 + %p0 = getelementptr inbounds i32, i32* %b, i32 0 + %p1 = getelementptr inbounds i32, i32* %b, i32 1 + %p2 = getelementptr inbounds i32, i32* %b, i32 2 + %p3 = getelementptr inbounds i32, i32* %b, i32 3 + %b0 = load i32, i32* %p0, align 4 + %b1 = load i32, i32* %p1, align 4 + %b2 = load i32, i32* %p2, align 4 + %b3 = load i32, i32* %p3, align 4 + %c0 = icmp eq i32 %a0, %b0 + %c1 = icmp eq i32 %b1, %a1 + %c2 = icmp eq i32 %b2, %a2 + %c3 = icmp eq i32 %a3, %b3 + %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0 + %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1 + %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2 + %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3 + %r = sext <4 x i1> %d3 to <4 x i32> + ret <4 x i32> %r +} + +define <4 x i32> @icmp_ne_v4i32(<4 x i32> %a, i32* %b) { +; CHECK-LABEL: @icmp_ne_v4i32( +; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x i32> [[A:%.*]], i32 0 +; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x i32> [[A]], i32 1 +; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x i32> [[A]], i32 2 +; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x i32> [[A]], i32 3 +; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 1 +; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2 +; CHECK-NEXT: [[P3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3 +; CHECK-NEXT: [[B0:%.*]] = load i32, i32* [[B]], align 4 +; CHECK-NEXT: [[B1:%.*]] = load i32, i32* [[P1]], align 4 +; CHECK-NEXT: [[B2:%.*]] = load i32, i32* [[P2]], align 4 +; CHECK-NEXT: [[B3:%.*]] = load i32, i32* [[P3]], align 4 +; CHECK-NEXT: [[C0:%.*]] = icmp ne i32 [[A0]], [[B0]] +; CHECK-NEXT: [[C1:%.*]] = icmp ne i32 [[B1]], [[A1]] +; CHECK-NEXT: [[C2:%.*]] = icmp ne i32 [[B2]], [[A2]] +; CHECK-NEXT: [[C3:%.*]] = icmp ne i32 [[A3]], [[B3]] +; CHECK-NEXT: [[D0:%.*]] = insertelement <4 x i1> undef, i1 [[C0]], i32 0 +; CHECK-NEXT: [[D1:%.*]] = insertelement <4 x i1> [[D0]], i1 [[C1]], i32 1 +; CHECK-NEXT: [[D2:%.*]] = insertelement <4 x i1> [[D1]], i1 [[C2]], i32 2 +; CHECK-NEXT: [[D3:%.*]] = insertelement <4 x i1> [[D2]], i1 [[C3]], i32 3 +; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[D3]] to <4 x i32> +; CHECK-NEXT: ret <4 x i32> [[R]] +; + %a0 = extractelement <4 x i32> %a, i32 0 + %a1 = extractelement <4 x i32> %a, i32 1 + %a2 = extractelement <4 x i32> %a, i32 2 + %a3 = extractelement <4 x i32> %a, i32 3 + %p0 = getelementptr inbounds i32, i32* %b, i32 0 + %p1 = getelementptr inbounds i32, i32* %b, i32 1 + %p2 = getelementptr inbounds i32, i32* %b, i32 2 + %p3 = getelementptr inbounds i32, i32* %b, i32 3 + %b0 = load i32, i32* %p0, align 4 + %b1 = load i32, i32* %p1, align 4 + %b2 = load i32, i32* %p2, align 4 + %b3 = load i32, i32* %p3, align 4 + %c0 = icmp ne i32 %a0, %b0 + %c1 = icmp ne i32 %b1, %a1 + %c2 = icmp ne i32 %b2, %a2 + %c3 = icmp ne i32 %a3, %b3 + %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0 + %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1 + %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2 + %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3 + %r = sext <4 x i1> %d3 to <4 x i32> + ret <4 x i32> %r +} + +define <4 x i32> @fcmp_oeq_v4i32(<4 x float> %a, float* %b) { +; CHECK-LABEL: @fcmp_oeq_v4i32( +; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> [[A:%.*]], i32 0 +; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> [[A]], i32 1 +; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x float> [[A]], i32 2 +; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x float> [[A]], i32 3 +; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 1 +; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2 +; CHECK-NEXT: [[P3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3 +; CHECK-NEXT: [[B0:%.*]] = load float, float* [[B]], align 4 +; CHECK-NEXT: [[B1:%.*]] = load float, float* [[P1]], align 4 +; CHECK-NEXT: [[B2:%.*]] = load float, float* [[P2]], align 4 +; CHECK-NEXT: [[B3:%.*]] = load float, float* [[P3]], align 4 +; CHECK-NEXT: [[C0:%.*]] = fcmp oeq float [[A0]], [[B0]] +; CHECK-NEXT: [[C1:%.*]] = fcmp oeq float [[B1]], [[A1]] +; CHECK-NEXT: [[C2:%.*]] = fcmp oeq float [[B2]], [[A2]] +; CHECK-NEXT: [[C3:%.*]] = fcmp oeq float [[A3]], [[B3]] +; CHECK-NEXT: [[D0:%.*]] = insertelement <4 x i1> undef, i1 [[C0]], i32 0 +; CHECK-NEXT: [[D1:%.*]] = insertelement <4 x i1> [[D0]], i1 [[C1]], i32 1 +; CHECK-NEXT: [[D2:%.*]] = insertelement <4 x i1> [[D1]], i1 [[C2]], i32 2 +; CHECK-NEXT: [[D3:%.*]] = insertelement <4 x i1> [[D2]], i1 [[C3]], i32 3 +; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[D3]] to <4 x i32> +; CHECK-NEXT: ret <4 x i32> [[R]] +; + %a0 = extractelement <4 x float> %a, i32 0 + %a1 = extractelement <4 x float> %a, i32 1 + %a2 = extractelement <4 x float> %a, i32 2 + %a3 = extractelement <4 x float> %a, i32 3 + %p0 = getelementptr inbounds float, float* %b, i32 0 + %p1 = getelementptr inbounds float, float* %b, i32 1 + %p2 = getelementptr inbounds float, float* %b, i32 2 + %p3 = getelementptr inbounds float, float* %b, i32 3 + %b0 = load float, float* %p0, align 4 + %b1 = load float, float* %p1, align 4 + %b2 = load float, float* %p2, align 4 + %b3 = load float, float* %p3, align 4 + %c0 = fcmp oeq float %a0, %b0 + %c1 = fcmp oeq float %b1, %a1 + %c2 = fcmp oeq float %b2, %a2 + %c3 = fcmp oeq float %a3, %b3 + %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0 + %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1 + %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2 + %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3 + %r = sext <4 x i1> %d3 to <4 x i32> + ret <4 x i32> %r +} + +define <4 x i32> @fcmp_uno_v4i32(<4 x float> %a, float* %b) { +; CHECK-LABEL: @fcmp_uno_v4i32( +; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> [[A:%.*]], i32 0 +; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> [[A]], i32 1 +; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x float> [[A]], i32 2 +; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x float> [[A]], i32 3 +; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 1 +; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2 +; CHECK-NEXT: [[P3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3 +; CHECK-NEXT: [[B0:%.*]] = load float, float* [[B]], align 4 +; CHECK-NEXT: [[B1:%.*]] = load float, float* [[P1]], align 4 +; CHECK-NEXT: [[B2:%.*]] = load float, float* [[P2]], align 4 +; CHECK-NEXT: [[B3:%.*]] = load float, float* [[P3]], align 4 +; CHECK-NEXT: [[C0:%.*]] = fcmp uno float [[A0]], [[B0]] +; CHECK-NEXT: [[C1:%.*]] = fcmp uno float [[B1]], [[A1]] +; CHECK-NEXT: [[C2:%.*]] = fcmp uno float [[B2]], [[A2]] +; CHECK-NEXT: [[C3:%.*]] = fcmp uno float [[A3]], [[B3]] +; CHECK-NEXT: [[D0:%.*]] = insertelement <4 x i1> undef, i1 [[C0]], i32 0 +; CHECK-NEXT: [[D1:%.*]] = insertelement <4 x i1> [[D0]], i1 [[C1]], i32 1 +; CHECK-NEXT: [[D2:%.*]] = insertelement <4 x i1> [[D1]], i1 [[C2]], i32 2 +; CHECK-NEXT: [[D3:%.*]] = insertelement <4 x i1> [[D2]], i1 [[C3]], i32 3 +; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[D3]] to <4 x i32> +; CHECK-NEXT: ret <4 x i32> [[R]] +; + %a0 = extractelement <4 x float> %a, i32 0 + %a1 = extractelement <4 x float> %a, i32 1 + %a2 = extractelement <4 x float> %a, i32 2 + %a3 = extractelement <4 x float> %a, i32 3 + %p0 = getelementptr inbounds float, float* %b, i32 0 + %p1 = getelementptr inbounds float, float* %b, i32 1 + %p2 = getelementptr inbounds float, float* %b, i32 2 + %p3 = getelementptr inbounds float, float* %b, i32 3 + %b0 = load float, float* %p0, align 4 + %b1 = load float, float* %p1, align 4 + %b2 = load float, float* %p2, align 4 + %b3 = load float, float* %p3, align 4 + %c0 = fcmp uno float %a0, %b0 + %c1 = fcmp uno float %b1, %a1 + %c2 = fcmp uno float %b2, %a2 + %c3 = fcmp uno float %a3, %b3 + %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0 + %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1 + %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2 + %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3 + %r = sext <4 x i1> %d3 to <4 x i32> + ret <4 x i32> %r +}