Index: lib/Transforms/Vectorize/BBVectorize.cpp =================================================================== --- lib/Transforms/Vectorize/BBVectorize.cpp +++ lib/Transforms/Vectorize/BBVectorize.cpp @@ -675,6 +675,14 @@ case Intrinsic::exp: case Intrinsic::exp2: case Intrinsic::pow: + case Intrinsic::round: + case Intrinsic::copysign: + case Intrinsic::ceil: + case Intrinsic::nearbyint: + case Intrinsic::rint: + case Intrinsic::trunc: + case Intrinsic::floor: + case Intrinsic::fabs: return Config.VectorizeMath; case Intrinsic::fma: case Intrinsic::fmuladd: Index: test/Transforms/BBVectorize/simple-int.ll =================================================================== --- test/Transforms/BBVectorize/simple-int.ll +++ test/Transforms/BBVectorize/simple-int.ll @@ -5,6 +5,14 @@ declare double @llvm.fmuladd.f64(double, double, double) declare double @llvm.cos.f64(double) declare double @llvm.powi.f64(double, i32) +declare double @llvm.round.f64(double) +declare double @llvm.copysign.f64(double, double) +declare double @llvm.ceil.f64(double) +declare double @llvm.nearbyint.f64(double) +declare double @llvm.rint.f64(double) +declare double @llvm.trunc.f64(double) +declare double @llvm.floor.f64(double) +declare double @llvm.fabs.f64(double) ; Basic depth-3 chain with fma define double @test1(double %A1, double %A2, double %B1, double %B2, double %C1, double %C2) { @@ -124,9 +132,218 @@ ; CHECK: ret double %R } +; Basic depth-3 chain with round +define double @testround(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.round.f64(double %X1) + %Y2 = call double @llvm.round.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testround +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.round.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with copysign +define double @testcopysign(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.copysign.f64(double %X1, double %A1) + %Y2 = call double @llvm.copysign.f64(double %X2, double %A1) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testcopysign +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1.v.i1.2 = insertelement <2 x double> %X1.v.i0.1, double %A1, i32 1 +; CHECK: %Y1 = call <2 x double> @llvm.copysign.v2f64(<2 x double> %X1, <2 x double> %Y1.v.i1.2) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with ceil +define double @testceil(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.ceil.f64(double %X1) + %Y2 = call double @llvm.ceil.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testceil +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.ceil.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with nearbyint +define double @testnearbyint(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.nearbyint.f64(double %X1) + %Y2 = call double @llvm.nearbyint.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testnearbyint +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with rint +define double @testrint(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.rint.f64(double %X1) + %Y2 = call double @llvm.rint.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testrint +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.rint.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with trunc +define double @testtrunc(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.trunc.f64(double %X1) + %Y2 = call double @llvm.trunc.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testtrunc +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.trunc.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with floor +define double @testfloor(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.floor.f64(double %X1) + %Y2 = call double @llvm.floor.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testfloor +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.floor.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + +; Basic depth-3 chain with fabs +define double @testfabs(double %A1, double %A2, double %B1, double %B2) { + %X1 = fsub double %A1, %B1 + %X2 = fsub double %A2, %B2 + %Y1 = call double @llvm.fabs.f64(double %X1) + %Y2 = call double @llvm.fabs.f64(double %X2) + %Z1 = fadd double %Y1, %B1 + %Z2 = fadd double %Y2, %B2 + %R = fmul double %Z1, %Z2 + ret double %R +; CHECK: @testfabs +; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0 +; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1 +; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0 +; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1 +; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2 +; CHECK: %Y1 = call <2 x double> @llvm.fabs.v2f64(<2 x double> %X1) +; CHECK: %Z1 = fadd <2 x double> %Y1, %X1.v.i1.2 +; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0 +; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1 +; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2 +; CHECK: ret double %R + +} + + ; CHECK: declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>) #0 ; CHECK: declare <2 x double> @llvm.fmuladd.v2f64(<2 x double>, <2 x double>, <2 x double>) #0 ; CHECK: declare <2 x double> @llvm.cos.v2f64(<2 x double>) #0 ; CHECK: declare <2 x double> @llvm.powi.v2f64(<2 x double>, i32) #0 - +; CHECK: declare <2 x double> @llvm.round.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.copysign.v2f64(<2 x double>, <2 x double>) #0 +; CHECK: declare <2 x double> @llvm.ceil.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.nearbyint.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.rint.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.trunc.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.floor.v2f64(<2 x double>) #0 +; CHECK: declare <2 x double> @llvm.fabs.v2f64(<2 x double>) #0 ; CHECK: attributes #0 = { nounwind readnone }