1 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128" 2 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -bb-vectorize -bb-vectorize-req-chain-depth=3 -instcombine -gvn -S | FileCheck %s 3 4 declare double @llvm.fma.f64(double, double, double) 5 declare double @llvm.fmuladd.f64(double, double, double) 6 declare double @llvm.cos.f64(double) 7 declare double @llvm.powi.f64(double, i32) 8 9 ; Basic depth-3 chain with fma 10 define double @test1(double %A1, double %A2, double %B1, double %B2, double %C1, double %C2) { 11 %X1 = fsub double %A1, %B1 12 %X2 = fsub double %A2, %B2 13 %Y1 = call double @llvm.fma.f64(double %X1, double %A1, double %C1) 14 %Y2 = call double @llvm.fma.f64(double %X2, double %A2, double %C2) 15 %Z1 = fadd double %Y1, %B1 16 %Z2 = fadd double %Y2, %B2 17 %R = fmul double %Z1, %Z2 18 ret double %R 19 ; CHECK: @test1 20 ; CHECK: ret double %R 21 } 22 23 ; Basic depth-3 chain with fmuladd 24 define double @test1a(double %A1, double %A2, double %B1, double %B2, double %C1, double %C2) { 25 %X1 = fsub double %A1, %B1 26 %X2 = fsub double %A2, %B2 27 %Y1 = call double @llvm.fmuladd.f64(double %X1, double %A1, double %C1) 28 %Y2 = call double @llvm.fmuladd.f64(double %X2, double %A2, double %C2) 29 %Z1 = fadd double %Y1, %B1 30 %Z2 = fadd double %Y2, %B2 31 %R = fmul double %Z1, %Z2 32 ret double %R 33 ; CHECK: @test1a 34 ; CHECK: ret double %R 35 } 36 37 ; Basic depth-3 chain with cos 38 define double @test2(double %A1, double %A2, double %B1, double %B2) { 39 %X1 = fsub double %A1, %B1 40 %X2 = fsub double %A2, %B2 41 %Y1 = call double @llvm.cos.f64(double %X1) 42 %Y2 = call double @llvm.cos.f64(double %X2) 43 %Z1 = fadd double %Y1, %B1 44 %Z2 = fadd double %Y2, %B2 45 %R = fmul double %Z1, %Z2 46 ret double %R 47 ; CHECK: @test2 48 ; CHECK: ret double %R 49 } 50 51 ; Basic depth-3 chain with powi 52 define double @test3(double %A1, double %A2, double %B1, double %B2, i32 %P) { 53 %X1 = fsub double %A1, %B1 54 %X2 = fsub double %A2, %B2 55 %Y1 = call double @llvm.powi.f64(double %X1, i32 %P) 56 %Y2 = call double @llvm.powi.f64(double %X2, i32 %P) 57 %Z1 = fadd double %Y1, %B1 58 %Z2 = fadd double %Y2, %B2 59 %R = fmul double %Z1, %Z2 60 ret double %R 61 ; CHECK: @test3 62 ; CHECK: ret double %R 63 } 64 65 ; Basic depth-3 chain with powi (different powers: should not vectorize) 66 define double @test4(double %A1, double %A2, double %B1, double %B2, i32 %P) { 67 %X1 = fsub double %A1, %B1 68 %X2 = fsub double %A2, %B2 69 %P2 = add i32 %P, 1 70 %Y1 = call double @llvm.powi.f64(double %X1, i32 %P) 71 %Y2 = call double @llvm.powi.f64(double %X2, i32 %P2) 72 %Z1 = fadd double %Y1, %B1 73 %Z2 = fadd double %Y2, %B2 74 %R = fmul double %Z1, %Z2 75 ret double %R 76 ; CHECK: @test4 77 ; CHECK: ret double %R 78 } 79 80
