44eca64224
Need to check the scalars if they can be vectorized before trying to schedule them. It may save compile time and improve vectorization on large functions/basic blocks. Differential Revision: https://reviews.llvm.org/D154891
47 lines
2.4 KiB
LLVM
47 lines
2.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
|
|
; RUN: opt -passes=slp-vectorizer -mtriple=x86_64 -mcpu=core-avx2 -S %s | FileCheck %s
|
|
|
|
define <4 x double> @test(ptr %ia, ptr %ib, ptr %ic, ptr %id, ptr %ie, ptr %x) {
|
|
; CHECK-LABEL: define <4 x double> @test
|
|
; CHECK-SAME: (ptr [[IA:%.*]], ptr [[IB:%.*]], ptr [[IC:%.*]], ptr [[ID:%.*]], ptr [[IE:%.*]], ptr [[X:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
; CHECK-NEXT: [[I4238:%.*]] = load double, ptr [[IA]], align 8
|
|
; CHECK-NEXT: [[I4252:%.*]] = load double, ptr [[IB]], align 8
|
|
; CHECK-NEXT: [[I4264:%.*]] = load double, ptr [[IC]], align 8
|
|
; CHECK-NEXT: [[I4275:%.*]] = load double, ptr [[ID]], align 8
|
|
; CHECK-NEXT: [[I4277:%.*]] = load double, ptr [[IE]], align 8
|
|
; CHECK-NEXT: [[I4326:%.*]] = load <4 x double>, ptr [[X]], align 8
|
|
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[I4326]], <4 x double> poison, <2 x i32> <i32 0, i32 poison>
|
|
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x double> [[TMP1]], double [[I4275]], i32 1
|
|
; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> poison, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
|
|
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x double> poison, double [[I4238]], i32 0
|
|
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> [[TMP4]], double [[I4252]], i32 1
|
|
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x double> [[TMP5]], double [[I4264]], i32 2
|
|
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x double> [[TMP6]], double [[I4277]], i32 3
|
|
; CHECK-NEXT: [[TMP8:%.*]] = fmul fast <4 x double> [[TMP3]], [[TMP7]]
|
|
; CHECK-NEXT: ret <4 x double> [[TMP8]]
|
|
;
|
|
%i4238 = load double, ptr %ia, align 8
|
|
%i4252 = load double, ptr %ib, align 8
|
|
%i4264 = load double, ptr %ic, align 8
|
|
%i4275 = load double, ptr %id, align 8
|
|
|
|
%i4277 = load double, ptr %ie, align 8
|
|
|
|
%i4326 = load <4 x double>, ptr %x, align 8
|
|
|
|
%i4330 = extractelement <4 x double> %i4326, i32 0
|
|
|
|
%i4332 = fmul fast double %i4330, %i4238
|
|
%i4333 = fmul fast double %i4330, %i4252
|
|
%i4334 = fmul fast double %i4330, %i4264
|
|
%i4278 = fmul fast double %i4275, %i4277
|
|
|
|
|
|
%i4425 = insertelement <4 x double> poison, double %i4332, i32 0
|
|
%i4426 = insertelement <4 x double> %i4425, double %i4333, i32 1
|
|
%i4427 = insertelement <4 x double> %i4426, double %i4334, i32 2
|
|
%i4428 = insertelement <4 x double> %i4427, double %i4278, i32 3
|
|
ret <4 x double> %i4428
|
|
}
|
|
|