7d6e8f2a96
If we vectorize a e.g. store, we leave around a bunch of getelementptrs for the individual scalar stores which we removed. We can go ahead and delete them as well. This is purely for test output quality and readability. It should have no effect in any sane pipeline. Differential Revision: https://reviews.llvm.org/D122493
96 lines
4.5 KiB
LLVM
96 lines
4.5 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt < %s -basic-aa -slp-vectorizer -S -mcpu=corei7-avx | FileCheck %s --check-prefixes=ANY,AVX
|
|
; RUN: opt < %s -basic-aa -slp-vectorizer -slp-max-reg-size=128 -S -mcpu=corei7-avx | FileCheck %s --check-prefixes=ANY,MAX128
|
|
|
|
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
|
target triple = "x86_64-unknown-linux-gnu"
|
|
|
|
define void @store_chains(double* %x) {
|
|
; AVX-LABEL: @store_chains(
|
|
; AVX-NEXT: [[TMP1:%.*]] = bitcast double* [[X:%.*]] to <4 x double>*
|
|
; AVX-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* [[TMP1]], align 8
|
|
; AVX-NEXT: [[TMP3:%.*]] = fadd <4 x double> [[TMP2]], [[TMP2]]
|
|
; AVX-NEXT: [[TMP4:%.*]] = fadd <4 x double> [[TMP3]], [[TMP2]]
|
|
; AVX-NEXT: [[TMP5:%.*]] = bitcast double* [[X]] to <4 x double>*
|
|
; AVX-NEXT: store <4 x double> [[TMP4]], <4 x double>* [[TMP5]], align 8
|
|
; AVX-NEXT: ret void
|
|
;
|
|
; MAX128-LABEL: @store_chains(
|
|
; MAX128-NEXT: [[TMP1:%.*]] = bitcast double* [[X:%.*]] to <2 x double>*
|
|
; MAX128-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
|
|
; MAX128-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[TMP2]], [[TMP2]]
|
|
; MAX128-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP3]], [[TMP2]]
|
|
; MAX128-NEXT: [[TMP5:%.*]] = bitcast double* [[X]] to <2 x double>*
|
|
; MAX128-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
|
|
; MAX128-NEXT: [[TMP6:%.*]] = getelementptr inbounds double, double* [[X]], i64 2
|
|
; MAX128-NEXT: [[TMP7:%.*]] = bitcast double* [[TMP6]] to <2 x double>*
|
|
; MAX128-NEXT: [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[TMP7]], align 8
|
|
; MAX128-NEXT: [[TMP9:%.*]] = fadd <2 x double> [[TMP8]], [[TMP8]]
|
|
; MAX128-NEXT: [[TMP10:%.*]] = fadd <2 x double> [[TMP9]], [[TMP8]]
|
|
; MAX128-NEXT: [[TMP11:%.*]] = bitcast double* [[TMP6]] to <2 x double>*
|
|
; MAX128-NEXT: store <2 x double> [[TMP10]], <2 x double>* [[TMP11]], align 8
|
|
; MAX128-NEXT: ret void
|
|
;
|
|
%1 = load double, double* %x, align 8
|
|
%2 = fadd double %1, %1
|
|
%3 = fadd double %2, %1
|
|
store double %3, double* %x, align 8
|
|
%4 = getelementptr inbounds double, double* %x, i64 1
|
|
%5 = load double, double* %4, align 8
|
|
%6 = fadd double %5, %5
|
|
%7 = fadd double %6, %5
|
|
store double %7, double* %4, align 8
|
|
%8 = getelementptr inbounds double, double* %x, i64 2
|
|
%9 = load double, double* %8, align 8
|
|
%10 = fadd double %9, %9
|
|
%11 = fadd double %10, %9
|
|
store double %11, double* %8, align 8
|
|
%12 = getelementptr inbounds double, double* %x, i64 3
|
|
%13 = load double, double* %12, align 8
|
|
%14 = fadd double %13, %13
|
|
%15 = fadd double %14, %13
|
|
store double %15, double* %12, align 8
|
|
ret void
|
|
}
|
|
|
|
define void @store_chains_prefer_width_attr(double* %x) #0 {
|
|
; ANY-LABEL: @store_chains_prefer_width_attr(
|
|
; ANY-NEXT: [[TMP1:%.*]] = bitcast double* [[X:%.*]] to <2 x double>*
|
|
; ANY-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
|
|
; ANY-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[TMP2]], [[TMP2]]
|
|
; ANY-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP3]], [[TMP2]]
|
|
; ANY-NEXT: [[TMP5:%.*]] = bitcast double* [[X]] to <2 x double>*
|
|
; ANY-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
|
|
; ANY-NEXT: [[TMP6:%.*]] = getelementptr inbounds double, double* [[X]], i64 2
|
|
; ANY-NEXT: [[TMP7:%.*]] = bitcast double* [[TMP6]] to <2 x double>*
|
|
; ANY-NEXT: [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[TMP7]], align 8
|
|
; ANY-NEXT: [[TMP9:%.*]] = fadd <2 x double> [[TMP8]], [[TMP8]]
|
|
; ANY-NEXT: [[TMP10:%.*]] = fadd <2 x double> [[TMP9]], [[TMP8]]
|
|
; ANY-NEXT: [[TMP11:%.*]] = bitcast double* [[TMP6]] to <2 x double>*
|
|
; ANY-NEXT: store <2 x double> [[TMP10]], <2 x double>* [[TMP11]], align 8
|
|
; ANY-NEXT: ret void
|
|
;
|
|
%1 = load double, double* %x, align 8
|
|
%2 = fadd double %1, %1
|
|
%3 = fadd double %2, %1
|
|
store double %3, double* %x, align 8
|
|
%4 = getelementptr inbounds double, double* %x, i64 1
|
|
%5 = load double, double* %4, align 8
|
|
%6 = fadd double %5, %5
|
|
%7 = fadd double %6, %5
|
|
store double %7, double* %4, align 8
|
|
%8 = getelementptr inbounds double, double* %x, i64 2
|
|
%9 = load double, double* %8, align 8
|
|
%10 = fadd double %9, %9
|
|
%11 = fadd double %10, %9
|
|
store double %11, double* %8, align 8
|
|
%12 = getelementptr inbounds double, double* %x, i64 3
|
|
%13 = load double, double* %12, align 8
|
|
%14 = fadd double %13, %13
|
|
%15 = fadd double %14, %13
|
|
store double %15, double* %12, align 8
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { "prefer-vector-width"="128" }
|