b10ecfa914
Currently SLP vectorizer tries to keep only GEPs as scalar, if they are vectorized but used externally. Same approach can be used for all scalar values. This patch tries to keep original scalars if all its operands remain scalar or externally used, the cost of the original scalar is lower than the cost of the extractelement instruction, or if the number of externally used scalars in the same entry is power of 2. Last criterion allows better revectorization for multiply used scalars. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: https://github.com/llvm/llvm-project/pull/100904
129 lines
5.5 KiB
LLVM
129 lines
5.5 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
|
|
|
|
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-n8:16:32:64-S128"
|
|
target triple = "x86_64-apple-macosx10.8.0"
|
|
|
|
; int foo(int * restrict B, int * restrict A, int n, int m) {
|
|
; B[0] = n * A[0] + m * A[0];
|
|
; B[1] = n * A[1] + m * A[1];
|
|
; B[2] = n * A[2] + m * A[2];
|
|
; B[3] = n * A[3] + m * A[3];
|
|
; return 0;
|
|
; }
|
|
|
|
define i32 @foo(ptr noalias nocapture %B, ptr noalias nocapture %A, i32 %n, i32 %m) #0 {
|
|
; CHECK-LABEL: @foo(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
|
|
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[A:%.*]], align 4
|
|
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 [[MUL238]], i32 0
|
|
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: [[TMP3:%.*]] = mul <4 x i32> [[TMP0]], [[TMP2]]
|
|
; CHECK-NEXT: store <4 x i32> [[TMP3]], ptr [[B:%.*]], align 4
|
|
; CHECK-NEXT: ret i32 0
|
|
;
|
|
entry:
|
|
%0 = load i32, ptr %A, align 4
|
|
%mul238 = add i32 %m, %n
|
|
%add = mul i32 %0, %mul238
|
|
store i32 %add, ptr %B, align 4
|
|
%arrayidx4 = getelementptr inbounds i32, ptr %A, i64 1
|
|
%1 = load i32, ptr %arrayidx4, align 4
|
|
%add8 = mul i32 %1, %mul238
|
|
%arrayidx9 = getelementptr inbounds i32, ptr %B, i64 1
|
|
store i32 %add8, ptr %arrayidx9, align 4
|
|
%arrayidx10 = getelementptr inbounds i32, ptr %A, i64 2
|
|
%2 = load i32, ptr %arrayidx10, align 4
|
|
%add14 = mul i32 %2, %mul238
|
|
%arrayidx15 = getelementptr inbounds i32, ptr %B, i64 2
|
|
store i32 %add14, ptr %arrayidx15, align 4
|
|
%arrayidx16 = getelementptr inbounds i32, ptr %A, i64 3
|
|
%3 = load i32, ptr %arrayidx16, align 4
|
|
%add20 = mul i32 %3, %mul238
|
|
%arrayidx21 = getelementptr inbounds i32, ptr %B, i64 3
|
|
store i32 %add20, ptr %arrayidx21, align 4
|
|
ret i32 0
|
|
}
|
|
|
|
|
|
; int extr_user(int * restrict B, int * restrict A, int n, int m) {
|
|
; B[0] = n * A[0] + m * A[0];
|
|
; B[1] = n * A[1] + m * A[1];
|
|
; B[2] = n * A[2] + m * A[2];
|
|
; B[3] = n * A[3] + m * A[3];
|
|
; return A[0];
|
|
; }
|
|
|
|
define i32 @extr_user(ptr noalias nocapture %B, ptr noalias nocapture %A, i32 %n, i32 %m) {
|
|
; CHECK-LABEL: @extr_user(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
|
|
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[A:%.*]], align 4
|
|
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[A]], align 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[MUL238]], i32 0
|
|
; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: [[TMP4:%.*]] = mul <4 x i32> [[TMP0]], [[TMP3]]
|
|
; CHECK-NEXT: store <4 x i32> [[TMP4]], ptr [[B:%.*]], align 4
|
|
; CHECK-NEXT: ret i32 [[TMP1]]
|
|
;
|
|
entry:
|
|
%0 = load i32, ptr %A, align 4
|
|
%mul238 = add i32 %m, %n
|
|
%add = mul i32 %0, %mul238
|
|
store i32 %add, ptr %B, align 4
|
|
%arrayidx4 = getelementptr inbounds i32, ptr %A, i64 1
|
|
%1 = load i32, ptr %arrayidx4, align 4
|
|
%add8 = mul i32 %1, %mul238
|
|
%arrayidx9 = getelementptr inbounds i32, ptr %B, i64 1
|
|
store i32 %add8, ptr %arrayidx9, align 4
|
|
%arrayidx10 = getelementptr inbounds i32, ptr %A, i64 2
|
|
%2 = load i32, ptr %arrayidx10, align 4
|
|
%add14 = mul i32 %2, %mul238
|
|
%arrayidx15 = getelementptr inbounds i32, ptr %B, i64 2
|
|
store i32 %add14, ptr %arrayidx15, align 4
|
|
%arrayidx16 = getelementptr inbounds i32, ptr %A, i64 3
|
|
%3 = load i32, ptr %arrayidx16, align 4
|
|
%add20 = mul i32 %3, %mul238
|
|
%arrayidx21 = getelementptr inbounds i32, ptr %B, i64 3
|
|
store i32 %add20, ptr %arrayidx21, align 4
|
|
ret i32 %0 ;<--------- This value has multiple users
|
|
}
|
|
|
|
; In this example we have an external user that is not the first element in the vector.
|
|
define i32 @extr_user1(ptr noalias nocapture %B, ptr noalias nocapture %A, i32 %n, i32 %m) {
|
|
; CHECK-LABEL: @extr_user1(
|
|
; CHECK-NEXT: entry:
|
|
; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
|
|
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 1
|
|
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[A]], align 4
|
|
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[MUL238]], i32 0
|
|
; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
|
|
; CHECK-NEXT: [[TMP4:%.*]] = mul <4 x i32> [[TMP0]], [[TMP3]]
|
|
; CHECK-NEXT: store <4 x i32> [[TMP4]], ptr [[B:%.*]], align 4
|
|
; CHECK-NEXT: ret i32 [[TMP1]]
|
|
;
|
|
entry:
|
|
%0 = load i32, ptr %A, align 4
|
|
%mul238 = add i32 %m, %n
|
|
%add = mul i32 %0, %mul238
|
|
store i32 %add, ptr %B, align 4
|
|
%arrayidx4 = getelementptr inbounds i32, ptr %A, i64 1
|
|
%1 = load i32, ptr %arrayidx4, align 4
|
|
%add8 = mul i32 %1, %mul238
|
|
%arrayidx9 = getelementptr inbounds i32, ptr %B, i64 1
|
|
store i32 %add8, ptr %arrayidx9, align 4
|
|
%arrayidx10 = getelementptr inbounds i32, ptr %A, i64 2
|
|
%2 = load i32, ptr %arrayidx10, align 4
|
|
%add14 = mul i32 %2, %mul238
|
|
%arrayidx15 = getelementptr inbounds i32, ptr %B, i64 2
|
|
store i32 %add14, ptr %arrayidx15, align 4
|
|
%arrayidx16 = getelementptr inbounds i32, ptr %A, i64 3
|
|
%3 = load i32, ptr %arrayidx16, align 4
|
|
%add20 = mul i32 %3, %mul238
|
|
%arrayidx21 = getelementptr inbounds i32, ptr %B, i64 3
|
|
store i32 %add20, ptr %arrayidx21, align 4
|
|
ret i32 %1 ;<--------- This value has multiple users
|
|
}
|