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
77 lines
3.0 KiB
LLVM
77 lines
3.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt -passes=slp-vectorizer -slp-vectorize-hor -slp-vectorize-hor-store -S < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=bdver2 | FileCheck %s
|
|
|
|
define void @i64_simplified(ptr noalias %st, ptr noalias %ld) {
|
|
; CHECK-LABEL: @i64_simplified(
|
|
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr [[LD:%.*]], align 8
|
|
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
|
|
; CHECK-NEXT: store <4 x i64> [[TMP2]], ptr [[ST:%.*]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%arrayidx1 = getelementptr inbounds i64, ptr %ld, i64 1
|
|
|
|
%t0 = load i64, ptr %ld, align 8
|
|
%t1 = load i64, ptr %arrayidx1, align 8
|
|
|
|
%arrayidx3 = getelementptr inbounds i64, ptr %st, i64 1
|
|
%arrayidx4 = getelementptr inbounds i64, ptr %st, i64 2
|
|
%arrayidx5 = getelementptr inbounds i64, ptr %st, i64 3
|
|
|
|
store i64 %t0, ptr %st, align 8
|
|
store i64 %t1, ptr %arrayidx3, align 8
|
|
store i64 %t0, ptr %arrayidx4, align 8
|
|
store i64 %t1, ptr %arrayidx5, align 8
|
|
ret void
|
|
}
|
|
|
|
define void @i64_simplifiedi_reversed(ptr noalias %st, ptr noalias %ld) {
|
|
; CHECK-LABEL: @i64_simplifiedi_reversed(
|
|
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr [[LD:%.*]], align 8
|
|
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0>
|
|
; CHECK-NEXT: store <4 x i64> [[TMP2]], ptr [[ST:%.*]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%arrayidx1 = getelementptr inbounds i64, ptr %ld, i64 1
|
|
|
|
%t0 = load i64, ptr %ld, align 8
|
|
%t1 = load i64, ptr %arrayidx1, align 8
|
|
|
|
%arrayidx3 = getelementptr inbounds i64, ptr %st, i64 1
|
|
%arrayidx4 = getelementptr inbounds i64, ptr %st, i64 2
|
|
%arrayidx5 = getelementptr inbounds i64, ptr %st, i64 3
|
|
|
|
store i64 %t1, ptr %st, align 8
|
|
store i64 %t0, ptr %arrayidx3, align 8
|
|
store i64 %t1, ptr %arrayidx4, align 8
|
|
store i64 %t0, ptr %arrayidx5, align 8
|
|
ret void
|
|
}
|
|
|
|
define void @i64_simplifiedi_extract(ptr noalias %st, ptr noalias %ld) {
|
|
; CHECK-LABEL: @i64_simplifiedi_extract(
|
|
; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i64, ptr [[LD:%.*]], i64 1
|
|
; CHECK-NEXT: [[T1:%.*]] = load i64, ptr [[ARRAYIDX1]], align 8
|
|
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr [[LD]], align 8
|
|
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> poison, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
|
|
; CHECK-NEXT: store <4 x i64> [[TMP2]], ptr [[ST:%.*]], align 8
|
|
; CHECK-NEXT: store i64 [[T1]], ptr [[LD]], align 8
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%arrayidx1 = getelementptr inbounds i64, ptr %ld, i64 1
|
|
|
|
%t0 = load i64, ptr %ld, align 8
|
|
%t1 = load i64, ptr %arrayidx1, align 8
|
|
|
|
%arrayidx3 = getelementptr inbounds i64, ptr %st, i64 1
|
|
%arrayidx4 = getelementptr inbounds i64, ptr %st, i64 2
|
|
%arrayidx5 = getelementptr inbounds i64, ptr %st, i64 3
|
|
|
|
store i64 %t0, ptr %st, align 8
|
|
store i64 %t0, ptr %arrayidx3, align 8
|
|
store i64 %t0, ptr %arrayidx4, align 8
|
|
store i64 %t1, ptr %arrayidx5, align 8
|
|
store i64 %t1, ptr %ld, align 8
|
|
ret void
|
|
}
|
|
|