llvm-project/llvm/test/Transforms/SLPVectorizer/RISCV/unsigned-icmp-signed-op.ll
Alexey Bataev a988821123
[SLP]Keep the original order in the reductions.
The patch tries to keep the original order of the instruction in the
reductions. Previously, two first instructions were switched, giving
reverse order.
The first step to support of the ordered reductions.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/98025
2024-07-09 12:26:42 -04:00

44 lines
1.6 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; RUN: opt -S --passes=slp-vectorizer -mtriple=riscv64-unknown-linux-gnu -mattr=+v < %s | FileCheck %s
define i32 @test(ptr %f, i16 %0) {
; CHECK-LABEL: define i32 @test(
; CHECK-SAME: ptr [[F:%.*]], i16 [[TMP0:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP1:%.*]] = load i16, ptr [[F]], align 2
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i16> <i16 poison, i16 0, i16 0, i16 0>, i16 [[TMP0]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i16> <i16 poison, i16 0, i16 0, i16 0>, i16 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP6:%.*]] = zext <4 x i16> [[TMP3]] to <4 x i32>
; CHECK-NEXT: [[TMP7:%.*]] = sext <4 x i16> [[TMP2]] to <4 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = icmp ule <4 x i32> [[TMP6]], [[TMP7]]
; CHECK-NEXT: [[TMP5:%.*]] = call i1 @llvm.vector.reduce.and.v4i1(<4 x i1> [[TMP4]])
; CHECK-NEXT: [[ZEXT_4:%.*]] = zext i1 [[TMP5]] to i32
; CHECK-NEXT: ret i32 [[ZEXT_4]]
;
entry:
%1 = load i16, ptr %f, align 2
%zext.0 = zext i16 %1 to i32
%sext.0 = sext i16 %0 to i32
%zext.1 = zext i16 0 to i32
%sext.1 = sext i16 0 to i32
%zext.2 = zext i16 0 to i32
%sext.2 = sext i16 0 to i32
%zext.3 = zext i16 0 to i32
%sext.3 = sext i16 0 to i32
%cmp.0 = icmp ule i32 %zext.0, %sext.0
%cmp.1 = icmp ule i32 %zext.1, %sext.1
%cmp.2 = icmp ule i32 %zext.2, %sext.2
%cmp.3 = icmp ule i32 %zext.3, %sext.3
%and.0 = and i1 %cmp.0, %cmp.1
%and.1 = and i1 %and.0, %cmp.2
%and.2 = and i1 %and.1, %cmp.3
%zext.4 = zext i1 %and.2 to i32
ret i32 %zext.4
}