File: load-scalarization-cost-1.ll

package info (click to toggle)
llvm-toolchain-16 1%3A16.0.6-15~deb12u1
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,634,792 kB
  • sloc: cpp: 6,179,261; ansic: 1,216,205; asm: 741,319; python: 196,614; objc: 75,325; f90: 49,640; lisp: 32,396; pascal: 12,286; sh: 9,394; perl: 7,442; ml: 5,494; awk: 3,523; makefile: 2,723; javascript: 1,206; xml: 886; fortran: 581; cs: 573
file content (28 lines) | stat: -rw-r--r-- 974 bytes parent folder | download | duplicates (12) 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
 
; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -passes=loop-vectorize \
; RUN:   -force-vector-width=4 -debug-only=loop-vectorize \
; RUN:   -enable-interleaved-mem-accesses=false -disable-output < %s 2>&1 \
; RUN:   | FileCheck %s
; REQUIRES: asserts
;
; Check that a scalarized load does not get a zero cost in a vectorized
; loop. It can only be folded into the add operand in the scalar loop.

define i32 @fun(ptr %data, i64 %n, i64 %s, ptr %Src) {
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %acc = phi i32 [ 0, %entry ], [ %acc_next, %for.body ]
  %gep = getelementptr inbounds i32, ptr %Src, i64 %iv
  %ld = load i32, ptr %gep
  %acc_next = add i32 %acc, %ld
  %iv.next = add nuw nsw i64 %iv, 2
  %cmp110.us = icmp slt i64 %iv.next, %n
  br i1 %cmp110.us, label %for.body, label %for.end

for.end:
  ret i32 %acc_next

; CHECK: Found an estimated cost of 4 for VF 4 For instruction:   %ld = load i32, ptr %gep
}