File: vector-alloca-addrspacecast.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.6-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,304 kB
  • sloc: cpp: 7,438,677; ansic: 1,393,822; asm: 1,012,926; python: 241,650; f90: 86,635; objc: 75,479; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (26 lines) | stat: -rw-r--r-- 1,201 bytes parent folder | download | duplicates (14) 
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
 
; RUN: opt -S -mtriple=amdgcn-- -data-layout=A5 -passes='amdgpu-promote-alloca,sroa,instcombine' < %s | FileCheck -check-prefix=OPT %s

; Should give up promoting alloca to vector with an addrspacecast.

; OPT-LABEL: @vector_addrspacecast(
; OPT: alloca [3 x i32]
; OPT: store i32 0, ptr addrspace(5) %alloca, align 4
; OPT: store i32 1, ptr addrspace(5) %a1, align 4
; OPT: store i32 2, ptr addrspace(5) %a2, align 4
; OPT: %tmp = getelementptr [3 x i32], ptr addrspace(5) %alloca, i64 0, i64 %index
; OPT: %ac = addrspacecast ptr addrspace(5) %tmp to ptr
; OPT: %data = load i32, ptr %ac, align 4
define amdgpu_kernel void @vector_addrspacecast(ptr addrspace(1) %out, i64 %index) {
entry:
  %alloca = alloca [3 x i32], addrspace(5)
  %a1 = getelementptr [3 x i32], ptr addrspace(5) %alloca, i32 0, i32 1
  %a2 = getelementptr [3 x i32], ptr addrspace(5) %alloca, i32 0, i32 2
  store i32 0, ptr addrspace(5) %alloca
  store i32 1, ptr addrspace(5) %a1
  store i32 2, ptr addrspace(5) %a2
  %tmp = getelementptr [3 x i32], ptr addrspace(5) %alloca, i64 0, i64 %index
  %ac = addrspacecast ptr addrspace(5) %tmp to ptr
  %data = load i32, ptr %ac
  store i32 %data, ptr addrspace(1) %out
  ret void
}