File: not-scalarize-volatile-load.ll

package info (click to toggle)
llvm-toolchain-15 1%3A15.0.6-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,554,644 kB
  • sloc: cpp: 5,922,452; ansic: 1,012,136; asm: 674,362; python: 191,568; objc: 73,855; f90: 42,327; lisp: 31,913; pascal: 11,973; javascript: 10,144; sh: 9,421; perl: 7,447; ml: 5,527; awk: 3,523; makefile: 2,520; xml: 885; cs: 573; fortran: 567
file content (15 lines) | stat: -rw-r--r-- 730 bytes parent folder | download 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
 
; RUN: llc -mtriple amdgcn--amdhsa -mcpu=fiji -amdgpu-scalarize-global-loads < %s | FileCheck -check-prefix=GCN %s

; GCN-LABEL: @volatile_load
; GCN:  s_load_dwordx2 s[[[LO_SREG:[0-9]+]]:[[HI_SREG:[0-9]+]]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
; GCN:  v_mov_b32_e32 v[[LO_VREG:[0-9]+]], s[[LO_SREG]]
; GCN:  v_mov_b32_e32 v[[HI_VREG:[0-9]+]], s[[HI_SREG]]
; GCN:  flat_load_dword v{{[0-9]+}}, v[[[LO_VREG]]:[[HI_VREG]]]

define amdgpu_kernel void @volatile_load(i32 addrspace(1)* %arg, [8 x i32], i32 addrspace(1)* nocapture %arg1) {
bb:
  %tmp18 = load volatile i32, i32 addrspace(1)* %arg, align 4
  %tmp26 = getelementptr inbounds i32, i32 addrspace(1)* %arg1, i64 5
  store i32 %tmp18, i32 addrspace(1)* %tmp26, align 4
  ret void
}