File: not-scalarize-volatile-load.ll

package info (click to toggle)
llvm-toolchain-6.0 1%3A6.0.1-10
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 598,080 kB
  • sloc: cpp: 3,046,253; ansic: 595,057; asm: 271,965; python: 128,926; objc: 106,554; sh: 21,906; lisp: 10,191; pascal: 6,094; ml: 5,544; perl: 5,265; makefile: 2,227; cs: 2,027; xml: 686; php: 212; csh: 117
file content (15 lines) | stat: -rw-r--r-- 739 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, 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
}