File: MachineSink-convergent.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 (25 lines) | stat: -rw-r--r-- 692 bytes parent folder | download | duplicates (10) 
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
 
; RUN: llc < %s | FileCheck %s
; RUN: %if ptxas %{ llc < %s | %ptxas-verify %}

target triple = "nvptx64-nvidia-cuda"

declare void @llvm.nvvm.barrier0()

; Load a value, then syncthreads.  Branch, and use the loaded value only on one
; side of the branch.  The load shouldn't be sunk beneath the call, because
; syncthreads is modeled as maystore.
define i32 @f(i32 %x, ptr %ptr, i1 %cond) {
Start:
  ; CHECK: ld.u32
  %ptr_val = load i32, ptr %ptr
  ; CHECK: bar.sync
  call void @llvm.nvvm.barrier0()
  br i1 %cond, label %L1, label %L2
L1:
  %ptr_val2 = add i32 %ptr_val, 100
  br label %L2
L2:
  %v4 = phi i32 [ %x, %Start ], [ %ptr_val2, %L1 ]
  %v5 = add i32 %v4, 1000
  ret i32 %v5
}