mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-25 13:35:10 +00:00
0ff4287fe2
IR for CUDA should use "nvptx[64]-nvidia-cuda", and IR for NV OpenCL should use "nvptx[64]-nvidia-nvcl" git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184579 91177308-0d34-0410-b5e6-96231b3b80d8
15 lines
458 B
LLVM
15 lines
458 B
LLVM
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
|
|
|
|
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"
|
|
target triple = "nvptx-nvidia-cuda"
|
|
|
|
define void @main(i1* %a1, i32 %a2, i32* %arg3) {
|
|
; CHECK: ld.u8
|
|
; CHECK-NOT: ld.u1
|
|
%t1 = getelementptr i1* %a1, i32 %a2
|
|
%t2 = load i1* %t1
|
|
%t3 = sext i1 %t2 to i32
|
|
store i32 %t3, i32* %arg3
|
|
ret void
|
|
}
|