mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-19 19:12:56 +02:00
dd9ab77318
Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
37 lines
1.0 KiB
LLVM
37 lines
1.0 KiB
LLVM
; RUN: opt -mtriple=amdgcn-- -loop-unroll -simplifycfg -sroa %s -S -o - | FileCheck %s
|
|
; RUN: opt -mtriple=r600-- -loop-unroll -simplifycfg -sroa %s -S -o - | FileCheck %s
|
|
|
|
|
|
; This test contains a simple loop that initializes an array declared in
|
|
; private memory. We want to make sure these kinds of loops are always
|
|
; unrolled, because private memory is slow.
|
|
|
|
; CHECK-LABEL: @test
|
|
; CHECK-NOT: alloca
|
|
; CHECK: store i32 5, i32 addrspace(1)* %out
|
|
define amdgpu_kernel void @test(i32 addrspace(1)* %out) {
|
|
entry:
|
|
%0 = alloca [32 x i32]
|
|
br label %loop.header
|
|
|
|
loop.header:
|
|
%counter = phi i32 [0, %entry], [%inc, %loop.inc]
|
|
br label %loop.body
|
|
|
|
loop.body:
|
|
%ptr = getelementptr [32 x i32], [32 x i32]* %0, i32 0, i32 %counter
|
|
store i32 %counter, i32* %ptr
|
|
br label %loop.inc
|
|
|
|
loop.inc:
|
|
%inc = add i32 %counter, 1
|
|
%1 = icmp sge i32 %counter, 32
|
|
br i1 %1, label %exit, label %loop.header
|
|
|
|
exit:
|
|
%2 = getelementptr [32 x i32], [32 x i32]* %0, i32 0, i32 5
|
|
%3 = load i32, i32* %2
|
|
store i32 %3, i32 addrspace(1)* %out
|
|
ret void
|
|
}
|