mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-11-13 06:06:27 +00:00
53a66d10f0
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@47296 91177308-0d34-0410-b5e6-96231b3b80d8
19 lines
731 B
Plaintext
19 lines
731 B
Plaintext
; The intrinsic lowering pass was lowering intrinsics like llvm.memcpy to
|
|
; explicitly specified prototypes, inserting a new function if the old one
|
|
; didn't exist. This caused there to be two external memcpy functions in
|
|
; this testcase for example, which caused the CBE to mangle one, screwing
|
|
; everything up. :( Test that this does not happen anymore.
|
|
;
|
|
; RUN: llvm-as < %s | llc -march=c | not grep _memcpy
|
|
|
|
declare void @llvm.memcpy.i32(i8*, i8*, i32, i32)
|
|
|
|
declare float* @memcpy(i32*, i32, i32)
|
|
|
|
define i32 @test(i8* %A, i8* %B, i32* %C) {
|
|
call float* @memcpy( i32* %C, i32 4, i32 17 ) ; <float*>:1 [#uses=0]
|
|
call void @llvm.memcpy.i32( i8* %A, i8* %B, i32 123, i32 14 )
|
|
ret i32 7
|
|
}
|
|
|