llvm/test/CodeGen/X86/fastcc-byval.ll
2009-09-08 23:54:48 +00:00

21 lines
636 B
LLVM

; RUN: llc < %s -tailcallopt=false | grep {movl\[\[:space:\]\]*8(%esp), %eax} | count 2
; PR3122
; rdar://6400815
; byval requires a copy, even with fastcc.
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
target triple = "i386-apple-darwin9.5"
%struct.MVT = type { i32 }
define fastcc i32 @bar() nounwind {
%V = alloca %struct.MVT
%a = getelementptr %struct.MVT* %V, i32 0, i32 0
store i32 1, i32* %a
call fastcc void @foo(%struct.MVT* byval %V) nounwind
%t = load i32* %a
ret i32 %t
}
declare fastcc void @foo(%struct.MVT* byval)