mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-22 18:54:02 +01:00
b3a8585bc1
`llc -march` is problematic because it only switches the target architecture, but leaves the operating system unchanged. This occasionally leads to indeterministic tests because the OS from LLVM_DEFAULT_TARGET_TRIPLE is used. However we can simply always use `llc -mtriple` instead. This changes all the tests to do this to avoid people using -march when they copy and paste parts of tests. This patch: - Removes -march if the .ll file already has a matching `target triple` directive or -mtriple argument. - In all other cases changes -march=ppc32/-march=ppc64 to -mtriple=ppc32--/-mtriple=ppc64-- See also the discussion in https://reviews.llvm.org/D35287 llvm-svn: 309754
31 lines
998 B
LLVM
31 lines
998 B
LLVM
; RUN: llc -relocation-model=pic < %s | FileCheck %s
|
|
|
|
; The load restoring r30 at the end of the function was placed out of order
|
|
; relative to its uses as the PIC base pointer.
|
|
; This was because the r30 operand was not marked as "def" which allowed
|
|
; the post-RA scheduler to move it over other uses of r30.
|
|
|
|
; CHECK-LABEL: fred
|
|
; CHECK: lwz 30, 24(1)
|
|
; R30 should not appear in an instruction after it's been restored.
|
|
; CHECK-NOT: 30,
|
|
|
|
target datalayout = "E-m:e-p:32:32-i64:64-n32"
|
|
target triple = "powerpc--"
|
|
|
|
define double @fred(i64 %a) #0 {
|
|
entry:
|
|
%0 = lshr i64 %a, 32
|
|
%conv = trunc i64 %0 to i32
|
|
%conv1 = sitofp i32 %conv to double
|
|
%mul = fmul double %conv1, 0x41F0000000000000
|
|
%and = and i64 %a, 4294967295
|
|
%or = or i64 %and, 4841369599423283200
|
|
%sub = fadd double %mul, 0xC330000000000000
|
|
%1 = bitcast i64 %or to double
|
|
%add = fadd double %sub, %1
|
|
ret double %add
|
|
}
|
|
|
|
attributes #0 = { norecurse nounwind readnone "target-cpu"="ppc" "use-soft-float"="false" }
|