Chromium Code Reviews| OLD | NEW |
|---|---|
| (Empty) | |
| 1 ; RUN: llc -mtriple=x86_64-nacl %s -o - | FileCheck %s | |
| 2 | |
| 3 ; CHECK-LABEL: @foo | |
|
jvoung (off chromium)
2015/05/12 23:53:05
nit: seems better to put this by the function (at
Derek Schuff
2015/05/13 00:44:50
Done.
| |
| 4 ; CHECK: .bundle_lock | |
| 5 ; CHECK: leal -16({{.*}}), %esp | |
| 6 ; CHECK: addq %r15, %rsp | |
| 7 ; CHECK: .bundle_unlock | |
| 8 | |
| 9 target datalayout = "e-p:32:32-i64:64-n32" | |
| 10 target triple = "le32-unknown-nacl" | |
| 11 | |
| 12 define hidden void @foo(i32 %agg.result, <16 x i8> %v) { | |
|
jvoung (off chromium)
2015/05/12 23:53:05
The parameters are unused -- is it important for t
Derek Schuff
2015/05/13 00:44:50
no, removed
| |
| 13 entry: | |
| 14 br label %bb1 | |
|
jvoung (off chromium)
2015/05/12 23:53:05
Is it required that the alloca not be in the entry
Derek Schuff
2015/05/13 00:44:50
yes it is required, comment added.
| |
| 15 | |
| 16 bb1: | |
| 17 %0 = alloca i8, i32 16, align 16 | |
| 18 %.asint = ptrtoint i8* %0 to i32 | |
| 19 %gep109 = add i32 %.asint, 0 | |
|
jvoung (off chromium)
2015/05/12 23:53:05
Would it be more resilient to optimization have ad
Derek Schuff
2015/05/13 00:44:50
actually the add isn't necessary at all.
| |
| 20 %gep109.asptr = inttoptr i32 %gep109 to i8* | |
| 21 %1 = load i8, i8* %gep109.asptr, align 1 | |
| 22 %conv.i = sext i8 %1 to i32 | |
| 23 %call5 = call i32 @bar(i32 undef, i32 %conv.i) | |
| 24 unreachable | |
| 25 } | |
| 26 | |
| 27 | |
| 28 declare hidden i32 @bar(i32, i32) | |
| 29 | |
| 30 | |
|
jvoung (off chromium)
2015/05/12 23:53:05
extra space
Derek Schuff
2015/05/13 00:44:50
Done.
| |
| 31 | |
| OLD | NEW |