diff options
| author | Guozhi Wei <carrot@google.com> | 2017-11-06 19:09:38 +0000 |
|---|---|---|
| committer | Guozhi Wei <carrot@google.com> | 2017-11-06 19:09:38 +0000 |
| commit | 416cdcce390af48f1c741ff91b26ade0beac3abd (patch) | |
| tree | b87113d11fea5ae46315eedb6a58f821be8deeff /test/CodeGen/PowerPC | |
| parent | 6e029979011433fef0a7cc11b7acc9e864affb87 (diff) | |
[PPC] Use xxbrd to speed up bswap64
Power doesn't have bswap instructions, so llvm generates following code sequence for bswap64.
rotldi 5, 3, 16
rotldi 4, 3, 8
rotldi 9, 3, 24
rotldi 10, 3, 32
rotldi 11, 3, 48
rotldi 12, 3, 56
rldimi 4, 5, 8, 48
rldimi 4, 9, 16, 40
rldimi 4, 10, 24, 32
rldimi 4, 11, 40, 16
rldimi 4, 12, 48, 8
rldimi 4, 3, 56, 0
But Power9 has vector bswap instructions, they can also be used to speed up scalar bswap intrinsic. With this patch, bswap64 can be translated to:
mtvsrdd 34, 3, 3
xxbrd 34, 34
mfvsrld 3, 34
Differential Revision: https://reviews.llvm.org/D39510
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@317499 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/PowerPC')
| -rw-r--r-- | test/CodeGen/PowerPC/bswap64.ll | 13 |
1 files changed, 13 insertions, 0 deletions
diff --git a/test/CodeGen/PowerPC/bswap64.ll b/test/CodeGen/PowerPC/bswap64.ll new file mode 100644 index 00000000000..0a78aa2dc54 --- /dev/null +++ b/test/CodeGen/PowerPC/bswap64.ll @@ -0,0 +1,13 @@ +; RUN: llc -verify-machineinstrs < %s -mtriple=ppc64le-- -mcpu=pwr9 | FileCheck %s + +declare i64 @llvm.bswap.i64(i64) + +; CHECK: mtvsrdd +; CHECK: xxbrd +; CHECK: mfvsrd +define i64 @bswap64(i64 %x) { +entry: + %0 = call i64 @llvm.bswap.i64(i64 %x) + ret i64 %0 +} + |