diff options
| author | Alan Modra <amodra@gmail.com> | 2013-08-17 18:47:59 +0930 |
|---|---|---|
| committer | Alan Modra <amodra@gmail.com> | 2013-10-04 10:41:35 +0930 |
| commit | 3be87c77d24c4456ccca4034363b6d1814cd0c84 (patch) | |
| tree | c81589a21d74b94d5ae20235bc6d81715eca858c /sysdeps/powerpc/powerpc64/power7/memset.S | |
| parent | 759cfef3ac4c07dba1ece0bbc1207e099348816d (diff) | |
| download | glibc-3be87c77d24c4456ccca4034363b6d1814cd0c84.tar.gz glibc-3be87c77d24c4456ccca4034363b6d1814cd0c84.tar.xz glibc-3be87c77d24c4456ccca4034363b6d1814cd0c84.zip | |
PowerPC LE memset
http://sourceware.org/ml/libc-alpha/2013-08/msg00104.html
One of the things I noticed when looking at power7 timing is that rlwimi
is cracked and the two resulting insns have a register dependency.
That makes it a little slower than the equivalent rldimi.
* sysdeps/powerpc/powerpc64/memset.S: Replace rlwimi with
insrdi. Formatting.
* sysdeps/powerpc/powerpc64/power4/memset.S: Likewise.
* sysdeps/powerpc/powerpc64/power6/memset.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power4/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power6/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/memset.S: Likewise.
Diffstat (limited to 'sysdeps/powerpc/powerpc64/power7/memset.S')
| -rw-r--r-- | sysdeps/powerpc/powerpc64/power7/memset.S | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/sysdeps/powerpc/powerpc64/power7/memset.S b/sysdeps/powerpc/powerpc64/power7/memset.S index b24cfa163a..8b081e87cb 100644 --- a/sysdeps/powerpc/powerpc64/power7/memset.S +++ b/sysdeps/powerpc/powerpc64/power7/memset.S @@ -32,8 +32,8 @@ L(_memset): mr 10,3 /* Replicate byte to word. */ - rlwimi 4,4,8,16,23 - rlwimi 4,4,16,0,15 + insrdi 4,4,8,48 + insrdi 4,4,16,32 ble cr6,L(small) /* If length <= 8, use short copy code. */ neg 0,3 @@ -321,7 +321,7 @@ L(medium): clrldi 0,0,62 beq L(medium_aligned) - /* Force 4-bytes alignment for SRC. */ + /* Force 4-bytes alignment for DST. */ mtocrf 0x01,0 subf 5,0,5 1: /* Copy 1 byte. */ |