Linus Nielsen Feltzing
|
5a0cb68e43
|
Moved some assembly optimizations to the target tree
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@10318 a1c6a512-1295-4272-9138-f99709370657
|
2006-07-25 11:16:03 +00:00 |
|
Thom Johansen
|
c85ad42c35
|
Didn't mean to commit this, revert.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@8794 a1c6a512-1295-4272-9138-f99709370657
|
2006-02-22 22:53:41 +00:00 |
|
Thom Johansen
|
e94d262795
|
ARM optimised memset from Linux.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@8785 a1c6a512-1295-4272-9138-f99709370657
|
2006-02-22 18:52:08 +00:00 |
|
Jens Arnold
|
d036e97d38
|
Added memmove() to the rockbox core. C implementation taken from newlib. Fully optimised ASM implementations for SH1 and coldfire, reusing the AMS memcpy code path for forward copying.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@8601 a1c6a512-1295-4272-9138-f99709370657
|
2006-02-06 16:00:58 +00:00 |
|
Jens Arnold
|
83d3dcfd88
|
Coldfire: Assembler optimised memcpy(), 2.4x...10x (avg. 6.8x) as fast as the C version when copying large blocks within SDRAM.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@7835 a1c6a512-1295-4272-9138-f99709370657
|
2005-11-12 23:42:47 +00:00 |
|
Jens Arnold
|
d8ad74de74
|
SH1 memcpy(): Slightly changed loop concept: saved 4 bytes, and the long+1 case is now as fast as the long+3 case.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@7696 a1c6a512-1295-4272-9138-f99709370657
|
2005-10-31 01:10:27 +00:00 |
|
Jens Arnold
|
c082bc42a2
|
Further optimised SH1 memcpy(): Lower latency for very small blocks, faster large block copying for odd destination alignment (+27% for long+1, +33% for long+3).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@7690 a1c6a512-1295-4272-9138-f99709370657
|
2005-10-30 20:48:52 +00:00 |
|
Christian Gmeiner
|
c6ff1f5eb5
|
Added CPU_COLDFIRE define - one step closer to iAudio-port
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@7186 a1c6a512-1295-4272-9138-f99709370657
|
2005-07-18 12:40:29 +00:00 |
|
Jens Arnold
|
0310f16005
|
Revived C implementations of memcpy() and memset() for platforms without asm optimized versions (gmini), replacing the intermediate strings.c. Moved the asm optimized versions 'out of the way' for the implicit 'make' rules by renaming them to *_a.S (for assembler/alternative).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@5628 a1c6a512-1295-4272-9138-f99709370657
|
2005-01-21 22:43:02 +00:00 |
|