| Re: [AD] Color convertors |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
At 19:30 06/09/2001 +0200, you wrote:
Here are some results for the MMX code:
[Warning: tons of numbers !!!]
[tons of numbers snipped]
step0: original Isaac's code
step0->step1: added _align_ at jump points
step1->step2: register permutation
step2->step3: replaced incl %ebp by decl %ecx
cmpl %ebp, %ecx jnz
jb ...
step3->step4: removed the three 'nop'
step4->step5: replaced '%ebp' by '%mm7'
It seems there are improvements only in step 1 and step 5, so why don't we
apply only this modifications? Also, we should test it in other CPUs (yours
is K6-3 isn't it?), afterall I wrote the code with the AMD optimization
guide in my hands and a k6-2 (now I own a Duron)
Also, I think we should target the most towards "slow" CPUs (Pentium MMX,
K6...) rather than P4 or athlons, because these won't have any problem in
maintaining a playable fps