[Mplayer-dev-eng] release today?

[Nick Kurshev]

Hi, Arpi!

On 2001-06-20 12:08:00 you wrote:

>> >> Nick,   SOLVED: disabled SSE code in fastmemcpy.h ;(
>> >> can you do better solution?
>> >i can: removed SSE code in fastmemcpy.h :)
>> >
>> >btw it was slower than non-sse code, so it doens't matter.
>> >
>> Could you describe me this problem in detail?
>> What's SSE-related problem in fastmemcpy? IMHO SSE stuff 
>> is disabled by default there. AFAIK mplayer contains a lot
>yes, there are no problems because it's disabled.
>but it's disabled because many users had SIGILl in it.
>you said that it's not faster than mmx version (because of
>memory dis-alignment) so it doesn't matter.
>
NO! Because SSE is too slow thing on P3.
As I wrote P3 has 1 SSE decoder. It's mean that it can execute only
1 SSE instruction per cpu clock. In the same time it can execute up to 3
MMX instructions per cpu clock, because has 3 MMX decoders.
Simple calculation shows us:
3*8=24 bytes (MMX register has length = 8 bytes)
1*16=16 bytes (XMMX register has length = 16 bytes)
So P3 can process 24 bytes through MMX unit and 16 through SSE.
IMHO 24 > 16 ;)

>> I have an idea: it would be better to check kernel version
>> at runtime - if it is 2.2.x and mplayer was optimized for SSE1(2)
>> then mplayer should cease of working.
>runtime? hmm. compiletime is better IMHO.
Does it matter what kernel I use during compilation?
Does it matter what .so files I have during compilation?
Anyone can compile mplayer on one system and use it on other.
Checking of gcc and as version does matter because these things
deal with source and object files. All other checking deserves only
WARNING during compile time and ceasing of working during runtime.
(Btw, I have different Linux kernels - 2.2.19 and 2.4.5. But I select
what kernel should be booted on power on).
>btw we should do some checks at runtime too, to avoid bugreports
>from cross-compiled mplayer users.
>(who compiles mplayer on the dual-p3 server and then use it on home 486)
>
IMHO correctly should be: Who compiles mplayer for P3 and then use it on 486.

>> ALSO: Could you tell me new primary URL of mlpayer where you plan
>> to locate project in the future (WAS: you leave sf.net), and what stops
>> you to release mplayer?
>
>[cvs|www|ftp].mplayerhq.hu
>(it still not work, we're waiting to domain registration)
>
>but the server is up, currently at mplayer.dev.hu (or mplayer.megesz.hu)
THANKS!

Best regards! Nick (2001-06-21 00:33:52)


_______________________________________________
Mplayer-dev-eng mailing list
Mplayer-dev-eng at lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/mplayer-dev-eng