Re: alpha: linux binary's ogr core very slow

I don't know the specifics of the linux or dec alpha ports,
but the speed differences usually break down to the effectiveness
of the compiler optimization. The OGR cores are ansi c++,
they are too complex to write in asm. Probably when they were built,
different compilers and/or optimizations were used.

There are newer versions of both the Dec and Linux ports you speak
of, you could try those and see if performance has been improved.

If you still see a large performance difference, post a bug
at www.distributed.net/bugs

On Wed, 23 Apr 2003, Tom Vier wrote:

> Date: Wed, 23 Apr 2003 19:31:48 -0400
> From: Tom Vier <[email protected]>
> Reply-To: hardware-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected]
> To: hardware-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected]
> Subject: [HARDWARE] alpha: linux binary's ogr core very slow
>
> using the dec unix client v2.8010-463-CPR-00080322, ogr core #0 (gasrp 5.13)
> i get about 4.8 million nodes per second.
>
> using the latest static linux/alpha binary, with core #0 (same core as the
> dec unix client) i get 2.6 million. using core #1 (gaspr 5.13-cix) i get 2.8
> million.
>
> the linux binary i'm using is the ev5 version, v2.9001-478-CTR-02112010.
>
> why is the old core 2 million nodes/sec faster than the newer core in the
> newer linux binary? aren't both cores written in asm, by hand? isn't core #0
> the same in both clients? this is a huge difference.
>
> (i'm using -bench on both)
>
> --
> Tom Vier <[email protected]>
> DSA Key ID 0xE6CB97DA
> --
> To unsubscribe, send 'unsubscribe hardware' to majordomo-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected]
>
>

steven_nikkel-WeBCG/[email protected]
-------------------------
ERROR:  Keyboard not attached.  Press F1 to continue.
--
To unsubscribe, send 'unsubscribe hardware' to majordomo-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected]