[Hardware] OGR-NG new x86 testing core

Konrad Rzepecki hannibal at astral.lodz.pl
Fri Aug 31 15:18:08 EDT 2012


  Hi

I try to write completely different core with use of AVX mnemonics and one 
pass shift. Unfortunately it won't work as well as I expected - it is 20% 
slower than fastest. However it pass all test without problems, so maybe 
someone other can improve it or take ideas from it to improve other cores. 

I attach 3 patches. Two simple to add AVX detection and AMD-FX cpu detection 
(this one I submit also in bugzilla). The third one is core itself. It's quite 
good commented - but can answer any questions about it's "design" if someone 
need deeply explanations.


PS. I give up with OpenCL core.

-- 
  Konrad Rzepecki (Hannibal)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dnetc_avx.diff
Type: text/x-patch
Size: 1711 bytes
Desc: not available
Url : http://lists.distributed.net/pipermail/hardware/attachments/20120831/c0f90871/attachment-0003.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dnetc_amdfx_2.diff
Type: text/x-patch
Size: 2190 bytes
Desc: not available
Url : http://lists.distributed.net/pipermail/hardware/attachments/20120831/c0f90871/attachment-0004.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dnetc_ogr_x86_8core.diff
Type: text/x-patch
Size: 24379 bytes
Desc: not available
Url : http://lists.distributed.net/pipermail/hardware/attachments/20120831/c0f90871/attachment-0005.bin 


More information about the Hardware mailing list