Ralph and SSEx

Message boards : Number crunching : Ralph and SSEx

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6031 - Posted: 30 Jan 2016, 15:08:58 UTC - in response to Message 6028.  

the commands are run in linux, but i'd guess u may have figured out what this means :D


You're right!
For first time I've analyzed the x86_64.exe with MS ProcessExplorer tool.
And it's a 32 bit native image.
ID: 6031 · Report as offensive    Reply Quote
rjs5

Send message
Joined: 5 Jul 15
Posts: 16
Credit: 79,724
RAC: 2
Message 6032 - Posted: 31 Jan 2016, 20:28:58 UTC - in response to Message 6029.  

on this topic, it may be also good to mention that modern compilers are sophisticated. even recent versions of open sourced compilers such as gcc and llvm has pretty advanced/sophisticated *auto-vectorization* features

https://gcc.gnu.org/projects/tree-ssa/vectorization.html
http://llvm.org/devmtg/2012-04-12/Slides/Hal_Finkel.pdf

while that may not produce the most tuned codes, it is probably an incorrect notion that r@h don't have SSEn/AVXn optimizations, the compiler may have embedded some of such SSEn/AVXn optimizations.

this may somewhat explain the somewhat higher performance of r@h in 64bits linux vs say 64 bits windows in the statistics. This is because the combination of optimised 64 bits binaries running in 64 bits linux would most likely have (possibly significantly) better performance compared to 32 bits (possibly less optimised) binaries running in 64 bits windows

i.e. windows platform may see (significant) performance gains just compiling and releasing 64 bit binaries targeting 64 bits windows platforms with a modern / recent sophisticated compiler


It depends on you mean by "significant" gains. I would guess that gains would be 10% to 20% over the current 32-bit binary.


Here is a link that how sensitive auto-vectorization is to the source code layout.
http://locklessinc.com/articles/vectorize/

Rosetta code does not have any AVX code but does have SSE scalar code. I think even in 32-bit Windows there is SSE code. All the applications still have 387 code. It will take some time and source code changes to generate any vector code.


For those who have time, you can install a Linux Guest environment on your Windows machine and test out the 32-bit Windows, 32-bit Linux and 64-bit Linux performance on the same hardware.


One of many ways would be to install Virtualbox:
https://www.virtualbox.org/wiki/Downloads

Download a prebuild Linux image that you are interested in.
http://www.osboxes.org/virtualbox-images/

Install BOINC package on that guest.

Run Rosetta application.



ID: 6032 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6053 - Posted: 4 Mar 2016, 16:31:55 UTC

GCC supports HSA

Heterogeneous Systems Architecture support [2016-01-27]
Heterogeneous Systems Architecture 1.0 support was added to GCC, contributed by Martin Jambor, Martin Liška and Michael Matz from SUSE.


https://gcc.gnu.org/gcc-6/changes.html#hsa
ID: 6053 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6186 - Posted: 9 May 2017, 20:33:14 UTC

Waiting for Godot....
ID: 6186 · Report as offensive    Reply Quote
Dr. Merkwürdigliebe

Send message
Joined: 12 Jun 15
Posts: 16
Credit: 23,473
RAC: 0
Message 6187 - Posted: 9 May 2017, 20:40:29 UTC

There will be no such changes to the code base. Posts concerning OpenCL or HSA are pointless.
ID: 6187 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6188 - Posted: 10 May 2017, 6:46:30 UTC - in response to Message 6187.  
Last modified: 10 May 2017, 6:47:34 UTC

There will be no such changes to the code base. Posts concerning OpenCL or HSA are pointless.


This thread is not about OpenCl (i posted HSA only for info), but about cpu optimizations, that are feasible. I know that gpu in this project is simply an utopia :-P
ID: 6188 · Report as offensive    Reply Quote
Dr. Merkwürdigliebe

Send message
Joined: 12 Jun 15
Posts: 16
Credit: 23,473
RAC: 0
Message 6189 - Posted: 10 May 2017, 14:12:29 UTC - in response to Message 6188.  

There will be no CPU optimizations either...there is either a lack of knowledge or a lack of interest, probably the former.
ID: 6189 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6190 - Posted: 11 May 2017, 12:47:56 UTC - in response to Message 6189.  

there is either a lack of knowledge or a lack of interest, probably the former.


The second one, for me...
ID: 6190 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6192 - Posted: 10 Jun 2017, 10:32:32 UTC

There is a volunteer (quantumethos) that posts some link to optimization to help developers. Here some examples:
Agner optimize
PGI Compiler
Boinc optimize
ID: 6192 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6246 - Posted: 26 Nov 2017, 18:41:15 UTC - in response to Message 5862.  

I just updated the minirosetta_beta app. I did not include SSE linux builds since it will require more testing. I did turn on SSE for the windows build. The latest linux SSE3 test was causing a significant amount of failures.


After over 2 years, have you try again these optimizations?
SSEx for Windows were good, i think, so why not use??
ID: 6246 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6551 - Posted: 7 May 2018, 10:39:50 UTC

GCC 8.1 ready to be public
AOCC 1.2 (Amd Optimizing Compiler for C/C++)
C++17 is ISO standard.
ID: 6551 · Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 9 Apr 08
Posts: 576
Credit: 993,059
RAC: 48
Message 6577 - Posted: 6 Jul 2018, 13:00:32 UTC

Interesting slides from DHPCC++ 2018
Visual Studio Code is at 1.25 version
ID: 6577 · Report as offensive    Reply Quote
Previous · 1 · 2 · 3

Message boards : Number crunching : Ralph and SSEx



©2018 University of Washington
http://www.bakerlab.org