Thanks Andrea Pellegrini for pointing me to his patch for SMT. I tried some 
SPEC 2006 benchmarks, it works for some of them, but I still met one problem as 
follows:

I run two benchmarks (bzip2 and gcc) with SMT enabled for the first 2B 
instructions, and it aborted after running for three hours:

gem5.opt: build/X86/cpu/o3/decode_impl.hh:392: void 
DefaultDecode<Impl>::skidInsert(ThreadID) [with Impl = O3CPUImpl; ThreadID = 
short int]: Assertion `skidBuffer[tid].size() <= skidBufferMax' failed.

Both program runs well separately. I also met similar error when running 
perlbench and hmmer with SMT. Does anyone have some idea on how to solve this 
problem? Thanks!

Best regards
Fangfei

________________________________
From: [email protected] [[email protected]] on behalf of 
Fangfei Liu [[email protected]]
Sent: Wednesday, January 16, 2013 8:07 PM
To: [email protected]
Subject: [gem5-users] SMT support for x86


Hi all,



I'm doing some experiment with SMT on x86, but it failed to run even for a 
simple hello world (with some abort error). I read from previous thread 
<http://www.mail-archive.com/[email protected]/msg04364.html> (ZeroReg 
problem) that someone has solved this problem and submitted a patch. I was 
wondering whether this patch has been committed to the repository and if not, 
how can I get this patch? Thanks in advance!



By the way, I can only succeed with SMT on alpha, the program is hanged when 
running on arm.



Best regards

Fangfei
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to