I tended to use 100M (which might have been slightly short). For Spec2k6, on average there are 13.5 L1 misses per 1000 instructions (stat taken from this graph - http://www.jaleels.org/ajaleel/workload/SPEC_CPU2006/all.FULL_RUN.DL1.jpg).
So to warm up a 1MB L2 (16384 cache lines) , you'd need to simulate at least (1000/13.5) * 16384 = ~1.2M instructions before the cache would usually be warmed up. This of course ignores things like locality of access, etc that would require more instructions be simulated. But given the noise already present in most academic studies, I didn't worry about it too much. On Wed, Sep 25, 2013 at 6:46 AM, Jagadish Kotra <[email protected]>wrote: > Thank you Mitch. What was your interval slice size ? I am using 1 Billion > simpoint interval slice. Do you think 1B slice is large enough to not warm > up the caches ? > > Regards, > Jagadish. > > > On Wed, Sep 25, 2013 at 7:17 AM, <[email protected]> wrote: > >> Send gem5-users mailing list submissions to >> [email protected] >> >> To subscribe or unsubscribe via the World Wide Web, visit >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> or, via email, send a message with subject or body 'help' to >> [email protected] >> >> You can reach the person managing the list at >> [email protected] >> >> When replying, please edit your Subject line so it is more specific >> than "Re: Contents of gem5-users digest..." >> >> >> Today's Topics: >> >> 1. gem5 with simpoint (Jagadish Kotra) >> 2. Re: gem5 with simpoint (Mitch Hayenga) >> 3. Re: Changes in cache resplacement policies >> (Roberto Rodr?guez-Rodr?guez) >> 4. Re: Changes in cache resplacement policies (Dongwoo Lee) >> >> >> ---------------------------------------------------------------------- >> >> Message: 1 >> Date: Tue, 24 Sep 2013 15:58:29 -0400 >> From: Jagadish Kotra <[email protected]> >> To: [email protected] >> Subject: [gem5-users] gem5 with simpoint >> Message-ID: >> < >> cacm86jogvdfqk5tvrkudvoxt9wfysy09lwfewjvn1gczsys...@mail.gmail.com> >> Content-Type: text/plain; charset="iso-8859-1" >> >> >> Hello, >> >> I am trying to run gem5 in SE mode with simpoints enabled. I am able to >> generate simpoints for some spec2006 benchmarks. I can see that after >> generating the simpoints, we have two options to simulate the simpoints >> using detailed cpu. >> >> (1) To do fastfwd using simpoint-mode=fastfwd - This will fast forward >> the >> non-simpoint regions using atomic cpu and then switch to detailed cpu in >> simpoint regions. >> >> Ex: >> ./build/ARM/gem5.opt configs/pharm/se2.py --cfg=configs/pharm/nehalem. >> cfg --cmd=../tests/libquantum_base.armv7a-O3-vfpv3d16-vect -o "15 2" >> --simpoint-mode=fastfwd --simpoint-interval=1000000 >> --simpoint-points=simpoints >> >> (2) To simulate a specific checkpoint using simpoint-mode=checkpoint. >> >> ./build/ARM/gem5.opt configs/pharm/se2.py --cfg=configs/pharm/nehalem. >> cfg --cmd=../tests/libquantum_base.armv7a-O3-vfpv3d16-vect -o "15 2" >> --simpoint-mode=checkpoint --simpoint-interval=1000000 >> --simpoint-points=simpoints --checkpoint-dir=ckpt --checkpoint-num=1 >> >> I am using classical caches in SE mode. Would like to know if using the >> corresponding commands in both the cases would warmup the caches >> automatically or do I need to mention some flag explicitly ? >> >> Regards, >> Jagadish. >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: < >> http://m5sim.org/cgi-bin/mailman/private/gem5-users/attachments/20130924/32f1176b/attachment-0001.html >> > >> >> ------------------------------ >> >> Message: 2 >> Date: Tue, 24 Sep 2013 16:31:51 -0500 >> From: Mitch Hayenga <[email protected]> >> To: gem5 users mailing list <[email protected]> >> Subject: Re: [gem5-users] gem5 with simpoint >> Message-ID: >> < >> canxmxn4dj+dg_ymjz+9fy9ui1ho1vzxpfjhemrj-pj+efp+...@mail.gmail.com> >> Content-Type: text/plain; charset="iso-8859-1" >> >> >> Hi, I'm the person who wrote the config scripts you are using. I don't >> have access to them right at this moment but if I remember correctly.... >> >> #1 would properly warm up the caches. It keeps the caches in the system, >> it just swaps the connection between the atomic or detailed cpu (depending >> on if its within a simpoint region). >> >> #2 does not warm up the caches (though the script could be modified to do >> so). I generally had a large enough simpoint interval for what I was >> studying that the time to warmup the caches had a negligible impact on my >> performance. >> >> >> On Tue, Sep 24, 2013 at 2:58 PM, Jagadish Kotra <[email protected] >> >wrote: >> >> > Hello, >> > >> > I am trying to run gem5 in SE mode with simpoints enabled. I am able >> to >> > generate simpoints for some spec2006 benchmarks. I can see that after >> > generating the simpoints, we have two options to simulate the simpoints >> > using detailed cpu. >> > >> > (1) To do fastfwd using simpoint-mode=fastfwd - This will fast forward >> > the non-simpoint regions using atomic cpu and then switch to detailed >> cpu >> > in simpoint regions. >> > >> > Ex: >> > ./build/ARM/gem5.opt configs/pharm/se2.py --cfg=configs/pharm/nehalem. >> > cfg --cmd=../tests/libquantum_base.armv7a-O3-vfpv3d16-vect -o "15 2" >> > --simpoint-mode=fastfwd --simpoint-interval=1000000 >> > --simpoint-points=simpoints >> > >> > (2) To simulate a specific checkpoint using simpoint-mode=checkpoint. >> > >> > ./build/ARM/gem5.opt configs/pharm/se2.py --cfg=configs/pharm/nehalem. >> > cfg --cmd=../tests/libquantum_base.armv7a-O3-vfpv3d16-vect -o "15 2" >> > --simpoint-mode=checkpoint --simpoint-interval=1000000 >> > --simpoint-points=simpoints --checkpoint-dir=ckpt --checkpoint-num=1 >> > >> > I am using classical caches in SE mode. Would like to know if using the >> > corresponding commands in both the cases would warmup the caches >> > automatically or do I need to mention some flag explicitly ? >> > >> > Regards, >> > Jagadish. >> > >> > _______________________________________________ >> > gem5-users mailing list >> > [email protected] >> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> > >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: < >> http://m5sim.org/cgi-bin/mailman/private/gem5-users/attachments/20130924/1a7884a4/attachment-0001.html >> > >> >> ------------------------------ >> >> Message: 3 >> Date: Wed, 25 Sep 2013 00:02:45 +0200 >> From: Roberto Rodr?guez-Rodr?guez <[email protected]> >> To: gem5 users mailing list <[email protected]> >> Subject: Re: [gem5-users] Changes in cache resplacement policies >> Message-ID: >> < >> caggl5hcj0uqakrymtf9mwefxzk5hoyzbgu41auh0frr-rqj...@mail.gmail.com> >> Content-Type: text/plain; charset="utf-8" >> >> Thank you Dongwoo Lee, it worked perfectly. >> >> Now I just need to add the tags selection as an option in Options.py for >> use it in CacheConfig.py. >> >> The problem is the asignation has to be done as tags=LRU() and not as >> tags="LRU()", and there is not data type in Options.py that supports this >> format (not a string) >> >> >> Saludos >> Roberto >> >> >> On Mon, Sep 23, 2013 at 5:55 AM, Dongwoo Lee <[email protected] >> >wrote: >> >> > Sorry, >> > src/mem/cache/BaseCache.py >> > >> > 2013-09-23(?) ?? 12:47, Dongwoo Lee ? ?: >> > >> > Try to check the following >> > >> > src/mem/cache/base.cc (at the bottom) >> > src/mem/cache/tags/Tags.py >> > src/mem/cache/tags/BaseCache.py >> > >> > and use 'configs/common/Caches.py' >> > >> > for example, >> > >> > class L1Cache(BaseCache): >> > ... >> > tags = LRU() >> > >> > >> > 2013-09-23(?) ?? 7:15, Roberto Rodr?guez-Rodr?guez ? ?: >> > >> > Hello everyone, >> > >> > I am working on cache replacement policies using the classic memory >> > model, I was using a gem5 version from April but I have a problem >> because >> > one Parsec benchmark (canneal) finish incorrectly with LRU based >> > policies. >> > >> > I tested an updated version of gem5 (the version from today) and the >> > benchmark works perfectly, I tried to implement all my policies into the >> > new version of the simulator but I see that the way the cache >> replacement >> > policies were definied have changed. >> > >> > There is not builder.cc, I the old versions I added a new variable in >> src >> > /mem/cache/BaseCache.py (replPolicy = Param.String("LRU", "")) and I >> used >> > that variable in src/mem/cache/builder.cc. Could somebody tell me how >> can >> > I add a new cache replacement policy in the new simulator version? >> > >> > ? Thanks in advance >> > >> > Roberto >> > >> > >> > _______________________________________________ >> > gem5-users mailing [email protected]:// >> m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> > >> > >> > >> > >> > _______________________________________________ >> > gem5-users mailing list >> > [email protected] >> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> > >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: < >> http://m5sim.org/cgi-bin/mailman/private/gem5-users/attachments/20130925/c9e14375/attachment-0001.html >> > >> >> ------------------------------ >> >> Message: 4 >> Date: Wed, 25 Sep 2013 20:17:36 +0900 >> From: Dongwoo Lee <[email protected]> >> To: gem5 users mailing list <[email protected]> >> Subject: Re: [gem5-users] Changes in cache resplacement policies >> Message-ID: <[email protected]> >> Content-Type: text/plain; charset="iso-8859-1"; Format="flowed" >> >> Use the python function 'eval()' >> >> For example, >> repl = "LRU()" >> tags = eval(repl) >> >> >> -Dongwoo Lee >> >> >> 2013-09-25(?) ?? 7:02, Roberto Rodr?guez-Rodr?guez ? ?: >> > Thank you Dongwoo Lee, it worked perfectly. >> > >> > Now I just need to add the tags selection as an option in Options.py >> > for use it in CacheConfig.py. >> > >> > The problem is the asignation has to be done as tags=LRU() and not as >> > tags="LRU()", and there is not data type in Options.py that supports >> > this format (not a string) >> > >> > >> > Saludos >> > Roberto >> > >> > >> > On Mon, Sep 23, 2013 at 5:55 AM, Dongwoo Lee <[email protected] >> > <mailto:[email protected]>> wrote: >> > >> > Sorry, >> > src/mem/cache/BaseCache.py >> > >> > 2013-09-23(?) ?? 12:47, Dongwoo Lee ? ?: >> >> Try to check the following >> >> >> >> src/mem/cache/base.cc (at the bottom) >> >> src/mem/cache/tags/Tags.py >> >> src/mem/cache/tags/BaseCache.py >> >> >> >> and use 'configs/common/Caches.py' >> >> >> >> for example, >> >> >> >> class L1Cache(BaseCache): >> >> ... >> >> tags = LRU() >> >> >> >> >> >> 2013-09-23(?) ?? 7:15, Roberto Rodr?guez-Rodr?guez ? ?: >> >>> Hello everyone, >> >>> >> >>> I am working on cache replacement policies using the classic >> >>> memory model, I was using a gem5 version from April but I have a >> >>> problem because one Parsec benchmark (canneal) finish >> >>> incorrectly with LRU based policies. >> >>> >> >>> I tested an updated version of gem5 (the version from today) and >> >>> the benchmark works perfectly, I tried to implement all my >> >>> policies into the new version of the simulator but I see that >> >>> the way the cache replacement policies were definied have changed. >> >>> >> >>> There is not builder.cc, I the old versions I added a new >> >>> variable in src/mem/cache/BaseCache.py (replPolicy = >> >>> Param.String("LRU", "")) and I used that variable in >> >>> src/mem/cache/builder.cc. Could somebody tell me how can I add a >> >>> new cache replacement policy in the new simulator version? >> >>> >> >>> Thanks in advance >> >>> >> >>> Roberto >> >>> >> >>> >> >>> _______________________________________________ >> >>> gem5-users mailing list >> >>> [email protected] <mailto:[email protected]> >> >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> >> > >> > >> > _______________________________________________ >> > gem5-users mailing list >> > [email protected] <mailto:[email protected]> >> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> > >> > >> > >> > >> > _______________________________________________ >> > gem5-users mailing list >> > [email protected] >> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: < >> http://m5sim.org/cgi-bin/mailman/private/gem5-users/attachments/20130925/bbf2a26e/attachment.html >> > >> >> ------------------------------ >> >> >> _______________________________________________ >> gem5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> >> End of gem5-users Digest, Vol 86, Issue 40 >> ****************************************** >> > > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
