The cache miss ratio statistics are already in marss. Couldn't you just take 1 - cache miss ratio to get the hit ratio?
On Fri, Apr 19, 2013 at 5:43 AM, beilei sun <[email protected]> wrote: > Hi All, > > Is MARSSx86 able to count the cache hit ratio？ I am not sure. > > And, here is the way I count the L2_CACHE hit ratio in MARSSx86: > *#########################################################* > *bool CacheController::cache_access_cb(void *arg) > * > *{* > * CacheQueueEntry *queueEntry = (CacheQueueEntry*)arg;* > * if(queueEntry->annuled)* > * return true;* > * queueEntry->eventFlags[CACHE_ACCESS_EVENT]--;* > * > * > * if(cacheLines_->get_port(queueEntry->request)) {* > * CacheLine *line = cacheLines_->probe(queueEntry->request);* > * bool hit = (line == NULL) ? false : line->state;* > * > * > * > * > * #ifdef CONFIG_CACHE_HIT_RATIO* > * stringbuf sb;* > * cacheAccessTimes++;* > * if(hit)* > * cacheHitTimes++;* > * else * > * cacheMissTimes++;* > * if(get_cache_type() == L2_CACHE) {* > * sb << "Cache: " << get_name() <<* > * " Access " << cacheAccessTimes <<* > * " Hit " << cacheHitTimes << * > * " Retry " << cacheRetryTimes <<* > * " Miss " << cacheMissTimes << endl;* > * ptl_logfile << sb;* > * }* > * #endif* > * .................* > *#########################################################* > * cacheAccessTimes, cacheHitTimes and cacheMissTimes are the variables I > added in class CacheController and get_cache_type() is to get the type of > cahe.* > > While, the data I got in this way seems unresonable.. > > CoresNum L2_CacheHitRatio 1 0.204246888 2 0.310282203 4 0.268968445 8 > 0.274661411 > * > * > *The above data shows the L2_cache hit ratio of a PARSEC benckmark.* > * > * > *Theoretically, the value of L2_CacheHitRatio should decrease as the > increasement* > *of the number of cores. While, the result I got was not the same as my > expectation...* > * > * > *Is the way I count the L2_CacheHitRatio wrong ?* > * > * > * > * > > _______________________________________________ > http://www.marss86.org > Marss86-Devel mailing list > [email protected] > https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel > >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
