caching - cpu cache performance. store misses vs load misses -


i'm using perf basic event counter. i'm working on program suffers data cache store misses. as high ratio of %80.

i know how caches in principle work. loads memory on various miss cases, removes data cache when pleases. don't understand , difference between store - load misses. how differ loading , storing. how can store-miss ?

a load-miss (as know) referring when processor needs fetch data main memory, data not exist in cache. whenever processor wants data main memory, esquires cache, , if data loaded load-hit , otherwise load-miss.

a store-miss related when processor wants write newly calculated data main memory.when wants write-back data main memory, hasto make sure content of cache , main memory in sync each other. can happen 2 different policies can find here: writing policies.

so no matter policy choose, first need check whether data in cache can store cache first (since it's faster), , if data block looking has been evicted cache, store-miss related cache.

you can check the applet here, better idea of happens in different scenarios.


Comments

Popular posts from this blog

Line ending issue with Mercurial or Visual Studio -

java - Jtable duplicate Rows -

java - Run a .jar on Heroku -