Date: | Thu, 27 Oct 2011 00:33:38 -0500 |
---|---|
From: | Monobrata Debnath <mono0908@xxxxxxxxx> |
Subject: | [Gems-users] How to get L1 miss statistics for SNUCA |
Hello gems Group: I have been running gems for a while.Initially I could not have managed to print any L1 miss statistics. What I had is given below: 1D_cache cache stats: L1D_cache_total_misses: 0 L1D_cache_total_demand_misses: 0 L1D_cache_total_prefetches: 0 L1D_cache_total_sw_prefetches: 0 L1D_cache_total_hw_prefetches: 0 L1D_cache_misses_per_transaction: 0 L1D_cache_misses_per_instruction: 0 L1D_cache_instructions_per_misses: NaN Then I reset the "REMOVE_SINGLE_CYCLE_DCACHE_FAST_PATH" flag to false. Now I have miss rate data for both L1D and L1I caches. However it is mentioned in gems documentation wiki /CMP-NUCA for SNUCA "REMOVE_SINGLE_CYCLE_DCACHE_FAST_PATH" must be set to true. Please let me know , if there is any easy way to get the statistics for L1 cache miss/hit choosing SNUCA as cache design without altering ruby's code. Also if the g_cache_design is CMP_TOKEN , what kind of cache design for this category? from the available information I can say CMP_token is a cache cohrency protocol. Any advise is deeply appreciated Thanks, Monobrata |
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | Re: [Gems-users] Ruby Make Error:-No rule to make target `MachineType.h, unni_c |
---|---|
Next by Date: | [Gems-users] How to parallelize RUBY, Jungho Park |
Previous by Thread: | [Gems-users] A NUCA based mesh CMP, lisa890608 |
Next by Thread: | [Gems-users] How to parallelize RUBY, Jungho Park |
Indexes: | [Date] [Thread] |