[gem5-users] Confusing performance result when tgts_per_mshr = 1 (O3 CPU model)

[yuhang liu via gem5-users]

Dear Sir/Madam,


With O3 CPU model, I test tgts_per_mshr = 1, 2, 4, 8, 16, 32, and found that  
the performance is best when tgts_per_mshr = 1. Could someone explain it ? It 
seems that  tgts_per_mshr = 1 equals to  tgts_per_mshr = Infinite large.


| tgts_per_mshr | 1 | 2 | 4 | 8 | 16 | 32 |
| 401.bzip2 running time | 0.0450 | 0.072 | 0.063 | 0.046 | 0.0453 | 0.0452 |
| 403.gcc   running time | 0.0141 | 0.015 | 0.015 | 0.014 | 0.0143 | 0.0142 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Best regards.


Yuhang


BaseCache::BaseCache(const Params *p)
    : MemObject(p),
      mshrQueue("MSHRs", p->mshrs, 4, MSHRQueue_MSHRs),
      writeBuffer("write buffer", p->write_buffers, p->mshrs+1000,
                  MSHRQueue_WriteBuffer),
      blkSize(p->system->cacheLineSize()),
      hitLatency(p->hit_latency),
      responseLatency(p->response_latency),
      numTarget(p->tgts_per_mshr),
      forwardSnoops(p->forward_snoops),
      isTopLevel(p->is_top_level),
      blocked(0),
      noTargetMSHR(NULL),
      missCount(p->max_miss_count),
      addrRanges(p->addr_ranges.begin(), p->addr_ranges.end()),
      system(p->system)
{
}


_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users