This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH v2 4/7] malloc: Ensure lower bound on chunk size in __libc_realloc.


"Carlos O'Donell" <carlos@redhat.com> writes:
>> I couldn't detect any measurable change from this patch.
>  
> How did you benchmark it? perf?

I time multiple runs of trace_run with all my non-stupidly-big
workloads, and average the results.  The raw results are:


----------  389ds ----------

PRISTINE:
Total call time: 7,365,545,225 cycles

PATCHED: 
Total call time: 7,443,954,619 cycles

----------  dj ----------

PRISTINE:
Total call time: 831,383,129 cycles

PATCHED: 
Total call time: 848,574,833 cycles

----------  dj2 ----------

PRISTINE:
Total call time: 7,391,093,601 cycles

PATCHED: 
Total call time: 7,389,861,196 cycles

----------  okular-1 ----------

PRISTINE:
Total call time: 3,175,103,662 cycles

PATCHED: 
Total call time: 3,190,124,319 cycles

----------  oocalc ----------

PRISTINE:
Total call time: 1,037,282,464 cycles

PATCHED: 
Total call time: 1,037,154,283 cycles

----------  proprietary-2 ----------

PRISTINE:
Total call time: 1,999,062,083 cycles

PATCHED: 
Total call time: 1,998,548,797 cycles

----------  qemu-virtio ----------

PRISTINE:
Total call time: 806,573,463 cycles

PATCHED: 
Total call time: 795,207,373 cycles

----------  qemu-win7 ----------

PRISTINE:
Total call time: 649,777,212 cycles

PATCHED: 
Total call time: 659,636,829 cycles


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]