I want to revive this patch for mainline and have some questions on
Honza's comments.
On Fri, Apr 29, 2011 at 1:48 PM, Jan Hubicka wrote:
>>
>> A known limitation is that value profiling is not yet sampled, but it
>> does not seem to cause problems.
>
> For the performance alone, we probably don'
Committed the attached patch to google/main. Will send a patch for trunk soon.
On Thu, Apr 28, 2011 at 10:03 PM, Xinliang David Li wrote:
> Please add regression test cases for the feature. Address the comments
> when available. Ok for google/main.
>
> Thanks,
>
> David
>
> On Thu, Apr 28, 2011
Hi,
> + Honza
>
> This patch may be a candidate for trunk as well. This feature not only
> allows profile collection with much less overhead (for multi-thread
> programs with hot regions, the slow down can be significant due to
> cache ping-pong effect of counter update) without sacrificing too mu
On Apr 29, 2011, at 9:12 AM, Silvius Rus wrote:
> When you build with
> -fprofile-generate=./fdoprof, the output gets dumped under
> ./fdoprof/..., but there seems to be no easy way to know this path
> within the profiling process. For Google servers this makes
> collection fragile. The user gen
On Fri, Apr 29, 2011 at 9:12 AM, Silvius Rus wrote:
>> How is code-size affected with this patch, non-instrumented vs.
>> regular-instrumented vs. sample-instrumented?
>
> I don't have the numbers, but the increase in code size from
> regular-instrumented to sample-instrumented is larger than that
> How is code-size affected with this patch, non-instrumented vs.
> regular-instrumented vs. sample-instrumented?
I don't have the numbers, but the increase in code size from
regular-instrumented to sample-instrumented is larger than that from
non-instrumented to sample-instrumented. Easwaran, co
On Fri, Apr 29, 2011 at 1:42 AM, Easwaran Raman wrote:
> This patch from Silvius Rus adds support for sampled edge profile collection
> to reduce instrumentation run overhead. Bootstraps and no test regressions.
> Ok for google/main?
>
> 2011-04-28 Silvius Rus
>
> * doc/invoke.texi: D
Please add regression test cases for the feature. Address the comments
when available. Ok for google/main.
Thanks,
David
On Thu, Apr 28, 2011 at 4:42 PM, Easwaran Raman wrote:
> This patch from Silvius Rus adds support for sampled edge profile collection
> to reduce instrumentation run overhe
On Apr 28, 2011, at 4:42 PM, Easwaran Raman wrote:
> This patch from Silvius Rus adds support for sampled edge profile collection
> to reduce instrumentation run overhead.
Sounds interesting for trunk...
On Thu, Apr 28, 2011 at 4:58 PM, Xinliang David Li wrote:
>
> + Honza
>
> This patch may be a candidate for trunk as well. This feature not only
> allows profile collection with much less overhead (for multi-thread
> programs with hot regions, the slow down can be significant due to
> cache ping-p
+ Honza
This patch may be a candidate for trunk as well. This feature not only
allows profile collection with much less overhead (for multi-thread
programs with hot regions, the slow down can be significant due to
cache ping-pong effect of counter update) without sacrificing too much
the performan
This patch from Silvius Rus adds support for sampled edge profile collection
to reduce instrumentation run overhead. Bootstraps and no test regressions. Ok
for google/main?
2011-04-28 Silvius Rus
* doc/invoke.texi: Document -fprofile-generate-sampling option.
* gcov-io.h (__
12 matches
Mail list logo
