Jon,
I looked into this further, and the logic to execute the loops is too
spread out, touching too many places, which will make things really hard to
maintain.
I'd like to see all the logic for all sort of looping (be it infinite loop,
last-frame, frame range, call-range etc) to be encapsulated in a single
class, "LoopParser", which acts as a decorator for vanilla Parsers:
// Parser interface
class BaseParser {
virtual parse_call() = 0;
};
// Concrete parser
class Parser : public BaseParser {
}
// Decorator for parser which loops
class LoopParser : public BaseParser {
}
LoopParser::parse_call() would internally rewind the stream, so that the
rest of retrace_main.cpp just calls parse_call() and doesn't even know if
its using a LoopParser or a Parser.
You could even have an hierrachy on top of LoopParser: LastFrameLoopParser,
CallRangeLoopParser, FrameRangeLoopParser, etc.
Jose
On Tue, Apr 22, 2014 at 7:53 PM, Jon Ashburn <j...@lunarg.com> wrote:
> glretrace has ability to loop the last frame of a trace continuosly. These
> two patches improve upon this feature in several ways. For analysis of
> the driver/GPU performance from a captured trace the existing looping can
> be
> improved. Firstly, statistics of the looping performance should be
> logged,
> in addition to the total retrace statistics which include non looping
> frames.
> Secondly, the looping should be generalized for N iterations and any
> arbitrary
> range of frames or API call numbers. Thirdly, time spent in apitrace
> parsing
> and executing calls should be minimized so any driver or GPU bottlenecks
> are
> not masked by glretrace looping replay time.
>
> To achieve these improvements, these features are added to glretrace by
> these
> patches:
>
> Add a numeric parameter to the --loop option which is the number of times
> to
> loop. Zero is continuous looping, default is 1.
>
> Add --framerange option that allows an arbitray range of frames to be
> looped
> and log the time spent in the framerange looping.
>
> Add --callrange option that allows an arbitrary range of call numbers to be
> looped and log the time spent in the callrange looping. Framerange and
> callrange
> are mutually exclusive options. --loop=N specifies the number of times to
> loop
> the range. The replay ends after the looping is finished.
>
> When executing a looping replay (non-continuos and iterations != 1) first
> save the parsed API calls. Thus, no need to reparse the calls on every
> iteration of a loop. This vastly improves the execution time of
> glretrace.
> The looping replay performance, in many cases now matches the performance
> for
> the frames of interest from the original app/benchmark prior to capture.
> This
> makes glretrace a very useful performance analysis tool and also
> performance
> regression tool.
> _______________________________________________
> apitrace mailing list
> apitrace@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/apitrace
>
_______________________________________________
apitrace mailing list
apitrace@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/apitrace
