On Fri, 4 Aug 2023, Matthew Malcomson wrote:
> Hopefully last update ...
>
> > Specifically, please try compiling with
> > -ftime-report -fdiagnostics-format=sarif-file
> > and have a look at the generated .sarif file, e.g. via
> > python -m json.tool foo.c.sarif
> > which will pretty-print the JSON to stdout.
>
> Rebasing onto the JSON output was quite simple -- I've inlined the only
> change in the patch below (to cast to floating point seconds before
> generating the json report).
>
> I have manually checked the SARIF output as you suggested and all looks
> good (an in fact better because we no longer save some strange times
> like the below due to avoiding the floating point rounding).
> "wall": -4.49516e-09,
>
>
> >
> > The patch looks OK to me if it passes bootstrap / regtest and the
> > output of -ftime-report doesn't change (too much).
> >
> > Thanks,
> > Richard.
>
> Though I don't expect you were asking for this, confirmation below that
> the output doesn't change. (Figured I may as well include that info
> since the rebase to include the JSON output that David had just added
> required re-sending an email anyway).
>
>
>
> ```
> hw-a20-8:checking-theory [10:07:01] $ ${old_build}/gcc/xgcc
> -B${old_build}/gcc/ -fdiagnostics-plain-output -Os -w -S test-sum.c -o
> /dev/null -ftime-report
>
> Time variable usr sys
> wall GGC
> phase setup : 0.01 ( 14%) 0.00 ( 0%) 0.02 (
> 18%) 3389k ( 74%)
> phase parsing : 0.03 ( 43%) 0.02 ( 67%) 0.06 (
> 55%) 982k ( 21%)
> phase opt and generate : 0.03 ( 43%) 0.01 ( 33%) 0.03 (
> 27%) 215k ( 5%)
> callgraph functions expansion : 0.02 ( 29%) 0.00 ( 0%) 0.03 (
> 27%) 162k ( 4%)
> callgraph ipa passes : 0.01 ( 14%) 0.01 ( 33%) -0.00 (
> -0%) 38k ( 1%)
> CFG verifier : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 0 ( 0%)
> preprocessing : 0.02 ( 29%) 0.02 ( 67%) 0.02 (
> 18%) 272k ( 6%)
> lexical analysis : 0.01 ( 14%) 0.00 ( 0%) 0.02 (
> 18%) 0 ( 0%)
> parser (global) : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 637k ( 14%)
> parser function body : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 18k ( 0%)
> tree CFG cleanup : 0.00 ( 0%) 0.01 ( 33%) 0.00 (
> 0%) 0 ( 0%)
> tree STMT verifier : 0.01 ( 14%) 0.00 ( 0%) 0.00 (
> 0%) 0 ( 0%)
> expand : 0.01 ( 14%) 0.00 ( 0%) 0.00 (
> 0%) 12k ( 0%)
> loop init : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 12k ( 0%)
> initialize rtl : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 15k ( 0%)
> rest of compilation : 0.01 ( 14%) 0.00 ( 0%) 0.00 (
> 0%) 6920 ( 0%)
> TOTAL : 0.07 0.03 0.11
> 4587k
> Extra diagnostic checks enabled; compiler may run slowly.
> Configure with --enable-checking=release to disable checks.
> hw-a20-8:checking-theory [10:06:44] $ ${new_build}/gcc/xgcc
> -B${new_build}/gcc/ -fdiagnostics-plain-output -Os -w -S test-sum.c -o
> /dev/null -ftime-report
>
> Time variable usr sys
> wall GGC
> phase setup : 0.01 ( 17%) 0.00 ( 0%) 0.02 (
> 18%) 3389k ( 74%)
> phase parsing : 0.02 ( 33%) 0.03 ( 75%) 0.05 (
> 45%) 982k ( 21%)
> phase opt and generate : 0.03 ( 50%) 0.01 ( 25%) 0.04 (
> 36%) 215k ( 5%)
> callgraph construction : 0.00 ( 0%) 0.01 ( 25%) 0.00 (
> 0%) 1864 ( 0%)
> callgraph functions expansion : 0.02 ( 33%) 0.00 ( 0%) 0.03 (
> 27%) 162k ( 4%)
> callgraph ipa passes : 0.01 ( 17%) 0.00 ( 0%) 0.01 (
> 9%) 38k ( 1%)
> ipa free lang data : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 0 ( 0%)
> CFG verifier : 0.02 ( 33%) 0.00 ( 0%) 0.00 (
> 0%) 0 ( 0%)
> preprocessing : 0.01 ( 17%) 0.03 ( 75%) 0.01 (
> 9%) 272k ( 6%)
> lexical analysis : 0.01 ( 17%) 0.00 ( 0%) 0.02 (
> 18%) 0 ( 0%)
> parser (global) : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 637k ( 14%)
> parser inl. func. body : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 19k ( 0%)
> tree STMT verifier : 0.01 ( 17%) 0.00 ( 0%) 0.00 (
> 0%) 0 ( 0%)
> expand : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 12k ( 0%)
> integrated RA : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 50k ( 1%)
> initialize rtl : 0.00 ( 0%) 0.00 ( 0%) 0.01 (
> 9%) 15k ( 0%)
> TOTAL : 0.06 0.04 0.11
> 4587k
> Extra diagnostic checks enabled; compiler may run slowly.
> Configure with --enable-checking=release to disable checks.
> ```
>
>
> N.b. in the change below it's worth mentioning that the
> `nanosec_to_floating_sec` and `percent_of` macros were already defined.
> I just moved them to below the `make_json_for_timevar_time_def`
> function.
LGTM.
Thanks for doing this,
Richard.
> ############### Changes to the previous patch.
> ###############
>
>
> @@ -832,12 +841,16 @@ json::object *
> make_json_for_timevar_time_def (const timevar_time_def &ttd)
> {
> json::object *obj = new json::object ();
> - obj->set ("user", new json::float_number (ttd.user));
> - obj->set ("sys", new json::float_number (ttd.sys));
> - obj->set ("wall", new json::float_number (ttd.wall));
> + obj->set ("user",
> + new json::float_number (nanosec_to_floating_sec (ttd.user)));
> + obj->set ("sys", new json::float_number (nanosec_to_floating_sec
> (ttd.sys)));
> + obj->set ("wall",
> + new json::float_number (nanosec_to_floating_sec (ttd.wall)));
> obj->set ("ggc_mem", new json::integer_number (ttd.ggc_mem));
> return obj;
> }
> +#undef nanosec_to_floating_sec
> +#undef percent_of
>
> /* Create a json value representing this object, suitable for use
> in SARIF output. */
>
>
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
