| Message ID | 20240813140338.143045-1-jdek@itanimul.li |
|---|---|
| State | New |
| Headers | show |
| Series | [FFmpeg-devel,1/7] checkasm: add csv/tsv bench output | expand |
| Context | Check | Description |
|---|---|---|
| yinshiyou/make_loongarch64 | success | Make finished |
| yinshiyou/make_fate_loongarch64 | success | Make fate finished |
Quoting J. Dekker (2024-08-13 16:03:30) > When collecting performance information from checkasm it is common > to parse the output for use in graphs to compare vs different > architectures. > > Signed-off-by: J. Dekker <jdek@itanimul.li> > --- > tests/checkasm/checkasm.c | 22 ++++++++++++++++++++-- > 1 file changed, 20 insertions(+), 2 deletions(-) Is TSV a standard acronym? I've never seen it before. And would it make sense to allow specifying an arbitrary separator string?
On Fri, Aug 16, 2024, at 13:03, Anton Khirnov wrote: > Quoting J. Dekker (2024-08-13 16:03:30) >> When collecting performance information from checkasm it is common >> to parse the output for use in graphs to compare vs different >> architectures. >> >> Signed-off-by: J. Dekker <jdek@itanimul.li> >> --- >> tests/checkasm/checkasm.c | 22 ++++++++++++++++++++-- >> 1 file changed, 20 insertions(+), 2 deletions(-) > > Is TSV a standard acronym? I've never seen it before. > And would it make sense to allow specifying an arbitrary separator > string? I think TSV is understood within the context of CSV. Initially I was going to just have only the ‘standard’ CSV but I would personally prefer a whitespace separator like <TAB> (I have seen the ASCII control characters used too). I don’t think that it would make sense to support any further arbitrary characters since you can s/\t/<sep>/g. jd
Le 16 août 2024 14:03:01 GMT+03:00, Anton Khirnov <anton@khirnov.net> a écrit : >Quoting J. Dekker (2024-08-13 16:03:30) >> When collecting performance information from checkasm it is common >> to parse the output for use in graphs to compare vs different >> architectures. >> >> Signed-off-by: J. Dekker <jdek@itanimul.li> >> --- >> tests/checkasm/checkasm.c | 22 ++++++++++++++++++++-- >> 1 file changed, 20 insertions(+), 2 deletions(-) > >Is TSV a standard acronym? I've never seen it before. Me neither. >And would it make sense to allow specifying an arbitrary separator >string? Sounds a bit overkill.
On Tue, Aug 13, 2024 at 4:03 PM J. Dekker <jdek@itanimul.li> wrote: > When collecting performance information from checkasm it is common > to parse the output for use in graphs to compare vs different > architectures. > > Signed-off-by: J. Dekker <jdek@itanimul.li> > --- When I redirect stdout to a csv file, the first two lines are: benchmarking with Linux Perf Monitoring API nop: 70.0 What about printing those to stderr when csv is enabled?
diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c index 58597d3888..f82ee0864f 100644 --- a/tests/checkasm/checkasm.c +++ b/tests/checkasm/checkasm.c @@ -362,6 +362,8 @@ static struct { const char *cpu_flag_name; const char *test_name; int verbose; + int csv; + int tsv; volatile sig_atomic_t catch_signals; } state; @@ -586,7 +588,14 @@ static void print_benchs(CheckasmFunc *f) CheckasmPerf *p = &v->perf; if (p->iterations) { int decicycles = (10*p->cycles/p->iterations - state.nop_time) / 4; - printf("%s_%s: %d.%d\n", f->name, cpu_suffix(v->cpu), decicycles/10, decicycles%10); + if (state.csv) { + const char sep = state.tsv ? '\t' : ','; + printf("%s%c%s%c%d.%d\n", f->name, sep, + cpu_suffix(v->cpu), sep, + decicycles / 10, decicycles % 10); + } else { + printf("%s_%s: %d.%d\n", f->name, cpu_suffix(v->cpu), decicycles/10, decicycles%10); + } } } while ((v = v->next)); } @@ -829,7 +838,12 @@ static void bench_uninit(void) static int usage(const char *path) { fprintf(stderr, - "Usage: %s [--bench] [--runs=<ptwo>] [--test=<pattern>] [--verbose] [seed]\n", + "Usage: %s [options...] [seed]\n" + " --test=<pattern> Run specific test.\n" + " --bench Run benchmark.\n" + " --csv, --tsv Output benchmark results in CSV or TSV format.\n" + " --runs=<ptwo> Manual number of benchmark iterations to run 2**<ptwo>.\n" + " --verbose Increase verbosity.\n", path); return 1; } @@ -877,6 +891,10 @@ int main(int argc, char *argv[]) state.bench_pattern = ""; } else if (!strncmp(arg, "--test=", 7)) { state.test_name = arg + 7; + } else if (!strcmp(arg, "--csv")) { + state.csv = 1; state.tsv = 0; + } else if (!strcmp(arg, "--tsv")) { + state.csv = 1; state.tsv = 1; } else if (!strcmp(arg, "--verbose") || !strcmp(arg, "-v")) { state.verbose = 1; } else if (!strncmp(arg, "--runs=", 7)) {
When collecting performance information from checkasm it is common to parse the output for use in graphs to compare vs different architectures. Signed-off-by: J. Dekker <jdek@itanimul.li> --- tests/checkasm/checkasm.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-)