Hm, would it be possible to benchmark at the end of the xenofarm build?
It would be interesting to be able to see performance graph versus revision (time). :)
(But maybe people don't want to spend the extra cpu-minute or so?)
Well, you could create a new xenofarm target so that those with lots of CPU time to waste can perform benchmarks after the other tests. Something like
xenofarm_benchmark: -rm -rf xenofarm_result mkdir xenofarm_result -CCACHE_LOGFILE="`pwd`/xenofarm_result/ccache.log.txt" \ MAKE="$(MAKE_CMD)" CONFIGUREARGS="$(CONFIGUREARGS)" \ BUILDDIR="$(BUILDDIR)" /bin/sh bin/xenofarm.sh @$(DO_MAKE) $(MAKE_FLAGS) "METATARGET=run_bench" > \ xenofarm_result/bench.txt cd xenofarm_result && tar cf - . > ../xenofarm_result.tar gzip -f9 xenofarm_result.tar
(but tested)
/ Martin Nilsson (ja till euro, nej till cent)
Previous text:
2003-09-12 15:44: Subject: benchmark in xenofarm
Hm, would it be possible to benchmark at the end of the xenofarm build?
It would be interesting to be able to see performance graph versus revision (time). :)
(But maybe people don't want to spend the extra cpu-minute or so?)
/ Mirar
Actually, the benchmarks would take a very, very insignificant amount of time compared to the other tests. :-)
/ Per Hedbor ()
Previous text:
2003-09-12 15:49: Subject: benchmark in xenofarm
Well, you could create a new xenofarm target so that those with lots of CPU time to waste can perform benchmarks after the other tests. Something like
xenofarm_benchmark: -rm -rf xenofarm_result mkdir xenofarm_result -CCACHE_LOGFILE="`pwd`/xenofarm_result/ccache.log.txt" \ MAKE="$(MAKE_CMD)" CONFIGUREARGS="$(CONFIGUREARGS)" \ BUILDDIR="$(BUILDDIR)" /bin/sh bin/xenofarm.sh @$(DO_MAKE) $(MAKE_FLAGS) "METATARGET=run_bench" > \ xenofarm_result/bench.txt cd xenofarm_result && tar cf - . > ../xenofarm_result.tar gzip -f9 xenofarm_result.tar
(but tested)
/ Martin Nilsson (ja till euro, nej till cent)
Probably true. Then it is bin/xenofarm.sh that needs extending.
/ Martin Nilsson (ja till euro, nej till cent)
Previous text:
2003-09-12 16:04: Subject: benchmark in xenofarm
Actually, the benchmarks would take a very, very insignificant amount of time compared to the other tests. :-)
/ Per Hedbor ()
In the last episode (Sep 12), Martin Nilsson (ja till euro, nej till cent) @ Pike (-) developers forum said:
Probably true. Then it is bin/xenofarm.sh that needs extending.
Could the benchmarks be run whether the verify/export succeeded or not? There are a lot of yellow builds in the 7.5 xenofarm and most of them generate pikes that function well enough to run the benchmark.
Sure.
/ Martin Nilsson (saturator)
Previous text:
2003-09-13 21:18: Subject: Re: benchmark in xenofarm
In the last episode (Sep 12), Martin Nilsson (ja till euro, nej till cent) @ Pike (-) developers forum said:
Probably true. Then it is bin/xenofarm.sh that needs extending.
Could the benchmarks be run whether the verify/export succeeded or not? There are a lot of yellow builds in the 7.5 xenofarm and most of them generate pikes that function well enough to run the benchmark.
-- Dan Nelson dnelson@allantgroup.com
/ Brevbäraren
I have a feeling that an elfwoodish nice graph view of such data would encourage some people to spend the extra cpu time. :-)
/ Johan Sundström, Lysator
Previous text:
2003-09-12 15:44: Subject: benchmark in xenofarm
Hm, would it be possible to benchmark at the end of the xenofarm build?
It would be interesting to be able to see performance graph versus revision (time). :)
(But maybe people don't want to spend the extra cpu-minute or so?)
/ Mirar
pike-devel@lists.lysator.liu.se