Hello Genodians
I have a JAVA application that I execute on request by starting it in a sub init. When debugging why it takes about 6 seconds on x86_64 to handle a request I made the discovery that when Java is included via `import_from_depot` in the runs script, that the execution time goes down to under one second. But when it is included via the depot mechanism, the execution is slow.
In my tries to speed up I modified the JDK package to take the content of the `classes.tar` from a ram_fs where I extracted the file to. But this did not improve the execution time in a significant way. I did this tests on sculpt.
Has someone an idea what causes this increases in execution time and if there could be a way to reduce it?
Regards, Pirmin
Hi Pirmin,
El 21/2/20 a las 14:33, Duss Pirmin escribió:
I have a JAVA application that I execute on request by starting it in a sub init. When debugging why it takes about 6 seconds on x86_64 to handle a request I made the discovery that when Java is included via `import_from_depot` in the runs script, that the execution time goes down to under one second. But when it is included via the depot mechanism, the execution is slow.
Just a wild guess - have you tried comparing the binaries used by import_from_depot (located in <GENODE>/depot/<DEPOT_USER>/bin/...) with those in the build dir (located in <BUILD_DIR>/bin) ? Maybe they come from different versions of the source code.
Apart from that, IMO it's hard to tell without more information about your scenario.
Cheers, Martin