I've extended filebench to write through a device specific API, and so far my rudimentary support works ok. But I've found a platform specific difference that raises a usage question of threads versus processes.
If USE_PROCESS_MODEL is defined, procflow_createproc() will fork/exec a process, otherwise it will spawn a thread. Is there a way to control the usage of threads vs. processes in the workload description? (on a platform that will support both, of course) I'd like to be able to measure behavior both ways. I ran into this issue when I found the code I'd added worked on a linux 2.4 system, but not on a Solaris 10 system, where I'd rather develop. I had added to the grammar to parse parameters my code needs, which I then put into the flowop structure. But after the fork/exec, the child process of course does not have a copy of the flowop structure from the parent, and the parameters are unavailable. Does that structure get put into shared memory so all spawned processes can access it? (It looks like it does, at least sometimes, in flowop_define_common) thanks, tim This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
