[parsec-users] A question regarding VIPS's behavior
jzz at mit.edu
Sun Jul 8 04:21:23 EDT 2012
ok, it seems it has some chaining functionality, that all functions are chained and executed at the latest moment. That's why the last function consumes most of the time...
在 2012-7-8，上午4:15， Justin Zhang 写道：
> I am encountering a strange behavior of VIPS that I am not sure if other people have the same issue. Basically, in my case, it seems most of the time was spent in the function im_LabQ2disp(...) which has disk access. The im_benchmark function in libsrc/other/im_benchmark.c calls 16 functions and finally calls im_LabQ2disp. I suppose those 16 functions are the main computing load. However a close look, in my case, shows that none of them does anything. They all rely on the im_wrapone, im_wrapmany, im_generate functions to handle their multi-threading. In im_generate, the situations was divided, according to im->dtype, into PARTIAL and FULL. The FULL situation will execute the function, and the PARTIAL situation will just return. In my case, they all evaluate to PARTIAL, so none of them does anything. I am not sure if I have explained this clear, but have anyone seen a similar problem?
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 1581 bytes
Desc: not available
More information about the parsec-users