Right now, output is single threaded to a single file/stdout.
It would be far superior for diffing and usability purposes to dump each thread into a separate file, so that the command sequence of each thread is easily isolated, and would make comparing output of multi-threaded applications much easier.
The main question is the naming scheme of files. Currently, API Dump uses 0 based indexing (adding a thread to the 'known thread' list on the first call it makes into Vulkan). This is okay, but would likely be more useful if the threadid mapped to an external value, as mentioned in #2005
Right now, output is single threaded to a single file/stdout.
It would be far superior for diffing and usability purposes to dump each thread into a separate file, so that the command sequence of each thread is easily isolated, and would make comparing output of multi-threaded applications much easier.
The main question is the naming scheme of files. Currently, API Dump uses 0 based indexing (adding a thread to the 'known thread' list on the first call it makes into Vulkan). This is okay, but would likely be more useful if the threadid mapped to an external value, as mentioned in #2005