Replies: 3 comments
-
I believe the two main options you have for decided where to split are call graphs analysis and profiling. The callgraph method is the likely the most reliable method but could end up pulling more than you would want into the main module. I would hope that you would automate it and therefore the cost of maintenance would not be significant. |
Beta Was this translation helpful? Give feedback.
-
The most robust solution would be to add a new option to the cc @dschuff, with whom I was recently discussing options in this space. |
Beta Was this translation helpful? Give feedback.
-
@tlively Yes, this is something useful that would help! |
Beta Was this translation helpful? Give feedback.
-
I'm building my application with threads and and with the
sPROXY_TO_PTHREAD
flag.When we split the modules (
-sSPLIT_MODULE
), there are some functions that have to be present in the primary module, mainly the proxied functions like syscalls, webgpu and pthread (the one I'm most interested in) operations.Is there a deterministic way to collect all these proxied functions and have them kept in the primary module?
I though of using a callgraph to find the functions that are needed to create a list, however there are too many workflows and this lists needs to be maintained.
cc @sbc100 @tlively
Beta Was this translation helpful? Give feedback.
All reactions