Comment # 56
on bug 97879
from Grazvydas Ignotas
Is the async I/O thread doing a lot of memory allocations? If so, perhaps the malloc implementation is slowed down by having large amount of small allocs/frees from shader compiles and/or fragmentation of it's free space. I wonder if it's possible to put mesa on a separate allocator without affecting the rest of the app. If the game is 32bit, it could also be that the address space is almost fully mapped and mmap is struggling to find holes to fit the new allocations in. Just some random guesses there.
You are receiving this mail because:
- You are on the CC list for the bug.
_______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
