You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When pasim is given the --print-stats some_fn it should print stats that are exclusively for that function.
For some programs, when this flag is given, no stats are printed at all.
An example program is cosf from the tacle benchmark when compiled with the options -O2 -mllvm --mpatmos-disable-vliw=false. I have not found a minimal example that triggers this.
I dug around in the code and found out that this loop is at fault.
From the comments it seems the enclosing blocks are trying to identify longjmps.
But, since they cannot be handled correctly, the debug stack is just deleted, meaning nothing gets printed at the end.
Because the printing is essential for performance measurements, I will be commenting out this block of code.
This issue will track the eventual fixing whatever the underlying issue is.
The text was updated successfully, but these errors were encountered:
When pasim is given the
--print-stats some_fn
it should print stats that are exclusively for that function.For some programs, when this flag is given, no stats are printed at all.
An example program is
cosf
from the tacle benchmark when compiled with the options-O2 -mllvm --mpatmos-disable-vliw=false
. I have not found a minimal example that triggers this.I dug around in the code and found out that this loop is at fault.
From the comments it seems the enclosing blocks are trying to identify longjmps.
But, since they cannot be handled correctly, the debug stack is just deleted, meaning nothing gets printed at the end.
Because the printing is essential for performance measurements, I will be commenting out this block of code.
This issue will track the eventual fixing whatever the underlying issue is.
The text was updated successfully, but these errors were encountered: