Displaying 3 results from an estimated 3 matches for "abababab".
2006 Jan 10
1
[LLVMdev] Re: passmanager, significant rework idea...
...n the second on each function bottom up.
>
> Instead of this, it would be better to have a CallGraphSCCPassBatcher
> thing, that is added to the ModulePass. Given this, for each function,
> bottom up, we can pipeline between then two passes. This gives us the
> nice ABABABAB ordering instead of AAAABBBB ordering which is nice for
> cache behavior of the compiler.
Yup, interleaving would be far more efficient.
> 2) Once 1) is implemented, if a Module PassManager currently has a
> "CallGraphSCCPassBatcher" active, it makes sense to use a new...
2006 Jan 10
0
[LLVMdev] passmanager, significant rework idea...
...n
bottom-up, then run the second on each function bottom up.
Instead of this, it would be better to have a CallGraphSCCPassBatcher
thing, that is added to the ModulePass. Given this, for each function,
bottom up, we can pipeline between then two passes. This gives us the
nice ABABABAB ordering instead of AAAABBBB ordering which is nice for
cache behavior of the compiler.
2) Once 1) is implemented, if a Module PassManager currently has a
"CallGraphSCCPassBatcher" active, it makes sense to use a new batcher
for the function passes. Since we don't need t...
2006 Jan 10
3
[LLVMdev] passmanager, significant rework idea...
The patch below basically hammers out some ideas as to where I'd like
to take the passmanager in LLVM. I've tried thinking things through,
but I'm still a n00b, so some criticism would be more than welcome. =)
Starting from line 191 down. If you're wondering why I created a
patch, well that's because I found thinking in passmanagert.h the most
productive.
--
Regards.