On 12/6/24 8:55 AM, Richard Biener wrote:
The following addresses slow bitmap operations for maintaining the
iteration order of df_worklist_dataflow_doublequeue for large number
of basic-blocks. One change is switching the worklist and pending
bitmaps to tree view, another change is avoiding the fully populated
initial bitmap for the first iteration and instead special-casing that
plus avoiding all forward worklist bitmap sets in that iteration.
Usually second or later iterations are sparse, so optimizing the first
iteration is worthwhile. In fact both changes in isolation achieve
the speedup below already, the combination accounts for a minor
additional speedup.
For PR117932 when isolating from ext-dce and fold-mem-offset issues
this results in a 10% compile-time reduction.
Bootstrap and regtest running on x86_64-unknown-linux-gnu. I plan
to push this in the next days unless there are comments suggesting
otherwise.
Sounds good. I haven't forgotten those compile-time issues, just been a
busier than usual couple weeks.
jeff
