It is creating useless work as the commuted nodes is the same as the node we are working on in that case.
Details
Diff Detail
- Build Status
Buildable 6948 Build 6948: arc lint + arc unit
Event Timeline
Do you have any stats?
lib/CodeGen/SelectionDAG/DAGCombiner.cpp | ||
---|---|---|
1565 | Why you're here, please can you change this to SDISel to make it clearer. |
Any compile time stats? Compiling clang before/after this patch should be good enough for an initial comparison.
Hi,
I got sidetracked into some other very urgent project during June. I'll do my best to provide that number as soon as I can. I don't expect the gain, as this, to be that great, but who knows ? What I'm after here is to reduce the impact of D33587 on performances.
OK, benchmarks. Compiling clang from a bc containing clang in its entierety. With the patch:
real 9m45.457s user 9m44.085s sys 0m1.384s
Without
real 9m45.521s user 9m44.016s sys 0m1.517s
The impact on perfs is not very significant. It's within the noise.
I think that as expected, this has a minimal impact on compile time. But of course, it can't hurt I suppose. I don't see why anyone would be opposed to this.
Why you're here, please can you change this to SDISel to make it clearer.