[PATCH] D54442: [llvm-exegesis] Optimize ToProcess in dbScan

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

That is more or less how it ends up looking in https://reviews.llvm.org/D54418, just without abstraction.



================
Comment at: tools/llvm-exegesis/lib/Clustering.cpp:134-135
       // Retrieve a point from the set.
-      const size_t Q = *ToProcess.begin();
-      ToProcess.erase(Q);
+      size_t Q = ToProcess.back();
+      ToProcess.pop_back();
 
----------------
I did try this, at different stages of the patchset.
It always measured to be slower than deque + pop front.


Repository:
  rL LLVM

https://reviews.llvm.org/D54442