[PATCH] D46227: [islpp] Do not abuse isl::stat::error as early-abort

[Philip Pfaffe via Phabricator via llvm-commits]

philip.pfaffe added a comment.

I think this makes sense, with some inline nits.



================
Comment at: lib/Analysis/ScopInfo.cpp:2358
 
-  if (isl_ctx_last_error(Ctx.get()) == isl_error_quota)
-    return isl::stat::error;
+  if (isl_ctx_last_error(Ctx.get()) == isl_error_quota) {
+    Aborted = true;
----------------
Shouldn't this still be an error?


================
Comment at: lib/Analysis/ScopInfo.cpp:2380
 
-  return isl::stat::ok;
+  return;
 }
----------------
This is now superfluous.


================
Comment at: lib/Analysis/ScopInfo.cpp:2412
+
+  return !Aborted;
 }
----------------
In case of `isl_error_quota`, this should also return false in that case.


================
Comment at: lib/Transform/FlattenAlgo.cpp:61
     if (isVariableDim(BMap))
-      return isl::stat::error;
+      IsVariableDim = true;
     return isl::stat::ok;
----------------
You could check  this condition before testing the BMap's out-dimension to save a little bit of compute.


Repository:
  rPLO Polly

https://reviews.llvm.org/D46227