[llvm] [LV] Don't trigger legacy/vplan assert when forcing costs (PR #156870)

[David Sherwood via llvm-commits]

================
@@ -7047,7 +7047,13 @@ VectorizationFactor LoopVectorizationPlanner::computeBestVF() {
   // Verify that the VPlan-based and legacy cost models agree, except for VPlans
   // with early exits and plans with additional VPlan simplifications. The
   // legacy cost model doesn't properly model costs for such loops.
+  // NOTE: If the user has forced a target instruction cost this assert is very
+  // likely to trigger because the VPlan recipes don't map 1:1 with the scalar
+  // instructions that the legacy cost model is based on. One example of this is
+  // for interleave groups - VPlan will use the forced cost for the whole group,
+  // whereas the legacy cost model will use it for each load.
   assert((BestFactor.Width == LegacyVF.Width || BestPlan.hasEarlyExit() ||
+          ForceTargetInstructionCost.getNumOccurrences() > 0 ||
----------------
david-arm wrote:

OK, I'm happy to look into that approach, but still it feels like a lot of effort to fix something that the developer probably doesn't care about. Obviously this is just my own example, but when I used the flag `-force-instruction-cost=1` I wasn't interested in knowing whether they agreed or not. What I really cared about was that all code paths have been tested, nothing crashes during compilation or execution. Sometimes the assert just gets in the way of making sure a patch works correctly and I frequently have to delete the assert and rebuild.

One alternative I'd also be happy with is an override flag that allows developers to disable the assert.

https://github.com/llvm/llvm-project/pull/156870