[GitHub] [tvm] masahi commented on pull request #10172: [TIR] Fix Ramp int32~64 mismatch in VectorizeLoop and NarrowDataType passes

[GitBox <gi...@apache.org>]

masahi commented on pull request #10172:
URL: https://github.com/apache/tvm/pull/10172#issuecomment-1047280891


   I've just hit a similar error, when compiling an int8 model with tensorized ops (VNNI):
   
   ```
     21: tvm::te::MakeTensorize(tvm::te::ComputeOpNode const*, tvm::te::Stage const&, std::unordered_map<tvm::tir::IterVar, tvm::Range, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::Range> > > const&, bool)
     20: tvm::te::VerifyTensorizeBody(tvm::te::ComputeOpNode const*, tvm::te::Stage const&, std::unordered_map<tvm::tir::IterVar, tvm::PrimExpr, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::PrimExpr> > > const&, std::unordered_map<tvm::tir::IterVar, tvm::Range, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::Range> > > const&, std::unordered_map<tvm::tir::IterVar, tvm::Range, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::Range> > > const&, std::unordered_map<tvm::te::Tensor, tvm::runtime::Array<tvm::Range, void>, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::runtime::Array<tvm::Range, void> > > > const&, tvm::te::TensorIntrin const&)
     19: tvm::te::MatchTensorizeBody(tvm::te::ComputeOpNode const*, tvm::te::Stage const&, std::unordered_map<tvm::tir::IterVar, tvm::Range, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::Range> > > const&, std::unordered_map<tvm::tir::IterVar, tvm::Range, std::hash<tvm::tir::IterVar>, std::equal_to<tvm::tir::IterVar>, std::allocator<std::pair<tvm::tir::IterVar const, tvm::Range> > > const&, std::unordered_map<tvm::te::Tensor, tvm::runtime::Array<tvm::Range, void>, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::runtime::Array<tvm::Range, void> > > > const&, tvm::te::TensorIntrin const&, tvm::runtime::Map<tvm::tir::Var, tvm::Range, void, void>*)
     18: non-virtual thunk to tvm::tir::StmtExprMutator::VisitExpr(tvm::PrimExpr const&)
     17: _ZZN3tvm3tir11ExprFunctorIFNS_8PrimExprERKS2_EE10InitVTableEvENUlRKNS_7runtime
     16: tvm::te::TensorIntrinMatcher::VisitExpr_(tvm::tir::ReduceNode const*)
   
     ...
   
     0: tvm::tir::FloorDiv::FloorDiv(tvm::PrimExpr, tvm::PrimExpr, tvm::Span)
     File "/home/masa/projects/dev/tvm/src/tir/ir/expr.cc", line 322
   TypeError: Check failed: (a.dtype() == b.dtype()) is false: mismatched types. int64 vs. int32
   ```
   
   I wonder if this is related.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org