[midend/lib/Conversion]TOSA:ReduceSumOP Vectorize Optimization #490
Conversation
There was a problem hiding this comment.
This analysis seems unrelated to the vectorization optimization. Please separate it into a different PR.
There was a problem hiding this comment.
This is only useful for analysis that counts the number of operations in different dialects.
Will re-issue a different PR
There was a problem hiding this comment.
Please use FileCheck to verify correctness and use rtclock to obtain performance results. See here as an example.
There was a problem hiding this comment.
This script is used to verify the correctness of using different Pass Pipelines for the same file and calculate the speedup ratio to verify the efficiency.
There was a problem hiding this comment.
Whether it is an mlir file, you need to use FileCheck to verify the correctness
There was a problem hiding this comment.
Please use English for comments.
| %a = memref.get_global @A : memref<12x40x40xf32> | ||
| call @kernel(%a) : (memref<12x40x40xf32>) -> () | ||
| return | ||
| } |
There was a problem hiding this comment.
Please add an empty line here.
There was a problem hiding this comment.
Name: next-reduce-sum-vec-manual.mlir
There was a problem hiding this comment.
The name '1' is unclear and lacks meaning. Please use a more descriptive and meaningful suffix.
| call @kernel(%c0) : (tensor<12x40x40xf32>) -> () | ||
|
|
||
| return | ||
| } |
There was a problem hiding this comment.
Please add an empty line here.
| call @kernel(%c0) : (tensor<1x40x1536xf32>) -> () | ||
|
|
||
| return | ||
| } |
There was a problem hiding this comment.
Please add an empty line here.
There was a problem hiding this comment.
The name '1' is unclear and lacks meaning. Please use a more descriptive and meaningful suffix.
| PassRegistration<ReduceSumVectorizationPass>(); | ||
| } | ||
| } // namespace buddy | ||
| } // namespace mlir |
There was a problem hiding this comment.
Please add an empty line here.
Hayden727
changed the title
A clean PR commit only have reduce_sum vectorize Pass