Support non-floating point operations in compute_flops for AutoTVM #2776
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #2628.
This is a simple modification that does two things:
a) Extends the unit tests in test_autotvm_flop_calculator to randomly sample various datatypes.
b) Modifies compute_flop to stop counting flops from indexing expressions (i.e.
X[a, 4 * b, 5 * c]
does not count as an operation, which is inline with expectations), and unconditionally accumulate flops from mathematical expressions (x * y + z, etc) without checking the datatype (which is what we expect for non-floating point operations).The tests now pass for all cases, which is what we expect.