Add NAN handling to convert() needed for some prefix routines with integer outputs.#502
Merged
rkarim2 merged 20 commits intonv-legate:branch-22.10from Sep 4, 2022
Merged
Conversation
…Ns when converting from complex/float to ints. Eager currently incomplete. On C++ side need to add OP handling to allow templatization on NAN identity values (for performance reasons).
Remove old comments.
…o identity based on operation. Not tested and likely buggy!
Bugfixes. Changed the prefix routines to use the convert with NAN handling. Remove unnecessary code from convert. Added nan handling to eager variant of convert. Intial testing seems to work correctly.
Modified the test code for scan routines, enabling int type outputs on float/complex inputs with NAN handling. Modified the scan test code's parametrization. Removed commented out unnecessary code from convert. pre-commit fixes.
Change input ranges to avoid overflows resulting in NANs (NANs are still tested based on n0)
…nge to avoid overflows.
magnatelee
requested changes
Aug 18, 2022
| type_dispatch(args.in.code(), SourceTypeDispatch<KIND>{}, args); | ||
| ConvertArgs args{ | ||
| context.outputs()[0], context.inputs()[0], context.scalars()[0].value<ConvertCode>()}; | ||
| op_dispatch(args.nan_op, ConvertDispatch<KIND>{}, args); |
Contributor
There was a problem hiding this comment.
Doing this dispatch upfront triples the number of template instantiations, even though the dispatch on nan_op is unnecessary when the source has an integer type (and please remember there are more integer types than floating point types and complex types combined). A more desirable implementation would be to instantiate a special conversion logic only for pairs of types that need it. You can express those pairs using a template like this:
template <LegateTypeCode SRC_TYPE, LegateTypeCode DST_TYPE>
using needs_dispatch_on_nan_op =
(legate::is_floating_type(SRC_TYPE)::value || legate::is_complex_type(SRC_TYPE)::value)
&& legate::is_integer_type(DST_TYPE)::value;
Then you move the dispatch on nan_op to the innermost template and do it only when needs_dispatch_on_nan_op is true.
Contributor
Author
There was a problem hiding this comment.
Resolved in multiple commits, completed fix in d20450a
…sabling unnecessary templates when input is not float/complex (to be disabled in a future commit)
Adjusted nancumsum/nancumprod implementation to switch to the faster cumsum/cumprod if NAN conversion is already handled by convert.
With the change to convert's templatization it's needed (and beneficial) to reroute nancumsum/nancumprod to cumsum/cumprod at python level before convert is called for non-float/complex types. Modified test to cover nancumsum/nancumprod for non-float/complex input types to catch any potential bugs (needed due to how convert's templatization is now done).
magnatelee
approved these changes
Sep 2, 2022
Contributor
|
LGTM. feel free to merge it once you fix the merge conflict |
manopapad
added a commit
that referenced
this pull request
Nov 17, 2024
* Initial pass * todos.rst * Address comments * Fix warnings * Update product positioning * Add supported platform info * Move all Jupyter instructions to Legate * more warnings * Remove todos --------- Co-authored-by: Manolis Papadakis <mpapadakis@nvidia.com>
XiaLuNV
pushed a commit
to XiaLuNV/cunumeric
that referenced
this pull request
Dec 13, 2024
* Initial pass * todos.rst * Address comments * Fix warnings * Update product positioning * Add supported platform info * Move all Jupyter instructions to Legate * more warnings * Remove todos --------- Co-authored-by: Manolis Papadakis <mpapadakis@nvidia.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
No description provided.