Dropped duplicated math function from Thrust

[viralbhadeshiya]

Description

closes #6104

• Removed duplicated math function from Thrust
• Moved all duplicated math function call from Thrust to cuda/std or cuda

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[bernhardmgruber]

Please remove

#include <thrust/detail/integer_math.h>

from files where we no longer need it.

[miscco]

Thanks a lot for working on this issue. It is already looking great.

I found one incorrect replacement and some potential conversion issues

[viralbhadeshiya]

@bernhardmgruber @miscco all changes that you guys suggested is done, lmk if there are more Changes. :)

[miscco]

@viralbhadeshiya This looks great 🎉

Thanks a lot for working on this.

I added a few minor changes:

• dropped unused is_odd
• improved is_negative
• dropped two more includes

[miscco]

/ok to test 07b811c

[miscco]

/ok to test 4261f1f

[miscco]

/ok to test 34fc385

[miscco]

@fbusato There is an issue with thrust::pool_options

I suspect there is a differing behavior for is_power_of_two could you please inevestigate why it differs?

[fbusato]

::cuda::is_power_of_two correctly evaluates 0 as non-power of two, while thrust::detail::is_power_of_2() misses the check x != 0.

This causes a different behavior in !::cuda::is_power_of_two(alignment), which is initialized to 0 by default.
https://github.com/NVIDIA/cccl/pull/6188/files#diff-d85bf89d70ec2f2a069c4c0ffeccd2722fe2a108a012566503392885900f40a1R115

[fbusato]

@viralbhadeshiya my suggestion is just to add the condition alignment != 0 && !::cuda::is_power_of_two(alignment)

[viralbhadeshiya]

::cuda::is_power_of_two correctly evaluates 0 as non-power of two, while thrust::detail::is_power_of_2() misses the check x != 0.

This causes a different behavior in !::cuda::is_power_of_two(alignment), which is initialized to 0 by default. https://github.com/NVIDIA/cccl/pull/6188/files#diff-d85bf89d70ec2f2a069c4c0ffeccd2722fe2a108a012566503392885900f40a1R115

@fbusato would this similar changes also required in !::cuda::is_power_of_two(smallest_block_size) & !::cuda::is_power_of_two(largest_block_size)?

[fbusato]

yes, but I'm not sure that this edge case is evaluated in the test case. You can also run the test locally to make sure it works

[viralbhadeshiya]

yes, but I'm not sure that this edge case is evaluated in the test case. You can also run the test locally to make sure it works

Okay made changes, unfortunately my remote server which had NVIDIA device is facing some issue so could not run tests. but changes are done.

[fbusato]

/ok to test 0ab0b72

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 4h 29m: Pass: 100%/118 | Total: 3d 10h | Max: 4h 28m | Hits: 90%/166971

See results here.