Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Develop Stream 2023-05-12 #243

Closed
wants to merge 26 commits into from
Closed

Develop Stream 2023-05-12 #243

wants to merge 26 commits into from

Conversation

Beanavil
Copy link
Contributor

This PR adds improvements to existing algorithms/tests and solves several small bugs.

  • BlockRadixRank and BlockRadixRankMatch are now implemented in terms of rocPRIM.
  • hipCUB now matches CUB's 2.0.0 and 2.0.1 releases, except for some breaking changes left out of this PR.
  • Extended floats are now present in more tests.
  • Documentation is now updated to the current state of hipCUB.
  • Fixed DeviceSegmentedReduce::ArgMin and DeviceSegmentedReduce::ArgMax for INF input .
  • References to hcc have been removed.
  • New overload of StableSortKeysCopy has been added to address part of CUB's 2.1.0 release.
  • Several bug fixes and optimizations.

Snektron and others added 26 commits May 10, 2023 13:33
@Snektron @Beanavil
use rocprim::block_radix_rank for BlockRadixRank
d237a1d
@Snektron @Beanavil
fix a comparsion warning in block radix rank test
47a9f7f
@Snektron @Beanavil
use rocprim::block_radix_rank for BlockRadixRankMatch
fe22067
@Snektron @Beanavil
use exchange to transform input to warp striped for match test
fe3880e 
load direct warp striped cannot load into warp striped if the
block size does not divide the hardware warp size.
@Snektron @Beanavil
re-enable rank match tests on amd
aeedb35
@Snektron @Beanavil
re-enable rank match benchmarks
ae11cec
@Snektron @Beanavil
factor out rank digit extractor adopter
ab0b16a
@Snektron @Beanavil
naming nit
1adace0
@Beanavil
CUB 2.0.1
5a90310
@Naraenda @Beanavil
Resolve "Add BFloat16 / half to more tests"
ed25df1
@MathiasMagnus @Beanavil
Add missing header
512e1b9
@MathiasMagnus @Beanavil
Alphabetical include order
ead86a1
@MathiasMagnus @Beanavil
Update copyright date
0a898b0
@Beanavil
Fix convert_to_fundamental util function name
a6b810b
@nolmoonen @Beanavil
workaround for failing benchmark upload
624b3e1
@Beanavil
Documentation updated and bug fix in grid_queue.hpp
18680d6
@nolmoonen @Beanavil
fix device segmented argmin/argmax
d095f73
@nolmoonen @Beanavil
add half types to segmented test, disable inf test on nvidia
d4c86e9
@nolmoonen @Beanavil
alternative segmented arg min max impl
29d3bcf
@nolmoonen @Beanavil
small optimization
d12d2a3
@Snektron @Beanavil
skip SortKeysOver4G if test device has less than 8 GB VRAM
db1e591 
The HipcubDeviceRadixSort.SortKeysOver4G requires more than 8 GB of
global memory to complete, and fails on devices with 8 GB or less. This
commit disables the test for devices which have 8 GB or less.
@Beanavil
Add support for single-threaded executions of WarpReduce
75503e2
@sbalint98 @Beanavil
Remove reference to hcc
230bb6e
@sbalint98 @Beanavil
Remove reference to hcc
ec3e0d4
@nolmoonen @Beanavil
add device merge sort stable sort keys copy
c404c23
@nolmoonen @Beanavil
disable test for nvidia platform
e9d7718
@Beanavil Beanavil closed this May 12, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
6 participants
@Beanavil @Snektron @Naraenda @MathiasMagnus @nolmoonen @sbalint98