Changes quicksort and quickselect to use template based sorting networks #61
Conversation
This file contains 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
sterrettm2
force-pushed
the
generic_nets
branch
from
6830beb to
70424a6
Compare
|
We will want to use |
r-devulap
reviewed
Aug 18, 2023
src/xss-network-qsort.hpp
Outdated
| int64_t num_to_write | ||
| = std::min((int64_t)std::max(0, N - i * vtype::numlanes), | ||
| (int64_t)vtype::numlanes); | ||
| typename vtype::opmask_t load_mask = ((0x1ull << num_to_write) - 0x1ull) |
There was a problem hiding this comment.
Use vtype::get_partial_loadmask instead. See
x86-simd-sort/src/avx512-32bit-qsort.hpp
Line 259 in 0890de5
You also won't need num_to_write to be a int64_t
There was a problem hiding this comment.
you might need to update that function. It won't work when num_to_write == numlanes
src/xss-network-qsort.hpp
Outdated
| int64_t num_to_write | ||
| = std::min((int64_t)std::max(0, N - i * vtype::numlanes), | ||
| (int64_t)vtype::numlanes); | ||
| typename vtype::opmask_t load_mask = ((0x1ull << num_to_write) - 0x1ull) |
There was a problem hiding this comment.
Ditto. Use vtype::get_partial_loadmask.
src/avx512-common-qsort.h
Outdated
| @@ -738,4 +819,37 @@ inline void avx512_partial_qsort_fp16(uint16_t *arr, | |||
| avx512_qselect_fp16(arr, k - 1, arrsize, hasnan); | |||
| avx512_qsort_fp16(arr, k - 1); | |||
| } | |||
|
|
|||
| template <typename vtype, typename type_t> | |||
| X86_SIMD_SORT_INLINE type_t get_pivot_64bit(type_t *arr, | |||
There was a problem hiding this comment.
Lets move these template specializations to their respective files.
src/avx512-common-qsort.h
Outdated
| const int64_t right); | ||
|
|
||
| template <typename vtype, typename type_t> | ||
| X86_SIMD_SORT_INLINE type_t get_pivot(type_t *arr, |
There was a problem hiding this comment.
We probably don't need this once you move the specialized functions to their own files.
r-devulap
requested changes
Aug 22, 2023
|
numpy/numpy#24498 was merged. Could you rebase with main and update? Also, use x86-simd-sort/src/avx512-common-qsort.h Line 98 in b9f9340 |
sterrettm2
force-pushed
the
generic_nets
branch
2 times, most recently
from
d5d5a90 to
4ff3e1d
Compare
…sks; also a few smaller changes
sterrettm2
force-pushed
the
generic_nets
branch
from
4ff3e1d to
70733ff
Compare
r-devulap
requested changes
Sep 7, 2023
r-devulap
approved these changes
Sep 7, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This changes quicksort and quickselect to use generic template based sorting networks, instead of the current implementations. It also changed the 32-bit qsort and qselect to use a 256 element sorting network instead of a 128 element sorting network. Here is a performance comparison for these changes: