Optimize merge algorithm for data sizes equal or greater then 4M items
#1933
Conversation
SergeyKopienko
requested review from
dmitriy-sobolev,
danhoeflinger,
mmichel11 and
MikeDvorskiy
SergeyKopienko
force-pushed
the
dev/skopienko/optimize_merge_to_main
branch
2 times, most recently
from
5a8ff9e
to
fedebda
Compare
…introduce new function __find_start_point_in Signed-off-by: Sergey Kopienko <[email protected]>
…introduce __parallel_merge_submitter_large for merge of biggest data sizes Signed-off-by: Sergey Kopienko <[email protected]>
SergeyKopienko
force-pushed
the
dev/skopienko/optimize_merge_to_main
branch
2 times, most recently
from
142ffa0
to
a6164fd
Compare
…using __parallel_merge_submitter_large for merge data equal or greater then 4M items Signed-off-by: Sergey Kopienko <[email protected]>
SergeyKopienko
force-pushed
the
dev/skopienko/optimize_merge_to_main
branch
from
a6164fd
to
d4721ca
Compare
Signed-off-by: Sergey Kopienko <[email protected]>
| auto __scratch_acc = __result_and_scratch.template __get_scratch_acc<sycl::access_mode::write>( | ||
| __cgh, __dpl_sycl::__no_init{}); | ||
|
|
||
| __cgh.parallel_for<_FindSplitPointsKernelOnMidDiagonal>( |
There was a problem hiding this comment.
@rarutyun Compile error is here:
https://github.com/oneapi-src/oneDPL/actions/runs/11722920053/job/32653481992?pr=1933
D:\a\oneDPL\oneDPL\include\oneapi\dpl\pstl\hetero\dpcpp\parallel_backend_sycl_merge.h(322,64): error: definition with same mangled name '...' as another definition
…fix compile error Signed-off-by: Sergey Kopienko <[email protected]>
|
|
||
| _PRINT_INFO_IN_DEBUG_MODE(__exec); | ||
|
|
||
| using _FindSplitPointsOnMidDiagonalKernel = |
There was a problem hiding this comment.
@rarutyun I have fixed the error here. Is it correct way?
I am using __kernel_name_generator here because I should have two Kernel names: one passed as template parameter pack and the second name I should create inside.
There was a problem hiding this comment.
I haven't yet looked at this in detail, but can't we just pass the _IdType to __kernel_name_generator directly, and use a single _find_split_points_kernel_on_mid_diagonal type?
In this PR we optimize
mergealgorithm for data sizes equal or greater then 4M items.The main idea - we doing two submits:
submitwe find split point in some"base"diagonal's subset.submitwe find split points in all other diagonal and runserial mergefor each diagonal (as before).But when we find split point on the current diagonal, we setup some indexes limits for
rng1and 'rng2'.For these limits we load split point's data from previous and next
"base"diagonals, calculated on the step (1).Applying this approach we have good perf profit for biggest data sizes with
floatandintdata types.As additional profit, we have sign performance boost for small and middle data sizes in the
merge_sortalgorithm.