chore: EXC: Pseudo-random embedders heap benchmarks results with THP changes #4795
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
The transparent huge pages improve host memory performance, resulting in a ~30% improvement across benchmarks without regressions. Some benchmarks (like `query_write_bwd_1gb_step_8_page_delta`) reduce their runtime by more than 700 ms (more than 60% of the runtime). Some benchmarks (like `query_read_bwd_1gb_step_4kb_new_allocation`) are improved by more than 90% from 226 ms down to 7 ms. ``` = 5d40b0f..4b15a2867f: Embedders Heap total time: 96426 ms (improved by -43022.6 ms / -30.9%) Top 10 by time: - query/wasm32_query_write_bwd_1gb_step_8_page_delta time improved by -729.9 ms (-63.1%) - query/wasm32_query_read_bwd_1gb_step_8_page_delta time improved by -729.0 ms (-63.8%) - query/wasm64_query_read_fwd_1gb_step_8_page_delta time improved by -728.8 ms (-63.0%) - query/wasm64_query_read_bwd_1gb_step_4kb_page_delta time improved by -725.2 ms (-70.9%) - query/wasm32_query_write_fwd_1gb_step_4kb_page_delta time improved by -724.0 ms (-70.8%) - query/wasm32_query_write_bwd_1gb_step_4kb_page_delta time improved by -721.7 ms (-70.7%) - query/wasm32_query_read_bwd_1gb_step_4kb_page_delta time improved by -718.4 ms (-70.5%) - query/wasm64_query_read_bwd_1gb_step_8_page_delta time improved by -717.2 ms (-62.3%) - query/wasm32_query_read_fwd_1gb_step_8_page_delta time improved by -716.7 ms (-63.4%) - query/wasm64_query_write_fwd_1gb_step_4kb_page_delta time improved by -709.9 ms (-70.4%) Top 10 by percentage: - update/wasm64_update_read_bwd_1gb_step_4kb_new_allocation time improved by -96.9% (-219.4 ms) - query/wasm64_query_read_bwd_1gb_step_4kb_new_allocation time improved by -96.9% (-220.5 ms) - query/wasm64_query_read_fwd_1gb_step_4kb_new_allocation time improved by -96.9% (-221.0 ms) - update/wasm64_update_read_fwd_1gb_step_4kb_new_allocation time improved by -96.8% (-219.1 ms) - update/wasm32_update_read_fwd_1gb_step_4kb_new_allocation time improved by -96.7% (-219.0 ms) - update/wasm32_update_read_bwd_1gb_step_4kb_new_allocation time improved by -96.7% (-219.5 ms) - query/wasm32_query_read_bwd_1gb_step_4kb_new_allocation time improved by -96.7% (-219.9 ms) - query/wasm32_query_read_fwd_1gb_step_4kb_new_allocation time improved by -96.7% (-222.5 ms) - update/wasm64_update_read_bwd_1gb_step_16kb_new_allocation time improved by -91.7% (-66.2 ms) - update/wasm32_update_read_bwd_1gb_step_16kb_new_allocation time improved by -91.6% (-66.6 ms) ```
For pseudo-random access, 9 benchmarks regressed, mostly `new_allocation`: ``` = 084e832..084e832: Embedders Heap total time: 1428 ms (regressed by 720.8 ms / 101.8%) Top 10 by time: + embedders:heap/query/wasm32_query_write_fwd_1gb_step_2mb_new_allocation time regressed by 131.1 ms (1666.3%) + embedders:heap/query/wasm64_query_write_fwd_1gb_step_2mb_new_allocation time regressed by 130.9 ms (1665.8%) + embedders:heap/query/wasm32_query_write_bwd_1gb_step_2mb_new_allocation time regressed by 130.6 ms (1725.5%) + embedders:heap/query/wasm64_query_write_bwd_1gb_step_2mb_new_allocation time regressed by 129.6 ms (1684.4%) + embedders:heap/query/wasm32_query_write_bwd_1gb_step_5mb_new_allocation time regressed by 54.5 ms (1738.0%) + embedders:heap/query/wasm32_query_write_fwd_1gb_step_5mb_new_allocation time regressed by 54.2 ms (1708.6%) + embedders:heap/query/wasm64_query_write_fwd_1gb_step_5mb_new_allocation time regressed by 53.9 ms (1712.7%) + embedders:heap/query/wasm64_query_write_bwd_1gb_step_5mb_new_allocation time regressed by 53.8 ms (1735.2%) + embedders:heap/update/wasm32_update_write_bwd_1gb_step_2mb_checkpoint time regressed by 1.5 ms (8.0%) - embedders:heap/query/wasm32_query_write_bwd_1gb_step_2mb_checkpoint time improved by -3.5 ms (-39.9%) - embedders:heap/query/wasm64_query_write_bwd_1gb_step_2mb_checkpoint time improved by -3.4 ms (-38.7%) - embedders:heap/query/wasm64_query_read_bwd_1gb_step_2mb_checkpoint time improved by -3.2 ms (-28.6%) - embedders:heap/query/wasm32_query_read_bwd_1gb_step_2mb_checkpoint time improved by -3.0 ms (-27.4%) - embedders:heap/query/wasm32_query_read_bwd_1gb_step_2mb_new_allocation time improved by -2.6 ms (-39.9%) - embedders:heap/query/wasm64_query_read_bwd_1gb_step_2mb_new_allocation time improved by -2.6 ms (-40.3%) - embedders:heap/update/wasm64_update_read_bwd_1gb_step_2mb_checkpoint time improved by -2.4 ms (-27.4%) - embedders:heap/update/wasm32_update_read_bwd_1gb_step_2mb_checkpoint time improved by -2.4 ms (-28.1%) - embedders:heap/update/wasm64_update_read_bwd_1gb_step_2mb_new_allocation time improved by -2.2 ms (-35.9%) - embedders:heap/update/wasm32_update_read_bwd_1gb_step_2mb_new_allocation time improved by -2.1 ms (-34.5%) ``` See `7dbe49b36a.log` for more details.
berestovskyy
force-pushed
the
andriy/exc-2012-random-benchmarks-thp
branch
from
076d8ca to
7966536
Compare
berestovskyy
changed the title
|
It's a draft PR is not supposed to be merged but just to run more benchmarks for the #4671 PR. |
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 PR runs pseudo-random heap benchmarks with 2 MiB, 5 MiB and 500 MiB steps on
thpbranch. There were 9 regressions detected, mostly querynew_allocationwrites. The top regression is from ~7 ms to ~130 ms:See details in
7dbe49b36a.logand rootEMBEDDERS_HEAP.min.