Add writeVarintSve for aarch64 - retry #9603
Open
Conversation
This file contains hidden or 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
|
This pull request was exported from Phabricator. Differential Revision: D73513003 |
Nicoshev
added a commit
to Nicoshev/hhvm
that referenced
this pull request
Apr 23, 2025
Summary: Implemented an explicit SVE version of writeVarint. Throughput for 64-bit types shows a ~15% improvement. 16-bit and 32-bit cases seem to show a small improvement as well. All three functions are branch-free, their disassembly can be seen here: https://godbolt.org/z/jG5d8Wfe8 before: bench_write(u16_any_branch_free) 110.66% 2.00us 500.10K bench_write(u32_any_branch_free) 126.90% 2.00us 499.37K bench_write(u64_any_branch_free) 193.56% 2.33us 429.37K bench_write(u16_1b_branch_free) 99.562% 1.91us 522.97K bench_write(u16_2b_branch_free) 114.92% 2.00us 500.59K bench_write(u16_3b_branch_free) 111.66% 2.00us 500.99K bench_write(u32_1b_branch_free) 97.918% 1.93us 518.38K bench_write(u32_2b_branch_free) 113.76% 1.99us 502.29K bench_write(u32_3b_branch_free) 111.14% 1.99us 503.03K bench_write(u32_4b_branch_free) 115.72% 1.97us 507.52K bench_write(u32_5b_branch_free) 122.05% 2.00us 498.82K bench_write(u64_1b_branch_free) 99.089% 1.95us 511.71K bench_write(u64_2b_branch_free) 90.484% 2.53us 396.00K bench_write(u64_3b_branch_free) 93.335% 2.38us 419.63K bench_write(u64_4b_branch_free) 100.61% 2.24us 446.86K bench_write(u64_5b_branch_free) 123.18% 2.37us 421.24K bench_write(u64_6b_branch_free) 120.10% 2.33us 429.84K bench_write(u64_7b_branch_free) 144.69% 2.36us 423.79K bench_write(u64_8b_branch_free) 149.44% 2.25us 443.92K bench_write(u64_9b_branch_free) 174.37% 2.31us 433.60K bench_write(u64_10b_branch_free) 176.81% 2.28us 438.61K bench_write(exponential_1b_branch_free) 108.05% 1.91us 522.52K bench_write(exponential_2b_branch_free) 118.34% 1.98us 504.37K bench_write(exponential_3b_branch_free) 114.22% 1.99us 501.87K after: bench_write(u16_any_branch_free) 115.30% 1.97us 507.43K bench_write(u32_any_branch_free) 130.06% 1.97us 508.40K bench_write(u64_any_branch_free) 226.45% 1.96us 509.18K bench_write(u16_1b_branch_free) 101.37% 1.84us 543.01K bench_write(u16_2b_branch_free) 116.65% 1.97us 508.51K bench_write(u16_3b_branch_free) 111.17% 1.96us 510.12K bench_write(u32_1b_branch_free) 99.679% 1.93us 519.42K bench_write(u32_2b_branch_free) 115.98% 1.98us 506.04K bench_write(u32_3b_branch_free) 111.45% 1.98us 503.85K bench_write(u32_4b_branch_free) 116.04% 1.95us 513.18K bench_write(u32_5b_branch_free) 124.59% 1.97us 508.35K bench_write(u64_1b_branch_free) 99.669% 1.91us 522.26K bench_write(u64_2b_branch_free) 117.53% 1.93us 518.86K bench_write(u64_3b_branch_free) 111.95% 1.95us 511.77K bench_write(u64_4b_branch_free) 111.29% 1.98us 504.98K bench_write(u64_5b_branch_free) 124.53% 1.96us 510.52K bench_write(u64_6b_branch_free) 145.48% 1.90us 526.18K bench_write(u64_7b_branch_free) 172.51% 1.97us 506.83K bench_write(u64_8b_branch_free) 174.92% 1.95us 514.13K bench_write(u64_9b_branch_free) 202.27% 1.97us 508.08K bench_write(u64_10b_branch_free) 205.43% 1.96us 510.44K bench_write(exponential_1b_branch_free) 105.67% 1.91us 523.63K bench_write(exponential_2b_branch_free) 116.10% 1.95us 512.64K bench_write(exponential_3b_branch_free) 119.08% 1.95us 513.34K Reviewed By: embg Differential Revision: D73513003
Summary: Implemented an explicit SVE version of writeVarint. Throughput for 64-bit types shows a ~15% improvement. 16-bit and 32-bit cases seem to show a small improvement as well. All three functions are branch-free, their disassembly can be seen here: https://godbolt.org/z/jG5d8Wfe8 before: bench_write(u16_any_branch_free) 110.66% 2.00us 500.10K bench_write(u32_any_branch_free) 126.90% 2.00us 499.37K bench_write(u64_any_branch_free) 193.56% 2.33us 429.37K bench_write(u16_1b_branch_free) 99.562% 1.91us 522.97K bench_write(u16_2b_branch_free) 114.92% 2.00us 500.59K bench_write(u16_3b_branch_free) 111.66% 2.00us 500.99K bench_write(u32_1b_branch_free) 97.918% 1.93us 518.38K bench_write(u32_2b_branch_free) 113.76% 1.99us 502.29K bench_write(u32_3b_branch_free) 111.14% 1.99us 503.03K bench_write(u32_4b_branch_free) 115.72% 1.97us 507.52K bench_write(u32_5b_branch_free) 122.05% 2.00us 498.82K bench_write(u64_1b_branch_free) 99.089% 1.95us 511.71K bench_write(u64_2b_branch_free) 90.484% 2.53us 396.00K bench_write(u64_3b_branch_free) 93.335% 2.38us 419.63K bench_write(u64_4b_branch_free) 100.61% 2.24us 446.86K bench_write(u64_5b_branch_free) 123.18% 2.37us 421.24K bench_write(u64_6b_branch_free) 120.10% 2.33us 429.84K bench_write(u64_7b_branch_free) 144.69% 2.36us 423.79K bench_write(u64_8b_branch_free) 149.44% 2.25us 443.92K bench_write(u64_9b_branch_free) 174.37% 2.31us 433.60K bench_write(u64_10b_branch_free) 176.81% 2.28us 438.61K bench_write(exponential_1b_branch_free) 108.05% 1.91us 522.52K bench_write(exponential_2b_branch_free) 118.34% 1.98us 504.37K bench_write(exponential_3b_branch_free) 114.22% 1.99us 501.87K after: bench_write(u16_any_branch_free) 115.30% 1.97us 507.43K bench_write(u32_any_branch_free) 130.06% 1.97us 508.40K bench_write(u64_any_branch_free) 226.45% 1.96us 509.18K bench_write(u16_1b_branch_free) 101.37% 1.84us 543.01K bench_write(u16_2b_branch_free) 116.65% 1.97us 508.51K bench_write(u16_3b_branch_free) 111.17% 1.96us 510.12K bench_write(u32_1b_branch_free) 99.679% 1.93us 519.42K bench_write(u32_2b_branch_free) 115.98% 1.98us 506.04K bench_write(u32_3b_branch_free) 111.45% 1.98us 503.85K bench_write(u32_4b_branch_free) 116.04% 1.95us 513.18K bench_write(u32_5b_branch_free) 124.59% 1.97us 508.35K bench_write(u64_1b_branch_free) 99.669% 1.91us 522.26K bench_write(u64_2b_branch_free) 117.53% 1.93us 518.86K bench_write(u64_3b_branch_free) 111.95% 1.95us 511.77K bench_write(u64_4b_branch_free) 111.29% 1.98us 504.98K bench_write(u64_5b_branch_free) 124.53% 1.96us 510.52K bench_write(u64_6b_branch_free) 145.48% 1.90us 526.18K bench_write(u64_7b_branch_free) 172.51% 1.97us 506.83K bench_write(u64_8b_branch_free) 174.92% 1.95us 514.13K bench_write(u64_9b_branch_free) 202.27% 1.97us 508.08K bench_write(u64_10b_branch_free) 205.43% 1.96us 510.44K bench_write(exponential_1b_branch_free) 105.67% 1.91us 523.63K bench_write(exponential_2b_branch_free) 116.10% 1.95us 512.64K bench_write(exponential_3b_branch_free) 119.08% 1.95us 513.34K Reviewed By: embg Differential Revision: D73513003
Nicoshev
force-pushed
the
export-D73513003
branch
from
0ea0b91 to
1563c67
Compare
|
This pull request was exported from Phabricator. Differential Revision: D73513003 |
Nicoshev
force-pushed
the
export-D73513003
branch
from
1563c67 to
dcdad0e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D73513003 |
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.
Summary:
Implemented an explicit SVE version of writeVarint.
Throughput for 64-bit types shows a ~15% improvement.
16-bit and 32-bit cases seem to show a small improvement as well.
All three functions are branch-free, their disassembly can be seen here: https://godbolt.org/z/jG5d8Wfe8
before:
bench_write(u16_any_branch_free) 110.66% 2.00us 500.10K
bench_write(u32_any_branch_free) 126.90% 2.00us 499.37K
bench_write(u64_any_branch_free) 193.56% 2.33us 429.37K
bench_write(u16_1b_branch_free) 99.562% 1.91us 522.97K
bench_write(u16_2b_branch_free) 114.92% 2.00us 500.59K
bench_write(u16_3b_branch_free) 111.66% 2.00us 500.99K
bench_write(u32_1b_branch_free) 97.918% 1.93us 518.38K
bench_write(u32_2b_branch_free) 113.76% 1.99us 502.29K
bench_write(u32_3b_branch_free) 111.14% 1.99us 503.03K
bench_write(u32_4b_branch_free) 115.72% 1.97us 507.52K
bench_write(u32_5b_branch_free) 122.05% 2.00us 498.82K
bench_write(u64_1b_branch_free) 99.089% 1.95us 511.71K
bench_write(u64_2b_branch_free) 90.484% 2.53us 396.00K
bench_write(u64_3b_branch_free) 93.335% 2.38us 419.63K
bench_write(u64_4b_branch_free) 100.61% 2.24us 446.86K
bench_write(u64_5b_branch_free) 123.18% 2.37us 421.24K
bench_write(u64_6b_branch_free) 120.10% 2.33us 429.84K
bench_write(u64_7b_branch_free) 144.69% 2.36us 423.79K
bench_write(u64_8b_branch_free) 149.44% 2.25us 443.92K
bench_write(u64_9b_branch_free) 174.37% 2.31us 433.60K
bench_write(u64_10b_branch_free) 176.81% 2.28us 438.61K
bench_write(exponential_1b_branch_free) 108.05% 1.91us 522.52K
bench_write(exponential_2b_branch_free) 118.34% 1.98us 504.37K
bench_write(exponential_3b_branch_free) 114.22% 1.99us 501.87K
after:
bench_write(u16_any_branch_free) 115.30% 1.97us 507.43K
bench_write(u32_any_branch_free) 130.06% 1.97us 508.40K
bench_write(u64_any_branch_free) 226.45% 1.96us 509.18K
bench_write(u16_1b_branch_free) 101.37% 1.84us 543.01K
bench_write(u16_2b_branch_free) 116.65% 1.97us 508.51K
bench_write(u16_3b_branch_free) 111.17% 1.96us 510.12K
bench_write(u32_1b_branch_free) 99.679% 1.93us 519.42K
bench_write(u32_2b_branch_free) 115.98% 1.98us 506.04K
bench_write(u32_3b_branch_free) 111.45% 1.98us 503.85K
bench_write(u32_4b_branch_free) 116.04% 1.95us 513.18K
bench_write(u32_5b_branch_free) 124.59% 1.97us 508.35K
bench_write(u64_1b_branch_free) 99.669% 1.91us 522.26K
bench_write(u64_2b_branch_free) 117.53% 1.93us 518.86K
bench_write(u64_3b_branch_free) 111.95% 1.95us 511.77K
bench_write(u64_4b_branch_free) 111.29% 1.98us 504.98K
bench_write(u64_5b_branch_free) 124.53% 1.96us 510.52K
bench_write(u64_6b_branch_free) 145.48% 1.90us 526.18K
bench_write(u64_7b_branch_free) 172.51% 1.97us 506.83K
bench_write(u64_8b_branch_free) 174.92% 1.95us 514.13K
bench_write(u64_9b_branch_free) 202.27% 1.97us 508.08K
bench_write(u64_10b_branch_free) 205.43% 1.96us 510.44K
bench_write(exponential_1b_branch_free) 105.67% 1.91us 523.63K
bench_write(exponential_2b_branch_free) 116.10% 1.95us 512.64K
bench_write(exponential_3b_branch_free) 119.08% 1.95us 513.34K
Reviewed By: embg
Differential Revision: D73513003