feat(crc): optimize crc32_iscsi with SVE2 fusion and improved PMULL path#410
Open
zwtao40 wants to merge 1 commit into
Open
feat(crc): optimize crc32_iscsi with SVE2 fusion and improved PMULL path#410zwtao40 wants to merge 1 commit into
zwtao40 wants to merge 1 commit into
Conversation
zwtao40
force-pushed
the
dev_crc_aarch64_optimize
branch
6 times, most recently
from
d1da083 to
ab23426
Compare
zwtao40
force-pushed
the
dev_crc_aarch64_optimize
branch
2 times, most recently
from
5e013ee to
00f5151
Compare
Optimize CRC32 iSCSI implementation for ARM platforms with two new accelerated code paths: 1. SVE2 Fusion Path (crc32_iscsi_fusion_p8_c10_asm): - Leverages SVE2 EOR3 instruction for parallel CRC computation - Processes 8 data lanes simultaneously with carry-less multiplication - Optimal for CPUs with SVE2 support (Kunpeng-950, etc.) 2. Optimized PMULL Path (crc32_iscsi_x6): - Replaces previous crc32_iscsi_3crc_fold implementation - Improved loop structure and register utilization - Better performance on CPUs with CRC32+PMULL but no SVE2 Dispatcher Logic: - Priority: SVE2 > PMULL(x6) > PMULL(base) > NEON - SVE2 path selected when HWCAP_CRC32 and HWCAP2_SVE2 are both present - PMULL path uses new x6 implementation for better throughput Performance Results: Kunpeng-920 (SVE, no SVE2): +------------+------------+------------+--------+ | Data Size | Before | After | Speedup| +------------+------------+------------+--------+ | 4KB | 33 GB/s | 37 GB/s | 1.12x | | 8KB | 33 GB/s | 38 GB/s | 1.15x | +------------+------------+------------+--------+ Kunpeng-950 (SVE2): +------------+------------+------------+--------+ | Data Size | Before | After | Speedup| +------------+------------+------------+--------+ | 4KB | 24 GB/s | 45 GB/s | 1.88x | | 8KB | 23 GB/s | 57 GB/s | 2.48x | +------------+------------+------------+--------+ Measured using crc32_iscsi_perf tool on Huawei Kunpeng platforms. Files modified: - crc/aarch64/Makefile.am - crc/aarch64/crc_aarch64_dispatcher.c - crc/aarch64/crc32_iscsi_p8c10.S (new, with macOS compatibility) - crc/aarch64/crc32_iscsi_p8c10_clmul_const.S (new, with macOS compatibility) - crc/aarch64/crc32_iscsi_x6.S (new, with macOS compatibility) Signed-off-by: Chenxuqiang <chenxuqiang3@hisilicon.com> Signed-off-by: Enigmo <guotaowei4@huawei.com>
zwtao40
force-pushed
the
dev_crc_aarch64_optimize
branch
from
00f5151 to
ea6a414
Compare
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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 patch optimizes the CRC32 ISCSI implementation for ARM aarch64 platforms with two new accelerated code paths.
For platforms with PMULL support (CRC32+PMULL but no SVE2):
For platforms with SVE2 support:
Performance Results:
Kunpeng-920 (SVE, no SVE2):
+------------+------------+------------+--------+
| Data Size | Before | After | Speedup|
+------------+------------+------------+--------+
| 4KB | 33 GB/s | 37 GB/s | 1.12x |
| 8KB | 33 GB/s | 38 GB/s | 1.15x |
+------------+------------+------------+--------+
Kunpeng-950 (SVE2):
+------------+------------+------------+--------+
| Data Size | Before | After | Speedup|
+------------+------------+------------+--------+
| 4KB | 24 GB/s | 45 GB/s | 1.88x |
| 8KB | 23 GB/s | 57 GB/s | 2.48x |
+------------+------------+------------+--------+
Measured using crc32_iscsi_perf tool on Huawei Kunpeng platforms.
Files modified: