Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20251113161518.57357-1-pincheng.plct@isrc.iscas.ac.cn>
Date: Fri, 14 Nov 2025 00:15:17 +0800
From: Pincheng Wang <pincheng.plct@...c.iscas.ac.cn>
To: musl@...ts.openwall.com
Cc: pincheng.plct@...c.iscas.ac.cn
Subject: [PATCH 0/1] riscv64: add optimized string functions

Hi all,

This patch series supersedes my previous RVV-optimized memset patch
series, now extending support to memmove and memcpy as well. It provides
optimized versions of these string functions for RISC-V with vector
extension, dispatching at runtime based on CPU capability (RVV support).

Implementation details:
- mem{set,cpy,move}.c are renamed internally to
  __mem{set,cpy,move}_scalar via macro, preserving the generic C
  implementation scalar fallback.
- mem{set,cpy,move}_vector.S provide the optimized
  __mem{set,cpy,move}_vector symbols.
- string_dispatch.c exports the public mem{set,cpy,move}() symbols,
  which dispatches via function pointer.
- __init_riscv_string_optimizations is called in __libc_start_main to
  initialize the function pointer based on AT_HWCAP.
- The vector implementation uses m8 register grouping for bulk fills.

Performance:
Function	Size	Improvement
memset		16B		0.06%
memset		64B		49.22%
memset		256B	127.81%
memset		1KB		58.12%
memset		4KB		47.95%
memset		64KB	2.56%

memcpy		16B		0.02%
memcpy		64B		35.94%
memcpy		256B	205.10%
memcpy		1KB		126.01%
memcpy		4KB		107.71%
memcpy		64KB	36.15%

memmove_bwd	16B		-0.67%
memmove_bwd	64B		47.03%
memmove_bwd	256B	207.32%
memmove_bwd	1KB		125.33%
memmove_bwd	4KB		106.72%
memmove_bwd	64KB	41.46%
Benchmarks are conducted on a Spacemit X60 CPU.

Functional behavior matches generic memset.

Thanks,
Pincheng Wang

Pincheng Wang (1):
  riscv64: add optimized memset, memcpy and memmove

 src/env/__libc_start_main.c          |  3 ++
 src/internal/libc.h                  |  3 ++
 src/string/riscv64/memcpy.c          |  4 ++
 src/string/riscv64/memcpy_vector.S   | 27 ++++++++++++++
 src/string/riscv64/memmove.c         |  4 ++
 src/string/riscv64/memmove_vector.S  | 51 ++++++++++++++++++++++++++
 src/string/riscv64/memset.c          |  4 ++
 src/string/riscv64/memset_dispatch.c | 36 ++++++++++++++++++
 src/string/riscv64/memset_vector.S   | 28 ++++++++++++++
 src/string/riscv64/string_dispatch.c | 55 ++++++++++++++++++++++++++++
 10 files changed, 215 insertions(+)
 create mode 100644 src/string/riscv64/memcpy.c
 create mode 100644 src/string/riscv64/memcpy_vector.S
 create mode 100644 src/string/riscv64/memmove.c
 create mode 100644 src/string/riscv64/memmove_vector.S
 create mode 100644 src/string/riscv64/memset.c
 create mode 100644 src/string/riscv64/memset_dispatch.c
 create mode 100644 src/string/riscv64/memset_vector.S
 create mode 100644 src/string/riscv64/string_dispatch.c

-- 
2.39.5

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.