diff options
| author | Wilco Dijkstra <wdijkstr@arm.com> | 2020-07-15 16:55:07 +0100 |
|---|---|---|
| committer | Wilco Dijkstra <wdijkstr@arm.com> | 2020-07-15 16:55:07 +0100 |
| commit | 4a733bf375238a6a595033b5785cea7f27d61307 (patch) | |
| tree | e649148bc3df2bf6ce1d256b5d9c8dc6d70b4195 /sysdeps/aarch64/multiarch/memcpy.c | |
| parent | 34f0d01d5e43c7dedd002ab47f6266dfb5b79c22 (diff) | |
| download | glibc-4a733bf375238a6a595033b5785cea7f27d61307.tar.gz glibc-4a733bf375238a6a595033b5785cea7f27d61307.tar.xz glibc-4a733bf375238a6a595033b5785cea7f27d61307.zip | |
AArch64: Add optimized Q-register memcpy
Add a new memcpy using 128-bit Q registers - this is faster on modern cores and reduces codesize. Similar to the generic memcpy, small cases include copies up to 32 bytes. 64-128 byte copies are split into two cases to improve performance of 64-96 byte copies. Large copies align the source rather than the destination. bench-memcpy-random is ~9% faster than memcpy_falkor on Neoverse N1, so make this memcpy the default on N1 (on Centriq it is 15% faster than memcpy_falkor). Passes GLIBC regression tests. Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Diffstat (limited to 'sysdeps/aarch64/multiarch/memcpy.c')
| -rw-r--r-- | sysdeps/aarch64/multiarch/memcpy.c | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/sysdeps/aarch64/multiarch/memcpy.c b/sysdeps/aarch64/multiarch/memcpy.c index 2fafefd5d2..e6f3ae1167 100644 --- a/sysdeps/aarch64/multiarch/memcpy.c +++ b/sysdeps/aarch64/multiarch/memcpy.c @@ -29,6 +29,7 @@ extern __typeof (__redirect_memcpy) __libc_memcpy; extern __typeof (__redirect_memcpy) __memcpy_generic attribute_hidden; +extern __typeof (__redirect_memcpy) __memcpy_simd attribute_hidden; extern __typeof (__redirect_memcpy) __memcpy_thunderx attribute_hidden; extern __typeof (__redirect_memcpy) __memcpy_thunderx2 attribute_hidden; extern __typeof (__redirect_memcpy) __memcpy_falkor attribute_hidden; @@ -36,11 +37,11 @@ extern __typeof (__redirect_memcpy) __memcpy_falkor attribute_hidden; libc_ifunc (__libc_memcpy, (IS_THUNDERX (midr) ? __memcpy_thunderx - : (IS_FALKOR (midr) || IS_PHECDA (midr) || IS_ARES (midr) || IS_KUNPENG920 (midr) + : (IS_FALKOR (midr) || IS_PHECDA (midr) || IS_KUNPENG920 (midr) ? __memcpy_falkor : (IS_THUNDERX2 (midr) || IS_THUNDERX2PA (midr) ? __memcpy_thunderx2 - : __memcpy_generic)))); + : (IS_ARES (midr) ? __memcpy_simd : __memcpy_generic))))); # undef memcpy strong_alias (__libc_memcpy, memcpy); |