AArch64: Improve enabling of SVE for libmvec

When using a -mcpu option in CFLAGS, GCC can report errors when building libmvec.
Fix this by overriding both -mcpu and -march with a generic variant with SVE added.
Also use a tune for a modern SVE core.

Reviewed-by: Yury Khrustalev <yury.khrustalev@arm.com>
This commit is contained in:
Wilco Dijkstra
2025-05-29 16:58:49 +00:00
parent da196e6134
commit aa18367c11
+4 -1
View File
@@ -56,8 +56,11 @@ libmvec-support = $(addsuffix f_advsimd,$(float-advsimd-funcs)) \
v_powf_data
endif
sve-cflags = -march=armv8-a+sve
# Enable SVE for building libmvec. Since CFLAGS may contain a -mcpu or -march,
# add a generic -mcpu and -march with SVE enabled. Also use a tune for a modern
# SVE core.
sve-cflags = -mcpu=generic+sve -march=armv8-a+sve -mtune=neoverse-v2
ifeq ($(build-mathvec),yes)
bench-libmvec = $(addprefix float-advsimd-,$(float-advsimd-funcs)) \