On Tue, May 28, 2024 at 09:43:54AM +0200, Mattias Rönnblom wrote: > Provide build option to have functions in <rte_memcpy.h> delegate to > the standard compiler/libc memcpy(), instead of using the various > traditional, handcrafted, per-architecture rte_memcpy() > implementations. > > A new meson build option 'use_cc_memcpy' is added. The default is > true. It's not obvious what should be the default, but compiler > memcpy() is enabled by default in this RFC so any tests run with this > patch use the new approach. > > One purpose of this RFC is to make it easy to evaluate the costs and > benefits of a switch. > > Only Loongarch, ARM and x86 is implemented. Only x86 is tested. > > RFC v2: > * Fix bug where rte_memcpy.h was not installed on x86. > * Made attempt to make Loongarch compile. > > Signed-off-by: Mattias Rönnblom <mattias.ronnb...@ericsson.com> > --- > config/meson.build | 1 + > lib/eal/arm/include/rte_memcpy.h | 10 +++++ > lib/eal/include/generic/rte_memcpy.h | 62 +++++++++++++++++++++++--- > lib/eal/loongarch/include/rte_memcpy.h | 52 ++------------------- > lib/eal/x86/include/meson.build | 1 + > lib/eal/x86/include/rte_memcpy.h | 11 ++++- > meson_options.txt | 2 + > 7 files changed, 82 insertions(+), 57 deletions(-) >
I really support the long-term goal here of eliminating the need for us to maintain our own memcpy. This looks a good idea to see how things perform. If we do decide to take this patch, having the default be regular memcpy should help with static analysis and other tooling, which would be aware of memcpy but not rte_memcpy. /Bruce
