Could you please send me source code, I will integrate it into memory testing suite and share results from Harmattan? Btw, if you have access to sources you can check sp-mem-throughput package or glibc itself.
With best wishes, Leonid -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of ext Thiago Macieira Sent: 25 January, 2011 11:41 To: [email protected] Subject: Re: [MeeGo-dev] NEON memcpy? Em terça-feira, 25 de janeiro de 2011, às 09:04:04, [email protected] escreveu: > We have tested version in our libc and Juha could share it. Having looked at the disassembly of that function from my Harmattan sysroot, I am not sure it's the best. It does the same thing that glibc tries to do: move as much data as possible on each step. That's great, except for when you have small memory chunks you want moved. My performance timings on real applications show that the overhead of selecting the best code path is worse than the benefit of the faster memory moving. See: http://labs.qt.nokia.com/2010/08/25/improving-the-string-performance-with- more-simd-or-not/ -- Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org Senior Product Manager - Nokia, Qt Development Frameworks PGP/GPG: 0x6EF45358; fingerprint: E067 918B B660 DBD1 105C 966C 33F5 F005 6EF4 5358 _______________________________________________ MeeGo-dev mailing list [email protected] http://lists.meego.com/listinfo/meego-dev
