Hi there, If only you had GCC 4.4.3 instead of 4.4.0 for platform (A)... I think it is that version that introduced __builtin_bswap32 and __builtin_bswap64. You can see the results of some benchmarks performed on another platform (including non-optimized code generated by GCC vs code that uses the above-mentioned intrinsics) in the link that I gladly respam here: http://adder.iworks.ro/Boost/RawMemory/#Benchmarks Regarding the importance of performance: what applications do you have in mind ? Loading a (portable) (binary) configuration file when the program is launched and saving it back when the program is closed might not require hand-written Assembler. A personal firewall application that enthusiastic power use for their Gigabit cable network and for their 300 Mbps wireless network might be a different thing. So is streaming OpenGL RGBA framebuffer data from the video game we are working on to the MPEG4 encoder which expects reversely formatted pixels in order to create a YouTube-uploadable demo. In such a case, the performance of the byte swapper made a significant difference. Thank you... (-: -- Yours truly, Adder On 9/6/11, Phil Endecott <spam_from_boost_dev@chezphil.org> wrote:
Dear All, [...] I've tested this on two systems: (A): Marvell ARMv5TE ("Feroceon") @ 1.2 GHz, g++ 4.4 (B): Freescale ARMv7 (i.MX53, Coretex A8) @ 1.0 GHz, g++ 4.6.1 [...] What do people see on other platforms? [...] How important do we consider performance to be for this library?