nRF51822 64 bitwise arithmetic

RE: nRF51822 64 bitwise arithmetic

Alev — Sat, 21 Mar 2015 20:43:27 GMT

Thank you for an answer. Sorry I cannot provide an example, because it's a huge application with a lot of code. When I try to use a simple test program only with 64 bit arithmetic (without my app) - all looks correct. But when I use it inside my app - trouble comes. I just want to know, that I am not missing something (in the datasheet or somewhere else) that state about NOT use a 64 bitwise. Now I know that it's completely possible. One question for you - how you declare a unsigned 64-bit wise? unsigned long long, uint64_t, QWORD (I know that it is near the same, but want to know). Thank you again.

RE: nRF51822 64 bitwise arithmetic

pabigot — Sun, 01 Mar 2015 21:19:54 GMT

Use uint64_t after including <inttypes.h>. If you're formatting it for output, use %llu or PRIu64 or similar macros. long long formatting is not available when using newlib nano.

Be aware that getting an accurate 64-bit timer value taking into account overflow can be tricky, and some details differ depending on whether you're reading from a HFCLK-synchronous TIMER or an LFCLK-synchronous RTC.

RE: nRF51822 64 bitwise arithmetic

Alev — Sat, 28 Feb 2015 12:40:25 GMT

[UPDATE] Comment removed to the bottom

RE: nRF51822 64 bitwise arithmetic

Clem Taylor — Sat, 28 Feb 2015 11:25:26 GMT

I use 64 bit fixed point on a nrf51822 quite heavily and haven't noticed any problems with gcc-arm-none-eabi-4_9-2014q4-20141203 from launchpad.net/gcc-arm-embedded

I'm mostly doing 32 * 32 multiplies with a 64 bit result and then >> the result right away. But looking at the compiler output it is using __aeabi_lmul() which is 'long long' * 'long long' -> 'long long'.

Can you provide a simple example that fails and I'll try it in my setup.