11:53 <braunr> hello

11:53 <braunr> i'm not sure where to ask this so trying here as people might know

11:54 <braunr> does anyone with experience with stm32f7 know how to best deal with ldr immediate instructions with regard to performance

11:55 <braunr> application note 4667 recmomends to separate code and data (so different busses can be used concurrently), but ldr immediate necessarily uses the same bus for instructions and the address being loaded

11:55 <braunr> or do you guys know a better place to ask this somewhere on irc ?

12:16 <PaulFertser> braunr: probably ##stm32

12:17 <braunr> oh there is such a channel

12:17 <braunr> thanks :)

13:05 <Haohmaru> i had the impression that the stm32 folks went <elsewhere> after the freenope incident

13:07 <karlp> there was already a telegram group, some people decided to just leave irc behind, but there's still lots of irc people

13:08 <Haohmaru> ahumz

13:09 <Haohmaru> real men won't quit IRC ;P~

13:42 <braunr> especially not for telegram :P

13:49 <borneoa_> PaulFertser: hi, would it be possible for you to make the rc1 release around the end of next week?

13:51 <borneoa_> PaulFertser: I will have a set of merge by Monday or Tuesday, then an email to ask for latest pending stuff, and at the end merge checkpatch series.

14:38 <PaulFertser> borneoa_: sure

14:43 <borneoa_> PaulFertser: thank you!

21:20 <FrancisWalton> Does anyone know why the maximum profile samples is hardcoded? For reference, line 4326 of https://openocd.org/doc/doxygen/html/target_8c_source.html

21:27 <Hawk777> I’ve locally patched OpenOCD to raise that number significantly in my own personal builds; I would personally lean towards suggesting that there be no limit, but I guess someone else would then argue that there should be a limit to prevent OpenOCD from exhausting a lot of RAM and crashing, and that just leads to bikeshedding about how high the limit should be.

21:28 <Hawk777> I can see points in favour of both sides.

21:29 <FrancisWalton> It looks like it has been set at 10000 for ~9 years, maybe an optional parameter would be good? If unspecified use the 10000 as it has been.

21:30 <Hawk777> The command already has a parameter to set the sample count, IIRC; this 10000 is just a cap on that parameter.

21:49 <FrancisWalton> My understanding of the code is that the function signature for the profiling function pointer in target_type takes a parameter for max_num_samples, but on line 4346 of target.c it is always called with MAX_PROFILE_SAMPLE_NUM. Also, I am not sure it is a safe assumption in the malloc that all program counters are at most 32 bit. Cortex-A can be 64

21:49 <FrancisWalton> bit, but it looks like that target doesn't implement profiling anyway.

21:51 <Hawk777> Oh, my mistake, it’s *seconds*, not *samples*, that’s taken on the command line.

21:59 <FrancisWalton> I'm not very familiar with sourceforge/mailing list style development, but I'll offer my changes once I get it working, if anyone wants them.

22:00 <Hawk777> OpenOCD has a git-backed code review system at <https://review.openocd.org/>.

22:00 <Hawk777> The HACKING file (IIRC) explains how to use it.

22:02 <FrancisWalton> Great, thanks for the help! Have a good weekend!