Øyvind Harboe wrote: > Here is a first cut of modifying arm7tdmi.c to reduce # of > roundtrips for reading registers. > > I'm not going to commit any patches before the weekend > that addresses performance problems only. I'm also > experimenting a little bit with the form that these changes > should take. > > The attached patch reduces # of flushes from from 62 > to 47 for a step in telnet(using flush_count). > > > ------------------------------------------------------------------------ > > _______________________________________________ > Openocd-development mailing list > [email protected] > https://lists.berlios.de/mailman/listinfo/openocd-development > Sure, this will work (if there are no silly mistakes) but what is actually happening here ? - One specific postprocess function for every variant of in_handler function used - The upper level user must know which postprocess function to use in every specific place, or it wont work, you are rebuilding the in_handler functionality in a different place - But this is really only a property of the arm7_9 debug interface, and the bit structure of the Scanchain 1 and Scanchain 2 registers. - When we do cortex-a8 we will have a target with almost the same register structure as the ARM7 and 9, but using the arm_adi scanchains. so it is more difficult to reuse the armv4_5 stuff since we want to use the same register reading logic and writing but we must use differentportprocess functions.
Regards Magnus _______________________________________________ Openocd-development mailing list [email protected] https://lists.berlios.de/mailman/listinfo/openocd-development
