> 2009/5/8 Magnus Lundin <lundin at mlu.mine.nu > <https://lists.berlios.de/mailman/listinfo/openocd-development>>: > >/ The situation is very different on embedded host with a bitbang connection > />/ to the target versus PC+USB, or PC + simple ethernet JTAG. > />/ > />/ Here are the numbers from the Swedish jury: > />/ > />/ Measured with the debug millsecond time output. > />/ STM32 + JLink adapter, Linux on Athlon dual core 2.4 GHz. > />/ Every test done 2 times > />/ > />/ old version rev 1606: > />/ > />/ single step: 59 ms > />/ flash 64k : 24kB/s > />/ mdb 0 128 : 44ms > />/ > />/ > />/ trunk rev 1662: > />/ > />/ single step: 99 ms > />/ flash 64k : 21.5kB/s > />/ mdb 0 128 : 72ms > />/ > />/ So, the new version does not improve performance, for writing there is > />/ almost no difference, but for reading the core state and bytes there is > 70% > />/ perfoirmance hit. > / > OK. So it's not crippling while I fix it. Phew! :-) > > I have to give a bit of thought on how to best to profile this, i.e. to > find the precise location of the culprit. > > Any ideas? > > Then I believe it will be a simple matter for me to fix it and we'll > see if we won't > actually end up with *better* numbe
WOOW ! You make me very crazy this morning ;-( ! You committed large changes -> all users lost performance -> added bugs -> after what you ask "Any ideas?" ... and you are a manager of a consulting company ? Amontec Team can give you the IDEA : -- BEGIN OF IDEA -- Come back with all the in_handler + the _mask concepts from the JTAG API. -- END OF IDEA -- There are really much other works/improvements to do on OpenOCD that just simplifying the API Layer! Regards, Laurent for Amontec Team _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
