<karlp>
so I can get openocd pretty fast, but the jlink gdb server is still ~twice as fast on this. which is much faster than I expected.... https://paste.jvnv.net/view/eKshN
<PaulFertser>
Impressive
<karlp>
yeah, I really wasn't expecting that much difference.
<karlp>
I thought something was wrong first, turns out it was defaulting to 500khz... :)
<Haohmaru>
1.5MB/sec?
<Haohmaru>
fancy
renrelkha has quit [Quit: bye]
renrelkha has joined #openocd
slobodan has quit [Read error: Connection reset by peer]
bvernoux has joined #openocd
wingsorc has quit [Ping timeout: 258 seconds]
PaulFertser has quit [Ping timeout: 252 seconds]
merethan has joined #openocd
merethan__ has joined #openocd
merethan__ has quit [Quit: Leaving]
slobodan has joined #openocd
PaulFertser has joined #openocd
<karlp>
hrm, works for the tm4c, doesn't work for this newish kinetis though. not sure I can fix this right now...
<borneoa___>
karlp: load in SRAM or in flash?
<karlp>
those speed numbers? those are to flash.
<karlp>
kinetis I've tweaked some detection, but it's not flashing properly yet. I suspected watchdog, but I can't see anything different vs the KL28 that should be supported already there,
<karlp>
"Error: error executing cortex_m crc algorithm" doesn't feel like a happy path either though...
<karlp>
i'm kinda poking this while I wait for access to do real work :)
<borneoa___>
Have you tried increasing the work area size? I remember some discussion with Tarek about speed improvement by increasing it
<karlp>
worth a try. I'm somewhat more interested in makin the kinetis k32l2a work at all, but the speed would be nice too.
<karlp>
for 20Mhz clock, that gets it from 868KB/sec -> 998 KB/sec. jlink it still up at 1500.
<karlp>
(on tm4c129x)
<Haohmaru>
what's the debugger you used for those speeds with SWD and openocd?
<Haohmaru>
for some reason i noticed that when i flash samd2x i get a speed rating printed by openocd, but i don't seem to get one when i flash same5x
<PaulFertser>
I'd try loading to SRAM to pinpoint the bottleneck.
<karlp>
well, setting swd speed to 20Mhz is enough for now, I'd be happy with openocd just like that really, but the kinetis part not flashign at all is more of a problem, I'll probably just end up on jlink with it :|
<karlp>
Haohmaru: this is a jlink pro.
<Haohmaru>
ah, fancy
<Haohmaru>
yet the jlink also had a built-in gdb server too, right?
<Haohmaru>
so you could use it without openocd too
<Haohmaru>
or was that something else
<PaulFertser>
karlp: have you tried scoping the rst pin? I'd expect it to pulse if it was a watchdog.
<PaulFertser>
karlp: cortex_m crc algo should just run on any cortex-m part...
<karlp>
load_image to ram just gives me ~983KB too.
<karlp>
not sure how to test load to ram with jlink itself.
<PaulFertser>
looks like "restore zero.bin binary 0" should work for both
<PaulFertser>
Just some random data to RAM location might make sense as a test.
<karlp>
restore works, but no timing info.
<PaulFertser>
This looks like a reasonable way to prepare an ELF which you can then "load zero.elf 0": powerpc-objcopy -I binary -O elf32-powerpc -B PowerPC -S zero.bin zero.elf
<karlp>
Info : [klx.cpu] external reset detected
<karlp>
I ge tthat when I try and run "compare-sections" to try and test whether flashing worked properly or not.
<karlp>
and hten it's lost the gdb remote protocol sync
<karlp>
it just gets stuck in "invalid hex digit" and protocol error in gdb land :|
<karlp>
isnt there a sync command somewhere?
<PaulFertser>
mon gdb_sync
<PaulFertser>
sii
<PaulFertser>
si
<karlp>
yeah, it's too far gone for that apparently :) (gdb) mon gdb_sync
<karlp>
Truncated register 16 in remote 'g' packet
<karlp>
(gdb)
<PaulFertser>
External reset really hints at either watchdog or probably some interrupt was enabled and you're not running after "reset init"?
<karlp>
yeah, external reaset really poitns to watchdog. going to chase that a bit further first
<karlp>
jlink_swd_run_queue(): Executing 2 queued transactions is _super_ spammy though :|