buzzmarshall has quit [Remote host closed the connection]
dddddd has quit [Ping timeout: 256 seconds]
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
<bshah>
yuq825: I tested latest patchset just now (sorry didn't find time yesterday)
<bshah>
I get a error which I am not sure if is related to pm patchset or changes in that branch though
<bshah>
[ 74.195872] lima 1c40000.gpu: mmu page fault at 0x2e36180 from bus id 0 of type read on gpmmu
<bshah>
[ 74.721670] lima 1c40000.gpu: fail to save task state: error task list is full
<bshah>
[ 74.721688] lima 1c40000.gpu: gp task error int_state=0 status=a
<bshah>
[ 74.727773] lima 1c40000.gpu: mmu resume
<bshah>
this happens when some apps exit
<bshah>
or opening more then ~2 apps :s
Barada has joined #lima
<bshah>
yuq825: I will maybe try reverting runtime_pm patch
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
<yuq825>
bshah: did you have all 10 patches on branch topic/add-pm especially drm/lima: always set page directory when switch vm ?
<yuq825>
oh, as the error message "lima 1c40000.gpu: fail to save task state: error task list is full", you should miss some patches on topic/add-pm
<yuq825>
note you should use all 10 patches from topic/add-pm because some of them are v2
psydread has left #lima [#lima]
psydread has joined #lima
<bshah>
ah
<bshah>
I picked only pm patches, (mistake I suppose) :)
<bshah>
yuq825: ah, no, I do have 10 patches as-is
<yuq825>
but if you have patch "drm/lima: print process name and pid when task error", the error message should be "lima 1c40000.gpu: fail to save task state from xxx pid xxx: error task list is full"
<bshah>
in my branch: drm/lima: save task info dump when task fail , drm/lima: power down ip blocks when pmu exit, drm/lima: add pm resume/suspend ops seems to be marked as v2
<bshah>
huh, git failed me seems
<bshah>
okay back to [drm/lima: fix recovering from PLBU out of memory] and then starting over picking patches
<bshah>
building kernel
<bshah>
yuq825: ah nice, this works fine :)
<bshah>
yuq825: sorry for false alarm. I will give it bit more testing before giving "Tested-by" :)
<yuq825>
thanks
<bshah>
v2 is not on ML yet right?
dllud_ has joined #lima
dllud has quit [Ping timeout: 256 seconds]
dllud_ is now known as dllud
<yuq825>
yes, I'll send out latter if the test result is ok
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
Barada has quit [Quit: Barada]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
psydread has left #lima [#lima]
psydread has joined #lima
enunes has quit [Ping timeout: 256 seconds]
<rellla>
is it possible to run glmark2 headless?
monstr has joined #lima
dddddd has joined #lima
enunes has joined #lima
<yuq825>
I don't think so
<rellla>
yuq825: i just did add some clock_gettime in a simple test before a loop of 100 iterations of a draw with 3000 indices.
<rellla>
new code is not better atm but also not much worse. i'm not sure if this is a suitable way for a quick but probably inaccurate measuring...
<rellla>
i have to think about the memcpy and qsort ...
<yuq825>
have you changed the shadow buffer transfer map?