 Luke Dashjr
|
5a00c2ac5d
Bugfix: minerloop_async: Always refer to real thread for select loop
|
13 years ago |
|
Luke Dashjr
|
f9ab7d16ba
Move new device API code to new deviceapi.c source file
|
13 years ago |
|
Luke Dashjr
|
1d06401a40
Make minerloop_async more async, using some callbacks to handle event completions
|
13 years ago |
|
Luke Dashjr
|
187abe9324
Split part of minerloop_async into do_get_results, and a bit other reorganization
|
13 years ago |
|
Luke Dashjr
|
7849bb1c11
Abstract select_timeout function to convert a realtime timeval to a timeout pointer for select()
|
13 years ago |
|
Luke Dashjr
|
32f07b8a3c
Split part of minerloop_async into do_process_results, and don't allow api->job_get_results to return hashes
|
13 years ago |
|
Luke Dashjr
|
e8090899bf
Split part of minerloop_async into do_job_prepare and do_job_start
|
13 years ago |
|
Luke Dashjr
|
da766f222b
Initialize thr->tv_poll to -1 (disabled)
|
13 years ago |
|
Luke Dashjr
|
cf13881ab3
Update the hashmeter one last time before disabling a device
|
13 years ago |
|
Luke Dashjr
|
297a30a9ec
minerloop_async: Break out of select for wakeup notifications
|
13 years ago |
|
Luke Dashjr
|
e4121d7c1f
Replace mining thread queues (which were only used for wakeup pings) with notifiers (which can be used with select and co)
|
13 years ago |
|
Luke Dashjr
|
399491f863
Bugfix: Unify all mining thread wakeup to mt_enable (simplifying and fixing code)
|
13 years ago |
|
Luke Dashjr
|
1aea99eae3
Merge commit 'dfbf189' into multiproc
|
13 years ago |
|
Luke Dashjr
|
dfbf1890b7
Bugfix: cpu: reinit_device hasn't worked since 93b284d, so just remove it entirely instead of letting it screw with thread 0
|
13 years ago |
|
Luke Dashjr
|
f2dfac373f
Bugfix: get_statline: Correct device summary status, only showing DEAD or OFF if it affects all processors
|
13 years ago |
|
Luke Dashjr
|
8be3fb102a
Working processor disable/enable with new async minerloop (currently gets stuck if all processors disabled)
|
13 years ago |
|
Luke Dashjr
|
f9456eace4
Bugfix: get_statline: Only care about the processor status if --show-processors is set
|
13 years ago |
|
Luke Dashjr
|
4a7a3929b0
Bugfix: watchdog: Use processor thr_info even if it isn't a real thread
|
13 years ago |
|
Luke Dashjr
|
709181de25
Only support thread-per-device or N-threads-per-processor; simplify work_restart check
|
13 years ago |
|
Luke Dashjr
|
f90616ed56
x6500: Remove mutex, since driver is single-threaded now
|
13 years ago |
|
Luke Dashjr
|
0b286b8f8b
Bugfix: Update utility every get_statline call, and include every processor involved
|
13 years ago |
|
Luke Dashjr
|
7cfa42ebbe
HACKING: New text file to document the internal workings of (currently) the device API
|
13 years ago |
|
Luke Dashjr
|
3875ec6931
Bugfix: mining_threads is now a total of thr_info objects, not necessarily actual running threads
|
13 years ago |
|
Luke Dashjr
|
eabdbe0897
x6500: Working (but incomplete) asynchronous/single-threaded driver
|
13 years ago |
|
Luke Dashjr
|
e07f3324f7
Incomplete (but workable) asynchronous minerloop
|
13 years ago |
|
Luke Dashjr
|
b604f8fd76
Core support for managing multiple processors from a single thread
|
13 years ago |
|
Luke Dashjr
|
2649cf7e3f
Allow device drivers to implement their own minerloop
|
13 years ago |
|
Luke Dashjr
|
cba42369d9
Move cgpu_info and thr_info initialization to main, and ensure all get initialized before starting any threads
|
13 years ago |
|
Luke Dashjr
|
0ba5203ad7
Refactor and simplify miner_thread (no major behavioural changes)
|
13 years ago |
|
Luke Dashjr
|
23a793ea6e
Move difficulties to end of share result message, so they can be made to line up nicely
|
13 years ago |
