Rusty Russell
8d25df3c72
Cleanup --cpu-threads/-t logic
...
Currently it gets negated which means the default printed is wrong.
Use an explicit flag to tell if the user has overridden it; if they
haven't, and they turn off the GPUs, reset it to num_processors.
14 years ago
Rusty Russell
3e8181216c
Use ccan/opt for commandline parsing.
...
This cleans up option handling, by using ccan/opt rather than
handcoded getopt_long. We still have to open-code some things, such
as json config file handling.
The main change is that the --config option causes a file to be parsed
during commandline parsing, so you can override the results, and
provide multiple of them.
Other improvements are that 'help' and 'ndevs' are not valid arguments
in the config file; we use a separate argument table for such
commandline-only flags.
14 years ago
Con Kolivas
5969d65a04
First work for each thread is now queued in advance.
14 years ago
Con Kolivas
e9a397d4b9
More space to wipe old text.
14 years ago
Con Kolivas
e99069808a
Fix the extra work queued every longpoll by queueing it all in the beginning.
14 years ago
Con Kolivas
719deb3aac
Create a wakeup thread that keeps the hashmeter going in case the mining threads all stall.
14 years ago
Con Kolivas
ddae5aa9d0
Remove unused variable.
14 years ago
Con Kolivas
837e7b1a85
Remove unused cgpu rolling_local variable.
14 years ago
Con Kolivas
93800e5790
More output tidy.
14 years ago
Con Kolivas
56eee5f007
Retry should get_work() fail for whatever reason.
14 years ago
Con Kolivas
07f6de8b34
Change ifdef order since batch is always defined if idle is.
14 years ago
Con Kolivas
cf1e4627e4
Give error messages at all points of failure.
14 years ago
Con Kolivas
ed7e5bf9a3
There is still potential for an uneven number of queued and dropped work items with this unnecessarily complex check. Flush work guarantees a symmetrical number of queue and discard already.
14 years ago
Con Kolivas
69d8df643a
More cleaning of output with detecting new block.
14 years ago
Con Kolivas
8333a2ff40
Improve output appearance +/- verbose mode.
14 years ago
Con Kolivas
1fb09eaa0e
Implement -n/--ndevs option.
14 years ago
Con Kolivas
0ebdc39a19
Make debug mode imply verbose mode.
14 years ago
Con Kolivas
413e6b8645
Fix potential divide by zero.
14 years ago
Con Kolivas
0a80d4b04d
Display the output as a refreshing line and only push continuous log to stderr if desired.
14 years ago
Con Kolivas
a93b22c630
Make it possible to build without GPU mining by picking up HAVE_OPENCL from config.h.
14 years ago
Con Kolivas
4172fec24a
Cope with invalid parameter passing.
14 years ago
Con Kolivas
bc7cde7b36
Implement short option parsing.
14 years ago
Con Kolivas
13c57a5779
Properly initialise qd_lock mutex.
14 years ago
Con Kolivas
f90e73af51
Queue and pop precisely the same amount of work in flush_request to ensure all work is cleared and there isn't an uneven add/remove possible.
14 years ago
Con Kolivas
af6e7937ca
Signal handling in dns lookup code is responsible for libcurl crashes when multithreading.
...
Disable signal handling and use many curl handles instead, thus making work more asynchronous.
Theoretically a curl can wait forever on a dns lookup with this but it's extremely unlikely.
14 years ago
Con Kolivas
91f9c527ba
Rename code and application binary to cgminer version 1.0.9 preempting version 1.1.0, moving main code to main.c.
14 years ago
Con Kolivas
8f098e4ab4
Put locking around the queued count to ensure we're never left without work.
14 years ago
Con Kolivas
122f0cf260
With pre-emptive queueing of work there is no need to have extra work in the queue by default. Make it a list of extra queue items instead of a total and default to 0.
14 years ago
Con Kolivas
5cbc33e756
Change default retry to 5 seconds.
14 years ago
Con Kolivas
de3d07d13b
Make the next work item queue 2/3 of the way through the work on the current work item.
...
Load the first queued extra work in the main function to avoid having a once-off variable in get_work().
Load an extra set of work for each function in advance once a longpoll is detected since every thread will need to get new work.
Discard requests with a separate function to ensure the right number is always queued.
14 years ago
Con Kolivas
ad787e1c2a
Move the flushing of old work into restart threads so it's done before any worker threads grab more work.
14 years ago
Con Kolivas
5007805d1f
Update CPU reentrant scan code to work properly for extended periods up to scantime and adjust rate properly.
...
Simplify locking in hashmeter since it's called relatively infrequently anyway.
Set gpuminer hashmeter interval update to log interval / 5 as well.
14 years ago
Con Kolivas
e4f9ea68a9
CPU mining is spending too much time in the work thread and not receiving restart messages.
...
Update the code to return at log interval / 5 and only get work at the right time.
14 years ago
Con Kolivas
8999bea110
People expect miners to run indefinitely instead of giving up by default so change to never giving up if json call fails.
14 years ago
Con Kolivas
9f726400d4
Logic error in the queueing of work ended up generating more stale blocks.
...
There is a small chance that a longpoll is signalled right at the start which would lead to a deadlock so check for first work before restart.
14 years ago
Con Kolivas
594b38b8cb
Fix redefinition of gnu source.
14 years ago
Ycros
9fe2106467
Fixed up using config.h instead of cpuminer-config.h.
14 years ago
Con Kolivas
edd0591e19
Make the number of queued work items configurable and default to 2.
14 years ago
Con Kolivas
131f60a5ee
Move queueing of one request to separate function in preparation for variable length queues.
14 years ago
Con Kolivas
ffdffe77d3
Make sure the work gets attributed to the correct gpu.
...
Add an fflush to stderr to minimise garbled output when multiple threads write at once.
14 years ago
Con Kolivas
86e40ed9c8
Stale block control makes it possible to make 2 threads per gpu the default again.
14 years ago
Con Kolivas
48caf248a7
Check for stale block after failed submission as well.
14 years ago
Con Kolivas
09104ce3e8
Flag the work back to just thread 0 used by all the threads to avoid lots of queued older work for each thread.
14 years ago
Con Kolivas
bed692152f
Get rid of the requirement for a static struct that needs locking to cache work.
...
Make it possible to use the thread id for getting work again.
Flag the getwork() function when we have a new block to explicitly discard any cached work when a new block is detected.
Store the header of each new work and compare it to blocks we're about to submit to decide if they're stale due to a new block and don't try to submit them.
This should significantly decrease the number of rejected blocks.
14 years ago
Con Kolivas
e2fb3e84cb
Queueing all kernel parameters dramatically reduces stale block rates.
14 years ago
ckolivas
7ae9afc40f
Profile points and warning clean ups.
14 years ago
ckolivas
b54a342529
Change default number of threads back to 1. The 2nd just increases the time taken to complete a work item thus increasing stale blocks, despite increasing the rate slightly.
14 years ago
Con Kolivas
88d9d631e3
Use two separate curl instances for submit and get and use separate threads for each to prevent one blocking the other.
14 years ago
Con Kolivas
72baac0889
Clearly delineate the cpus from the gpus for their local data.
14 years ago
Con Kolivas
142576a961
We already have gpu/cpu from id, so use that. Likely the current convoluted code is wrong and leading to segfaults!
14 years ago