|
8f098e4a
|
2011-07-04T15:31:12
|
|
Put locking around the queued count to ensure we're never left without work.
|
|
|
122f0cf2
|
2011-07-04T13:59:24
|
|
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.
|
|
|
5007805d
|
2011-07-03T19:10:59
|
|
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.
|
|
|
e4f9ea68
|
2011-07-03T09:32:15
|
|
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.
|
|
9fe21064
|
2011-06-25T08:18:29
|
|
Fixed up using config.h instead of cpuminer-config.h.
|
|
|
5cbc33e7
|
2011-07-03T22:36:56
|
|
Change default retry to 5 seconds.
|
|
|
8999bea1
|
2011-07-03T00:39:02
|
|
People expect miners to run indefinitely instead of giving up by default so change to never giving up if json call fails.
|
|
|
edd0591e
|
2011-07-02T13:29:31
|
|
Make the number of queued work items configurable and default to 2.
|
|
|
131f60a5
|
2011-07-02T13:06:51
|
|
Move queueing of one request to separate function in preparation for variable length queues.
|
|
|
de3d07d1
|
2011-07-03T22:14:32
|
|
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.
|
|
|
9f726400
|
2011-07-02T21:42:14
|
|
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.
|
|
|
ffdffe77
|
2011-07-02T12:12:35
|
|
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.
|
|
|
86e40ed9
|
2011-07-02T09:44:29
|
|
Stale block control makes it possible to make 2 threads per gpu the default again.
|
|
|
594b38b8
|
2011-07-02T13:46:17
|
|
Fix redefinition of gnu source.
|
|
|
48caf248
|
2011-07-02T09:39:43
|
|
Check for stale block after failed submission as well.
|
|
|
09104ce3
|
2011-07-02T00:13:13
|
|
Flag the work back to just thread 0 used by all the threads to avoid lots of queued older work for each thread.
|
|
|
ad787e1c
|
2011-07-03T19:39:47
|
|
Move the flushing of old work into restart threads so it's done before any worker threads grab more work.
|
|
|
bed69215
|
2011-07-01T23:45:15
|
|
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.
|
|
|
e2fb3e84
|
2011-07-01T20:34:22
|
|
Queueing all kernel parameters dramatically reduces stale block rates.
|
|
|
7ae9afc4
|
2011-07-01T14:16:41
|
|
Profile points and warning clean ups.
|
|
|
b54a3425
|
2011-07-01T13:58:43
|
|
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.
|
|
|
88d9d631
|
2011-06-30T23:36:57
|
|
Use two separate curl instances for submit and get and use separate threads for each to prevent one blocking the other.
|
|
|
72baac08
|
2011-06-30T21:55:39
|
|
Clearly delineate the cpus from the gpus for their local data.
|
|
|
142576a9
|
2011-06-30T20:50:52
|
|
We already have gpu/cpu from id, so use that. Likely the current convoluted code is wrong and leading to segfaults!
|
|
|
18f8b0f9
|
2011-06-30T16:30:05
|
|
Submit work async is still unreliable and only used for cpu mining, so back it out for now.
|
|
|
d5d4d1da
|
2011-06-30T14:41:01
|
|
Don't want to free the work data out of the transient structs.
|
|
|
d100281d
|
2011-06-30T14:21:34
|
|
Make sure correct thread id is in work struct and correct cpu is set for per-cpu data.
|
|
|
998d8d45
|
2011-06-30T11:30:37
|
|
Postcalc hash is already its own thread so work can be submitted synchronously from that.
|
|
|
2b6e8416
|
2011-06-29T23:38:16
|
|
Use a buffer of up to 512 * 4 integers when retrieving work from the GPU.
This allows each local thread id to have one slot to put any positive results into, thus making overlapping results far less likely.
Thus races will be much rarer, allowing more threads.
It should also pick up blocks close to each other more reliably and hopefully decrease the number of rejects and opencl errors.
Do the search over the buffer entirely in a separate thread to allow the GPU to stay as busy as possible.
Detach threads from themselves to prevent unlucky even where dereferencing occurs by freeing the data that stores the thread info.
|
|
|
6af84770
|
2011-06-29T11:30:06
|
|
Add spaces to make output clearer.
|
|
|
a45c54aa
|
2011-06-27T11:31:05
|
|
Make postcalc_hash asynchronous as well.
|
|
|
378d18f8
|
2011-06-27T10:15:03
|
|
Submit all work asynchronously via a submit_work thread.
|
|
|
612c3a45
|
2011-06-27T09:32:12
|
|
Curl doesn't like multiple instances so go back to one instance.
|
|
|
f0dcd127
|
2011-06-27T09:17:13
|
|
Show which cpu mining thread when giving affinity message.
|
|
|
58f6bf42
|
2011-06-26T16:21:58
|
|
Prevent 32bit overflow of local_mhashes as well.
|
|
|
00de8225
|
2011-06-26T15:28:33
|
|
Upper limit should be -hashes.
|
|
|
c29a4322
|
2011-06-26T13:45:38
|
|
Only update the hashmeter once per second from gpu mining threads.
|
|
|
063adc64
|
2011-06-26T12:59:15
|
|
Implement runtime selectable numbers of GPU threads and rename CPU threads option.
|
|
|
b6ae1db8
|
2011-06-26T10:53:16
|
|
The submit_lock is not required nor helpful.
|
|
|
d1c0cccd
|
2011-06-26T09:09:07
|
|
Show correct GPU from thread number.
|
|
|
f1114992
|
2011-06-26T09:07:52
|
|
Implement a potentially variable number of threads per gpu, setting it to 2 for now.
|
|
|
08f56f5f
|
2011-06-26T08:55:53
|
|
Set default CPU threads to 0 if GPU mining.
|
|
|
295ef0f9
|
2011-06-25T21:47:16
|
|
Discard accumulated work when longpoll indicates a new block.
|
|
|
f44e8fac
|
2011-06-25T20:56:17
|
|
Curl appears to be not thread safe so only have one curl open at a time.
|
|
|
343ae851
|
2011-06-25T20:38:40
|
|
Intensity 5 is too high for a normal desktop causing unacceptable lag so change the default to 4.
|
|
|
88e2cf7b
|
2011-06-25T20:22:23
|
|
Initialise libcurl properly.
|
|
|
656b485d
|
2011-06-25T18:58:59
|
|
Make the worksize and vector width configurable.
|
|
|
ead1281b
|
2011-06-25T18:27:56
|
|
Cleanup of return codes.
|
|
|
f6486efb
|
2011-06-25T13:40:42
|
|
Make the getting of work asynchronous from the mining threads requests by always having one work item queued.
This prevents drops in hash rates when getting work from a pool that is slow to respond.
Use a local static struct work in get_work that is used to queue one extra work item.
|
|
|
0cef8f8d
|
2011-06-25T12:50:15
|
|
Default scan timeout of 5 seconds is way too short leading to abandoning blocks too early and being seen as an "inefficient" miner. Increase it to 60.
|
|
|
d2cb012f
|
2011-06-25T10:07:29
|
|
Detach the thread once created so we don't have to explicitly try and join it.
|
|
|
e1dd27c5
|
2011-06-29T11:19:43
|
|
Ensure that we don't overflow due to 32 bit limitations.
|
|
|
b38a02bd
|
2011-06-29T11:14:16
|
|
Make the log time hash rate a rolling exponential average so it doesn't fluctuate so dramatically.
|
|
|
08a78210
|
2011-06-29T10:12:00
|
|
Make the log show what the thread is: cpu or gpu and what number.
|
|
|
f490143a
|
2011-06-29T09:22:21
|
|
Add local thread count to info, store hw error count, and make share submission debug only.
|
|
|
e016d0c8
|
2011-06-28T23:41:57
|
|
Increase maximum intensity configurable to 14.
|
|
|
dfc52fd5
|
2011-06-28T21:46:09
|
|
Make sure we can have gpu and cpu threads running.
|
|
|
24a28e29
|
2011-06-28T21:28:50
|
|
Make it possible to run as a pure cpu miner by setting gpu threads to 0.
|
|
|
e1d01d06
|
2011-06-28T11:18:26
|
|
Minor fixes.
|
|
|
6374e0fa
|
2011-06-28T21:11:04
|
|
Import the phatk kernel. Enable it only for hardware with amd media ops for now since it crashes nvidia et. al.
Fallback to the poclbm kernel for the rest. Try harder to avoid stale blocks around longpoll detecting new blocks.
|
|
|
948b514c
|
2011-06-27T12:02:47
|
|
The buffer needs to be flushed before enqueueing the kernel again.
Further optimise the mining loop by removing the need_work bool.
|
|
|
b7a17753
|
2011-06-25T09:56:37
|
|
Make a separate thread for work submission that returns immediately so that miner threads aren't kept waiting when submitting results to slow pools.
|
|
|
e8f4eead
|
2011-06-24T16:24:53
|
|
Use total mhashes as a counter to prevent 32 bit overflows.
|
|
|
910e6943
|
2011-06-19T22:21:51
|
|
Increase baseline threads to 1<<22. Make total counter regularly update every 5 seconds.
Only write the blank buffer when it needs to be blanked.
|
|
|
6b77d850
|
2011-06-17T14:00:41
|
|
Fixes.
|
|
|
f7926088
|
2011-06-24T10:55:05
|
|
Limit intensity to 10. Anything larger overflows.
Simplify test for new work.
|
|
|
feb8cfc8
|
2011-06-24T09:51:54
|
|
applog fixes.
|
|
|
b19ee2f5
|
2011-06-24T09:39:33
|
|
Make sure a GPU doesn't work on a block longer than opt_scantime.
|
|
|
26546ad5
|
2011-06-24T09:17:09
|
|
Make the optimisations per-gpu card and update code to work properly with multiple cards.
|
|
|
852e78e7
|
2011-06-23T22:09:49
|
|
Fix mutex unlocking with only one thread and opt_log_interval.
|
|
|
70f73576
|
2011-06-23T21:58:46
|
|
Make the output display the 5 second and total average Mhash/s.
Make the log interval configurable.
|
|
|
debe7776
|
2011-06-23T21:23:46
|
|
Use cpu_from_thr_id when binding threads.
|
|
|
4cd5f47e
|
2011-06-23T21:09:22
|
|
Revert "Multiple compiler warning fixes."
This reverts commit a5cbfbde2610e9f60e14b41a4e0595bcb34c772a.
Broke.
|
|
|
88761e6c
|
2011-06-23T21:04:29
|
|
Multiple compiler warning fixes.
|
|
|
19eea906
|
2011-06-23T17:50:37
|
|
Implement code detecting max work size and optimal vector width.
Use this to patch the kernel to suit the idea values for the card.
Then use these values when invoking the kernel.
|
|
|
237a5067
|
2011-06-23T15:28:12
|
|
Skip trying to start thread of GPUs that don't successfully initcl().
|
|
|
14ca8883
|
2011-06-23T14:59:17
|
|
Update help.
|
|
|
c08be809
|
2011-06-23T14:56:27
|
|
Fix the setting of number of processors.
Add scan intensity variable.
|
|
|
2ab6180d
|
2011-06-23T10:34:40
|
|
Reset count once all threads are started to avoid slow rate being shown initially.
Update copyright notice and comments.
|
|
|
932ff72f
|
2011-06-22T23:35:23
|
|
The gpuminer thread uses very little cpu and needs to keep the gpu busy with as few delays as possible. Don't nice it.
|
|
|
f54d2cc0
|
2011-06-22T23:07:30
|
|
Make poclbm use 4 vectors and decrease worksize to keep pipelines fullish.
Make it possible to have 0 CPU threads and update docs.
Fix counter with no cpu threads.
|
|
|
66240481
|
2011-06-22T15:39:27
|
|
Fix deref.
|
|
|
fa2f6b19
|
2011-06-22T13:54:06
|
|
Unwind.
|
|
|
79fec01a
|
2011-06-22T12:27:57
|
|
Remove the input buffer and just pass args to the kernel as per plugin design.
|
|
|
f117675a
|
2011-06-22T10:15:23
|
|
Optimise work loop to make cl calls asynchronous where possible.
|
|
|
f05270b8
|
2011-06-22T01:19:19
|
|
Optimise loop and make debug debug only.
|
|
|
91e5cef3
|
2011-06-22T00:13:46
|
|
Actually get first BFI_INT patch working.
|
|
|
ce3382ca
|
2011-06-14T16:26:34
|
|
Don't run gpu thread idle prio.
|
|
|
dde70397
|
2011-06-14T10:32:54
|
|
Merge gpumining from oclmine. Unstable.
|
|
|
51817422
|
2011-06-14T14:09:10
|
|
Cope with older linux kernel headers that don't have the newer scheduling policies defined.
|
|
|
0a8ac14c
|
2011-06-12T14:26:26
|
|
Forgot the else.
|
|
|
4f8045c2
|
2011-06-12T11:40:15
|
|
Only increase solutions count when confirmed true.
|
|
|
ce750e42
|
2011-06-12T01:35:08
|
|
Add a solution counter to the output.
|
|
|
7062557e
|
2011-06-12T01:19:04
|
|
Implement global hash rate counter through mutex lock protected data.
Make the output easier to read.
Don't do hashmeter updates if no output is requested.
Remove redundant output when using a single thread.
|
|
4d6be0c1
|
2011-06-09T03:47:07
|
|
Fix number-of-threads init logic on Windows
|
|
|
8e0e2493
|
2011-06-08T22:30:10
|
|
only read processor count via sysconf on non-Windows platforms
|
|
|
262b98ca
|
2011-06-09T11:45:06
|
|
Linux + x86_64 optimisations.
Add likely() macro.
Optimise a few obvious code paths with likely/unlikely.
Change algo to sse2_amd64 by default.
Move priority change to worker threads only.
Detect number of CPUs and set default number of threads == CPUs.
Add scheduling policy change to worker threads to SCHED_IDLE first and fallback to SCHED_BATCH on linux.
Don't error when failing to set priority.
Add CPU affinity and bind worker threads to CPUs when number of threads is a multiple of number of CPUs.
Update NEWS with changes.
|
|
|
69529c38
|
2011-03-24T14:09:49
|
|
Support full URL, in X-Long-Polling header
|
|
|
46819af3
|
2011-03-21T20:50:59
|
|
--user/--pass fixes
Also, some newline fixes (applog callers do not need newlines in strings)
|
|
|
81352ca4
|
2011-03-21T04:27:02
|
|
Support --user and --pass, as alternative to --userpass
|