6107b5ddcc
MH/s output
9 years ago
adf32e82f8
...
9 years ago
a5be252960
nvidia maxregs
9 years ago
9276235da6
amd_bitalign
9 years ago
322534d26e
cleanup of chunks and msperbatch host code
9 years ago
83537733c8
work in progress
9 years ago
ff1928657a
Add more info on CL device info printing
10 years ago
a80dadeda1
Synchronise windows logging in ethminer.
10 years ago
a0b7159ce6
Avoid boost dependency for libethash-cl.
10 years ago
5904badec7
Use boost counterparts. Apparently std::random and boost::random are not
reliably interoperable.
10 years ago
09e5dbfa06
Use boost's guaranteed non-deterministic random number generator. If it
compiles, it's safe.
10 years ago
c429b03ddb
Disable chunking until further notice
10 years ago
3a61c2577e
CL: switch to chunks if clEnqueuWriteBuffer fails
Probably fixes #2559 even though chunking is not stable at the moment.
10 years ago
8c5247af5d
Style.
10 years ago
9b45b6b571
Use external prebuilt OpenCL ICD DLL.
10 years ago
fa2f3f2e14
I really need to get some sleep.
10 years ago
201d2fff0d
Ping buildbot.
10 years ago
53f8d86589
Ping buildbot.
10 years ago
63730439b8
Disable OpenCL_ICD on OSX.
10 years ago
8dd3ccb88a
Handle no devices and no platforms better
- Properly catch the exception thrown by getDevices() and if it's a no
devices found error just return an empty vector.
- Replace C macro for getPlatforms() with a proper function
10 years ago
abe869600b
Query Platforms checks for special not found value
This value is provided by the Khronos ICD
10 years ago
a5e8a6a760
Fix style issues
10 years ago
af6f5c3a13
simplify clminer current block code
10 years ago
1863def3a1
ethminer no longer depends on boost
Introducing an extra boolean argument to remove the boost::optional
dependency. Too bad we have to wait for C++17 to get an std::optional
10 years ago
35174e2798
More style: change const location
10 years ago
4eeb90e813
Style regarding public class constants
10 years ago
7450b8a0bc
Quick discovery of optimal global work size
Use a dichotomic algo to discover optimal m_globalWorkSize:
- m_wayWorkSizeAdjust is the direction steps are done (-1 or +1)
- m_stepWorkSizeAdjust is the steps of adjustment (added or substracted
to m_globalWorkSize)
- when a change of direction is needed, step is divided by 2
10 years ago
5ef3677ceb
Fix global work size logic adjustment
When batch duration is too long, m_globalWorkSize must decrease. The other
way if batch duration is too small, m_globalWorkSize should increase ...
10 years ago
020ef4b811
ethashcl doesnt depend on ethcore
At least not anymore. Moved global constants into ethash_cl_miner class
instead of being in Ethash class.
10 years ago
09d091b8d0
Default values of args are now class constants
Default values of some mining related arguments are no longer C constant
macros but are instead C++ Ethash class constants
10 years ago
8689bc8164
msvc compilation/warnings
10 years ago
db54ff3b4a
CL Argument for local work size
- Now the user can also set the local work size (workgroup size)
- In addition the global work size is specified in the command line only
as a multiplier of the local work size.
10 years ago
15fc63d6a2
New OpenCL arguments
- Adding an argument to specify OpenCL global work size.
- Adding an argument to specify milliseconds per global work
size (msPerBatch). If this is 0 then no adjustment of the global work
size happens.
10 years ago
7501191b40
GlobalWork size should never be less than local size
10 years ago
f51033dc75
CL Global and local work size adjustments
- Giving names to the variables that properly reflect the API
- Making sure that the limitations that are stated in
clEnqueueNDRangeKernel() documentation are adhered to
10 years ago
f03c12ca24
Adjust CL miner work batch size properly
We are now propery adjusting the batch size of the OpenCL miner
properly depending on the execution time of the last search.
10 years ago
057e374bce
Remove unneeded dummy use.
10 years ago
7a5b4db728
Minor renaming in BlockQueue, remove unnecessary invariant check, add
additional one.
10 years ago
81ca67dd4c
configureGPU() should respect given platformID
Related to #2314
10 years ago
9935fe75da
Fix a warning in ethash_cl_miner
10 years ago
a37d37a9c2
don't try to alter the batch size until we can make timing work.
10 years ago
d0dd66ac18
One more tryout.
10 years ago
a142e968d0
Greater timing scope.
10 years ago
84dd9dd35d
More diagnostics for batch size evolving.
10 years ago
a0e8512bed
Build fixes.
10 years ago
0ee5a5572e
Minor miner fix.
10 years ago
a8eb96755c
Miner targets a restart time of 100ms by default, reducing inter-block
"pauses".
10 years ago
8e5e2f4a9e
Catch OpenCL exceptions by const&
10 years ago
25d2fa1607
single chunk test: catch only cl::Error
10 years ago
3302539a11
OpenCL: Always try single chunk DAG upload
- Removed the `--force-single-chunk` option
- Always attempt to create a single chunk DAG buffer in the GPU. If that
fails then and only then switch to multiple chunks.
This change is motivated by the fact that many GPUs appear to be able to
actually allocate a lot more than what CL_DEVICE_MAX_MEM_ALLOC_SIZE
returns which proves that the results of querying the CL API on this
basically can't be trusted.
10 years ago
Genoil
Lefteris Karapetsas
Gav Wood
Paweł Bylica
bargst
arkady.paronyan@gmail.com