Commits

Branch selector

 master

User selector

cbuchner1

Datepicker

All time

Commit History

Commits on Apr 13, 2014

• Fixing the parenthesis of doom.

  cbuchner1

  committed
  746a773

Commits on Apr 11, 2014

• according to Alexey, "this condition should not be commented out".

  cbuchner1

  committed
  b082f65

• Another contribution by Alexey, fixing the high CPU utilization in scrypt and the multi-GPU support (-d flag)

  cbuchner1

  committed
  a9a2972

Commits on Apr 9, 2014

• here's a big fat code contribution from Alexey Panteleev (nVidia). Failover support, overlapping computation and memory transfers, better error handling, up to 32 GPUs supported would be the most i…

  Show description for 6a0c357
  cbuchner1

  committed
  6a0c357

Commits on Mar 20, 2014

• commit NVML wrapper for Windows. This one has an issue when running x86 binaries on x64 Windows. It won't find a 32 bit nvml.dll. Any workarounds?

  cbuchner1

  committed
  6bfc1e4

Commits on Mar 12, 2014

• call timeBeginPeriod() in cudaminer on WIN32 to boost thread scheduling resolution (mimicks Google Chrome trick?)

  cbuchner1

  committed
  c8188a9

Commits on Mar 11, 2014

• allow CudaMiner to also use GPU cbuchner1#8 and cbuchner1#9 and above, however the total number of GPUs in one instance is still limited to 8...

  cbuchner1

  committed
  c4630df

• fix stratum for BlakeCoin pools. Needs SHA256 instead of sha256d for merkle root hash apparently

  cbuchner1

  committed
  8498982

Commits on Mar 8, 2014

• add blake256 support to K and T kernel. Note that the warp limit for T has been relaxed to 32 when running blake.

  cbuchner1

  committed
  87160f7

• blake: change endianity of pdata[] array for CUDA (saves a lot of endianness conversions inn the kernel). Replace several constant memory arrays with in-code constants

  cbuchner1

  committed
  3d458a8

• removing the small footprint blake code (runs much slower)

  cbuchner1

  committed
  893dca5

• blake: Move the input[] array to registers, get rid of the buf[] character array

  cbuchner1

  committed
  60cf9be

• blake: get rid of separate state structures, and storing/restoring of said state

  cbuchner1

  committed
  3777e39

• manual inlining of blake functions into the main kernel. explicit loop unrolling, removal of unused code paths.

  cbuchner1

  committed
  4f7f100

• Visual C++ project file changes to include the sph blake code for CPU validation

  cbuchner1

  committed
  81d2c5a

Commits on Mar 6, 2014

• Difficulty fix for blake... pools were rejecting most shares.

  cbuchner1

  committed
  ec99908

• First working blake implementation in CUDA. Generic compute 1.0 kernel only - but already 50 MHash/s on GTX 750Ti. Optimization pending...

  cbuchner1

  committed
  d8f3355

Commits on Mar 2, 2014

• further optimization of Compute 3.5 keccak kernel (but very small improvements, if any...)

  cbuchner1

  committed
  35984c7

Commits on Feb 28, 2014

• speed-ups for Yacoin, Maxcoin, Helixcoin... roll-up to 2014-02-28 version. PLEASE USE TITAN KERNEL for keccak hashing on Compute 3.5 or better!

  cbuchner1

  committed
  f1562b0

Commits on Feb 27, 2014

• add a high N-factor scrypt-jane optimization that works best for higher lookup gap values (L=5,6,7...). GTX 780 and better gain 1 kHash/s or more.

  cbuchner1

  committed
  d2c6261

Commits on Feb 18, 2014

• don't currently consider the T kernel for -l auto with the keccak algorithm.

  cbuchner1

  committed
  32586d9

• committing some fixes (autotune for keccak, etc...) and updating version info to 2014-02-18.

  cbuchner1

  committed
  bd8ecee

Commits on Feb 17, 2014

• I heard through the grapevine that there will be a Compute 3.7 capability. Fixing the code to not crash on it. Also I fix the long time it takes after hitting the time-limit or CTRL-C to stop the s…

  Show description for 9e02a6a
  cbuchner1

  committed
  9e02a6a

Commits on Feb 14, 2014

• make cudaminer not crash when it encounters a Compute 5.0 device like Maxwell. Instead use the Titan kernel

  cbuchner1

  committed
  c254eb2

• make cudaminer *not* crash if it finds a Compute 5.0 part like Maxwell

  cbuchner1

  committed
  d1e7f2b

Commits on Feb 11, 2014

• BLAHBLAH detected new block - when I specified -q I meant quiet please! Also Alexey Panteleev (nVidia) is working in their Moscow office.

  cbuchner1

  committed
  9eb44d0

Commits on Feb 9, 2014

• final fix for duplicate share submission, I think. Tested and found good.

  cbuchner1

  committed
  9cb5f81

• okay, now how is this commit at preventing duplicate shares? I try to prevent the overflow of pdata[19] so the if (work.data[19] >= end_nonce) in the miner thread can detect that we've reached the …

  Show description for ef265aa
  cbuchner1

  committed
  ef265aa

• Posting a maxcoin fix in the scanhash function that seems to prevent duplicate share generation. Brings down the reject rate a lot.

  cbuchner1

  committed
  5cba5b3

Commits on Feb 8, 2014

• This commit should fix CPU utilization issues and non-validating hashes that some people were seeing. cudaMemcpyAsync was not excecuted in the correct stream, but rather in the default stream.

  cbuchner1

  committed
  1f5ee2e

• small fix for terminating condition check against max_nonce

  cbuchner1

  committed
  6a6cf20

• start using CUDA streams for maxcoin hashing. Might boost performance slightly.

  cbuchner1

  committed
  8a39043

• this commit allows to use larger launch configurations without overflowing the h_V/h_V_extra arrays in salsa_kernel.cu. I also commented out some MyStreamSynchronize() calls -> more speedin keccak …

  Show description for 3d90ed0
  cbuchner1

  committed
  3d90ed0

Commits on Feb 7, 2014

• commit fix for configure.ac file (add OpenSSL) and update the README.txt file to match 2014-02-07 release.

  cbuchner1

  committed
  d90235c

• I have received two proposed fixes for stratum issues. Could you test these out for me?

  cbuchner1

  committed
  9d03ad6

Pagination

Previous

Next