Skip to content
Snippets Groups Projects
  1. Apr 01, 2019
  2. Mar 25, 2019
  3. Mar 10, 2019
    • psychocrypt's avatar
      fix opencl and cuda cryptonight_r caching · 3ebf66a3
      psychocrypt authored
      The original implementation of the cache release and create always new
      kernel, this can lead into performance issues and my crashes.
      3ebf66a3
    • psychocrypt's avatar
      fix masari · be2144d6
      psychocrypt authored
      Since masari increased the block size the miner crashed each time it
      gets connected with a masari pool.
      
      This PR extent the possible size of a block to 128 byte and updated the
      kernel.
      be2144d6
  4. Mar 07, 2019
    • EDDragonWolf's avatar
      Support of CryptoNight v8 ReverseWaltz · 2d9087c7
      EDDragonWolf authored
      rebased version of #2261
      
      Added support of CryptoNight v8 Reverse Waltz (named cryptonight_v8_reversewaltz here) - equal to CryptoNight v8 but with 3/4 iterations of CryptoNight v8 and with reversed shuffle operation
      
      We plan to use CryptoNight v8 Reverse Waltz as new PoW algorithm for Graft (graft-project/GraftNetwork#234).
      2d9087c7
    • psychocrypt's avatar
      fix CUDA compile · cd36058e
      psychocrypt authored
      - fix linker issues with CUDA8
      - fix device selection
      cd36058e
  5. Mar 03, 2019
  6. Feb 13, 2019
  7. Feb 10, 2019
  8. Feb 07, 2019
    • psychocrypt's avatar
      remove cn_turtle as native POW · 1033dc28
      psychocrypt authored
      cryptonight_turtle is only cryptonight_v8 with a different scratchpad,
      iteration and mask value.
      We are using now the new machanism to describe such derived POWs.
      1033dc28
    • psychocrypt's avatar
      refactor POW definition · 3426e185
      psychocrypt authored
      A POW is now defined by a function `f` and three degrees of freedom `f(iteration, scratchpad, mask)`.
      `f` is the base algorithm like `cryptonight, cryptonight_gpu`
      An easy to pars snytax to write the full POW definition down is: `cryptonight_gpu:0x0000c000:0x00200000:0x001fffc0`
      
      This change make it very easy to integrate the new trend of variate the
      number of iteations or the scratchpad size without modifying the full
      code.
      3426e185
  9. Feb 04, 2019
  10. Feb 01, 2019
  11. Jan 30, 2019
  12. Jan 25, 2019
  13. Dec 29, 2018
    • psychocrypt's avatar
      improve POW algorithm selection · 758dbfb1
      psychocrypt authored
      - add helper method `GetAllAlgorithms()` to get all active POW
      algorithms
      - select max scratchpad memory size based on the dev pool and user
      algorithms
      758dbfb1
  14. Nov 29, 2018
  15. Nov 19, 2018
  16. Oct 24, 2018
    • psychocrypt's avatar
      NVIDIA: fix wrong number of threads · 954296ed
      psychocrypt authored
      In the cuda backend for monero we start always twice as much threads as needed.
      Those threads are than removed after the AES matrix is copied to the shared memory.
      Never the less it is the result of an copy past bug.
      
      - start correct number of threads for `monero`
      954296ed
  17. Oct 10, 2018
    • SChernykh's avatar
      NVIDIA: tweak `get_reciprocal` · b1504b36
      SChernykh authored
      - remove helper array to perform division
      - tweak `get_reciprocal`
      b1504b36
    • psychocrypt's avatar
      NVIDIA: rename config option `comp_mode` · bd4a4c94
      psychocrypt authored
      The name `comp_mode` for a memoy load pattern if a bad choosen name.
      Therefore I changed it to `mem_mode` which also gives use the possibility
      to add new mode later if needed.
      
      - rename `comp_mode` to `mem_mode`
      - fix documentation
      bd4a4c94
    • psychocrypt's avatar
      CUDA: fix invalid results · ed2168b4
      psychocrypt authored
      If `comp_mode` is false the results on a windows platform will be invalid.
      The reason for that is that `ulong4` is in windows 16byte and in linux 32byte.
      
      thx @xmrig for finding and solving the issue
      
      fix #1873
      ed2168b4
  18. Oct 08, 2018
  19. Oct 05, 2018
    • psychocrypt's avatar
      CUDA: tine cryptonight_v8 · 99a12cb6
      psychocrypt authored
      Read memory in bigger chunks per thread to increase the used memory bandwith.
      Use for Kepla and Fermi GPUs the old autosuggestion instead of the new settings for cryptonight_v8.
      99a12cb6
  20. Oct 04, 2018
  21. Oct 01, 2018
  22. Sep 30, 2018
    • psychocrypt's avatar
      cuda: implement cryptonight_v8 · 5db405c2
      psychocrypt authored
      - introduce a new schema where two threads work together on one hash
      - update autoadjustment
      - remove an mistake where shared memory was shrinked for gpus < sm_70
      5db405c2
  23. Sep 21, 2018
  24. Sep 19, 2018
    • psychocrypt's avatar
      NVIDIA: optimze v8 · fd27561b
      psychocrypt authored
      - fix that shared memory for fast div is always used even if an algorithm is not using it
      - optimize fast div algo
      - store `division_result` (64_bit) per thread instead of shuffle around and store it as 32bit
      fd27561b
    • psychocrypt's avatar
      NVIDIA: optimize div and sqrt · 659918f2
      psychocrypt authored
      - use optimzed div and sqrt
      - reduce memory footprint
      659918f2
Loading