unzvfu/cuda-fixnum

Issues

• Benchmark against nVidia's XMP library

  #46 opened 5 years ago by unzvfu
  2

• Cases not found in tests

  #56 opened 4 years ago by duanbing
  0

• Use only unsigned values for RHS of compile-time div and mod

  #55 opened 5 years ago by unzvfu
  0

• Start wiki page for bug post-mortem analyses

  #54 opened 5 years ago by unzvfu
  0

• Automate memcheck/address sanitisation checks

  #53 opened 5 years ago by unzvfu
  0

• Implement systematic profiling

  #52 opened 5 years ago by unzvfu
  0

• Consider what can be learned from the CalcCrypto library

  #51 opened 5 years ago by unzvfu
  0

• Investigate use of other PTX instructions in arithmetic implementations

  #12 opened 5 years ago by unzvfu
  1

• Implement the NTT

  #40 opened 5 years ago by unzvfu
  0

• Consider advantages to Shoup's modular multiplication

  #50 opened 5 years ago by unzvfu
  0

• Use recurrency shallowing to remove carry dependencies

  #49 opened 5 years ago by unzvfu
  0

• Use AND for multiplying a carry flag

  #48 opened 5 years ago by unzvfu
  0

• Remove final subtraction from Monty multiplication

  #47 opened 5 years ago by unzvfu
  0

• Implement Wallace tree multiplier

  #45 opened 5 years ago by unzvfu
  0

• Support 'signed-digit' representations

  #44 opened 5 years ago by unzvfu
  0

• Load modulus into shared memory

  #43 opened 5 years ago by unzvfu
  0

• Consider improvements from "Outsourcing Modular Exponentiation" paper

  #42 opened 5 years ago by unzvfu
  0

• Overhaul benchmarking system

  #30 opened 5 years ago by unzvfu
  1

• Implement a register cache

  #41 opened 5 years ago by unzvfu
  0

• Review consequences of Cuda 7 Independent Thread Scheduling on warp-synchronicity assumptions

  #27 opened 5 years ago by unzvfu
  1

• Incorporate test vectors from Wycheproof

  #39 opened 5 years ago by unzvfu
  0

• Exploit concurrency of integer and floating point operations

  #38 opened 5 years ago by unzvfu
  0

• Finalise and incorporate the build improvements from jkruska

  #37 opened 5 years ago by unzvfu
  0

• Implement faster Newton-Raphson

  #36 opened 5 years ago by unzvfu
  0

• Investigate specialised mulmod for base*cuml in modexp

  #35 opened 5 years ago by unzvfu
  0

• Rewrite test suite to run via Python interface

  #34 opened 5 years ago by unzvfu
  0

• Re-jigger the API to ease writing HLL interfaces

  #33 opened 5 years ago by unzvfu
  0

• Implement Python interface

  #32 opened 5 years ago by unzvfu
  0

• Allow checking for overflow without penalising fast path

  #31 opened 5 years ago by unzvfu
  0

• Use warp votes to branch on argument size to select fastest algo

  #29 opened 5 years ago by unzvfu
  0

• Work out why 32-bit digits is slower than 64-bit digits with same fixnum size

  #28 opened 5 years ago by unzvfu
  0

• Benchmark against AccelerateHS "bignum" implementation

  #26 opened 5 years ago by unzvfu
  0

• Understand why CLNW sliding-window is faster than k-ary in the tests

  #25 opened 5 years ago by unzvfu
  0

• Implement VLNW and compare to the current CLNW sliding-window method

  #24 opened 5 years ago by unzvfu
  0

• Enforce exponent sharing in sliding-window modexp function

  #23 opened 5 years ago by unzvfu
  0

• Assess potential utility of CUDA warp matrix functions

  #22 opened 5 years ago by unzvfu
  0

• Consider reëxpressing fixnum layout in terms of CUDA's "coöperative groups"

  #21 opened 5 years ago by unzvfu
  0

• Investigate, document and incorporate "secure" arithmetic implementations

  #20 opened 5 years ago by unzvfu
  0

• Clean up warp_fixnum division code

  #19 opened 5 years ago by unzvfu
  0

• Ensure arguments from user functions are always read into registers

  #18 opened 5 years ago by unzvfu
  0

• Implement bignum arithmetic in terms of fixnum arithmetic

  #17 opened 5 years ago by unzvfu
  0

• Speed up test case generation using multiprocessing

  #16 opened 5 years ago by unzvfu
  0

• Store test cases in compressed format

  #15 opened 5 years ago by unzvfu
  0

• Investigate using Montgomery multiplication in the division algorithms

  #14 opened 5 years ago by unzvfu
  0

• Consider storing precomputed values in shared memory

  #13 opened 5 years ago by unzvfu
  0

• Specialise multi_modexp to case where all exponents are small

  #11 opened 5 years ago by unzvfu
  0

• Ensure that device-side assertions are disabled in non-debug builds

  #10 opened 5 years ago by unzvfu
  0

• Implement support for even moduli

  #9 opened 5 years ago by unzvfu
  0

• Correctly handle errors from device code

  #8 opened 5 years ago by unzvfu
  0

• Specialise modexp for word-sized base

  #7 opened 5 years ago by unzvfu
  0