srivathsanmurali

Interesting blogs

• Andrew Bosworth
• Benjamin Kovach on Haskell
• Bartek's coding blog

srivathsanmurali

GPU

• Parallel-forall

C++

• range.hpp

srivathsanmurali

Dynamic parallelism using CUDA and CMAKE

Requirements:

• compute_35 or higher cards. I am using Pascal (compute_61)

CMAKE MAGIC

• set(CUDA_SEPARABLE_COMPILATION TRUE)
• list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_61,code=sm_51;)
• target_link_libraries(yourCudaLib ${CUDA_cudadevrt_LIBRARY}

srivathsanmurali

My Awesome List

• Algorithms
• Art
• Career
• Computer Vision
• Elixir
• Functional Programming
• Git
• Linux
• Machine Learning

srivathsanmurali

with import <nixpkgs> {};
stdenv.mkDerivation rec {
  name = "openmp-${version}";
  version = "24.6.17";

  src = fetchgit  {
    url = "http://llvm.org/git/openmp";
    rev = "81a7c91728a2c763d6ecbb9374a0bbf413e43ac4";
#date: "2017-04-25T19:04:07+00:00";
    sha256  = "06641656wckpg4va1fn1dz3b6irr525vw699qallhf095h16gskm";

srivathsanmurali

#include <random>
float randFloat(float low, float high) {
    thread_local static std::random_device rd;
    thread_local static std::mt19937 rng(rd());
    thread_local std::uniform_real_distribution<float> urd;
    return urd(rng, decltype(urd)::param_type{low,high});
}

srivathsanmurali

#!/usr/bin/python

import sys
import json
from collections import OrderedDict
from operator import itemgetter

def getWordMap(filename):
    wordMap = dict()
    totalWords = 0