random123

Counter-based Random Number Generators
git clone git://git.meso-star.com/random123.git
Log | Files | Refs | README | LICENSE

pi_cuda.cu (3990B)

      1 /*
      2 Copyright 2010-2011, D. E. Shaw Research.
      3 All rights reserved.
      4 
      5 Redistribution and use in source and binary forms, with or without
      6 modification, are permitted provided that the following conditions are
      7 met:
      8 
      9 * Redistributions of source code must retain the above copyright
     10   notice, this list of conditions, and the following disclaimer.
     11 
     12 * Redistributions in binary form must reproduce the above copyright
     13   notice, this list of conditions, and the following disclaimer in the
     14   documentation and/or other materials provided with the distribution.
     15 
     16 * Neither the name of D. E. Shaw Research nor the names of its
     17   contributors may be used to endorse or promote products derived from
     18   this software without specific prior written permission.
     19 
     20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     21 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     22 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
     23 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
     24 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
     25 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
     26 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
     27 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
     28 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
     29 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
     30 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     31 */
     32 // Simple CUDA device kernel and host main program to
     33 // compute pi via random darts at a square
     34 
     35 // functions for boilerplate CUDA init and done
     36 #include "../tests/util_cuda.h"
     37 
     38 #include <Random123/philox.h>
     39 
     40 int debug = 0;
     41 const char *progname;
     42 
     43 // CUDA Kernel:
     44 // generates n x,y points and returns hits[tid] with the count of number
     45 // of those points within the unit circle on each thread.
     46 __global__ void counthits(unsigned n, unsigned useed, uint2 *hitsp)
     47 {
     48     unsigned tid = blockDim.x * blockIdx.x + threadIdx.x;
     49     unsigned hits = 0, tries = 0;
     50     philox4x32_key_t k = {{tid, useed}};
     51     philox4x32_ctr_t c = {{}}; // start counter from 0
     52 
     53     while (tries < n) {
     54 	union {
     55 	    philox4x32_ctr_t c;
     56 	    int4 i;
     57 	}u;
     58 	c.v[0] = tries;
     59 	u.c = philox4x32(c, k);
     60 	int64_t x1 = u.i.x, y1 = u.i.y;
     61 	int64_t x2 = u.i.z, y2 = u.i.w;
     62 	if ((x1*x1 + y1*y1) < (1LL<<62)) {
     63 	    hits++;
     64 	}
     65 	tries++;
     66 	if ((x2*x2 + y2*y2) < (1LL<<62)) {
     67 	    hits++;
     68 	}
     69 	tries++;
     70     }
     71     hitsp[tid] = make_uint2(hits, tries);
     72 }
     73 
     74 #include "pi_check.h"
     75 #include "example_seeds.h"
     76 
     77 int
     78 main(int argc, char **argv)
     79 {
     80     unsigned seed = example_seed_u32(EXAMPLE_SEED9_U32); // example user-settable seed
     81     CUDAInfo *infop;
     82     uint2 *hits_host, *hits_dev;
     83     size_t hits_sz;
     84     unsigned nthreads;
     85     unsigned count = argc > 1 ? atoi(argv[1]) : 0;
     86     double d  = 0.;
     87 
     88     d = timer(&d);
     89     progname = argv[0];
     90     debug = argc > 2 ? atoi(argv[2]): 0;
     91 
     92     infop = cuda_init(argc > 3 ? argv[3] : NULL);
     93     nthreads =  infop->blocks_per_grid * infop->threads_per_block;
     94     if (count == 0)
     95 	count = NTRIES/nthreads;
     96 
     97     hits_sz = nthreads * sizeof(hits_host[0]);
     98     CHECKCALL(cudaMalloc(&hits_dev, hits_sz));
     99     CHECKNOTZERO((hits_host = (uint2 *)malloc(hits_sz)));
    100 
    101     printf("starting %u blocks with %u threads/block for %u points each with seed 0x%x\n",
    102 	   infop->blocks_per_grid, infop->threads_per_block, count, seed);
    103     fflush(stdout);
    104 
    105     counthits<<<infop->blocks_per_grid, infop->threads_per_block>>>(count, seed, hits_dev);
    106 
    107     CHECKCALL(cudaDeviceSynchronize());
    108     CHECKCALL(cudaMemcpy(hits_host, hits_dev, hits_sz, cudaMemcpyDeviceToHost));
    109 
    110     unsigned long long hits = 0, tries = 0;
    111     for (unsigned i = 0; i < nthreads; i++) {
    112 	if (debug)
    113 	    printf("%u %u %u\n", i, hits_host[i].x, hits_host[i].y);
    114 	hits += hits_host[i].x;
    115 	tries += hits_host[i].y;
    116     }
    117     CHECKCALL(cudaFree(hits_dev));
    118     free(hits_host);
    119     cuda_done(infop);
    120     return pi_check(hits, tries);
    121 }