mirror-github-bspeice/cuburn
1
0
Fork 0
mirror of https://github.com/stevenrobertson/cuburn.git synced 2025-11-03 18:00:55 -05:00
Code Issues Actions Packages Projects Releases Wiki Activity

Use new mad.cc instruction in MWC

Browse Source
This commit is contained in:
Steven Robertson
2011-12-08 11:49:31 -05:00
parent 094df0ae21
commit df8100d1f4
1 changed files with 11 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
cuburn/code/mwc.py
View File

@ -4,7 +4,7 @@ The multiply-with-carry random number generator.
import numpy as np
from cuburn.code.util import *
from util import *
class MWC(HunkOCode):
decls = """
@ -17,11 +17,12 @@ typedef struct {
defs = r"""
__device__ uint32_t mwc_next(mwc_st &st) {
asm("{\n\t.reg .u64 val;\n\t"
"cvt.u64.u32 val, %0;\n\t"
"mad.wide.u32 val, %1, %2, val;\n\t"
"mov.b64 {%1, %0}, val;\n\t}\n\t"
: "+r"(st.carry), "+r"(st.state) : "r"(st.mul));
asm("{\n\t"
".reg .u32 tmp;\n\t"
"mad.lo.cc.u32 tmp, %2, %1, %0;\n\t"
"madc.hi.u32 %0, %2, %1, 0;\n\t"
"mov.u32 %1, tmp;\n\t"
"}" : "+r"(st.carry), "+r"(st.state) : "r"(st.mul));
return st.state;
}
@ -114,3 +115,7 @@ __global__ void test_mwc(mwc_st *msts, uint64_t *sums, float nrounds) {
print sums
print dsums
if __name__ == "__main__":
import pycuda.autoinit
MWCTest.test_mwc()