devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Made the blocking computation aware of the l3 cache

Browse Source 
Also optimized the blocking parameters to take into account the number of threads used for a computation
This commit is contained in:
Benoit Steiner
2014-10-15 15:32:59 -07:00
parent dba55041ab
commit bfdd9f3ac9
13 changed files with 117 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
test/product_large.cpp
View File

@@ -39,15 +39,16 @@ void test_product_large()
// check the functions to setup blocking sizes compile and do not segfault
// FIXME check they do what they are supposed to do !!
std::ptrdiff_t l1 = internal::random<int>(10000,20000);
std::ptrdiff_t l2 = internal::random<int>(1000000,2000000);
setCpuCacheSizes(l1,l2);
std::ptrdiff_t l2 = internal::random<int>(100000,200000);
std::ptrdiff_t l3 = internal::random<int>(1000000,2000000);
setCpuCacheSizes(l1,l2,l3);
VERIFY(l1==l1CacheSize());
VERIFY(l2==l2CacheSize());
std::ptrdiff_t k1 = internal::random<int>(10,100)*16;
std::ptrdiff_t m1 = internal::random<int>(10,100)*16;
std::ptrdiff_t n1 = internal::random<int>(10,100)*16;
// only makes sure it compiles fine
internal::computeProductBlockingSizes<float,float>(k1,m1,n1);
internal::computeProductBlockingSizes<float,float>(k1,m1,n1,1);
}
{