Increase unit-test L1 cache size to ensure we are doing at least 2 peeled loop within product kernel.

This commit is contained in:
Gael Guennebaud 2015-02-27 22:55:12 +01:00
parent b10cd3afd2
commit 9aee1e300a

View File

@ -155,7 +155,7 @@ void computeProductBlockingSizes(Index& k, Index& m, Index& n, Index num_threads
// In unit tests we do not want to use extra large matrices,
// so we reduce the cache size to check the blocking strategy is not flawed
#ifdef EIGEN_DEBUG_SMALL_PRODUCT_BLOCKS
l1 = 4*1024;
l1 = 9*1024;
l2 = 32*1024;
l3 = 512*1024;
#endif