On Monday 30 August 2010 09:47:23 you wrote: > - Use a maximal-length LFSR-type PRNG (or Shrinking Generator PRNG) to > fill the memory with data. (...) > Should be pretty simple to do this in assembler on an LM32...
Already did that (and other tests), but using a linear congruencial generator instead of a LFSR, and always testing the memory linearly (ie. write the entire array from the first address to the last, then read back). The bug (if it's the memory subsystem) was too subtle to manifest itself during those tests. Is it better to use a LFSR instead of the LCG? I will also try to implement a hardware-based memory tester that stresses the memory subsystem with high bandwidths and low transaction-to-transaction times that cannot be acheived in software. S. _______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkym...@freenode Webchat: www.milkymist.org/irc.html Wiki: www.milkymist.org/wiki
