Round 3: This week Jonathan Halliday was able to access a machine which has both an NVM DIMM and CPU that implements the clflush and clwb instructions. He is currently preparing some benchmark figures for running transactions using a log stored in NVM. However, in order to get to that point we had to exercise and then fixed a few things that were unexercised on DRAM/CPU without clflush and clwb.
So, here is a new webrev which fixes the errors found in the previous drop: New webrev: http://cr.openjdk.java.net/~adinn/pmem/webrev.03 The changes are all fairly trivial: MappedByteBuffer.java - fixed a copy-paste error which meant force(from to) was passing an incorrect address range length vm_version_x86.hpp - corrected an error in the bitwise tests that detect presence of hw flush insns assembler_x86.cpp - ensured a register prefix is generated when the address is encoded in register r8 and upwards regards, Andrew Dinn ----------- Senior Principal Software Engineer Red Hat UK Ltd Registered in England and Wales under Company Registration No. 03798903 Directors: Michael Cunningham, Michael ("Mike") O'Neill, Eric Shander