Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Tue, 27 Aug 2013 17:23:34 +0200
From: Dániel Bali <>
Subject: Daniel's weekly report #11


Last week I was working on finding out why patching fails sometimes.
I still don't understand why patching fails when the ISA used is simpler
than the one that was used to make the dummy.

Then I started worked on md4 format, because it's simple and testing with
an actual format makes more sense. I created a wrapper for it, which is a
modified version of "binary_gen.c" from the repository.

I loaded the ISA generated by the KernelAnalyzer into gcnasm to get the
microcode. I realized I can't check the output byte-by-byte because gcnasm
uses different formats sometimes (for example VOPC is always represented
with VOP3B).

I created a dummy using a completely unrelated OpenCL code. I changed the
VGPR and SGPR count to higher values in the binary.

Then to my absolute surprise, patching worked. So the issue seems to come
up only when patching with simpler ISA, which is really strange. At least
this worked though.

I have to discuss my priorities for this week with Lukas. I'll try to keep
reporting every 1-2 days.


Content of type "text/html" skipped

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.