I can trigger 3 bugs in the OpenCL compiler for which I can supply code:
$ md5sum libamdocl64.so
d55c853a7b8c892576eb2355298ca59f libamdocl64.so
is 30% faster on a piece of code (when "-cl-opt-disable" is given) than the new compiler (with or without the "-cl-opt-disable"):
$ md5sum libamdocl64.so
7d2a234f375b693d9d2e71bf93f28edf libamdocl64.so
I have separate codes that triggers every one of these. Unfortunately, it's not public code, but I would be happy to share them with any AMD developer. I have been using the old compiler from 2011 for some while, but it crashes when compiling for 7970's, so I am forced to use 7d2a234f375 (13.4), but that miscompiles, so I am in a deadlock. I would really appreciate if the miscompilation issue would be fixed. I think it's very little to ask, to be honest.
Thanks for sharing your views.
1. I will probably try to check it myself. In case you know more specific situation for this bug, let us know.
2. Can you tell the catalyst version of the 2011 driver that is 30% faster than 13.4. also for what application is it that fast? Can you share some repro-case.
3. Again, I have seen many similar wrong compilation issues with 13.4. Can you check with 12.10 driver here? Also a repro-case would be helpful in fixing the invalid compilation. Probably you can send the testcase via private messages. I will add you as friend.
I have sent you a private message. Please check.
Do you still observe these issues? If so, could you provide us with the testcase(s)?