[Mesa-dev] [Bug 101952] OpenCL enqueueReadBuffer returns trash
bugzilla-daemon at freedesktop.org
bugzilla-daemon at freedesktop.org
Thu Jul 27 18:53:18 UTC 2017
https://bugs.freedesktop.org/show_bug.cgi?id=101952
--- Comment #5 from Jan Vesely <jan.vesely at rutgers.edu> ---
(In reply to Janpieter Sollie from comment #4)
> yes, a simpler kernel works perfectly. Though I think it only makes the
> search more difficult:
yes, it might be miscompilation on clover/llvm part, or the program relies on
undefined behaviour that is just implemented differently in mesa/clover.
> the combination of these functions does not work (no 255 is visible):
> in external_function_call:
> while (left > 0) {
> count++;
> extern2(2048, count, digest);
> plen = left > SHA1_MAC_LEN ? SHA1_MAC_LEN : left;
> // result = get_global_id(0) | get_global_id(1) |
> get_global_id(2);
> result += os_memcmp(digest, pos, plen);
Is the 'result' variable initialized at the beginning of this function?
> pos += plen;
> left -= plen;
> }
> return result;
> in startTest:
> if(result == 2) {
> output[3] = 255;
> output[0] = get_global_id(0);
> output[1] = get_global_id(1);
> output[2] = get_global_id(2);
> }
> ----------------------------------------
> while
> result = get_global_id(0) | get_global_id(1) |
> get_global_id(2);
> // result += os_memcmp(digest, pos, plen);
>
> and
> if(result == 0) {
>
> does work! (255 int included) :s
>
> I suspect an overflow happens somewhere, but is it possible oclgrind does
> not detect it?
--
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20170727/36e772e2/attachment.html>
More information about the mesa-dev
mailing list