[Pkg-opencl-devel] Bug#768090: beignet: pow(n), erf(c), tgamma, rootn give wrong results

[Rebecca N. Palmer]

In 0.8 rootn is also affected, but the constant bug isn't present; fix 
attached. (Fix-pow-erf-tgamma.patch is the actual fix, the rest are for 
testing it.)

The remaining test failures are:
-cospi/sinpi/tanpi, powr/pown/pow, tgamma are less accurate than the 
OpenCL spec requires (at least the first group explicitly use the 
fast-but-inaccurate path, 2e-5 instead of 1e-8 typical relative error 
but 10-20x faster; one reason 0.9 is "faster" is that it has sin/cos/tan 
do the same by default).
-sub_bufffer_check sometimes crashes.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Fix-pow-erf-tgamma.patch
Type: text/x-patch
Size: 18663 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-opencl-devel/attachments/20141112/1b40cbc8/attachment-0004.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: add_constexpr_test.patch
Type: text/x-patch
Size: 2307 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-opencl-devel/attachments/20141112/1b40cbc8/attachment-0005.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fix_tests_enable_debug.patch
Type: text/x-patch
Size: 15713 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-opencl-devel/attachments/20141112/1b40cbc8/attachment-0006.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: accuracy_speed_test.py
Type: text/x-python
Size: 2635 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-opencl-devel/attachments/20141112/1b40cbc8/attachment-0001.py>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: beignet_0.8_test.log
Type: text/x-log
Size: 3193769 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-opencl-devel/attachments/20141112/1b40cbc8/attachment-0007.bin>