The Beignet test suite contains two groups of files that do not appear
consistent with the project's LGPL license:
1. Derived from the Len(n)a standard test image:
kernels/lenna128x128.bmp kernels/compiler_box_blur_float_ref.bmp
kernels/compiler_box_blur_ref.bmp
This is not generally
kernels/compiler_mandelbrot.cl kernels/compiler_mandelbrot_alternate.cl
kernels/compiler_mandelbrot_alternate_ref.bmp
kernels/compiler_mandelbrot_ref.bmp were included in my previous report
by mistake: they are not actually part of the Shadertoy set, so can be kept.
To keep the test suite
In current beignet (commit 4caba11ce88b3c1ec80ee119ceaf4bf778b96471):
-pow/pown ignore the sign of their first argument (e.g. pow(-2,3) gives
8 instead of -8)
-erf/erfc diverge (instead of converging to 1 or 0) for arguments above
about 2
-tgamma is actually lgamma, a related but very different
(Referring to
https://bugs.debian.org/cgi-bin/bugreport.cgi?msg=5;filename=Fix-pow-erf-tgamma.patch;att=3;bug=768090
)
And it triggers some other bugs related to constant expression
handling when run all unit test cases.
I didn't see that when I ran the tests in 0.9.3 on IvyBridge M GT2 (only
And it triggers some other bugs related to constant expression
handling when run all unit test cases.
I didn't see that when I ran the tests in 0.9.3 on IvyBridge M GT2
(only builtin_pow and builtin_tgamma failed, both due to the
absolute error checking you've already fixed) but your patch says
https://bugs.launchpad.net/ubuntu/+source/pyopencl/+bug/1354086
It may help to wait for one run of the kernel to finish before starting
the next: for me, repeated pyopencl.enqueue_copy(cq0,bCL.data,aCL.data)
leaks memory (sometimes ending in a crash, sometimes a whole-system
hang) but
Now working, thanks.
___
Beignet mailing list
Beignet@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/beignet
...due to a stray . at utests/builtin_pow.cpp:79:112.
Also, the strict-conformance spec for pow and tgamma is 16ulp, not 1ulp.
___
Beignet mailing list
Beignet@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/beignet
As previously noted, when cl_mem_allocate fails, its error handling then
calls cl_mem_delete on the incompletely-set-up buffer, which aborts at
assert(mem-ctx).
This patch appears to fix the problem, but be warned I don't know this
code well enough to know what else it might break.
The dmesg output from a failure is:
[ 427.521907] [drm:drm_ioctl] pid=776, dev=0xe200, auth=1,
I915_GEM_THROTTLE
[ 427.533899] [drm:drm_ioctl] pid=2340, dev=0xe200, auth=1,
I915_GEM_MADVISE
[ 427.533905] [drm:drm_ioctl] pid=2340, dev=0xe200, auth=1, I915_GEM_PWRITE
[ 427.533916]
Both [3.18 kernel] and the 3.16 kernel have a different intermittent failure I
have
yet to investigate:
compiler_fill_image_1d_array()[FAILED]
Error: dst[j*w + i] == 0
at file
/home/rnpalmer/Debian/builds/stackbuild/beignet/utests/compiler_fill_image_1d_array.cpp,
function
Two patches from Debian's packaging of 1.0.1:
Prefer LLVM 3.5 (as recently suggested here):
https://anonscm.debian.org/cgit/pkg-opencl/beignet.git/tree/debian/patches/llvm-search-order.patch
Don't use -D__$(USER)__ (it breaks on usernames containing special
characters):
runtime_marker_list()[FAILED]
Error: ((int*)buf_data[0])[i] == (int)value + 0x3
at file /tmp/buildd/beignet-1.0.2/utests/runtime_marker_list.cpp,
function runtime_marker_list, line 66
As this is only the second time I've seen this (the first was back in
0.9.3), it appears to be
That the file gets named intel-beignet-.icd rather than
intel-beignet.icd is a purely cosmetic issue: it should work either way,
but won't support multiarch, as the names of .icd files need to be
distinct for co-installability but are otherwise irrelevant.
If you want working multiarch,
If /dev/dri/cardX is inaccessible, return CL_DEVICE_NOT_FOUND,
don't assert-fail.
Signed-off-by: Rebecca Palmer rebecca_pal...@zoho.com
diff --git a/src/x11/dricommon.c b/src/x11/dricommon.c
index 03f542c..16f50e4 100644
--- a/src/x11/dricommon.c
+++ b/src/x11/dricommon.c
@@ -284,7 +284,6 @@
Compile the OpenCL standard library with the same version of clang
as will compile OpenCL user code, not plain clang (i.e. the
system default version, which may be different).
Signed-off-by: Rebecca Palmer rebecca_pal...@zoho.com
diff --git a/CMake/FindLLVM.cmake b/CMake/FindLLVM.cmake
index
As to the multi-arch support for beignet, I assume the requirement is from
running 32-bit binary on a 64-bit system. Right?
Yes.
Then the 32bit application will call to ocl-icd to search available icd files.
I never test it and don't know whether the ocl-icd library will choose the
32bit
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=779213
Summary: On hardware where the Intel GPU is disabled, beignet was found
to assert-fail on load, taking the application down with it before it
can do anything (including checking for hardware via clGetDeviceIDs).
This fixes this crash,
It is currently not possible to have 32- and 64-bit builds of beignet
installed on the same system, as the path in intel-beignet.icd
can only be one of the two installations. This fixes this by giving
this file a different name when beignet is installed in a multiarch
directory:
In beignet 1.0.1 on i5-3230M Debian 8 (libdrm 2.4.58, kernel
3.16.7-ckt2-1 with the whitelist OpenCL registers patch), the recently
added runtime_alloc_host_ptr_buffer (CL_MEM_ALLOC_HOST_PTR) test often
fails when run as part of the whole set, but not when run by itself.
The backtrace points
Should use tgamma instead of tgammaf when generating reference.
That I agree with: we don't want the error of libc tgammaf (~4ulp)
appearing to be our error.
And the ulp bound is = 16ulp
That I wouldn't: it's defined in terms of the infinitely precise value,
not the correctly-rounded-float
I will raise this issue to Linux i915 kernel team. And hope we can get a
clean fix in Linux kernel side finally.
I agree that's a good idea, but given how long it is likely to take, I'd
like to at least give a clear user-friendly error in the meantime.
-Is it likely that
When I run ./utest_run, I get 18 errors,
Those tests use the __local memory space, which does not work on Haswell
without a kernel patch:
https://01.org/zh/beignet/downloads/linux-kernel-patch-hsw-support
When I run any OpenCL programs, I need root privilege
You shouldn't; you may have
After revert the beignet's atomic in l3 patch [9e8874c ?], the
i915.enable_ppgtt=2 should work for HSW platform.
If it's that easy to really fix this, I agree we should do so; I'd taken
the absence of anything other than patch the kernel in the
documentation to mean that no other fix was
On 29/04/15 08:54, Zhigang Gong wrote:
The patch LGTM too, but the patch seems broken for me, could you check
and send a new version?
Don't know what's wrong with this one: what error do you get?
Compare with tgamma instead of tgammaf for better accuracy.
Include negative inputs, and handle
-61,[8] +62[,8] means the following block adds one more line:
No, that means the _preceding_ block(s) add one more line, which they do.
raise
all_vector = 1,2,3,4,8,16
There was a space-only line between these: has the email system lost it?
Make the build scripts work in both Python
On 29/04/15 10:11, Zhigang Gong wrote:
Actually, almost all the patches you sent out have similar problem and
I had manually applied them line by line before. I think it's better to
find out the root cause.
From http://www.freedesktop.org/wiki/Software/Beignet/ :
Based on our test result, LLVM 3.5 has best pass rate on all the test suites.
Compare to LLVM 3.5, LLVM 3.6 has slightly lower pass rate(caused by one front
end bug at clang 3.6) but has better performance (3% to 5% up).
When I run the
Given that really fixing shared local memory on Haswell is likely
to take some time, at least make it an explicit error that points
the user to the kernel patch, instead of returning success
without actually doing the computation.
As I can't find an error code to check for this problem (if there
Run a small __local-using kernel in clGetDeviceIDs; if this returns
the wrong result, return CL_DEVICE_NOT_FOUND.
---
just check kernel version is not
an ideal method for those unofficial kernels with back porting patches. Then
we have the
following open questions in my mind:
How do we
Sorry, both of those should have been
Signed-off-by: Rebecca Palmer rebecca_pal...@zoho.com
As usual, I can only test on Ivy Bridge, so someone should
probably check that they actually catch the
no-__local-on-Haswell bug.
On 16/05/15 18:48, Rebecca N. Palmer wrote:
Run a small __local-using
Since commit d47f6dd8f308323919d2acb0c1b9f562c084866c, beignet has
included backend/src/ir/structural_analysis.* based on gpuocelot.
gpuocelot's license was then BSD plus you must obey US export
restrictions (even if you are not in the US). This is non-free by
Debian's definition[1,2], and
Nothing works on Haswell-ULT is a known bug, fixed by
http://cgit.freedesktop.org/beignet/commit/?id=83f8739b6fc4893fac60145326052ccb5cf653dc
(Should that also be applied to the 1.0 branch?)
If you prefer not to recompile beignet, a workaround is:
sudo echo 0
The old tgamma=exp(lgamma) implementation had high rounding error on
large outputs, exceeding the 16ulp specification for approx. x8
(hence the test failure in strict conformance mode).
Replace this with an implementation based on glibc's
(Resend: the first copy appears to have gone missing)
0.01**20.5 is denormal; at least Ivy Bridge does not support
denormals and hence returns 0. As this is allowed by the
OpenCL standard, it shouldn't fail the test.
Signed-off-by: Rebecca Palmer rebecca_pal...@zoho.com
diff --git
http://cgit.freedesktop.org/beignet/commit/?h=Release_v1.0id=9e8874c7a16b029d5b857a11ff9b23102752b4db
was intended to catch the silently does nothing on large arrays bug,
and does so (with the message drm_intel_gem_bo_context_exec() failed:
No space left on device).
However, a user reports
Control: retitle -1 beignet: silently does nothing on large arrays
(previous discussion:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=781875 )
This isn't a Haswell-specific problem (and might not be ICD-specific
either: did you test that at these sizes, or only with the testsuite?),
On 05/04/15 00:08, David Couturier wrote:
This looks like the problem I experienced on my i7-4770. I fixed the issue by
applying this patch to the kernel:
https://01.org/zh/beignet/downloads/linux-kernel-patch-hsw-support
That's the long-standing no shared local memory on Haswell problem,
with a 4.0.0-rc6 kernel
...which includes that fix as standard.
Has anyone else tried beignet on Haswell-ULT, and if so, did it work?
___
Beignet mailing list
Beignet@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/beignet
-Original Message-
From: Beignet [mailto:beignet-boun...@lists.freedesktop.org] On Behalf Of
Rebecca N. Palmer
Sent: Saturday, November 07, 2015 6:06 AM
To: beignet@lists.freedesktop.org
Subject: Re: [Beignet] Unrecoverable system lockup when allocating too much
memory
(The example in
https
On 03/11/15 01:50, Pan, Xiuli wrote:
LGTM, but the commit line is too long,
When it was written, the longest line was 66, but something appears to
have joined it back together; does
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/email-clients.txt#n208
it.
The bug was exposed by a subcase in Khronos OpenCL conformance test, we will
try to isolate the bug to report to llvm.
Thanks
Zou Nanhai
-Original Message-
From: Beignet [mailto:beignet-boun...@lists.freedesktop.org] On Behalf Of
Rebecca N. Palmer
Sent: Thursday, October 01, 2015 1:32 AM
Reading p+1 when p==end is an out of bounds read.
Signed-off-by: Rebecca Palmer
---
(Found by valgrind while investigating #90472; probably not the
actual cause of that crash, but still a bug.)
diff --git a/backend/src/llvm/llvm_printf_parser.cpp
Specifically (and assuming it is the same bug on Ivy Bridge and
Haswell), fixed in git master by
http://cgit.freedesktop.org/beignet/commit/?id=f48b4f6766fcaa193652fcbe6ea0bb29f92e45aa
, still present in 1.1.x.
___
Beignet mailing list
This has been the case for some time
(https://bugs.launchpad.net/ubuntu/+source/pyopencl/+bug/1354086); given
that I've also occasionally seen it in graphics-only use, the root cause
may be further down the stack.
___
Beignet mailing list
When p == end (the null terminator byte), don't try to read p + 1:
as this is outside the string, it might be a '%' from a different
object (causing __parse_printf_state(end + 2, end, ...) to be called,
which will fail), or an invalid address.
Signed-off-by: Rebecca Palmer
Would this be better if you turn off the overcommit via proc fs?
Only if you also disable any swap space ( sudo swapoff -a && sudo sh -c
"echo -n 2 > /proc/sys/vm/overcommit_memory" #warning, this may itself
crash your desktop); if I disable overcommit but leave swap on, I get a
hang with
When p==end (the null terminator byte), don't try to read p+1
(outside the string, so might be an invalid address or a '%' from
a different object).
Signed-off-by: Rebecca Palmer
diff --git a/backend/src/llvm/llvm_printf_parser.cpp
(The example in
https://bugs.launchpad.net/ubuntu/+source/pyopencl/+bug/1354086 no
longer hangs, so the "rapidly allocating and freeing pyopencl objects
doesn't actually free the memory" aspect has evidently been fixed, but
keeping too many objects for the available memory still does hang.
On
Debian are planning to switch their default LLVM/Clang to 3.6 soon. Is
it still the case that
The recommended LLVM/CLANG version is 3.5 and/or 3.6. Based on our test result,
LLVM 3.5 has best pass rate on all the test suites. Compare to LLVM 3.5, LLVM
3.6 has slightly lower pass rate(caused
I also have an i5-3230M, and get the same error in my build chroots
(Debian sid = libdrm 2.4.64) but not my base system (Debian jessie =
libdrm 2.4.58); I have not attempted to determine whether the problem is
the version or being in a chroot.
___
The documentation says "an unsupported feature which is to store/load
pointers to/from memory[...]We plan to fix it in next major release
1.1.0" (docs/Beignet.mdwn:216-219)
Given that 1.1.0 has now been released, and does appear to fix this
(creating an array of pointers works in 1.1.2 but
As noted at llvm_gen_backend:94, we currently lack a mechanism for
reporting failures in backend (beignet-managed) compiler passes to the
build log, and instead print the error to stderr and assert-fail.
This patch creates such a mechanism, and uses it for "function not found".
Please note that
CMake Error: The following variables are used in this project, but they
are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the
CMake files:
CLANG_LIB
[...]
Please let me know what I'm missing.
libclang-3.7-dev ?
As beignet now works with LLVM/Clang 3.7, accept this version
when searching for llvm-config.
Signed-off-by: Rebecca Palmer
--- a/CMake/FindLLVM.cmake
+++ b/CMake/FindLLVM.cmake
@@ -8,12 +8,12 @@
# LLVM_FOUND - True if llvm found.
if (LLVM_INSTALL_DIR)
otype cannot use the
spir_function calling convention"), but I don't know whether this
applies to all missing-function errors.
On 06/06/16 23:37, Rebecca N. Palmer wrote:
As noted at llvm_gen_backend:94, we currently lack a mechanism for
reporting failures in backend (beignet-managed) compi
cl_get_gt_device(): error, unknown device: 5916
[...] Is there a support missing for
Kabylake/.?
Yes - that error means "Beignet does not recognize this device, so will
not try to use it" (from
https://cgit.freedesktop.org/beignet/tree/src/cl_device_id.c#n597 ; the
supported devices list is
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
> -The release notes say cl_khr_gl_sharing is now supported, but
> docs/Beignet.mdwn still says it isn't.
It exists but is off by default, probably because it isn't finished:
clEnqueueAcquireGLObjects / clEnqueueRele
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
(Warning: has not been tested)
diff --git a/src/cl_khr_icd.c b/src/cl_khr_icd.c
index 7b3600c..e4daf79 100644
--- a/src/cl_khr_icd.c
+++ b/src/cl_khr_icd.c
@@ -18,10 +18,14 @@
#include "cl_platform_id.h"
#includ
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--- a/backend/src/llvm/llvm_gen_backend.cpp
+++ b/backend/src/llvm/llvm_gen_backend.cpp
@@ -308,7 +308,7 @@ namespace gbe
if(StrTy)
return getTypeByteSize(unit,StrTy);
}
- GBE_ASSERTM(false, "
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--
(Possibly should be higher in the priority order, since it's required for
OpenCL 2.0?)
--- a/CMake/FindLLVM.cmake
+++ b/CMake/FindLLVM.cmake
@@ -8,12 +8,12 @@
# LLVM_FOUND - True if llvm found.
if (LLVM_INSTA
On 21/01/17 15:40, Bruno Pagani wrote:
Le 21/01/2017 à 16:20, Rebecca N. Palmer a écrit :
is there any downside in compiling with
OpenCL 2.0 support,
Yes - on older (Ivybridge/Haswell - no emitUntypedReadA64Instruction)
hardware, a 2.0-enabled beignet won't work, at all.
That’s what I’ve
is there any downside in compiling with
OpenCL 2.0 support,
Yes - on older (Ivybridge/Haswell - no emitUntypedReadA64Instruction)
hardware, a 2.0-enabled beignet won't work, at all.
I see two possible approaches:
-Build two beignet packages, with and without 2.0 enabled, and make the
I have a first attempt at "enable/disable 2.0 at run time" written, but
haven't yet tested it.
On 21/01/17 16:08, Bruno Pagani wrote:
Le 21/01/2017 à 16:55, Rebecca N. Palmer a écrit :
On 21/01/17 15:40, Bruno Pagani wrote:
Le 21/01/2017 à 16:20, Rebecca N. Palmer a écrit :
is
test that) would be appreciated.
(Upstream may not want to take this as-is, as it doesn't scale very
well to more than one 2.0-supporting hardware generation: it's
intended as a quick relatively-low-risk fix for packagers.)
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--- beignet-1.3
> regarding whether OpenCL 2.0
> is compiled or not I would rather think of something like that:
>
> – if llvm39 and libdrm2466 are found, compile with OpenCL 2.0 by default.
> – if they’re not, compile without it. Issue an error message if
> -DENABLE_OPENCL_20=1 was passed.
Agreed - we actually
[New recipients: this is a reply to
https://lists.freedesktop.org/archives/beignet/2017-January/008516.html ]
On 24/01/17 08:58, Pan, Xiuli wrote:
> And we have a conformance test with these patches, and the only problem is
> that the sizeof(void *):
> On GEN9 machine, the device will return 32
As for the "OpenCL 2.0 = worse performance" problem, the OpenCL spec
actually _requires_ 2.0 compilers to default to 1.2, and only enable
2.0 if explicitly asked to:
https://www.khronos.org/registry/OpenCL/specs/opencl-2.0.pdf page 203
beignet currently defaults to 2.0 if it is enabled, but this
ime-detection.patch/#L351
is what I used, though you'll want to remove the (Debian-specific)
jessie-backports reference.
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
___
Beignet mailing list
Beignet@lists.freedesktop.
to be faster). Update documentation.
Unconditionally build 2.0-only tests, but skip them on non-2.0
hardware. Add -cl-std=2.0 to tests that need it.
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -231,19 +231,8 @@ IF (EXPERIMENTAL_
When beignet 1.2.1 is built with LLVM 3.9,
OCL_STRICT_CONFORMANCE=0 utest_run -c vload_test_uchar
(and other vload_test_*) hang during kernel compile, deep inside LLVM.
This does not happen in master; a bisect finds the fix to be 75cb7ca,
and applying this to 1.2.1 (at least in the Debian
AppStream is a standard for software metadata,
including what hardware a driver supports:
https://www.freedesktop.org/software/appstream/docs/chap-Metadata.html
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
Before this is pushed, needs to be filled in:
AppStream
utf-8')
has an error in python2.x:
Sorry, and that fix looks OK (though I haven't tried it).
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
___
Beignet mailing list
Beignet@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/beignet
powr(x,y) is explicitly undefined for negative x; on my hardware,
it happens to be correct in default mode, but acts like pow(abs(x),y)
in fast (OCL_STRICT_CONFORMANCE=0) mode, failing the test.
Signed-off-by: Rebecca Palmer
diff --git a/utests/utest_math_gen.py
A patch set for supporting LLVM 3.9 was posted here last month
https://lists.freedesktop.org/archives/beignet/2016-August/007843.html
, but not pushed; why?
Debian mesa plans to switch to 3.9
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=836913
and we would normally follow.
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
docs/Beignet.mdwn | 20 ++--
docs/howto/cross-compiler-howto.mdwn| 4 ++--
docs/howto/stand-alone-utest-howto.mdwn | 8
utests/builtin_global_linear_id.cpp | 2 +-
4
This is required to find KHR extensions via ICD, as the loader needs to
know which ICD to send the request to.
If the function is not found, fail the test instead of crashing.
---
utests/utest_helper.cpp | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
utests/builtin_global_linear_id.cpp | 2 +-
utests/builtin_global_size.cpp| 2 +-
utests/builtin_kernel_block_motion_estimate_intel.cpp | 4 ++--
utests/builtin_local_si
+ if (drm_intel_get_pooled_eu(driver->fd) >= 0) {
Shouldn't this be >0 , not >=0? The libdrm commit message (
https://cgit.freedesktop.org/mesa/drm/commit/intel/intel_bufmgr_gem.c?id=98887140e343493f01be7a1dec721c024bcf72c7
) says 0 means not supported.
(>=0 works for me with libdrm
This extends 4be3664 to the test suite.
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
utests/CMakeLists.txt | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/utests/CMakeLists.txt b/utests/CMakeLists.txt
index a5d8d40..8b33666 100644
--- a/
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
utests/CMakeLists.txt | 4 ++--
utests/builtin_kernel_block_motion_estimate_intel.cpp | 4
2 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/utests/CMakeLists.txt b/
I just built beignet in a chroot with no X11 installed; it didn't have
that string, and worked (when run as root - for ordinary users, it
failed with "/dev/dri/card0 not authenticated", but this is expected).
Debian's (with-X) beignet has that string and a few more (__cxx11,
x11_display,
Weird - setenv.sh has export OCL_IGNORE_SELF_TEST=1, which is supposed
to turn that off (does setting environment variables that way not work
in FreeBSD? if so, you'll also need to set the rest of setenv.sh in
whatever way does work). Does passing that explicitly (i.e.
OCL_IGNORE_SELF_TEST=1
> builtin_kernel_block_motion_estimate_intel()[SUCCESS]
> runtime_climage_from_boname()Unresolved symbol:
> _Z22__gen_ocl_write_imagef11ocl_image2dDv2_iDv4_f
> Aborting...
That message is from backend/src/llvm/llvm_gen_backend.hpp:97, and means
it can't find that function (in its internal
That looks like abi::__cxa_demangle is sometimes failing completely
(and possibly inconsistently, given that it's failing on a different
symbol this time). Given that it only happens on FreeBSD, possibly
a libc++ (LLVM) vs libstdc++ (GCC) issue??
___
v3: Use extension check, not beignet check. Treat claiming
to have the extension but not having the kernel as a failure.
---
(v2 was the un-numbered 10/10/16 08:07 version...which I subsequently
noticed was broken in that it assumed a non-NULL
clGetExtensionFunctionAddressForPlatform result
> [ Chuanbo ] This utest test intel extension cl_intel_accelerator and
> cl_intel_motion_estimation, so it's not beignet specific,
> but it's intel specific. You can add code such as:
> if (!cl_check_accelerator() && !cl_check_motion_estimation())
> The implementation of
This patch causes a warning:
/home/yr /beignet/src/cl_khr_icd.c:187:3: warning: initialization from
incompatible pointer type
clGetKernelSubGroupInfoKHR,
/home/yr/maintain/beignet/src/cl_khr_icd.c:187:3: warning: (near initialization
for ‘cl_khr_icd_dispatch.clUnknown136’)
I see that as
Now I see that the problem is the API cannot be used through ICD, but in
OpenCL1.2 it is a vender extension.
Beignet may consider it a vendor extension, but ocl-icd doesn't:
anything with a KHR or EXT name goes through the "check the dispatch
table first" lookup process, whether or not it
> Beignet: self-test failed: (3, 7, 5) + (5, 7, 3) returned (3, 7, 5)
> See README.md or http://www.freedesktop.org/wiki/Software/Beignet/
> Beignet: disabling non-working device
That looks like the "no __local on Haswell" issue (Linux fixed this in 4.2,
but FreeBSD might not have): try
2baff9c moved mem->magic to cl_base_object.
---
(Or should this be CL_OBJECT_IS_MEM(mem), i.e. also checking the reference
count?)
--- a/src/cl_cmrt.cpp
+++ b/src/cl_cmrt.cpp
@@ -256,7 +256,7 @@ cl_int cmrt_set_kernel_arg(cl_kernel k,
result = cmrt_kernel->SetKernelArg(index, sz, value);
As the "do we have any usable devices?" check uses this,
it needs to not crash even when we don't.
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
---
The user who reported a crash here ( https://bugs.debian.org/848792 )
was using unsupported hardware, but I do
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--- a/docs/howto/gl-buffer-sharing-howto.mdwn
+++ b/docs/howto/gl-buffer-sharing-howto.mdwn
@@ -1,16 +1,16 @@
GL Buffer Sharing HowTo
=
-Beignet now support cl_khr_gl_sharing partially(the most commonl
Signed-off-by: Rebecca N. Palmer <rebecca_pal...@zoho.com>
--- a/docs/Beignet.mdwn
+++ b/docs/Beignet.mdwn
@@ -283,7 +283,7 @@ Documents for OpenCL application developers
- [[OpenGL Buffer Sharing|Beignet/howto/gl-buffer-sharing-howto]]
- [[Video Motion Estimation|Beignet/howto/video-
T 2017 x86_64 GNU/Linux
2017-04-10 9:21 GMT+02:00 Rebecca N. Palmer <rebecca_pal...@zoho.com>:
If you can, revert to the jessie (not -backports) kernel: that fixes this
issue for me (on Ivybridge).
___
Beignet mailing list
Beignet@lists.freedeskt
+#ifdef ENABLE_OPENCL_20
+.address_bits = 64,
+#else
+.address_bits = 32,
+#endif
Should that be 32 unconditionally, like it now is in master?
https://lists.freedesktop.org/archives/beignet/2017-February/008565.html
___
Beignet mailing list
If you can, revert to the jessie (not -backports) kernel: that fixes
this issue for me (on Ivybridge).
___
Beignet mailing list
Beignet@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/beignet
I think a better choice is to make OpenCL 1.2 also use 64bit address. So that
we have a unified address_bits across 1.2 and 2.0.
But using 64bit address would hurt some performance. I think we need more work on 64bit
optimization to make "unified address_bits across different ocl version"
As beignet's Debian maintainer, I am considering whether to enable CL-GL
sharing in our package, given its incomplete state.
Fedora do enable it, and have a Blender crash bug
https://bugs.freedesktop.org/show_bug.cgi?id=101779 , though as I can't
reproduce this I can't tell whether disabling
The *missing* functions aren't likely to be a problem: as far as I can
tell,
http://sources.debian.net/src/forge/0.9.2-2/examples/opencl/cl_helpers.h/?hl=72#L72
is the only place in Debian that tries to use one of them, and it
correctly handles its absence.
I'm more worried about the
1 - 100 of 132 matches
Mail list logo