Re: [Beignet] [PATCH] API: Fix device type bugs

2017-02-09 Thread Yang, Rong R
Sorry, if support multiple type device later, can't check the type in the api function, should check the device type when get device. I will send a patch to let cl_get_gt_device function return device when type is CL_DEVICE_TYPE_GPU and CL_DEVICE_TYPE_DEFAULT. > -Original Message- > From

Re: [Beignet] [PATCH] API: Fix device type bugs

2017-02-09 Thread Yang, Rong R
There are two types of relative return error, one is CL_INVALID_DEVICE_TYPE, for invalid value, another is CL_DEVICE_NOT_FOUND, for no matched device. The original check for CL_INVALID_DEVICE_TYPE is right, but miss the CL_DEVICE_NOT_FOUND check. Giuseppe's patch will return CL_DEVICE_NOT_FOUND,

Re: [Beignet] [PATCH] CMAKE: if user disable ICD support, must pass cmake option explicitly.

2017-02-09 Thread Pan, Xiuli
LGTM! -Original Message- From: Beignet [mailto:beignet-boun...@lists.freedesktop.org] On Behalf Of Yang Rong Sent: Wednesday, February 8, 2017 12:53 PM To: beignet@lists.freedesktop.org Cc: Yang, Rong R Subject: [Beignet] [PATCH] CMAKE: if user disable ICD support, must pass cmake optio

[Beignet] [PATCH] API: Fix device type bugs

2017-02-09 Thread Xiuli Pan
From: Pan Xiuli Beignet only support GPU now, we should return CL_INVALID_DEVICE_TYPE for CPU and ACCELERATOR. Contributor: Giuseppe Bilotta Signed-off-by: Pan Xiuli --- src/cl_api_context.c | 3 +-- src/cl_api_device_id.c | 4 +--- 2 files changed, 2 insertions(+), 5 deletions(-) diff --g

Re: [Beignet] [PATCH 1/2] GBE: use shl instead of multiply as possible.

2017-02-09 Thread Song, Ruiling
The patchset LGTM Ruiling ___ Beignet mailing list Beignet@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/beignet

[Beignet] [PATCH] Runtime: add a warning when load gen binary fail.

2017-02-09 Thread Yang Rong
Some applications use program's binary by default, if load the former's gen binary, because the fields of gen binary has changed, and lack of version checking, will lead to clCreateProgramWithBinary fail, may cause applications fail silently. Add a warning to hint user. Signed-off-by: Yang Rong -

Re: [Beignet] [PATCH] Only provide device IDs if the type matches

2017-02-09 Thread Pan, Xiuli
Hi Giuseppe, You point about the device type is right, but the git commit you found is not right the function is just copy form somewhere else. And the this patch is not follow the OpenCL spec. As the Beignet does not have a CPU device, we should just return CL_INVALID_DEVICE_TYPE. I will send

Re: [Beignet] [PATCH] Free context devices on context release

2017-02-09 Thread Yang, Rong R
LGTM, thanks, pushed. > -Original Message- > From: Beignet [mailto:beignet-boun...@lists.freedesktop.org] On Behalf Of > Giuseppe Bilotta > Sent: Friday, February 10, 2017 6:13 > To: Beignet ML > Cc: Giuseppe Bilotta > Subject: [Beignet] [PATCH] Free context devices on context release >

Re: [Beignet] [PATCH] Fix obvious copy-paste

2017-02-09 Thread Yang, Rong R
Yes, always NULL. LGTM, thanks, pushed. > -Original Message- > From: Beignet [mailto:beignet-boun...@lists.freedesktop.org] On Behalf Of > Giuseppe Bilotta > Sent: Friday, February 10, 2017 6:13 > To: Beignet ML > Cc: Giuseppe Bilotta > Subject: [Beignet] [PATCH] Fix obvious copy-paste >

[Beignet] [PATCH] Only provide device IDs if the type matches

2017-02-09 Thread Giuseppe Bilotta
If the user is asking for (say) CPU devices only, we should not return a GPU device instead. This fixes a regression introduced with the refactoring done in 494081299ac3f6970b906cb4d7d1eb1fc2fb23c8, by reinstating the check that used to be in cl_check_device_type() (except that instead of hard-codi

[Beignet] [PATCH] Fix obvious copy-paste

2017-02-09 Thread Giuseppe Bilotta
The conditional was equal to the one before, and would never be hit because internal kernels were reset after release. Instead, since the body is resetting built-in kernels, it appears obvious that the conditional should be on the existence of built-in kernels. Signed-off-by: Giuseppe Bilotta ---

[Beignet] [PATCH] Free context devices on context release

2017-02-09 Thread Giuseppe Bilotta
The context owns the array of devices passed to cl_context_new, so it's its duty to free it. Signed-off-by: Giuseppe Bilotta --- src/cl_context.c | 1 + 1 file changed, 1 insertion(+) diff --git a/src/cl_context.c b/src/cl_context.c index cbe2e017..1ba23024 100644 --- a/src/cl_context.c +++ b/s

Re: [Beignet] VME support for Intel 4th generation (Haswell)

2017-02-09 Thread Weng, Chuanbo
Hi Sateesh, Yes, vme now is just supported on 3rd generation. And we are implementing vme on 6th generation (SKYLAKE) platform firstly, because it’s a common used platform currently. 4th generation is also a little old. Thanks, Chuanbo Weng From: Beignet [mailto:bei

[Beignet] VME support for Intel 4th generation (Haswell)

2017-02-09 Thread sat iiitb
Hi I am interested in using opencl to extract motion vectors between two frames. With in Beignet notes, it was mentioned that as of version 1.3.0, only 3rd generation chips are supported. I would like some advise on how to extend the support for 4th generation as I am using a Intel core i7 (4th g