Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Pan, Xiuli
Hi all, Thank you for pay close attention to our latest release. We are very sorry that we are not very thoughtful about the package making things. We did not have a performance test for OCL2.0 so we did not set it as a default option and we did not think about how to run a binary version on

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Rebecca N. Palmer
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

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Igor Gnatenko
On Sat, 2017-01-21 at 17:39 +, Rebecca N. Palmer wrote: > > I have a first attempt at "enable/disable 2.0 at run time" written, > > but haven't yet tested it.  > > It worked for me, and here it is; testing on other hardware (in > particular, hardware that supports 2.0: the last 15 tests of

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Bruno Pagani
Le 21/01/2017 à 18:01, Rebecca N. Palmer a écrit : > I have a first attempt at "enable/disable 2.0 at run time" written, > but haven't yet tested it. OK. I have both an Haswell and Skylake machine at hand if this can help in testing. > On 21/01/17 16:08, Bruno Pagani wrote: >> Le 21/01/2017 à

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Rebecca N. Palmer
> I have a first attempt at "enable/disable 2.0 at run time" written, but > haven't yet tested it. It worked for me, and here it is; testing on other hardware (in particular, hardware that supports 2.0: the last 15 tests of the test suite appear to be the ones that test that) would be

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Rebecca N. Palmer
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 there any

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Bruno Pagani
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 there any downside in compiling with OpenCL 2.0 support, >>> Yes - on older (Ivybridge/Haswell - no emitUntypedReadA64Instruction) >>>

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Rebecca N. Palmer
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

Re: [Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Rebecca N. Palmer
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

[Beignet] Building beignet with OpenCL 2.0 support in distros

2017-01-21 Thread Bruno Pagani
Hi there, First post here, so first of all thanks for building Beignet and this 1.3 release! I’m ArchLinux packager for beignet[0], and I’m trying to figure out how/if I should package Beignet with OpenCL 2.0 support. From what I understand, compiling with -DENABLE_OPENCL_20=1 just results in