And here are the CUDA tests that fail on the Tegra X2 with CUDA backend. FYI.

The following tests FAILED:
         26 - kernel/test_shuffle_char (Failed)
         27 - kernel/test_shuffle_short (Failed)
         28 - kernel/test_shuffle_ushort (Failed)
         30 - kernel/test_shuffle_int (Failed)
         31 - kernel/test_shuffle_uint (Failed)
         32 - kernel/test_shuffle_float (Failed)
         33 - kernel/test_shuffle_long (Failed)
         34 - kernel/test_shuffle_ulong (Failed)
         37 - kernel/test_shuffle_double (Failed)
         42 - regression/phi_nodes_not_replicated_REPL (Failed)
         44 - regression/barrier_between_two_for_loops_REPL (Failed)
         45 - regression/simple_for-loop_with_a_barrier_inside_REPL (Failed)
         46 - regression/for-loop_with_computation_after_the_brexit_REPL 
(Failed)
         47 - regression/for-loop_with_a_variable_iteration_count_REPL (Failed)
         48 - regression/early_return_before_a_barrier_region_REPL (Failed)
         49 - regression/id-dependent_computation_before_kernel_exit_REPL 
(Failed)
         50 - regression/barrier_just_before_return_REPL (Failed)
         52 - 
regression/undominated_variable_from_conditional_barrier_handling_REPL (Failed)
         56 - regression/phi_nodes_not_replicated_LOOPS (Failed)
         58 - regression/barrier_between_two_for_loops_LOOPS (Failed)
         59 - regression/simple_for-loop_with_a_barrier_inside_LOOPS (Failed)
         60 - regression/for-loop_with_computation_after_the_brexit_LOOPS 
(Failed)
         61 - regression/for-loop_with_a_variable_iteration_count_LOOPS (Failed)
         62 - regression/early_return_before_a_barrier_region_LOOPS (Failed)
         63 - regression/id-dependent_computation_before_kernel_exit_LOOPS 
(Failed)
         64 - regression/barrier_just_before_return_LOOPS (Failed)
         66 - 
regression/undominated_variable_from_conditional_barrier_handling_LOOPS (Failed)
         70 - regression/setting_a_buffer_argument_to_NULL_causes_a_segfault 
(Failed)
         71 - regression/clSetKernelArg_overwriting_the_previous_kernel's_args 
(Failed)
         73 - 
regression/case_with_multiple_variable_length_loops_and_a_barrier_in_one 
(Failed)
Errors while running CTest



> On Aug 21, 2018, at 11:44 AM, Michal Babej <[email protected]> wrote:
> 
> Hi,
> 
> > void cpu_setup_vector_widths(cl_device_id): 
> 
> This code sets up "preferred vector widths" OpenCL device information. It 
> uses LLVM's sys::getHostCPUFeatures call which apparently fails for your CPU.
> 
> You can either fill them manually, or just comment that code out, it's 
> usually not that important.
> 
> Cheers,
>  -- mb
> From: Gaurav Khanna <[email protected]>
> Sent: Tuesday, August 21, 2018 3:11:27 PM
> To: Portable Computing Language development discussion
> Subject: Re: [pocl-devel] POCL on Power9, Tegra
>  
> 
> Hi Folks 
> 
> An update. I got POCL built and running on Tegra (ARM) perfectly with CUDA 
> backend. Out of the box. OpenCL apps run fine too. Thats great. 
> 
> Now, I’m trying on Power9 with CUDA backend using LLVM 6.0.1. I got 
> everything built last night. And I can certainly compile simple OpenCL apps. 
> However, if I try to run any of them .. I get 
> 
> [root@centos7-ppc64le ~]# ./a.out 
> a.out: /root/pocl/lib/CL/pocl_llvm_utils.cc:179: void 
> cpu_setup_vector_widths(cl_device_id): Assertion `res' failed.
> Aborted (core dumped)
> 
> Its not much info .. but maybe someone on this list can help me troubleshoot. 
> 
> Cheers
> Gaurav
> 
> 
> 
> > On Aug 15, 2018, at 1:28 PM, Gaurav Khanna <[email protected]> wrote:
> > 
> >> 
> >> Are you interested in using Power8 with the Pocl's CPU backend (as a 
> >> compute device), or just CPU as a host -  used by the applications to 
> >> handle the runtime part of OpenCL, with CUDA backend doing the computation 
> >> ?
> >> 
> > 
> > Thanks for the response. Mostly interested in the Power CPU serving as host 
> > and using the CUDA GPUs as the compute-device. 
> > 
> > 
> >> 
> >> IIRC i've seen some commits which dealt with ARM in the CUDA backend, so 
> >> it's quite possible.
> > 
> > 
> > Thats great news. Who do I ask about this? 
> > 
> > 
> > 
> > 
> 
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> pocl-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/pocl-devel
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! 
> http://sdm.link/slashdot_______________________________________________
> pocl-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/pocl-devel

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
pocl-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/pocl-devel

Reply via email to