Bug#1031080: nvidia-driver: OpenCL compiler dorsn't wok since version 520

[Andreas Beckmann]

On 11/02/2023 11.51, krzysztof wrote:
> The problem stated sice upgrade nvidia-driver to version 520

There is now a 525 driver in sid. Note that you will need to enable the
new non-free-firmware archive area to use that.

https://www.debian.org/releases/bookworm/amd64/release-notes/ch-information.en.html#non-free-split

Andreas

[Krzysztof Marczak]

As I know from other users the problem is not resolved in 525. Of course I'm going to check it by myself as soon as possible.

[Krzysztof Marczak]

After upgrading the driver to version 525 the problem still exists. OpenCL compiler doesn't work at all. Still it returns error code 11 and there is no build log output.

[Pascal Obry]

Hello,

Just to say that I have the same issue. This is on darktable when
trying to build the OpenCL programs.

I'm using:

$ apt-cache policy nvidia-driver
nvidia-driver:
Installed: 525.85.12-1
Candidate: 525.85.12-1

Thanks,

--
  Pascal Obry /  Magny Les Hameaux (78)

  The best way to travel is by means of imagination

  http://photos.obry.net

  gpg --keyserver keys.gnupg.net --recv-key F949BD3B

[Andreas Beckmann]

Control: found -1 525.85.12-1

On 12/02/2023 22.47, Pascal Obry wrote:
> Just to say that I have the same issue. This is on darktable when
> trying to build the OpenCL programs.

Are there bug reports about this issue elsewhere? The nvidia forum,
other distributions, ... i.e. is this a Debian specific packaging bug
(perhaps a missing dependency) or a general issue?

Does clinfo work?

In case this issue looks Debian specific, and clinfo fails it might be
helpful to have the output from 'strace -s128 clinfo' (ideally with a
reference from a working driver version to compare against, might also
come from a different machine)

Andreas

[Pascal Obry]

Hi Andreas;

> Does clinfo work?

Yes, no problem with clinfo:

$ clinfo
Number of platforms 1
Platform Name NVIDIA CUDA
Platform Vendor NVIDIA Corporation
Platform Version OpenCL 3.0 CUDA 12.0.139
Platform Profile FULL_PROFILE
Platform Extensions cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_copy_opts cl_khr_gl_event cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_nv_kernel_attribute cl_khr_device_uuid cl_khr_pci_bus_info cl_khr_external_semaphore cl_khr_external_memory cl_khr_external_semaphore_opaque_fd cl_khr_external_memory_opaque_fd
Platform Extensions with Version cl_khr_global_int32_base_atomics 0x400000 (1.0.0)
cl_khr_global_int32_extended_atomics 0x400000 (1.0.0)
cl_khr_local_int32_base_atomics 0x400000 (1.0.0)
cl_khr_local_int32_extended_atomics 0x400000 (1.0.0)
cl_khr_fp64 0x400000 (1.0.0)
cl_khr_3d_image_writes 0x400000 (1.0.0)
cl_khr_byte_addressable_store 0x400000 (1.0.0)
cl_khr_icd 0x400000 (1.0.0)
cl_khr_gl_sharing 0x400000 (1.0.0)
cl_nv_compiler_options 0x400000 (1.0.0)
cl_nv_device_attribute_query 0x400000 (1.0.0)
cl_nv_pragma_unroll 0x400000 (1.0.0)
cl_nv_copy_opts 0x400000 (1.0.0)
cl_khr_gl_event 0x400000 (1.0.0)
cl_nv_create_buffer 0x400000 (1.0.0)
cl_khr_int64_base_atomics 0x400000 (1.0.0)
cl_khr_int64_extended_atomics 0x400000 (1.0.0)
cl_nv_kernel_attribute 0x400000 (1.0.0)
cl_khr_device_uuid 0x400000 (1.0.0)
cl_khr_pci_bus_info 0x400000 (1.0.0)
cl_khr_external_semaphore 0x9000 (0.9.0)
cl_khr_external_memory 0x9000 (0.9.0)
cl_khr_external_semaphore_opaque_fd 0x9000 (0.9.0)
cl_khr_external_memory_opaque_fd 0x9000 (0.9.0)
Platform Numeric Version 0xc00000 (3.0.0)
Platform Extensions function suffix NV
Platform Host timer resolution 0ns
Platform External memory handle types Opaque FD
Platform External semaphore import types Opaque FD
Platform External semaphore export types Opaque FD

Platform Name NVIDIA CUDA
Number of devices 1
Device Name Quadro T1000
Device Vendor NVIDIA Corporation
Device Vendor ID 0x10de
Device Version OpenCL 3.0 CUDA
Device UUID acba1b3a-c3af-4662-d17f-ed215bf60c4d
Driver UUID acba1b3a-c3af-4662-d17f-ed215bf60c4d
Valid Device LUID No
Device LUID 6d69-637300000000
Device Node Mask 0
Device Numeric Version 0xc00000 (3.0.0)
Driver Version 525.85.12
Device OpenCL C Version OpenCL C 1.2
Device OpenCL C all versions OpenCL C 0x400000 (1.0.0)
OpenCL C 0x401000 (1.1.0)
OpenCL C 0x402000 (1.2.0)
OpenCL C 0xc00000 (3.0.0)
Device OpenCL C features __opencl_c_fp64 0xc00000 (3.0.0)
__opencl_c_images 0xc00000 (3.0.0)
__opencl_c_int64 0xc00000 (3.0.0)
__opencl_c_3d_image_writes 0xc00000 (3.0.0)
Latest conformance test passed v2022-10-05-00
Device Type GPU
Device Topology (NV) PCI-E, 0000:01:00.0
Device PCI bus info (KHR) PCI-E, 0000:01:00.0
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 14
Max clock frequency 1725MHz
Compute Capability (NV) 7.5
Device Partition (core)
Max number of sub-devices 1
Supported partition types None
Supported affinity domains (n/a)
Max work item dimensions 3
Max work item sizes 1024x1024x64
Max work group size 1024
Preferred work group size multiple (device) 32
=== CL_PROGRAM_BUILD_LOG ===
Preferred work group size multiple (kernel) <getWGsizes:1980: create kernel : error -45>
Warp size (NV) 32
Max sub-groups per work group 0
Preferred / native vector sizes
char 1 / 1
short 1 / 1
int 1 / 1
long 1 / 1
half 0 / 0 (n/a)
float 1 / 1
double 1 / 1 (cl_khr_fp64)
Half-precision Floating-point support (n/a)
Single-precision Floating-point support (core)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations Yes
Double-precision Floating-point support (cl_khr_fp64)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Address bits 64, Little-Endian
External memory handle types Opaque FD
External semaphore import types Opaque FD
External semaphore export types Opaque FD
Global memory size 4084924416 (3.804GiB)
Error Correction support No
Max memory allocation 1021231104 (973.9MiB)
Unified memory for Host and Device No
Integrated memory (NV) No
Shared Virtual Memory (SVM) capabilities (core)
Coarse-grained buffer sharing Yes
Fine-grained buffer sharing No
Fine-grained system sharing No
Atomics No
Minimum alignment for any data type 128 bytes
Alignment of base address 4096 bits (512 bytes)
Preferred alignment for atomics
SVM 0 bytes
Global 0 bytes
Local 0 bytes
Atomic memory capabilities relaxed, work-group scope
Atomic fence capabilities relaxed, acquire/release, work-group scope
Max size for global variable 0
Preferred total size of global vars 0
Global Memory cache type Read/Write
Global Memory cache size 458752 (448KiB)
Global Memory cache line size 128 bytes
Image support Yes
Max number of samplers per kernel 32
Max size for 1D images from buffer 268435456 pixels
Max 1D or 2D image array size 2048 images
Base address alignment for 2D image buffers 0 bytes
Pitch alignment for 2D image buffers 0 pixels
Max 2D image size 32768x32768 pixels
Max 3D image size 16384x16384x16384 pixels
Max number of read image args 256
Max number of write image args 32
Max number of read/write image args 0
Pipe support No
Max number of pipe args 0
Max active pipe reservations 0
Max pipe packet size 0
Local memory type Local
Local memory size 49152 (48KiB)
Registers per block (NV) 65536
Max number of constant args 9
Max constant buffer size 65536 (64KiB)
Generic address space support No
Max size of kernel argument 4352 (4.25KiB)
Queue properties (on host)
Out-of-order execution Yes
Profiling Yes
Device enqueue capabilities (n/a)
Queue properties (on device)
Out-of-order execution No
Profiling No
Preferred size 0
Max size 0
Max queues on device 0
Max events on device 0
Prefer user sync for interop No
Profiling timer resolution 1000ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
Non-uniform work-groups No
Work-group collective functions No
Sub-group independent forward progress No
Kernel execution timeout (NV) Yes
Concurrent copy and kernel execution (NV) Yes
Number of async copy engines 3
IL version (n/a)
ILs with version (n/a)
printf() buffer size 1048576 (1024KiB)
Built-in kernels (n/a)
Built-in kernels with version (n/a)
Device Extensions cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_copy_opts cl_khr_gl_event cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_nv_kernel_attribute cl_khr_device_uuid cl_khr_pci_bus_info cl_khr_external_semaphore cl_khr_external_memory cl_khr_external_semaphore_opaque_fd cl_khr_external_memory_opaque_fd
Device Extensions with Version cl_khr_global_int32_base_atomics 0x400000 (1.0.0)
cl_khr_global_int32_extended_atomics 0x400000 (1.0.0)
cl_khr_local_int32_base_atomics 0x400000 (1.0.0)
cl_khr_local_int32_extended_atomics 0x400000 (1.0.0)
cl_khr_fp64 0x400000 (1.0.0)
cl_khr_3d_image_writes 0x400000 (1.0.0)
cl_khr_byte_addressable_store 0x400000 (1.0.0)
cl_khr_icd 0x400000 (1.0.0)
cl_khr_gl_sharing 0x400000 (1.0.0)
cl_nv_compiler_options 0x400000 (1.0.0)
cl_nv_device_attribute_query 0x400000 (1.0.0)
cl_nv_pragma_unroll 0x400000 (1.0.0)
cl_nv_copy_opts 0x400000 (1.0.0)
cl_khr_gl_event 0x400000 (1.0.0)
cl_nv_create_buffer 0x400000 (1.0.0)
cl_khr_int64_base_atomics 0x400000 (1.0.0)
cl_khr_int64_extended_atomics 0x400000 (1.0.0)
cl_nv_kernel_attribute 0x400000 (1.0.0)
cl_khr_device_uuid 0x400000 (1.0.0)
cl_khr_pci_bus_info 0x400000 (1.0.0)
cl_khr_external_semaphore 0x9000 (0.9.0)
cl_khr_external_memory 0x9000 (0.9.0)
cl_khr_external_semaphore_opaque_fd 0x9000 (0.9.0)
cl_khr_external_memory_opaque_fd 0x9000 (0.9.0)

NULL platform behavior
clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...) NVIDIA CUDA
clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...) Success [NV]
clCreateContext(NULL, ...) [default] Success [NV]
clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT) No platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU) No platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM) Invalid device type for platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL) No platform

ICD loader properties
ICD loader Name OpenCL ICD Loader
ICD loader Vendor OCL Icd free software
ICD loader Version 2.3.1
ICD loader Profile OpenCL 3.0

So it seems that only the CL compiler is broken.

[Andreas Beckmann]

clinfo does build and run a minimal OpenCL program (to get parameters
not available elsewhere), and this fails too (with no output in the
program build log):

On 13/02/2023 07.55, Pascal Obry wrote:
...

> Preferred work group size multiple (device) 32
> === CL_PROGRAM_BUILD_LOG ===
> Preferred work group size multiple (kernel) <getWGsizes:1980: create kernel : error -45>
> Warp size (NV) 32

...

Please send me the strace output for that.


Andreas

[Andreas Beckmann]

On 13/02/2023 11.59, Pascal Obry wrote:
> openat(AT_FDCWD, "/usr/lib/x86_64-linux-gnu/libnvidia-nvvm.so.4", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory)

Please try OpenCL again install after installing libnvidia-nvvm4.


Andreas

[Krzysztof Marczak]

I have also tested it with libnvidia-nvvm4 installed. I got exactly the same
result.
It could be a problem with wrong the path for libnvidia-nvvm.so.4. After
installation of the package it is located here:

/usr/lib/x86_64-linux-gnu/nvidia/current/libnvidia-nvvm.so.4
/usr/lib/x86_64-linux-gnu/nvidia/current/libnvidia-nvvm.so.525.85.12

But the nvidia compiller is looking of the library here:

/lib/x86_64-linux-gnu/libnvidia-nvvm.so.525.85.12
/usr/lib/x86_64-linux-gnu/libnvidia-nvvm.so.525.85.12
/lib/libnvidia-nvvm.so.525.85.12
/usr/lib/libnvidia-nvvm.so.525.85.12

When in created symlinks to the files libnvidia-nvvm.so.4 and libnvidia-
nvvm.so.525.85.12 in /usr/lib/x86_64-linux-gnu/ the OpenCL compiler started to
work properly.

So installation of libnvidia-nvvm4 and creating symlinks is actual
workaround.

[Andreas Beckmann]

On 13/02/2023 15.31, Krzysztof Marczak wrote:
> When in created symlinks to the files libnvidia-nvvm.so.4 and libnvidia-
> nvvm.so.525.85.12 in /usr/lib/x86_64-linux-gnu/ the OpenCL compiler started to
> work properly.

OK, I'll take care of the long version variant of the
libnvidia-nvvm.so.#VERSION# symlink as well.

Once there is a new package version out, don't forget to remove these
manually created links again (there is no guarantee that the package
upgrade will "overwrite them"), otherwise you might run into problems
later on ...

Did one of you try the 515 driver? I'd expect that the OpenCL problems
started back then, at least there were some changes w.r.t.
libnvidia-nvvm.so* ...

Andreas

PS: since most libraries in the nvidia driver are not linked against all
their dependencies but use dlopen for some optional libraries (that are
only needed for certain operations), it's not trivial to get the
dependencies in the Debian packages right ... especially if there are
driver internal changes in the usage of these things ... or these libs
are not used that often ...

[Krzysztof Marczak]

I have worked with 515 driver for a quite long time and it worked without any issue. All problems started with 520.

Thank you very much for working on this issue and of course for maintaining these difficult packages.

    Krzysztof