Support WSLg GPU acceleration inside containers

[peci1]

Is your feature request related to a problem? Please describe.
When running a container inside a WSL 2 console, GPU is not HW accelerated.

Describe the solution you'd like
Add a flag --wsl smilar to --nv.

Describe alternatives you've considered
The current workaround is to add these bind mounts (I added them to /etc/singularity/singularity.conf):

# Support WSLg
bind path = /usr/lib/wsl/lib/libd3d12core.so:/.singularity.d/libs/libd3d12core.so
bind path = /usr/lib/wsl/lib/libd3d12.so:/.singularity.d/libs/libd3d12.so
bind path = /usr/lib/wsl/lib/libdxcore.so:/.singularity.d/libs/libdxcore.so
bind path = /mnt/wsl
bind path = /mnt/wslg
bind path = /usr/lib/wsl/drivers
bind path = /usr/lib/wsl/lib

Additional context
With this config, I can now run glxgears with the following output, using D3D12 driver instead of the llvmpipe software renderer:

$ singularity exec image.simg glxgears -info
GL_RENDERER   = D3D12 (AMD Radeon(TM) Renoir Graphics D1)
GL_VERSION    = 3.1 Mesa 21.2.6
GL_VENDOR     = Microsoft Corporation
GL_EXTENSIONS = GL_ARB_multisample GL_EXT_abgr GL_EXT_bgra GL_EXT_blend_color GL_EXT_blend_minmax GL_EXT_blend_subtract GL_EXT_copy_texture GL_EXT_subtexture GL_EXT_texture_object GL_EXT_vertex_array GL_EXT_compiled_vertex_array GL_EXT_texture GL_EXT_texture3D GL_IBM_rasterpos_clip GL_ARB_point_parameters GL_EXT_draw_range_elements GL_EXT_packed_pixels GL_EXT_point_parameters GL_EXT_rescale_normal GL_EXT_separate_specular_color GL_EXT_texture_edge_clamp GL_SGIS_generate_mipmap GL_SGIS_texture_border_clamp GL_SGIS_texture_edge_clamp GL_SGIS_texture_lod GL_ARB_framebuffer_sRGB GL_ARB_multitexture GL_EXT_framebuffer_sRGB GL_IBM_multimode_draw_arrays GL_IBM_texture_mirrored_repeat GL_ARB_texture_cube_map GL_ARB_texture_env_add GL_ARB_transpose_matrix GL_EXT_blend_func_separate GL_EXT_fog_coord GL_EXT_multi_draw_arrays GL_EXT_secondary_color GL_EXT_texture_env_add GL_EXT_texture_filter_anisotropic GL_EXT_texture_lod_bias GL_INGR_blend_func_separate GL_NV_blend_square GL_NV_light_max_exponent GL_NV_texgen_reflection GL_NV_texture_env_combine4 GL_S3_s3tc GL_SUN_multi_draw_arrays GL_ARB_texture_border_clamp GL_ARB_texture_compression GL_EXT_framebuffer_object GL_EXT_texture_compression_s3tc GL_EXT_texture_env_combine GL_EXT_texture_env_dot3 GL_MESA_window_pos GL_NV_packed_depth_stencil GL_NV_texture_rectangle GL_ARB_depth_texture GL_ARB_occlusion_query GL_ARB_shadow GL_ARB_texture_env_combine GL_ARB_texture_env_crossbar GL_ARB_texture_env_dot3 GL_ARB_texture_mirrored_repeat GL_ARB_window_pos GL_ATI_fragment_shader GL_EXT_stencil_two_side GL_EXT_texture_cube_map GL_NV_copy_depth_to_color GL_NV_depth_clamp GL_NV_fog_distance GL_NV_half_float GL_APPLE_packed_pixels GL_ARB_draw_buffers GL_ARB_fragment_program GL_ARB_fragment_shader GL_ARB_shader_objects GL_ARB_vertex_program GL_ARB_vertex_shader GL_ATI_draw_buffers GL_ATI_texture_env_combine3 GL_ATI_texture_float GL_EXT_shadow_funcs GL_EXT_stencil_wrap GL_MESA_pack_invert GL_NV_primitive_restart GL_ARB_depth_clamp GL_ARB_fragment_program_shadow GL_ARB_half_float_pixel GL_ARB_occlusion_query2 GL_ARB_point_sprite GL_ARB_shading_language_100 GL_ARB_sync GL_ARB_texture_non_power_of_two GL_ARB_vertex_buffer_object GL_ATI_blend_equation_separate GL_EXT_blend_equation_separate GL_OES_read_format GL_ARB_color_buffer_float GL_ARB_pixel_buffer_object GL_ARB_texture_compression_rgtc GL_ARB_texture_float GL_ARB_texture_rectangle GL_EXT_packed_float GL_EXT_pixel_buffer_object GL_EXT_texture_compression_dxt1 GL_EXT_texture_compression_rgtc GL_EXT_texture_rectangle GL_EXT_texture_sRGB GL_EXT_texture_shared_exponent GL_ARB_framebuffer_object GL_EXT_framebuffer_blit GL_EXT_framebuffer_multisample GL_EXT_packed_depth_stencil GL_ARB_vertex_array_object GL_ATI_separate_stencil GL_EXT_draw_buffers2 GL_EXT_draw_instanced GL_EXT_gpu_program_parameters GL_EXT_gpu_shader4 GL_EXT_texture_array GL_EXT_texture_integer GL_EXT_texture_sRGB_decode GL_EXT_timer_query GL_OES_EGL_image GL_EXT_texture_buffer_object GL_ARB_copy_buffer GL_ARB_depth_buffer_float GL_ARB_draw_instanced GL_ARB_half_float_vertex GL_ARB_instanced_arrays GL_ARB_map_buffer_range GL_ARB_texture_buffer_object GL_ARB_texture_rg GL_ARB_texture_swizzle GL_ARB_vertex_array_bgra GL_EXT_texture_swizzle GL_EXT_vertex_array_bgra GL_NV_conditional_render GL_AMD_conservative_depth GL_AMD_draw_buffers_blend GL_AMD_shader_stencil_export GL_ARB_ES2_compatibility GL_ARB_blend_func_extended GL_ARB_compatibility GL_ARB_debug_output GL_ARB_draw_buffers_blend GL_ARB_draw_elements_base_vertex GL_ARB_explicit_attrib_location GL_ARB_fragment_coord_conventions GL_ARB_provoking_vertex GL_ARB_sampler_objects GL_ARB_seamless_cube_map GL_ARB_shader_stencil_export GL_ARB_shader_texture_lod GL_ARB_texture_buffer_object_rgb32 GL_ARB_texture_multisample GL_ARB_texture_query_lod GL_ARB_texture_rgb10_a2ui GL_ARB_uniform_buffer_object GL_ARB_vertex_type_2_10_10_10_rev GL_EXT_provoking_vertex GL_EXT_texture_snorm GL_MESA_texture_signed_rgba GL_NV_copy_image GL_ARB_get_program_binary GL_ARB_robustness GL_ARB_separate_shader_objects GL_ARB_shader_bit_encoding GL_ARB_shader_subroutine GL_ARB_timer_query GL_EXT_direct_state_access GL_ANGLE_texture_compression_dxt3 GL_ANGLE_texture_compression_dxt5 GL_ARB_compressed_texture_pixel_storage GL_ARB_conservative_depth GL_ARB_internalformat_query GL_ARB_map_buffer_alignment GL_ARB_shading_language_420pack GL_ARB_shading_language_packing GL_ARB_texture_storage GL_EXT_framebuffer_multisample_blit_scaled GL_EXT_transform_feedback GL_AMD_shader_trinary_minmax GL_ARB_ES3_compatibility GL_ARB_arrays_of_arrays GL_ARB_clear_buffer_object GL_ARB_explicit_uniform_location GL_ARB_invalidate_subdata GL_ARB_program_interface_query GL_ARB_stencil_texturing GL_ARB_texture_buffer_range GL_ARB_texture_query_levels GL_ARB_texture_storage_multisample GL_ARB_vertex_attrib_binding GL_KHR_debug GL_KHR_texture_compression_astc_ldr GL_ARB_internalformat_query2 GL_ARB_multi_bind GL_ARB_shading_language_include GL_ARB_texture_stencil8 GL_ARB_vertex_type_10f_11f_11f_rev GL_EXT_shader_integer_mix GL_ARB_direct_state_access GL_ARB_get_texture_sub_image GL_KHR_context_flush_control GL_ARB_parallel_shader_compile GL_KHR_no_error GL_KHR_texture_compression_astc_sliced_3d GL_MESA_shader_integer_functions GL_ARB_texture_filter_anisotropic GL_KHR_parallel_shader_compile GL_EXT_EGL_image_storage GL_EXT_EGL_sync
VisualID 370, 0x172
757 frames in 5.0 seconds = 150.927 FPS
717 frames in 5.0 seconds = 143.338 FPS
654 frames in 5.0 seconds = 130.602 FPS

[peci1]

Relevant links I used when setting up the WSL 2 GPU support:

Install newest GPU drivers: https://learn.microsoft.com/en-us/windows/wsl/tutorials/gui-apps

Resolve a problem with clock inside the WSL 2 container: https://stackoverflow.com/questions/65086856/wsl2-clock-is-out-of-sync-with-windows

Install singularity from PPA https://launchpad.net/~peci1/+archive/ubuntu/singularity-ce-v4 .

Also enable Virtual Machine Platform and Hyper-V when enabling WSL: https://learn.microsoft.com/en-us/answers/questions/1187339/resolving-error-0x80370114-the-operation-could-not

[dtrudg]

Thanks for opening this issue. We do have NVIDIA GPU compute support (https://sylabs.io/2022/03/wsl2-gpu/) in WSL2, and expanding WSL support to DirectX<->GL acceleration is certainly of interest.

Unless as a third party contribution, this is unlikely to make it for SingularityCE 4.1 because the roadmap for a Dec/Jan release is full vs developer time available.

We'd welcome contributions that implement (including in the new OCI-mode). Otherwise it'll be considered for a future version.

[peci1]

Thanks. For us, the workaround described here is enough. So I meant this issue only as a kind of kickstarter for somebody else who would like to work on it.