v5.0.1 (Beta)

[osy]

Highlights

• Improved graphics acceleration for Linux: Vulkan 1.3 is now supported on Linux guests with VirtIO Venus drivers in Mesa. OpenGL 4.1 is supported with the new Apple Core OpenGL backend (macOS only).

Known Issues

• (macOS) Apple CoreGL backend does not support Vulkan.
• DXVK does not work due to missing features (Support DXVK #7575)
• KosmicKrisp is provided as a WIP but the current upstream build is incomplete so MoltenVK is the recommended driver

Changes (v5.0.1)

• CocoaSpice: Reworked Metal renderer to be faster and have lower latency
• Fixed an issue where a BIOS file with a comma in the filename will fail to load
• Default FPS will now be the maximum refresh rate of the display on macOS and iPadOS (iPhone with ProMotion will still have 60Hz as the default rate but can be overridden to 120Hz through settings)
• (macOS) Fixed issue causing HV_UNSUPPORTED on startup with Vulkan enabled (5.0.0. beta wont start any VMs: HV_UNSUPPORTED #7579)
• (iOS) Fixed resolution issue on external displays (iPad Pro External Monitor Display Issue with Windows 11 Virtual Machine #6040)
• (iOS) Pinch to zoom will turn off automatically updating zoom from the guest (until reset with the resize button)
• (iOS) Fixed external display menu not updating
• (iOS) Fixed incorrect scale on external displays and connecting an external display will now automatically scale to fit
• (iOS) Fixed an issue where powering off the VM will show the home screen but not terminate UTM in the background

Changes (v5.0.0)

• Add Venus support for Vulkan (Explore Venus + MoltenVK for GPU acceleration #4551)
• Add Apple Core OpenGL support for OpenGL 4.1 (OpenGL 3.3 with GPU acceleration? #4285) (thanks @startergo)

Issues

Please check the full list on Github and help report any bug you find that is not listed.

Installation

• iOS
• macOS

File	Description	Installation	JIT	Hypervisor	USB
UTM.deb	Jailbroken iOS version	Open in Cydia, dpkg, or Sileo	Yes	Yes(1)	Yes
UTM.dmg	macOS version	Mounting and copying UTM.app to /Applications	Yes	Yes	Yes
UTM.ipa	Non-jailbroken iOS version (sideloading)	AltStore, etc (see guide)	Yes(2)	No	No
UTM-HV.ipa	Non-jailbroken iOS version (TrollStore)	TrollStore	Yes	Yes(1)	Yes
UTM-SE.ipa	Non-jailbroken iOS version (sideloading)	AltStore, enterprise signing, etc	No	No	No
UTM-Remote.ipa	Remote client	Any	No	No	No

1. Hypervisor on iOS requires an M1 iPad or newer.
2. Enabling JIT may require a separate JIT enabler such as Jitterbug or Jitstreamer.

---

This discussion was created from the release v5.0.1 (Beta).

[startergo]

Not too bad

vkmark
=======================================================
    vkmark 2025.01
=======================================================
    Vendor ID:      0x106B
    Device ID:      0xF060209
    Device Name:    Virtio-GPU Venus (Apple M4 Pro)
    Driver Version: 105263203
    Device UUID:    470aad6dd4211c3ccf5e237974687930
=======================================================
[vertex] device-local=true: FPS: 1059 FrameTime: 0.944 ms
[vertex] device-local=false: FPS: 1056 FrameTime: 0.947 ms
[texture] anisotropy=0: FPS: 1072 FrameTime: 0.933 ms
[texture] anisotropy=16: FPS: 1045 FrameTime: 0.957 ms
[shading] shading=gouraud: FPS: 1054 FrameTime: 0.949 ms
[shading] shading=blinn-phong-inf: FPS: 1126 FrameTime: 0.888 ms
[shading] shading=phong: FPS: 1088 FrameTime: 0.919 ms
[shading] shading=cel: FPS: 1062 FrameTime: 0.942 ms
[effect2d] kernel=edge: FPS: 1089 FrameTime: 0.918 ms
[effect2d] kernel=blur: FPS: 1097 FrameTime: 0.912 ms
[desktop] <default>: FPS: 1102 FrameTime: 0.907 ms
[cube] <default>: FPS: 1087 FrameTime: 0.920 ms
[clear] <default>: FPS: 1017 FrameTime: 0.983 ms
=======================================================
                                   vkmark Score: 1073
=======================================================

glmark2
=======================================================
    glmark2 2023.01
=======================================================
    OpenGL Information
    GL_VENDOR:      Mesa
    GL_RENDERER:    virgl
    GL_VERSION:     2.1 Mesa 26.0.0-devel (git-c9e523be97)
    Surface Config: buf=32 r=8 g=8 b=8 a=8 depth=24 stencil=0 samples=0
    Surface Size:   800x600 windowed
=======================================================
[build] use-vbo=false: FPS: 2071 FrameTime: 0.483 ms
[build] use-vbo=true: FPS: 2148 FrameTime: 0.466 ms
[texture] texture-filter=nearest: FPS: 2244 FrameTime: 0.446 ms
[texture] texture-filter=linear: FPS: 2534 FrameTime: 0.395 ms
[texture] texture-filter=mipmap: FPS: 2496 FrameTime: 0.401 ms
[shading] shading=gouraud: FPS: 2056 FrameTime: 0.487 ms
[shading] shading=blinn-phong-inf: FPS: 1984 FrameTime: 0.504 ms
[shading] shading=phong: FPS: 1836 FrameTime: 0.545 ms
[shading] shading=cel: FPS: 1821 FrameTime: 0.549 ms
[bump] bump-render=high-poly: FPS: 1659 FrameTime: 0.603 ms
[bump] bump-render=normals: FPS: 1830 FrameTime: 0.547 ms
[bump] bump-render=height: FPS: 2319 FrameTime: 0.431 ms
[effect2d] kernel=0,1,0;1,-4,1;0,1,0;: FPS: 1933 FrameTime: 0.517 ms
[effect2d] kernel=1,1,1,1,1;1,1,1,1,1;1,1,1,1,1;: FPS: 1717 FrameTime: 0.583 ms
[pulsar] light=false:quads=5:texture=false: FPS: 2168 FrameTime: 0.461 ms
[desktop] blur-radius=5:effect=blur:passes=1:separable=true:windows=4: FPS: 1154 FrameTime: 0.867 ms
[desktop] effect=shadow:windows=4: FPS: 2191 FrameTime: 0.456 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 711 FrameTime: 1.407 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=subdata: FPS: 1734 FrameTime: 0.577 ms
[buffer] columns=200:interleave=true:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 764 FrameTime: 1.310 ms
[ideas] speed=duration: FPS: 1107 FrameTime: 0.904 ms
[jellyfish] <default>: FPS: 1953 FrameTime: 0.512 ms
[terrain] <default>: FPS: 867 FrameTime: 1.154 ms
[shadow] <default>: FPS: 1840 FrameTime: 0.544 ms
[refract] <default>: FPS: 1071 FrameTime: 0.934 ms
[conditionals] fragment-steps=0:vertex-steps=0: FPS: 2438 FrameTime: 0.410 ms
[conditionals] fragment-steps=5:vertex-steps=0: FPS: 2325 FrameTime: 0.430 ms
[conditionals] fragment-steps=0:vertex-steps=5: FPS: 2388 FrameTime: 0.419 ms
[function] fragment-complexity=low:fragment-steps=5: FPS: 2413 FrameTime: 0.415 ms
[function] fragment-complexity=medium:fragment-steps=5: FPS: 2196 FrameTime: 0.456 ms
[loop] fragment-loop=false:fragment-steps=5:vertex-steps=5: FPS: 2384 FrameTime: 0.420 ms
[loop] fragment-steps=5:fragment-uniform=false:vertex-steps=5: FPS: 2392 FrameTime: 0.418 ms
[loop] fragment-steps=5:fragment-uniform=true:vertex-steps=5: FPS: 2331 FrameTime: 0.429 ms
=======================================================
                                  glmark2 Score: 1910 
=======================================================

Can't get null descriptor to test zink:

VK_ICD_FILENAMES=/usr/local/share/vulkan/icd.d/virtio_icd.aarch64.json vulkaninfo | grep -i nullDescriptor -n
MESA: error: Use of VkSurfacePresentModeCompatibilityKHR without a VkSurfacePresentModeKHR set. This is an application bug.
1179:	nullDescriptor      = false

[osy]

I believe Zink works on an older version of Ubuntu (before Mesa made nullDescriptor required)

[startergo]

@osy you may try:

--- a/ui/cocoa.m	2026-01-13 20:17:13
+++ b/ui/cocoa.m	2026-01-13 20:16:42
@@ -2328,6 +2328,8 @@ static void cocoa_gl_render(void)
     if (cocoa_gl_trace_enabled()) {
         glFlush();
         cocoa_gl_log_error("render end");
+    /* Clear dirty flag after render completes to ensure clean handoff */
+    gl_dirty = false;
     }
 }
 
@@ -2336,7 +2338,6 @@ static void cocoa_gl_refresh(DisplayChangeListener *dcl)
     cocoa_refresh(dcl);
 
     if (gl_dirty) {
-        gl_dirty = false;
 
         if (cocoa_gl_trace_enabled()) {
             cocoa_gl_log("refresh: triggering render dirty scanout=%d egl_surface=%d",
             

It should reduce the CPU usage.

[osy]

We don't use the Cocoa backend in UTM

[startergo]

I switched to kosmickrisp in UTM settings:

vkmark
=======================================================
    vkmark 2025.01
=======================================================
    Vendor ID:      0x106B
    Device ID:      0x64
    Device Name:    Virtio-GPU Venus (Apple M4 Pro)
    Driver Version: 105263203
    Device UUID:    e572dff82304700b856a555ac3a4558d
=======================================================
[vertex] device-local=true: FPS: 823 FrameTime: 1.215 ms
[vertex] device-local=false: FPS: 839 FrameTime: 1.192 ms
[texture] anisotropy=0: FPS: 840 FrameTime: 1.190 ms
[texture] anisotropy=16: FPS: 837 FrameTime: 1.195 ms
[shading] shading=gouraud: FPS: 874 FrameTime: 1.144 ms
[shading] shading=blinn-phong-inf: FPS: 860 FrameTime: 1.163 ms
[shading] shading=phong: FPS: 840 FrameTime: 1.190 ms
[shading] shading=cel: FPS: 799 FrameTime: 1.252 ms
[effect2d] kernel=edge: FPS: 752 FrameTime: 1.330 ms
[effect2d] kernel=blur: FPS: 767 FrameTime: 1.304 ms
[desktop] <default>: FPS: 858 FrameTime: 1.166 ms
[cube] <default>: FPS: 908 FrameTime: 1.101 ms
[clear] <default>: FPS: 685 FrameTime: 1.460 ms
=======================================================
                                   vkmark Score: 821
=======================================================

My guess is Device ID: 0x64 reflects kosmickrisp?

[osy]

Do you have benchmark results for GL and Vulkan running natively on macOS to compare with?

[startergo]

only GLMARK2

[osy]

Can you post that and also MoltenVK native?

[startergo]

glmark2 results macOS

export DISPLAY=:0
/Users/macbookpro/glmark2/build-x11/src/glmark2 --data-path /Users/macbookpro/glmark2/data
=======================================================
    glmark2 2023.01
=======================================================
    OpenGL Information
    GL_VENDOR:      Apple
    GL_RENDERER:    Apple M4 Pro
    GL_VERSION:     2.1 Metal - 89.4
    Surface Config: buf=32 r=8 g=8 b=8 a=8 depth=32 stencil=0 samples=0
    Surface Size:   800x600 windowed
=======================================================
[build] use-vbo=false: FPS: 10900 FrameTime: 0.092 ms
[build] use-vbo=true: FPS: 11730 FrameTime: 0.085 ms
[texture] texture-filter=nearest: FPS: 12827 FrameTime: 0.078 ms
[texture] texture-filter=linear: FPS: 13191 FrameTime: 0.076 ms
[texture] texture-filter=mipmap: FPS: 13235 FrameTime: 0.076 ms
[shading] shading=gouraud: FPS: 11085 FrameTime: 0.090 ms
[shading] shading=blinn-phong-inf: FPS: 7710 FrameTime: 0.130 ms
[shading] shading=phong: FPS: 10529 FrameTime: 0.095 ms
[shading] shading=cel: FPS: 11721 FrameTime: 0.085 ms
[bump] bump-render=high-poly: FPS: 9793 FrameTime: 0.102 ms
[bump] bump-render=normals: FPS: 12908 FrameTime: 0.077 ms
[bump] bump-render=height: FPS: 12095 FrameTime: 0.083 ms
[effect2d] kernel=0,1,0;1,-4,1;0,1,0;: FPS: 12891 FrameTime: 0.078 ms
[effect2d] kernel=1,1,1,1,1;1,1,1,1,1;1,1,1,1,1;: FPS: 12310 FrameTime: 0.081 ms
[pulsar] light=false:quads=5:texture=false: FPS: 15087 FrameTime: 0.066 ms
[desktop] blur-radius=5:effect=blur:passes=1:separable=true:windows=4: FPS: 2961 FrameTime: 0.338 ms
[desktop] effect=shadow:windows=4: FPS: 6552 FrameTime: 0.153 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 292 FrameTime: 3.434 ms
[buffer] columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=subdata: FPS: 237 FrameTime: 4.233 ms
[buffer] columns=200:interleave=true:update-dispersion=0.9:update-fraction=0.5:update-method=map: FPS: 511 FrameTime: 1.957 ms
[ideas] speed=duration: FPS: 5030 FrameTime: 0.199 ms
[jellyfish] <default>: FPS: 10579 FrameTime: 0.095 ms
[terrain] <default>: FPS: 1308 FrameTime: 0.765 ms
[shadow] <default>: FPS: 8927 FrameTime: 0.112 ms
[refract] <default>: FPS: 2216 FrameTime: 0.451 ms
[conditionals] fragment-steps=0:vertex-steps=0: FPS: 12106 FrameTime: 0.083 ms
[conditionals] fragment-steps=5:vertex-steps=0: FPS: 11732 FrameTime: 0.085 ms
[conditionals] fragment-steps=0:vertex-steps=5: FPS: 11726 FrameTime: 0.085 ms
[function] fragment-complexity=low:fragment-steps=5: FPS: 11559 FrameTime: 0.087 ms
[function] fragment-complexity=medium:fragment-steps=5: FPS: 12721 FrameTime: 0.079 ms
[loop] fragment-loop=false:fragment-steps=5:vertex-steps=5: FPS: 11743 FrameTime: 0.085 ms
[loop] fragment-steps=5:fragment-uniform=false:vertex-steps=5: FPS: 11377 FrameTime: 0.088 ms
[loop] fragment-steps=5:fragment-uniform=true:vertex-steps=5: FPS: 11526 FrameTime: 0.087 ms
=======================================================
                                  glmark2 Score: 9426 
=======================================================

[startergo]

Vkmark results macOS

macbookpro@macbookpros-MacBook-Pro vkmark % DYLD_LIBRARY_PATH="/opt/homebrew/lib:$DYLD_LIBRARY_PATH" \
./build-out/bin/vkmark \           
--winsys-dir build-out/lib/vkmark \
--present-mode immediate \
-b clear:duration=5 -b cube:duration=5 -b texture:duration=5 -b vertex:duration=5
[mvk-info] MoltenVK version 1.4.0, supporting Vulkan version 1.4.323.
	The following 145 Vulkan extensions are supported:
	VK_KHR_16bit_storage v1
	VK_KHR_8bit_storage v1
	VK_KHR_bind_memory2 v1
	VK_KHR_buffer_device_address v1
	VK_KHR_calibrated_timestamps v1
	VK_KHR_copy_commands2 v1
	VK_KHR_create_renderpass2 v1
	VK_KHR_dedicated_allocation v3
	VK_KHR_deferred_host_operations v4
	VK_KHR_depth_stencil_resolve v1
	VK_KHR_descriptor_update_template v1
	VK_KHR_device_group v4
	VK_KHR_device_group_creation v1
	VK_KHR_driver_properties v1
	VK_KHR_dynamic_rendering v1
	VK_KHR_dynamic_rendering_local_read v1
	VK_KHR_external_fence v1
	VK_KHR_external_fence_capabilities v1
	VK_KHR_external_memory v1
	VK_KHR_external_memory_capabilities v1
	VK_KHR_external_semaphore v1
	VK_KHR_external_semaphore_capabilities v1
	VK_KHR_format_feature_flags2 v2
	VK_KHR_fragment_shader_barycentric v1
	VK_KHR_get_memory_requirements2 v1
	VK_KHR_get_physical_device_properties2 v2
	VK_KHR_get_surface_capabilities2 v1
	VK_KHR_global_priority v1
	VK_KHR_image_format_list v1
	VK_KHR_imageless_framebuffer v1
	VK_KHR_incremental_present v2
	VK_KHR_index_type_uint8 v1
	VK_KHR_line_rasterization v1
	VK_KHR_load_store_op_none v1
	VK_KHR_maintenance1 v2
	VK_KHR_maintenance2 v1
	VK_KHR_maintenance3 v1
	VK_KHR_maintenance4 v2
	VK_KHR_maintenance5 v1
	VK_KHR_maintenance6 v1
	VK_KHR_maintenance7 v1
	VK_KHR_maintenance8 v1
	VK_KHR_map_memory2 v1
	VK_KHR_multiview v1
	VK_KHR_portability_subset v1
	VK_KHR_present_id v1
	VK_KHR_present_id2 v1
	VK_KHR_present_wait v1
	VK_KHR_present_wait2 v1
	VK_KHR_push_descriptor v2
	VK_KHR_relaxed_block_layout v1
	VK_KHR_robustness2 v1
	VK_KHR_sampler_mirror_clamp_to_edge v3
	VK_KHR_sampler_ycbcr_conversion v14
	VK_KHR_separate_depth_stencil_layouts v1
	VK_KHR_shader_draw_parameters v1
	VK_KHR_shader_expect_assume v1
	VK_KHR_shader_float_controls v4
	VK_KHR_shader_float_controls2 v1
	VK_KHR_shader_float16_int8 v1
	VK_KHR_shader_integer_dot_product v1
	VK_KHR_shader_maximal_reconvergence v1
	VK_KHR_shader_non_semantic_info v1
	VK_KHR_shader_quad_control v1
	VK_KHR_shader_relaxed_extended_instruction v1
	VK_KHR_shader_subgroup_extended_types v1
	VK_KHR_shader_subgroup_rotate v2
	VK_KHR_shader_subgroup_uniform_control_flow v1
	VK_KHR_shader_terminate_invocation v1
	VK_KHR_spirv_1_4 v1
	VK_KHR_storage_buffer_storage_class v1
	VK_KHR_surface v25
	VK_KHR_surface_protected_capabilities v1
	VK_KHR_swapchain v70
	VK_KHR_swapchain_mutable_format v1
	VK_KHR_synchronization2 v1
	VK_KHR_timeline_semaphore v2
	VK_KHR_uniform_buffer_standard_layout v1
	VK_KHR_variable_pointers v1
	VK_KHR_vertex_attribute_divisor v1
	VK_KHR_vulkan_memory_model v3
	VK_KHR_zero_initialize_workgroup_memory v1
	VK_EXT_4444_formats v1
	VK_EXT_buffer_device_address v2
	VK_EXT_calibrated_timestamps v2
	VK_EXT_debug_marker v4
	VK_EXT_debug_report v10
	VK_EXT_debug_utils v2
	VK_EXT_depth_clip_control v1
	VK_EXT_descriptor_indexing v2
	VK_EXT_extended_dynamic_state v1
	VK_EXT_extended_dynamic_state2 v1
	VK_EXT_extended_dynamic_state3 v2
	VK_EXT_external_memory_host v1
	VK_EXT_external_memory_metal v1
	VK_EXT_fragment_shader_interlock v1
	VK_EXT_global_priority v2
	VK_EXT_global_priority_query v1
	VK_EXT_hdr_metadata v3
	VK_EXT_headless_surface v1
	VK_EXT_host_image_copy v1
	VK_EXT_host_query_reset v1
	VK_EXT_image_2d_view_of_3d v1
	VK_EXT_image_robustness v1
	VK_EXT_index_type_uint8 v1
	VK_EXT_inline_uniform_block v1
	VK_EXT_layer_settings v2
	VK_EXT_line_rasterization v1
	VK_EXT_load_store_op_none v1
	VK_EXT_memory_budget v1
	VK_EXT_metal_objects v2
	VK_EXT_metal_surface v1
	VK_EXT_pipeline_creation_cache_control v3
	VK_EXT_pipeline_creation_feedback v1
	VK_EXT_pipeline_robustness v1
	VK_EXT_post_depth_coverage v1
	VK_EXT_private_data v1
	VK_EXT_robustness2 v1
	VK_EXT_sample_locations v1
	VK_EXT_scalar_block_layout v1
	VK_EXT_separate_stencil_usage v1
	VK_EXT_shader_atomic_float v1
	VK_EXT_shader_demote_to_helper_invocation v1
	VK_EXT_shader_stencil_export v1
	VK_EXT_shader_subgroup_ballot v1
	VK_EXT_shader_subgroup_vote v1
	VK_EXT_shader_viewport_index_layer v1
	VK_EXT_subgroup_size_control v2
	VK_EXT_surface_maintenance1 v1
	VK_EXT_swapchain_colorspace v5
	VK_EXT_swapchain_maintenance1 v1
	VK_EXT_texel_buffer_alignment v1
	VK_EXT_texture_compression_astc_hdr v1
	VK_EXT_tooling_info v1
	VK_EXT_vertex_attribute_divisor v3
	VK_AMD_gpu_shader_half_float v2
	VK_AMD_negative_viewport_height v1
	VK_AMD_shader_image_load_store_lod v1
	VK_AMD_shader_trinary_minmax v1
	VK_GOOGLE_display_timing v1
	VK_IMG_format_pvrtc v1
	VK_INTEL_shader_integer_functions2 v1
	VK_MVK_macos_surface v3
	VK_MVK_moltenvk v37
	VK_NV_fragment_shader_barycentric v1
[mvk-info] GPU device:
	model: Apple M4 Pro
	type: Integrated
	vendorID: 0x106b
	deviceID: 0xf060209
	pipelineCacheUUID: 823EF213-0F06-0209-0000-000100000000
	GPU memory available: 16384 MB
	GPU memory used: 0 MB
	Metal Shading Language 3.2
	supports the following GPU Features:
		GPU Family Metal 3
		GPU Family Apple 9
		GPU Family Mac 2
		Read-Write Texture Tier 2
[mvk-info] Created VkInstance for Vulkan version 1.0.323, as requested by app, with the following 2 Vulkan extensions enabled:
	VK_KHR_surface v25
	VK_EXT_metal_surface v1
[mvk-info] Destroyed VkPhysicalDevice for GPU Apple M4 Pro with 0 MB of GPU memory still allocated.
[mvk-info] Destroying VkInstance for Vulkan version 1.0.323 with 2 Vulkan extensions enabled.
MetalNativeSystem: Created macOS window 800x600
[mvk-info] MoltenVK version 1.4.0, supporting Vulkan version 1.4.323.
	The following 145 Vulkan extensions are supported:
	VK_KHR_16bit_storage v1
	VK_KHR_8bit_storage v1
	VK_KHR_bind_memory2 v1
	VK_KHR_buffer_device_address v1
	VK_KHR_calibrated_timestamps v1
	VK_KHR_copy_commands2 v1
	VK_KHR_create_renderpass2 v1
	VK_KHR_dedicated_allocation v3
	VK_KHR_deferred_host_operations v4
	VK_KHR_depth_stencil_resolve v1
	VK_KHR_descriptor_update_template v1
	VK_KHR_device_group v4
	VK_KHR_device_group_creation v1
	VK_KHR_driver_properties v1
	VK_KHR_dynamic_rendering v1
	VK_KHR_dynamic_rendering_local_read v1
	VK_KHR_external_fence v1
	VK_KHR_external_fence_capabilities v1
	VK_KHR_external_memory v1
	VK_KHR_external_memory_capabilities v1
	VK_KHR_external_semaphore v1
	VK_KHR_external_semaphore_capabilities v1
	VK_KHR_format_feature_flags2 v2
	VK_KHR_fragment_shader_barycentric v1
	VK_KHR_get_memory_requirements2 v1
	VK_KHR_get_physical_device_properties2 v2
	VK_KHR_get_surface_capabilities2 v1
	VK_KHR_global_priority v1
	VK_KHR_image_format_list v1
	VK_KHR_imageless_framebuffer v1
	VK_KHR_incremental_present v2
	VK_KHR_index_type_uint8 v1
	VK_KHR_line_rasterization v1
	VK_KHR_load_store_op_none v1
	VK_KHR_maintenance1 v2
	VK_KHR_maintenance2 v1
	VK_KHR_maintenance3 v1
	VK_KHR_maintenance4 v2
	VK_KHR_maintenance5 v1
	VK_KHR_maintenance6 v1
	VK_KHR_maintenance7 v1
	VK_KHR_maintenance8 v1
	VK_KHR_map_memory2 v1
	VK_KHR_multiview v1
	VK_KHR_portability_subset v1
	VK_KHR_present_id v1
	VK_KHR_present_id2 v1
	VK_KHR_present_wait v1
	VK_KHR_present_wait2 v1
	VK_KHR_push_descriptor v2
	VK_KHR_relaxed_block_layout v1
	VK_KHR_robustness2 v1
	VK_KHR_sampler_mirror_clamp_to_edge v3
	VK_KHR_sampler_ycbcr_conversion v14
	VK_KHR_separate_depth_stencil_layouts v1
	VK_KHR_shader_draw_parameters v1
	VK_KHR_shader_expect_assume v1
	VK_KHR_shader_float_controls v4
	VK_KHR_shader_float_controls2 v1
	VK_KHR_shader_float16_int8 v1
	VK_KHR_shader_integer_dot_product v1
	VK_KHR_shader_maximal_reconvergence v1
	VK_KHR_shader_non_semantic_info v1
	VK_KHR_shader_quad_control v1
	VK_KHR_shader_relaxed_extended_instruction v1
	VK_KHR_shader_subgroup_extended_types v1
	VK_KHR_shader_subgroup_rotate v2
	VK_KHR_shader_subgroup_uniform_control_flow v1
	VK_KHR_shader_terminate_invocation v1
	VK_KHR_spirv_1_4 v1
	VK_KHR_storage_buffer_storage_class v1
	VK_KHR_surface v25
	VK_KHR_surface_protected_capabilities v1
	VK_KHR_swapchain v70
	VK_KHR_swapchain_mutable_format v1
	VK_KHR_synchronization2 v1
	VK_KHR_timeline_semaphore v2
	VK_KHR_uniform_buffer_standard_layout v1
	VK_KHR_variable_pointers v1
	VK_KHR_vertex_attribute_divisor v1
	VK_KHR_vulkan_memory_model v3
	VK_KHR_zero_initialize_workgroup_memory v1
	VK_EXT_4444_formats v1
	VK_EXT_buffer_device_address v2
	VK_EXT_calibrated_timestamps v2
	VK_EXT_debug_marker v4
	VK_EXT_debug_report v10
	VK_EXT_debug_utils v2
	VK_EXT_depth_clip_control v1
	VK_EXT_descriptor_indexing v2
	VK_EXT_extended_dynamic_state v1
	VK_EXT_extended_dynamic_state2 v1
	VK_EXT_extended_dynamic_state3 v2
	VK_EXT_external_memory_host v1
	VK_EXT_external_memory_metal v1
	VK_EXT_fragment_shader_interlock v1
	VK_EXT_global_priority v2
	VK_EXT_global_priority_query v1
	VK_EXT_hdr_metadata v3
	VK_EXT_headless_surface v1
	VK_EXT_host_image_copy v1
	VK_EXT_host_query_reset v1
	VK_EXT_image_2d_view_of_3d v1
	VK_EXT_image_robustness v1
	VK_EXT_index_type_uint8 v1
	VK_EXT_inline_uniform_block v1
	VK_EXT_layer_settings v2
	VK_EXT_line_rasterization v1
	VK_EXT_load_store_op_none v1
	VK_EXT_memory_budget v1
	VK_EXT_metal_objects v2
	VK_EXT_metal_surface v1
	VK_EXT_pipeline_creation_cache_control v3
	VK_EXT_pipeline_creation_feedback v1
	VK_EXT_pipeline_robustness v1
	VK_EXT_post_depth_coverage v1
	VK_EXT_private_data v1
	VK_EXT_robustness2 v1
	VK_EXT_sample_locations v1
	VK_EXT_scalar_block_layout v1
	VK_EXT_separate_stencil_usage v1
	VK_EXT_shader_atomic_float v1
	VK_EXT_shader_demote_to_helper_invocation v1
	VK_EXT_shader_stencil_export v1
	VK_EXT_shader_subgroup_ballot v1
	VK_EXT_shader_subgroup_vote v1
	VK_EXT_shader_viewport_index_layer v1
	VK_EXT_subgroup_size_control v2
	VK_EXT_surface_maintenance1 v1
	VK_EXT_swapchain_colorspace v5
	VK_EXT_swapchain_maintenance1 v1
	VK_EXT_texel_buffer_alignment v1
	VK_EXT_texture_compression_astc_hdr v1
	VK_EXT_tooling_info v1
	VK_EXT_vertex_attribute_divisor v3
	VK_AMD_gpu_shader_half_float v2
	VK_AMD_negative_viewport_height v1
	VK_AMD_shader_image_load_store_lod v1
	VK_AMD_shader_trinary_minmax v1
	VK_GOOGLE_display_timing v1
	VK_IMG_format_pvrtc v1
	VK_INTEL_shader_integer_functions2 v1
	VK_MVK_macos_surface v3
	VK_MVK_moltenvk v37
	VK_NV_fragment_shader_barycentric v1
[mvk-info] GPU device:
	model: Apple M4 Pro
	type: Integrated
	vendorID: 0x106b
	deviceID: 0xf060209
	pipelineCacheUUID: 823EF213-0F06-0209-0000-000100000000
	GPU memory available: 16384 MB
	GPU memory used: 0 MB
	Metal Shading Language 3.2
	supports the following GPU Features:
		GPU Family Metal 3
		GPU Family Apple 9
		GPU Family Mac 2
		Read-Write Texture Tier 2
[mvk-info] Created VkInstance for Vulkan version 1.0.323, as requested by app, with the following 2 Vulkan extensions enabled:
	VK_KHR_surface v25
	VK_EXT_metal_surface v1
[mvk-info] Vulkan semaphores using MTLEvent.
[mvk-info] Descriptor sets binding resources using Metal3 argument buffers.
[mvk-info] Created VkDevice to run on GPU Apple M4 Pro with the following 1 Vulkan extensions enabled:
	VK_KHR_swapchain v70
MetalNativeSystem: Created Metal surface
[mvk-info] Created 3 swapchain images with size (800, 600) and contents scale 1.0 in layer (null) (0x600000baccc0) on screen Built-in Retina Display.
=======================================================
    vkmark 2025.01
=======================================================
    Vendor ID:      0x106B
    Device ID:      0xF060209
    Device Name:    Apple M4 Pro
    Driver Version: 10400
    Device UUID:    823ef2130f0602090000000100000000
=======================================================
[clear] duration=5: FPS: 1260 FrameTime: 0.794 ms
[cube] duration=5: FPS: 1240 FrameTime: 0.806 ms
[texture] duration=5: FPS: 1220 FrameTime: 0.820 ms
[vertex] duration=5: FPS: 1170 FrameTime: 0.855 ms
=======================================================
                                   vkmark Score: 1222
=======================================================
[mvk-info] Destroyed VkDevice on GPU Apple M4 Pro with 1 Vulkan extensions enabled.
[mvk-info] Destroyed VkPhysicalDevice for GPU Apple M4 Pro with 4 MB of GPU memory still allocated.
[mvk-info] Destroying VkInstance for Vulkan version 1.0.323 with 2 Vulkan extensions enabled.

[startergo]

So for the Firefox (likely Chrome too) in gl=core mode:
Firefox on Wayland forces EGL compositor when using gl=core mode. Firefox creates EGL context requesting Desktop GL Core Profile.
Virglrenderer receives this and uses OpenGL.framework (CGL) for rendering, but eglSwapBuffers() doesn't present the rendered frame to the display.

[osy]

Can you explain why you think this is the case? eglSwapBuffers isn't called from virglrenderer or QEMU (ie I don't see any reference to it) Also, I tried launching FireFox on Ubuntu 25.10 fresh install and it works fine with gl=core

[startergo]

Firefox calls eglSwapBuffers() on its EGL surface (but fails there), Mesa DRI layer through virgl driver sends commands to host via virtio-gpu. No actual swap buffers happens on the host. So that is not the problem. I was trying Firefox on immutable Fedora and Debian. I can't get any UI when I start Firefox. Are you sure though the Firefox is not rendering through the software? Can you run aquarium webgl test and get real hardware acceleration?
I remember once I could get chromium through Flatpack and it worked fine.

[osy]

I actually did reproduce the firefox issue on Ubuntu 24.04 so maybe it's a Mesa issue? I'm not sure. Do you see the same issue with QEMU + cocoa + your patches?

[startergo]

yes. same issue. It is always present in gl=core

[startergo]

How is this even getting OpenGL ES in gl=core mode?

It does not appear your computer supports WebGL.
Click here for more information.

Status: Could not create a WebGL context, VENDOR = 0x1af4, DEVICE = 0x1050, GL_VENDOR = Google Inc. (Mesa), GL_RENDERER = ANGLE (Mesa, virgl (Apple M4 Pro), OpenGL ES 3.0 Mesa 25.3.3 (git-a9c6420d8c)), GL_VERSION = 25.3.3, Sandboxed = no, Optimus = no, AMD switchable = no, Reset notification strategy = 0x0000, ErrorMessage = BindToCurrentSequence failed: .

[osy]

Guest ES != Host ES

Basically you have the pipeline: Guest API (ES, Core, etc) -> Mesa -> Gallium (VirGL driver) -> VirtIO -> Host (virglrenderer) -> Host Graphics API

What's being passed from guest to host are NOT GL commands but Gallium internal commands which gets translated back to host graphics APIs (can be ES, Core, etc)

[startergo]

Found that Desktop GL Core Profile uses GL_RGBA format for BGRA textures (while GLES uses GL_BGRA).

[patham9]

@osy @startergo Thank you so much, I owe you a beer! UTM v5.0.1 is AMAZING, running robotics Gazebo simulations fluidly now in my VM at real-time speed. UTM now (on MacOS 15.7.3) delivers better guest graphics acceleration than VMWare Fusion and Parallels for my M4 Mac for both Ubuntu 22 and Ubuntu 24 guests.

[startergo]

delivers better guest graphics acceleration than VMWare Fusion and Parallels

How did you test? I personally think Parallels performance is outstanding.

[patham9]

@startergo My test involved only Gazebo, which I could only make run with Parallels in flickering, crashing, or software-rendering modes.

[startergo]

@osy Upon further investigation for the rendering issues found in gnome for gl=core mode (particularly UEFI os installation mode or os updates mode), where the text is transparent and does not work properly.
In OpenGL Core Profile, GL_ALPHA format was removed:

OpenGL 3.2 core profile doesn't support either texture alpha format or texture swizzling. A possible workaround is to expand the source data to RGBA8"

For texture alpha, you must use texture red format (GL_R8) with texture swizzle:

glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_SWIZZLE_A, GL_RED);

https://www.g-truc.net/post-0734.html

The problem is that A8_UNORM format (alpha-only textures used for text glyphs) is marked as view_class_unsupported in virglrenderer:

{ VIRGL_FORMAT_A8_UNORM, GL_R8, GL_RED, GL_UNSIGNED_BYTE, OOOR_SWIZZLE, view_class_unsupported }

[osy]

Interesting, if you get it working in QEMU, send me the patch and I'll integrate it. My focus is on Vulkan and hopefully we won't need any GL once that's mature and we have Zink

[andyleiserson]

It seems like this is actually related to how virglrenderer handles FS_COLOR0_WRITES_ALL_CBUFS combined with dual-source blending -- it puts an explicit layout(location=1) in the GLSL for the second blend source, which is not correct. The change in https://gitlab.freedesktop.org/virgl/virglrenderer/-/merge_requests/1668 fixes the problem for me.

[startergo]

So this fixed the transparency in gl=core mode on UEFI selector screen? By any chance did it also fix firefox? Can you render firefox in gl=core?

[andyleiserson]

Sorry, no, I had transparent text in X11, which sounded like a similar problem, so I started investigating with the A8/R8 format issues. But given that the root cause was elsewhere and I'm using ANGLE/Metal, it probably isn't actually the same problem. Sorry if I gave false hope.

[K0bin]

For some reason, v5 seems to break Linux VMs. I'm trying to create a virtualized VM running ARM64 Fedora using Qemu. And upon first start, the first thing I notice is that Grub is insanely slow. Every key press is delays by like 10 seconds and I can see it render character by character. After I've somehow made it past that, into what should be the Fedora Live Iso installer, all I'm getting is a black screen. I've waited a minute and nothing happened in that time so I gave up.

Doing the same thing in UTM 4.7.5 works fine. Grub actually performs as expected and is usable and after that it's actually booting inside the Fedora Live Installer as expected.

[startergo]

I am running Arch and Debian and there are no issues. Perhaps it is related to Fedora?

[K0bin]

Interesting, Ubuntu does actually work. Grub is still weirdly slow but after that it works fine.

[K0bin]

Okay now that it actually works I'm unfortunately I'm seeing severe sync issues in vkcube.

[startergo]

Did you patch and reinstall mesa?
https://gist.github.com/osy/a8f705050eed1c8421ad1a0855a8faa9

[K0bin]

No. I guess that makes it impressive that it worked at all.

I assume that's also one of the reasons why 5.0.1 is still in beta?

[startergo]

@osy I have assembled an endeavouros live ISO with KDE/Plasma Wayland, which works out of the box in Parallels, but in UTM the login screen is black in Angle-Metal renderer, partially ok ( the background is visible but the pop-up Welcome screen is black) in Angle-OpenGl and it works fine in Apple core OpenGL. What gives?
https://github.com/startergo/EndeavourOS-ISO-arm64/releases

[osy]

There's a couple of fixes in the pipeline.

[startergo]

I tried the latest artifacts and the behavior is the same.

[PowerRangers1993]

how do you upgrade the VM the UTM?