Repository navigation
build: update graphics stack and QEMU for Neptune fixes - #7963
Merged
Merged
Conversation
…talls Pick up QEMU patches needed for the Neptune graphics work: - hvf: use the Hypervisor.framework in-kernel GICv3 on macOS 15+ so SGIs, timer PPIs and GIC register accesses no longer each cost a VM exit, which is expensive for guests whose job systems wake threads across CPUs. On by default (kernel-irqchip=on); the hypervisor GIC has no ITS, so the board exposes a GICv2m MSI frame instead. Its state is saved and restored, so suspend and migration keep working. - edid: carry the host refresh rate into the standard timings and range limits. Windows built every mode at 60 Hz, and a host display past 125 Hz produced a self-contradicting EDID that made Windows drop the adapter and fail D3D device creation. Timings the base block cannot express get a DisplayID 2.0 section. - virtio-gpu: park a create whose id still has a deferred unref instead of rejecting it; the Windows viogpu3d KMD reuses ids this way and came up with no display. - virtio-gpu: initialise the per-command deferred/suspended flags in every handle_ctrl loop. Leftover heap bytes left commands unanswered, hanging Windows on shutdown and stalling 3DMark on Linux. Assisted-by: Claude:claude-fable-5-1 Assisted-by: Claude:claude-opus-5 Assisted-by: Claude:claude-opus-5-5
Pick up the latest fixes across the accelerated graphics stack: - virglrenderer: Neptune ring and wait fixes (park an idle ring until its doorbell, a heartbeat in the ring status word, monotonic-clock timed waits), D3D11 query feedback fixes, a decoder lookup cache sized for a frame's working set, resource updates up to 2 GB, and neptune-protocol wire version 2. Guest shmems/dmabufs can now back DXMT DYNAMIC buffer renames and large D3D12 heap blobs. - MoltenVK: query copy-results and availability fixes, a NULL multisample-state guard for mesh pipelines, and a temp-buffer pool lifetime fix. - DXMT: complete occlusion queries that end in a draw-less chunk, UpdateSubresource into STAGING resources, and embedder-provided shared memory for DYNAMIC buffer renames. - D3DMetal: accurate format-support and multisample-quality answers that refuse what D3DMetal cannot create, DXIL subroutine inlining for every stage, and a pinned shared-texture POD ABI. Assisted-by: Claude:claude-opus-5-5
…pple GL Apple's native OpenGL 4.1 core profile rejected shaders that required GL_ARB_draw_instanced for gl_InstanceID (core since GLSL 1.40), so those draws silently did nothing. Integer vertex attributes were also declared as vec4 on Apple's native GL, so the guest saw the bits of their float conversion. Resolves #7944 Assisted-by: Claude:claude-opus-5-5
osy
force-pushed
the
dev/update-graphics
branch
4 times, most recently
from
October 15, 2026 16:05
7ceea08 to
fec4c10
Compare
brunocastello
added a commit
to brunocastello/UTM
that referenced
this pull request
Oct 15, 2026
Assisted-by: Claude:claude-opus-5-5
This was referenced Oct 18, 2026
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update the accelerated graphics stack and pick up the QEMU fixes needed for the Neptune (Windows GPU acceleration) work.
QEMU (
patches/qemu-10.0.12-utm.patch)kernel-irqchip=on); the hypervisor GIC has no ITS, so the board exposes a GICv2m MSI frame instead. The in-kernel GIC state is saved and restored, so suspend and migration keep working.handle_ctrlloop; leftover heap bytes left commands unanswered, hanging Windows on shutdown and stalling 3DMark on Linux.Dependencies (
patches/sources)gl_InstanceIDand integer vertex attributes on Apple's native OpenGL.UpdateSubresourceinto STAGING resources, shared memory for DYNAMIC buffer renames.Resolves #7944
Resolves #7820
Testing: Tested by a human on macOS 15.6, Mac mini (M4 Pro) with 3DMark Fire Strike, 3DMark Time Spy, GravityMark and Cyberpunk 2077.
The author acknowledges that this change has been tested and/or reviewed by a human in accordance with UTM's AI contribution guidelines.