Mesa 24.0.4 Release Notes / 2024-03-27

Mesa 24.0.4 is a bug fix release which fixes bugs found since the 24.0.3 release.

Mesa 24.0.4 implements the OpenGL 4.6 API, but the version reported by glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) / glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used. Some drivers don’t support all the features required in OpenGL 4.6. OpenGL 4.6 is only available if requested at context creation. Compatibility contexts may report a lower version depending on each driver.

Mesa 24.0.4 implements the Vulkan 1.3 API, but the version reported by the apiVersion property of the VkPhysicalDeviceProperties struct depends on the particular driver being used.

SHA256 checksum

90febd30a098cbcd97ff62ecc3dcf5c93d76f7fa314de944cfce81951ba745f0  mesa-24.0.4.tar.xz

New features

  • None

Bug fixes

  • nvk: dota 2 crashes after ~5 seconds in game

  • VAAPI: Incorrect HEVC block size reported with radeonsi

  • radv: WWE 2K24 has very quirky DCC issues on RDNA2

  • RUSTICL creating a shared reference to mutable static is discouraged and will become a hard error

  • KiCAD 3D Viewer - rounded pads rendered incorrectly (texture mapping or stencil test error)

  • OpenSCAD rendering incorrect and inconsistent on radeonsi

  • [radv] Half-Life Alyx renders solid black for reflective surfaces

  • [RX 7900 XTX] Helldivers 2 cause GPU reset

  • radeon: Crash in radeon_bo_can_reclaim_slab

  • RV530 renders improperly at non 4:3 resolutions.

  • anv: new cooperative matrix failures with CTS 1.3.8.0

  • `[gfxhub0] no-retry page fault` triggered by `AMD_TEST=testdmaperf` on gfx90c APU

Changes

Boris Brezillon (1):

  • panvk: Disable global offset on varying and non-VS attribute descriptors

Caio Oliveira (2):

  • intel/brw: Use helper to create accumulator register

  • intel/brw: Fix validation of accumulator register

Charlie Turner (1):

  • {vulkan,radv,anv}/video: fix issue in H264 scaling lists derivation

Corentin Noël (2):

  • st_pbo/compute: Use the correct structure type when allocating a specialized key

  • zink: Make sure to initialize all the fields of VkMemoryBarrier

Dave Airlie (1):

  • radv/video: fix h265 decode with unaligned w/h

David Rosca (1):

  • radv/video: Set maxActiveReferencePictures to 16 for H264/5

Eric Engestrom (5):

  • docs: add sha256sum for 24.0.3

  • .pick_status.json: Update to 9b6d6c1d2d0c8a517e974abbf7b75a47a607f6ec

  • .pick_status.json: Update to eac703f69128d5aa6879c9becbad627ce08a7920

  • .pick_status.json: Update to 912e203a534be8b70b3ef8bf00294e9c962e385a

  • .pick_status.json: Update to c0875d21563257442fd91aab5740248b0fd96a5c

Faith Ekstrand (2):

  • nir/builder: Correctly handle decl_reg or undef as the first instruction

  • nir/gather_types: Support unstructured control-flow

Francisco Jerez (1):

  • intel/eu/xe2+: Translate brw_reg fields in REG_SIZE units to physical 512b GRF units during codegen.

Friedrich Vock (2):

  • radv: Only enable SEs that the device reports

  • radeonsi: Only enable SEs that the device reports

Gert Wollny (2):

  • nir-to-spirv: Cast SSBO input pointer when needed

  • nir_to_spirv: Allow LOD for external images

Hyunjun Ko (1):

  • anv/video: fix scan order for scaling lists on H265 decoding.

Iván Briano (2):

  • compiler/types: fix serialization of cooperative matrix

  • intel/cmat: fix stride calculation in cmat load/store

Jordan Justen (1):

  • intel/compiler/fs: Restore SIMD32 restriction for ray_queries on Xe2

Karol Herbst (2):

  • rusticl/kernel: assign sampler locations before DCEing variables

  • nouveau: call glsl_type_singleton_init_or_ref earlier

Kenneth Graunke (1):

  • intel/brw: Fix opt_split_sends() to allow for FIXED_GRF send sources

Konstantin Seurer (1):

  • zink: Handle aoa derefs of images

Lionel Landwerlin (6):

  • intel/fs: fixup sampler header message

  • anv: return unsupported for FSR images on Gfx12.0

  • anv: ignore descriptor alignment for inline uniforms

  • blorp: handle a few allocation failure cases

  • anv: fix block pool allocation failure

  • anv: fix bitfield checks in gfx runtime flushing

Lucas Stach (1):

  • etnaviv: fix fixpoint conversion of negative values

Marek Olšák (8):

  • amd/registers: add correct gfx11.x enums for BINNING_MODE

  • radeonsi: disable binning correctly on gfx11.5

  • radeonsi/gfx11: fix programming of PA_SC_BINNER_CNTL_1.MAX_ALLOC_COUNT

  • radeonsi/gfx10.3: add a GPU hang workaround for legacy tess+GS

  • radeonsi/gfx11: add missing DCC_RD_POLICY setting

  • ac/llvm: fix SSBO bounds checking by using raw instead of struct opcodes

  • radeonsi: fix the DMA compute shader

  • r300: port scanout pitch alignment from the DDX to fix DRI3

Mary Guillemard (1):

  • nvk: Always copy conditional rendering value before compare

Matthew Waters (1):

  • teximage: allow glCopyTex{Sub}Image[123]D into R/RG textures with OpenGL ES 2.0

Mike Blumenkrantz (13):

  • zink: destroy batch states after copy context

  • mesa: force rendertarget usage on required-renderable formats

  • zink: try getting sparse page size again without storage bit on fail

  • zink: set the sparse format usage flags directly based on queried props

  • zink: rename optimal_key in update_gfx_program_optimal()

  • zink: use the sanitized key in update_gfx_program_optimal()

  • zink: always sync and replace separable progs even with ZINK_DEBUG=noopt

  • zink: add even more strict checks for separate shader usage

  • glx: only print zink failure-to-load messages if explicitly requested

  • zink: iterate all the modes when doing separate shader fixups

  • zink: do io fixup on patch variables too

  • zink: defer present barrier to flush if a clear is pending

  • zink: clamp swapchain renderarea instead of asserting

Patrick Lerda (1):

  • ac/llvm,radeonsi: fix memory leaks triggered by ac_nir_translate() errors

Paulo Zanoni (1):

  • anv: don’t leak device->vma_samplers

Philipp Zabel (1):

  • rusticl: work around reference-to-mutable-static warnings

Pierre-Eric Pelloux-Prayer (2):

  • winsys/radeon: pass priv instead NULL to radeon_bo_can_reclaim

  • radeonsi: preserve alpha if needed in kill_ps_outputs_cb

Rhys Perry (4):

  • aco: don’t reuse misaligned attribute destination VGPRs in VS prologs

  • radv: use dual_color_blend_by_location with Half-Life Alyx

  • aco/cssa: reset equal_anc_out if merging fails

  • aco/gfx11: fix scratch ST mode assembly

Ruijing Dong (3):

  • radeonsi/vcn: add enc surface alignment caps

  • frontends/va: add surface alignment attribute

  • radeonsi/vcn: update to use correct padding size.

Samuel Pitoiset (7):

  • ac/nir: fix exporting NGG streamout outputs with implicit PrimId from VS/TES

  • radv: disable binning correctly on GFX11.5

  • radv: fix programming of PA_SC_BINNER_CNTL_1.MAX_ALLOC_COUNT on GFX11

  • radv: fix occlusion queries with MSAA and no attachments

  • radv: add radv_force_pstate_peak_gfx11_dgpu and enable it for Helldivers 2

  • radv: add a workaround for null IBO on GFX6

  • radv: invalidate L2 metadata for VK_ACCESS_2_MEMORY_READ_BIT

Yusuf Khan (1):

  • nvk: fix valve segfault from setting a descriptor set from NULL