Mesa 22.3.7 Release Notes / 2023-03-08

Mesa 22.3.7 is a bug fix release which fixes bugs found since the 22.3.6 release.

Mesa 22.3.7 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 22.3.7 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

894ce2f4a1c2e76177cdd2284620192d0da3066b243eec2fbb1d7cf37f13042c  mesa-22.3.7.tar.xz

New features

  • None

Bug fixes

  • radv: A Plague Tale: Requiem black “flash” on 7900XTX

  • 7900 XTX: Graphical corruption / artifacts in Cyberpunk

  • radv: CmdCopyQueryPoolResults broken for VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT with queryCount > 1

  • rusticl over llvmpipe + ffmpeg’s Opencl filter = error -51

  • System freeze when playing some h264 videos with VA-API on Rembrandt

  • radv: GTA IV graphical artifacts on 7900XTX

  • radv: Resident Evil Revelations 2 artifacts on 7900XTX with DCC

  • radv: Prototype 2 black textures on RDNA 3 when DCC is enabled

  • hasvk: Black pixels with 8xMSAA and fast clears on Intel(R) HD Graphics 4400 (HSW GT2)

  • ACO: dEQP-VK.binding_model.descriptor_buffer.multiple.graphics_geom_buffers1_sets3_imm_samplers hangs on NAVI10

  • Build failures with recent lld

  • [gen9atom] Vulkan tests cause gpu hang: dEQP-VK.memory_model.*

  • mesa: “Fragmented” dynamic lights in IronWail with `r_fsaamode 1` on

  • GL_SHADER_BINARY_FORMAT_SPIR_V is not added to the list of GL_SHADER_BINARY_FORMATS even if GL_ARB_gl_spirv is supported.

  • [ANV/DG2] Vertex explosion in nvpro-samples/vk_raytracing_tutorial_KHR/ray_tracing_gltf

  • CUEtools FLACCL hit assert in rusticl

  • Assertion Failed on Intel HD 5500 with Linux / Mesa 22.3.1 / OpenGL

Changes

Caio Oliveira (2):

  • iris, crocus: Align workaround address to 32B

  • gallivm: Fix handling of nir_intrinsic_scoped_barrier

Dave Airlie (1):

  • llvmpipe: fix compute address bits to return native pointer size.

Dylan Baker (1):

  • util: rzalloc and free hash_table_u64

Eric Engestrom (6):

  • docs/relnotes: add sha256sum for 22.3.6

  • .pick_status.json: Update to 41ae2d0725b3ecd2018a6105981a6db7b87c3e79

  • radv: split linker script for android since it requires different symbols

  • glapi/meson: drop duplicate line in deps

  • .pick_status.json: Mark 646cff13bca8a92b846984d782ef00e57d34d7a1 as denominated

  • .pick_status.json: Mark bb796f70a144e9c88ab46e6fb08be5895ca7a689 as denominated

Friedrich Vock (1):

  • mesa: Report GL_SHADER_BINARY_FORMAT_SPIR_V as supported

Gert Wollny (1):

  • r600/sfn: Fix atomic lowering

Iván Briano (3):

  • vulkan: track the right value on CmdSetColorWriteMasks

  • anv: fix testing for dynamic color blend bits

  • anv: stop tracking color blend state in the pipeline

Karol Herbst (2):

  • nir/deref: don’t replace casts with deref_struct if we’d lose the stride

  • rusticl/kernel: Images arg sizes also have to match the host pointer size

Lionel Landwerlin (7):

  • genxml: Fix STATE_BASE_ADDRESS::BindlessSurfaceStateSize field size

  • genxml: fix border color offset field on Gfx12+

  • anv: fixup condition for Wa_14016118574

  • anv: remove pre hasvk split assert

  • anv: pull Wa_14016118574 out of some loop not changing state

  • anv: fix incorrect parameter

  • nir: fix nir_ishl_imm

Marek Olšák (6):

  • amd: add missing gfx11 register definitions

  • radeonsi/gfx11: fix the CU_EN clear mask for RSRC4_GS

  • radeonsi/gfx11: don’t set non-existent VGT_STRMOUT_BUFFER_CONFIG

  • radeonsi/gfx11: set CB_COLORi_INFO.MAX_COMP_FRAG on GFX1103_R2

  • radeonsi: disable Smart Access Memory because CPU access has large overhead

  • radeonsi: don’t merge SET_* packets that have a different index in si_pm4_state

Mike Blumenkrantz (9):

  • zink: fix shader read access removal for barrier generation

  • zink: avoid adding ubo/ssbo bindings multiple times for different bitsizes

  • zink: remove suspended queries from list before resuming

  • zink: disable queries for clear_texture()

  • zink: fix zink_resource_access_is_write()

  • zink: unbind fb on context destroy

  • zink: only add deferred barrier on fb unbind when layout needs to change

  • zink: fix descriptor update flagging on null ssbo set

  • zink: propagate valid_buffer_range when replacing buffer storage

Pierre-Eric Pelloux-Prayer (1):

  • radeonsi: fix incorrect vgpr indices in the ps_prolog

Rhys Perry (2):

  • aco: fix pathological case in LdsDirectVALUHazard

  • aco: always update orig_names in get_reg_phi()

Ruijing Dong (1):

  • raseonsi/vcn: fix a h264 decoding issue

Sam Edwards (1):

  • nouveau: Fix null dereference in nouveau_pushbuf_destroy

Samuel Pitoiset (9):

  • radv: fix flushing non-coherent images inside secondaries on GFX9+

  • radv: fix flushing non-coherent images in EndCommandBuffer()

  • radv: fix draw calls with 0-sized index buffers and robustness on NAVI10

  • amd,ac/rgp: fix SQTT memory types

  • radv: fix DCC decompress on GFX11

  • radv: fix incorrect stride for primitives generated query with GDS

  • radv: fix border color swizzle for stencil-only format on GFX9+

  • radv: fix defining RADV_USE_WSI_PLATFORM

  • radv: disable DCC with signedness reinterpretation on GFX11

Sil Vilerino (3):

  • d3d12: Fix VP9 Decode - Checking 0xFF instead of 0x7F for invalid frame_ref[i].Index7Bits

  • frontend/va: Keep track of some VP9 previous frame data for current frame use_prev_in_find_mvs_refs

  • d3d12: VP9 Decode - Fix use_prev_in_find_mvs_refs calculation

Sviatoslav Peleshko (3):

  • iris: Avoid creating uncompressed view with unaligned tile offsets on BDW

  • anv: Handle all fields in VkAccelerationStructureBuildRangeInfoKHR

  • anv: Move WA MEDIA_VFE_STATE after stalling PIPE_CONTROL

Tatsuyuki Ishi (2):

  • radeonsi: SDMA v4 size field is size - 1

  • radv: SDMA v4 size field is size - 1

Väinö Mäkelä (2):

  • hasvk: Mark VK_IMAGE_LAYOUT_ATTACHMENT_OPTIMAL as stencil write optimal

  • hasvk: Disable non-zero fast clears for 8xMSAA images

Yiwei Zhang (1):

  • venus: fix VK_EXT_image_view_min_lod feature query

Yogesh Mohan Marimuthu (2):

  • wsi/display: check alloc failure in wsi_display_alloc_connector()

  • ac/surface: only adjust pitch if surf_pitch was modified