Mesa 20.3.5 Release Notes / 2021-03-24

Mesa 20.3.5 is a bug fix release which fixes bugs found since the 20.3.4 release.

Mesa 20.3.5 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 20.3.5 implements the Vulkan 1.2 API, but the version reported by the apiVersion property of the VkPhysicalDeviceProperties struct depends on the particular driver being used.

SHA256 checksum

909a72df63dfa3c0844ccf44a26ac028d148297ed333ab51560893923d7691ce  mesa-20.3.5.tar.xz

New features

  • None

Bug fixes

  • [RADV][BISECTED] The Surge 2 (644830) - In-game assets do not render correctly since 20.3.4.

  • Metal Gear Solid V: The Phantom Pain: texture issues and vertex stretches

  • Rage 2: Visual corruption on in-game menu with ACO.

  • radv: Some MSAA tests fail when DCC is forced.

  • Rendering artifacts in Barn Finders specifically on Radeon Vega

  • Mafia III Demo: Artifacts around barrels

  • subgroupBallotFindMSB() broken in RADV/ACO 20.3.4

  • Texture views on blits ignore formats

  • “unknown intrinsic” assertion triggered by multiview shader in non-multiview renderpass in Vulkan on intel

  • ANV: Weird jitter in Witcher 1

  • DXVK is broken in latest master

  • mesa/st: Uniforms are not updated after lowering alpha test

  • [RADV] Nioh 2 - The Complete Edition: “Bloom” on lights

  • RADV: robustBufferAccessUpdateAfterBind is not exposed

  • ANV: Weird jitter in Witcher 1

  • ANV: Weird jitter in Witcher 1

  • meson: meson-built libraries have inconsistent compatability / current versions compared to older autotools-built libraries

  • device select layer breaks other layers

  • RADV: Extreme overhead in vkQueueSubmit

  • Graphical glitch of popupping missing texture on Mesa version >18.0.5 (Padoka Stable + Unstable/Oibaf/ubuntu-x-swat PPAs)

  • occasional corruption issue with RADV in multiple games, disappears after using amdvlk

  • OpenGL on GMA4500MHD

  • [RADV][REGRESSION][BISECTED] radv_GetMemoryFdPropertiesKHR returns no valid memory types for vaapi drmbuf

  • anv: vkQueueSubmit with waitSemaphore value of 0 hangs CPU

  • ttn: invalid base/range triggering nir_validate assertion

  • [RADV][ACO] Overwatch game crash: amd/compiler/aco_insert_exec_mask.cpp: Failed Assertion

  • Use out encoding for float immediates

  • [RADV] Severe performance drop when exceeding VRAM compared to AMDVLK

  • RADV: Occlusion query hangs Big Navi GPU

  • miscompiled compute shader loop on llvmpipe (and Iris)

  • Graphics glitches after upgrade to mesa 20.3 on Khadas VIM3 Pro (Mali G52 GPU)

  • glthread crash in _mesa_glthread_upload

  • Iris driver causing graphics glitch in QEMU spice egl DMA-BUF

  • [RADV/ACO] Death Stranding cause a GPU hung (ERROR Waiting for fences timed out!)

Changes

Alyssa Rosenzweig (3):

  • pan/bi: Fix M1/M2 decoding in disassembler

  • panfrost: Don’t advertise OES_copy_image

  • panfrost/lcra: Fix constraint counting

Anuj Phogat (2):

  • intel/anv: Fix condition to set MipModeFilter for YUV surface

  • intel/anv: Fix condition for planar yuv surface

Bas Nieuwenhuizen (8):

  • radv: Do not hash vk_object_base in descriptor set layout.

  • radv: Improve spilling on discrete GPUs.

  • radv: Fix vram override with fully visible VRAM.

  • radv: Ignore WC flags for VRAM.

  • radv: Don’t use dedicated memory info to indicate sharing.

  • vulkan/device_select: Stop using device properties 2.

  • radv: Expose robustBufferAccessUpdateAfterBind correctly.

  • radv: Use stricter HW resolve swizzle compat check.

Boris Brezillon (2):

  • panfrost: Fix tiler job injection (again)

  • panfrost: Fix a polygon list corruption in the multi-context case

Caio Marcelo de Oliveira Filho (2):

  • spirv: Allow variable pointers pointing to an array of blocks

  • anv: Lower ViewIndex to zero when multiview is disabled

Daniel Schürmann (2):

  • aco: fix nir_intrinsic_ballot with wave32

  • aco: fix shared VGPR allocation on RDNA2

Dave Airlie (3):

  • zink: don’t pick a cpu device ever.

  • glsl: fix leak in gl_nir_link_uniform_blocks

  • glx: proposed fix for setSwapInterval

Dylan Baker (10):

  • docs: Add sha256sum for 20.3.4

  • .pick_status.json: Update to e346f61a91266bde48cc0857f82ca6348e38dc03

  • .pick_status.json: Update to ea27f2bf092f462171fe14a44619565d14f43fb8

  • .pick_status.json: Update to 38b2e137665f36f1b0fe58e40971548d9eafa865

  • .pick_status.json: Update to 4fdbc22c42dd4672e1660e7598dcd7874880b2f1

  • .pick_status.json: Update to b1ab69f467b7b10e383f3b6bb957c7e6669a3772

  • .pick_status.json: Update to 5d2c9fd161eb1c6b29852e22431fa8bc804bf5af

  • .pick_status.json: Update to 2407952ec99ab1c2a16e01a9e300f799e4d53320

  • .pick_status.json: Update to 60ea60ec4deaf42c55bd8367a15cb0ab885d662e

  • ci: mark a couple of LLVMPIPE tests as expected pass

Erico Nunes (1):

  • lima: fix max sampler views

Georg Lehmann (1):

  • vulkan/device_select: Only call vkGetPhysicalDeviceProperties2 if the device supports it.

Giovanni Mascellani (2):

  • disk_cache: Fail creation when cannot inizialize queue.

  • anv: Allow null handle in DestroyDescriptorUpdateTemplate.

Hyunjun Ko (1):

  • ir3: fix has_src() to return correctly in ir3_nir_lower_tex_prefetch

Ian Romanick (6):

  • i965: Don’t parse driconf again

  • nir/algebraic: Fix a >> #b << #b for sizes other than 32-bit

  • intel/compiler: Enable the ability to emit CMPN instructions

  • intel/compiler: Make the CMPN builder work like the CMP builder

  • intel/compiler: Use CMPN for min / max on Gen4 and Gen5

  • nir/algebraic: Fix some min/max of b2f replacements

Icecream95 (2):

  • panfrost: Add the tiler heap to fragment jobs

  • st/mesa: Update constants on alpha test change if it’s lowered

James Park (1):

  • gallium/tessellator: Fix warning suppression

Faith Ekstrand (4):

  • nir: Don’t optimize bcsel-of-shuffle across blocks

  • intel/fs: Shuffle can’t handle source modifiers

  • anv/formats: Advertise linear sampling on depth formats

  • intel/fs: Handle payload node interference in destinations

Jeremy Huddleston (1):

  • Adjust dylib compatibility versions to match what was set by mesa-18.3’s autotools-based builds

Jordan Justen (1):

  • i965/gen11: Fix must-be-ones bit positions in 3D_MODE

Lionel Landwerlin (7):

  • anv: only signal wsi fence BO on last command buffer

  • anv: discard all timeline wait/signal value=0

  • anv: don’t wait for completion of work on vkQueuePresent()

  • anv: Fix wait_count missing increment

  • anv: reset binary syncobj to be signaled before submission

  • anv: fix MI_PREDICATE_RESULT write

  • intel/fs/vec4: add missing dependency in write-on-write fixed GRFs

Marek Olšák (2):

  • glthread: fix interpreting vertex size == GL_BGRA for vertex attribs

  • mesa: flush glBegin/End before changing GL_DEPTH_STENCIL_TEXTURE_MODE

Matt Turner (1):

  • turnip: Remove unused TU_DEBUG_IR3 flag

Mauro Rossi (1):

  • android: pan/bi: reorder static dependencies in gallium/dri

Michel Dänzer (1):

  • intel/tools: Use subprocess.Popen to read output directly from a pipe

Mike Blumenkrantz (7):

  • radv: null bo list pointer for null descriptors on update

  • radv: zero the bo descriptor array when allocating a new set

  • zink: handle 1bit undef values in ntv

  • gallium/trace: add a pipe_screen::get_compiler_options method

  • mesa/st: clamp scissored clear regions to fb size

  • zink: fix streamout emission for super-enhanced layouts

  • compiler/spirv: fix image sample queries

Nanley Chery (3):

  • gallium: Map _DRI_IMAGE_FORMAT_NONE to NULL

  • gallium: Flush GL API resources in eglCreateImage

  • iris: Disable aux as needed in iris_flush_resource

Pierre Moreau (1):

  • spirv: Ignore WorkgroupSize in non-compute stages

Pierre-Eric Pelloux-Prayer (5):

  • radeonsi: properly set SPI_SHADER_PGM_HI_ES

  • frontends/va: fix protected slice data buffer read size

  • st/mesa: use the correct src format in ReadPixels

  • mesa/fbo: don’t check_end_texture_render on fb read change

  • st/mesa: consider texture view format for fbo blits

Rhys Perry (14):

  • radv: correctly enable WGP_MODE for tessellation control

  • aco: always set exec_live=false

  • aco: do not flag all blocks WQM to ensure we enter all nested loops in WQM

  • aco/lower_phis: fix all_preds_uniform with continue_or_break

  • aco: add missing usable_read2 check

  • radv: don’t set sx_blend_opt_epsilon for V_028C70_COLOR_10_11_11

  • aco: set compr for fp16 exports

  • aco: implement 64-bit VGPR {u,i}find_msb

  • radv,aco: don’t use MUBUF for multi-channel loads on GFX8 with robustness2

  • radv: correctly enable WGP_MODE for NGG and GS

  • radv: round-up num_records division in radv_flush_vertex_descriptors

  • aco: calculate all p_as_uniform and v_readfirstlane_b32 sources in WQM

  • nir/opt_shrink_vectors: add option to skip shrinking image stores

  • radv: don’t shrink image stores for The Surge 2

Samuel Pitoiset (9):

  • nir/algebraic: mark more optimization with fsat(NaN) as inexact

  • radv: fix centroid with VRS coarse shading

  • radv: fix waiting on the last enabled RB for occlusion queries

  • radv: set correct value for OFFCHIP_BUFFERING on GFX10+

  • radv: make sure FMASK compression is enabled for MSAA copies

  • radv,aco: fix shifting input VGPRs for the LS VGPR init bug on GFX9

  • radv: fix separate depth/stencil layout in render pass

  • radv: fix color resolves if the dest image has DCC

  • radv: only apply the MRT output NaN fixup to non-meta shaders

Simon Ser (2):

  • nouveau/nvc0: fix linear buffer alignment for scan-out/cursors

  • nouveau/nv50: fix linear buffer alignment for scan-out/cursors

Timothy Arceri (1):

  • glsl: fix declarations of gl_MaxVaryingFloats

Timur Kristóf (5):

  • tgsi_to_nir: Fix uniform ranges.

  • radv/llvm: Fix reporting LDS stats of tess control shaders.

  • aco: Disallow LSHS temp-only I/O when VS output is written indirectly.

  • aco: Fix LDS statistics of tess control shaders.

  • aco: Fix constant address offset calculation for ds_read2 instructions.

Tony Wasserka (2):

  • aco: Fix vector::reserve() being called with the wrong size

  • aco/ra: Fix register allocation for subdword operands

Vinson Lee (2):

  • etnaviv: Fix memory leak in etna_vertex_elements_state_create.

  • aco: Initialize ds_state.front.writeMask.

Yevhenii Kharchenko (1):

  • st/mesa: fix PBO download for TEXTURE_1D_ARRAY textures

Yevhenii Kolesnikov (1):

  • nir/from_ssa: consider defs in sibling blocks