Mesa 21.3.4 Release Notes / 2022-01-12 ====================================== Mesa 21.3.4 is a bug fix release which fixes bugs found since the 21.3.3 release. Mesa 21.3.4 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 21.3.4 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 --------------- :: TBD. New features ------------ - None Bug fixes --------- - i965: gen5 exposes EXT_texture_integer incorrectly - [radeonsi, regression, bisected]: Rendering issues with Factorio - mesa >= 21.x (with llvmpipe) will break the game JVGS - Panfrost - RK3399 - FACTORIO - glitches everywhere - Baldurs Gate 3 (Patch 6) - ribbon-like artifacts on textures - Game Starsector crashes under certain circumstances with mesa 21.3.+ - FreeSpace models incorrectly rendered on Polaris cards, causing system freeze - Incomplete evaluation of nested DEFINE macros - [r300g, bisected] piglit glsl-fs-discard-04 fails - Panfrost G52 Firefox terrible glitches on youtube playback Changes ------- Alyssa Rosenzweig (3): - pan/bi: Fix load_const of 1-bit booleans - gallium/util: Add pixel->blocks box helper - lima,panfrost: Correct pixel vs block mismatches Bas Nieuwenhuizen (1): - radv: Set optimal copy alignment to 1. Boris Brezillon (1): - microsoft/compiler: Fix dxil_nir_create_bare_samplers() Connor Abbott (2): - ir3/ra: Fix logic bug in compress_regs_left - ir3: Bump type mismatch penalty to 3 Daniel Schürmann (2): - aco: don't allow SDWA on VOP3P instructions - aco: validate VOP3P opsel correctly Danylo Piliaiev (1): - tu: fix workaround for depth bounds test without depth test Dave Airlie (2): - intel/genxml/gen4-5: fix more Raster Operation in BLT to be a uint - crocus: fail resource allocation properly. Emma Anholt (4): - freedreno/afuc: Disable the disassembler on 32-bit builds. - i915g: Turn off FP16 in the vertex shaders. - r300: Fix omod failing to increase the number of channels stored. - nir_to_tgsi: Fix a bug in TXP detection after backend lowering. Eric Engestrom (3): - .pick_status.json: Update to 8a78706643ecad8a1f303cc9358873abc29978b4 - .pick_status.json: Mark 00bea38242d97e7ace1954f1bc7d32cbf0ce3ee0 as denominated - .pick_status.json: Mark 2a0253b9b5d21c7571555abe3a1d851468a18740 as denominated Filip Gawin (1): - r300: fix handling swizzle in transform_source_conflicts Henry Goffin (1): - intel/compiler/test: Fix build with GCC 7 Ian Romanick (1): - i965: Disable EXT_texture_integer on Gfx4 and Gfx5 Jason Ekstrand (1): - Revert "anv: Stop doing too much per-sample shading" Lionel Landwerlin (3): - anv: don't leave anv_batch fields undefined - anv: limit compiler valid color outputs using NIR variables - intel/dev: fixup chv workaround Lucas Stach (3): - etnaviv: initialize vertex attributes on context reset - etnaviv: drm: fix size limit in etna_cmd_stream_realloc - etnaviv: drm: properly handle reviving BOs via a lookup Michel Zou (1): - zink: fix -Warray-bounds warning Mike Blumenkrantz (4): - radv: fix xfb query copy param ordering - zink: always unset vertex shader variant key data when changing last vertex stage - zink: add extra synchronization for buffer descriptor binds - zink: use device-local heap for sparse backing allocations Pavel Ondračka (1): - r300: Remove broken optimization in rc_transform_KILL Pierre-Eric Pelloux-Prayer (4): - radeonsi/gfx8: use the proper dcc clear size - vbo/dlist: fix loopback crash - vbo/dlist: add vertices to incomplete primitives - radeonsi/gfx10: fix si_texture_get_offset for mipmapped tex Qiang Yu (1): - glapi: should not add alias function to static_data.py Rhys Perry (1): - aco: remove pack_half_2x16(a, 0) optimization Rohan Garg (1): - intel/fs: OpImageQueryLod does not support arrayed images as an operand Roman Stratiienko (1): - v3dv: Hotfix: Rename remaining V3DV_HAS_SURFACE->V3DV_USE_WSI_PLATFORM Samuel Pitoiset (1): - radv: add drirc radv_disable_htile_layers and enable it for F1 2021 Tapani Pälli (3): - iris: unref syncobjs and free r/w dependencies array for slab entries - mesa: free idalloc storage for display lists - mesa: free vbo_save_vertex_list store prims Timothy Arceri (1): - glsl/glcpp: make sure to expand new token after concatenation Yiwei Zhang (1): - venus: subtract appended header size in vn_CreatePipelineCache satmandu (1): - Fix compilation on armv7l with gcc 11.2.0