Mesa 26.1.0 Release Notes / 2026-05-06

Mesa 26.1.0 is a new development release. People who are concerned with stability and reliability should stick with a previous release or wait for Mesa 26.1.1.

Mesa 26.1.0 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 26.1.0 implements the Vulkan 1.4 API, but the version reported by the apiVersion property of the VkPhysicalDeviceProperties struct depends on the particular driver being used.

SHA checksums

SHA256: a5095e6dc2986c78f0cef4c5555dc803e93b6bfe5670e991f9e8bd49395bae19  mesa-26.1.0.tar.xz
SHA512: 02972b1a2e6a2d10fa9c970ab579ddd6f0d3acfb782d82616b91cca7a28dd494b15fb14748fff333b1078898cf990a25666e9122d7fd41cb212a151caf933786  mesa-26.1.0.tar.xz

New features

  • GL_NV_timeline_semaphore on radeonsi

  • VK_QCOM_image_processing on Turnip

  • VK_EXT_present_timing on RADV, NVK, Turnip, ANV, Honeykrisp, panvk

  • VK_KHR_sampler_ycbcr_conversion on pvr

  • VK_EXT_image_drm_format_modifier on pvr

  • VK_KHR_internally_synchronized_queues on RADV

  • VK_EXT_blend_operation_advanced on lavapipe

  • VK_KHR_get_surface_capabilities2 on panvk

  • VK_KHR_get_display_properties2 on panvk

  • VK_EXT_acquire_drm_display on panvk

  • VK_KHR_present_id on panvk, v3dv

  • VK_KHR_present_wait on panvk, v3dv

  • VK_KHR_pipeline_executable_properties on pvr

  • VK_EXT_zero_initialize_device_memory on panvk

  • GL_EXT_shader_image_load_store on panfrost

  • VK_KHR_swapchain_mutable_format on panvk

  • VK_EXT_astc_decode_mode on panvk

  • VK_KHR_copy_memory_indirect on nvk, RADV/GFX8+

  • VK_EXT_color_write_enable on panvk

  • VK_EXT_hdr_metadata on v3dv

  • VK_EXT_image_view_min_lod on panvk

  • VK_EXT_depth_clamp_control on panvk

  • VK_VALVE_shader_mixed_float_dot_product on RADV (Vega20, Navi14, RDNA2+)

  • VK_EXT_legacy_dithering on panvk

  • GL_ARB_sample_shading on v3d

  • VK_KHR_maintenance4 on pvr

  • VK_ARM_scheduling_controls on panvk

  • cl_khr_subgroup_ballot on asahi, iris, llvmpipe, radeonsi and zink

  • cl_khr_subgroup_clustered_reduce on asahi, llvmpipe, radeonsi and zink

  • cl_khr_subgroup_extended_types on asahi, iris, llvmpipe, radeonsi and zink

  • cl_khr_subgroup_non_uniform_arithmetic on asahi, llvmpipe, radeonsi and zink

  • cl_khr_subgroup_non_uniform_vote on asahi, iris, llvmpipe, radeonsi and zink

  • cl_khr_subgroup_rotate on asahi, llvmpipe and zink

  • VK_EXT_nested_command_buffer on panvk

  • VK_VALVE_mutable_descriptor_type on panvk

  • VK_EXT_shader_stencil_export on panvk

  • VK_EXT_map_memory_placed on panvk

  • VK_EXT_conditional_rendering on panvk

  • VK_{KHR,EXT}_{surface,swapchain}_maintenance1 on panvk

  • VK_EXT_shader_atomic_float on panvk

  • VK_KHR_device_address_commands on RADV

  • VK_EXT_non_seamless_cube_map on pvr

  • fragmentStoresAndAtomics on panvk/v6-7

  • VK_KHR_shader_untyped_pointers on panvk

  • VK_EXT_primitive_restart_index on RADV

  • VK_EXT_attachment_feedback_loop_layout on panvk

  • VK_EXT_attachment_feedback_loop_dynamic_state on panvk

  • VK_KHR_shader_integer_dot_product on pvr

  • VK_EXT_descriptor_heap on RADV (with ‘export RADV_EXPERIMENTAL=heap’)

  • fullDrawIndexUint32 on pvr

  • multiDrawIndirect on pvr

  • depthBiasClamp on pvr

  • wideLines on pvr

  • VK_EXT_rgba10x6_formats on panvk

  • static C++ stdlib required on rusticl to workaround applications using their own C++ stdlib

Bug fixes

  • 26.0.1 fails to build: `create_context.c: error: ‘struct glx_screen’ has no member named ‘frontend_screen’`

  • A770: Counter-Strike 2 visual glitches (regression)

  • ACO: assertion in insert_exec_mask()

  • ANV: SkQP regression in Android CTS

  • Accumulation of black squares with OpenGL applications at high resolutions (hiz-related)

  • After updating GStreamer, all videos in Showtime are green/purple

  • Anv/Blorp: implement RT clear operations without changing the binding table

  • Anv: implement transfer operations on buffers without render targets

  • Bisected regression: Assertion texObj->pt == view->texture failed.

  • Compile failure on gentoo since !40305

  • Intel BDW regression due to load_push_data_intel intrinsic

  • Is maxFragmentCombinedOutputResources=16 in Honeykrisp reflects an actual HW limit?

  • KHR-GL46.geometry_shader.limits.max_output_components

  • KHR_display: plane size limits

  • Kodi regression with panthor >= 1.7 after updating to Linux 7.0-rc1

  • MDK2 HD (opengl) has most elements rendered as black

  • Mesa 25.3 amdgpu memory issue

  • Mesa LLVMpipe Memory Leak

  • Missing Haswell case after a097a3d214eda7fb7b9ff63176754b7260e09e03 leads to bogus assert in intel_perf_mdapi.c

  • OpenGL 4.1 VRAM Memory Leak with setting uniform variables

  • Panfrost Bifrost compiler assertion failure: wrong vectorization in bi_alu_src_index (Mesa 26.0.0)

  • Portal hard locks the machine on rv350.

  • Question: Does building Lavapipe on Windows require building “microsoft-experimental” as well?

  • RADV: Invalid hitAttributeEXT value when using function-call RT pipelines

  • RADV: RDNA4 visual corruption in DX11 (DXVK) – Mafia III character model glitches, AMDVLK renders correctly (9070XT)

  • Segmentation fault in gm200_validate_sample_locations with Firefox on GTX 1070 Ti (nouveau)

  • Sekiro: Shadows Die Twice lighting corruption on RX 7900 XT

  • Shader inputs/outputs for vertex/pixel shaders that have the integer (int) type are broken on RDNA 3 and 4 graphics cards

  • Support for timeline semaphores in radeonsi

  • The End is Nigh (Wine): No lighting in The Hollows

  • Transcoding mpeg2video with ffmpeg h264_vulkan on Intel cause Conversion failed!

  • Turnip crash with lazy depth textures: GPUMEM_BIND_RANGES failed (Not a typewriter)

  • Unroll loops before lowering indirects

  • VK_KHR_display fails on PowerVR on Mesa master

  • Vulkan CTS regression bisected to 5d2c17a5fdce (“vtn: skip make-available/visible for shared”)

  • [26.0.0~rc1] d3d12_screen.cpp:1165:(.text._ZL31d3d12_interop_query_device_infoP11pipe_screenjPv+0x4b): undefined reference to `d3d12_video_encoder_get_last_slice_completion_fence(pipe_video_codec*, void*, pipe_fence_handle**)’

  • [ANV]: Regression in dxvk Greedfall

  • [ANV][A770] Deep Rock Galactic assert fails

  • [ANV][BMG] Building Mesa with Clang causes Missing Skin Textures in UE games - Tekken 8

  • [ANV][BMG] Dying Light The Beast graphics artifacts

  • [ANV][BMG] Regression - Broken lighting and flickering in Kingdom Come: Deliverance II

  • [ANV][DG2][Regression]: Flickering water “boxes” in Civilization VII

  • [ANV][LNL] - Elden Ring (1245620) - Vertex explosions with ray tracing enabled

  • [ANV][PTL] R.E.P.O. GPU Hang

  • [ANV][Xe2+] Resource Barriers Invalid Signal Stage

  • [NVL-S] deqp-vk failures within 2d_array and cubemap tests

  • [RADV] Killer7 has a blue tint with RDNA3/4

  • [anv] Intel ARC B390 | Horizon Forbidden West | DX12 | Flashing effects

  • [anv][ptl] valheim gpu hang and visual corruption

  • [bisected] Xe3 regression with piglit tess/barrier-patch.shader_test after cmod prop change

  • [docs] add documentation on how to install debug symbols in various distros

  • [hasvk] Regressions from enabling VK_KHR_maintenance6

  • [radeonsi] Missing ground texture in Lethis Path of the Progress

  • [radeonsi] Regression: GL_FEEDBACK returns 0.0 for X-coordinates (Legacy GL)

  • [radv] Regression causes glitches in Strange Brigade (Vulkan renderer)

  • [radv][bisected][regression] GhostwireTokyo RT gpu hangs with HPLOC commit

  • [regression] Left 4 Dead 2 crashing when joining or starting survival with “Official Dedicated” servers

  • amdgpu reset/crash when simulating stereo camera

  • amdgpu_device_initialize: amdgpu_query_info(ACCEL_WORKING) failed (-13)

  • anv, bisected: Genshin Impact wrong shadows, flickering grass

  • anv: Implement BTI switching for fast-clears

  • anv: Implement multi-layer fast-clear

  • anv: Suballocate indirect clear colors

  • anv: missing implementation of vkGetPhysicalDeviceCooperativeMatrixFlexibleDimensionsPropertiesNV

  • anv: optimize descriptor buffer binding

  • anv: ycbcr CTS tests asserts

  • building mesa_clc on ubutu-26.04 with gcc-16 fails link

  • ci: Wayland issues after 6641c891fdaa15923f0b61a5fef4b9d9ed91ac0e

  • freedreno/avx2 build failure

  • freedreno/decode: Usage based register summaries

  • glcpp: incorrect macro expansion in token pasting

  • glsl: spec@glsl-es-1.00@linker@glsl-mismatched-uniform-precision-unused broken

  • gnome-control-center hitting assert

  • intel/isl: Support CCS on Ys-tiled images

  • ir3: ir3_get_predicate() vs &ctx->build

  • isl: Prefer Tile64 when it saves space

  • lavapipe: crash in caselist

  • mesa: deleting a buffer bound only to an index also undoes the associated general target binding

  • nir/rusticl: optimize 64b sys vals

  • nir: possible exactness bug in reassociate

  • nvk: Enable ZCULL

  • nvk: shader does unnecesary move to local

  • panvk: VK_KHR_present_id, VK_KHR_present_wait not implemented

  • r300 , regression , bisected : Glitches with Sauerbraten

  • r300: HiZ related dEQP failures

  • radv, regression : Crysis 2 Remastered raytracing blocky reflections

  • radv: Port si_emit_guardband to RADV

  • radv: enable rdna3 bfloat16 support

  • regression, bisect: commit a8272bf0f1f9229d73252b03d0fb32d563396a9c breaks KWin through NVK & Zink

  • static linking regression since !37495 - spirv-tools shared library required at runtime if exists at build time

  • tu,ir3: Lowering IO before unrolling loops broke forced loop unrolling heuristic, breaking at least multiview

  • tu: GPU faults during LRZ clears on unallocated transient attachments in gmem mode

  • turnip: VK_EXT_device_memory_report unsupported

  • turnip: inconsistent host allocator use for command buffers

  • turnip: llama.cpp: Running test-backend-ops results in segmentation fault

  • util/hash_table: regression: use after free on 32-bit platforms

  • util: Build regression with MSYS2 MinGW-W64 x64 clang 21 on 26.0.0-rc3

  • venus crashes in vn_CreateDevice() with latest mesa/main [bisected]

  • virgl: Trace scripts timeout with Linux 6.17

  • virtio/vulkan: vkcube fails to run when calling the libvulkan_virtio dynamic library

  • wsi: `assert(chain->dxgi);` may failed under venus for win32

  • zink, turnip: compilation is failing when compiling zink and turnip with only kgsl support

  • zink: mesh shaders broken

Changes

Adam Jackson (2):

  • zink: use VK_EXT_pci_bus_info for PCI address

  • venus: advertise VK_KHR_shader_fma

Adam Simpkins (1):

  • iris: fix a crash in disable_rb_aux_buffer

Aditya Swarup (11):

  • anv: Add helper macros for address binding report extension

  • anv: Report address binding events for memory buffers

  • anv: Report address binding events for images

  • anv: Add BO helper macros for binding report extension

  • anv: Report address binding events for VkQueryPool

  • anv: Report address binding events for VkDescriptorPool

  • anv: Report address binding events for VkCommandPool

  • anv: Report bind/unbind events for sparse VA range

  • anv: Report addr bind events for opaque/non-opaque sparse allocations

  • anv: Enable support for VK_EXT_device_address_binding_report

  • anv: Report bind events for image private binding

Adrián Larumbe (2):

  • pan/kmod: Fix minor version number check for USER_MMIO_OFFSET ioctl

  • pan/kmod: fix double syncop count sum when populating vm_bind syncs

Agate, Jesse (3):

  • amd/vpelib: Add RGB 601 Primaires to BG Color

  • amd/vpelib: Predication fix

  • amd/vpelib: Visual Confirm Fix

Ahmed Hesham (3):

  • rusticl: return correct error from clCreateSubBuffer

  • rusticl: fix flag validation when creating an image

  • pan/bi: Restore b3210 as a valid swizzle

Aitor Camacho (45):

  • nvk: Handle unbound sets that contain dynamic buffers

  • hk: Handle unbound sets that contain dynamic buffers

  • kk: Update kk_bind_descriptor_sets comment to reflect updated binding

  • kk: Bundle nir_to_msl options into a struct for easier option addition

  • kk: Force frag output component count to match render targets’

  • kk: Use nir_opt_shrink_stores

  • kk: Fill pipelineUUID

  • kk: Fix shader uint32_t value serialization

  • kk: Correctly release pipeline handles at shader destroy

  • kk: Fix compute pipeline cache

  • kk: Move gfx pipeline data to the info struct within kk_shader

  • kk: Fix graphics pipeline serialization

  • kk: Remove primitive type from pipeline and rely on dynamic one

  • kk: Enable vertexPipelineStoresAndAtomics

  • kk: Move nir_opt_shrink_stores after nir_opt_remove_phis for correct shrink

  • kk: Fix disabling workaround 4

  • kk: Expose pushDescriptor from 1.4

  • kk: Expose VK_EXT_extended_dynamic_state2

  • kk: Expose VK_EXT_texel_buffer_alignment

  • kk: Expose depthBiasClamp

  • kk: Expose largePoints

  • kk: Expose VK_EXT_image_2d_view_of_3d

  • kk: Expose sampleRateShading

  • wsi/metal: Expose additional color spaces if instance extension enabled

  • kk: Remove helper invocation flag in read system values

  • kk: Fix crash in PositiveShaderImageAccess.UndefImage

  • kk: Assign type to load_frag_coord

  • kk: Fix push descriptor set layout when rebinding

  • kk: Force fragment output matches render targets’

  • kk: Fix image access issues

  • kk: Default to max descriptor size if mutable list is empty

  • kk: Increase push constant size to 256 from 128

  • docs/kk: Update build instructions to add –prefer-static

  • docs: Add KosmicKrisp to the list of layered drivers

  • kk: Set command buffer state to 0 when reset

  • kk: Add clc in a similar fashion to other drivers like HK

  • kk: Rework draw recording for easier addition of stages like tessellation

  • kk: Implement VK_KHR_draw_indirect_count as HK does

  • kk: Rework command buffers’ compute shader state tracking

  • kk: Rework command buffers’ graphics shader state tracking

  • kk: Clean up gfx state flushing

  • kk: Reset queries through compute dispatch instead of queue writes

  • kk: Remove buffer arg from queue writes

  • kk: Demote events, query availabilities and queue writes to 32 bits

  • kk: Fix pre-gfx encoder dependency with gfx encoder

Alejandro Piñeiro (3):

  • v3dv/meson: fix missing headers and duplicate entry

  • v3dv: split v3dv_private.h into smaller headers

  • broadcom/vulkan: remove v3dv_private.h

Aleksi Sapon (8):

  • llvmpipe: pass explicit derivatives to sampling codegen

  • llvmpipe: elliptical derivative transform for anisotropic filtering

  • llvmpipe: implement per-fragment anisotropic rho

  • llvmpipe: add GALLIVM_PERF=no_lod_ellipse

  • llvmpipe: add stride argument to lp_build_swizzle_aos_n

  • llvmpipe: update traces

  • llvmpipe: fix incorrect image 64bit fetch return value type

  • lavapipe: update fails

Alexander Koskovich (1):

  • freedreno/common: add support for the Adreno 810

Ali, Nawwar (4):

  • amd/vpelib: update 3dlut and shaper FL

  • amd/vpelib: coding style rectify

  • amd/vpelib: Fix crash during encoding test

  • amd/vpelib: Move shaper and 3D LUT updates to vpe_color_update_movable_cm

Allen Ballway (1):

  • vulkan: update ALLOWED_ANDROID_VERSION for api level 37

Alyssa Milburn (1):

  • nv50,nvc0: Avoid uninitialized cbuf reads in blits

Alyssa Rosenzweig (92):

  • brw: move nir_opt_memcpy OOTL

  • brw: remove redundant nir_opt_combine_stores

  • brw: hoist lower_pack OOTL

  • brw: unloop post-mem vectorize opts

  • brw: run opt_deref only once

  • brw: only optimize ray queries once

  • brw: only optimize ray queries if there are any

  • brw: run nir_opt_idiv_const only once

  • brw: optimize bfi only late

  • brw: combine more peephole select

  • brw: remove a redundant DCE

  • util: add linear_memdup

  • util: add BITSET_LINEAR_ZALLOC

  • brw: use BITSET_LINEAR_ZALLOC

  • brw: move fsign lower OOTL

  • brw: hoist fsat lower OOTL

  • pvr,pan,agx: drop cargo-culted nir_opt_loop calls

  • util: allow string shader “statistics”

  • util: hide hashes from GL shader stats

  • intel: add scheduling mode statistic

  • intel: simplify shader stats names

  • intel: report code size in shader stats

  • nir: disable fast-math for lowering conversions

  • vtn: fix wait_group_events memory scope

  • brw: drop buggy SLM optimization

  • nir: add missing ssbo atomics to nir_get_io_index_src_number

  • util/bitset: add an assert for big BITSET_EXTRACT

  • nir/opt_constant_folding: optimize ballot(false)

  • nir: add nir_get_io_data_src

  • nir/lower_atomics: use data helper

  • nir/opt_uniform_atomics: use data helper

  • nir/opt_fragdepth: use data helper

  • nir/opt_intrinsics: use data helpers

  • brw: use data helper

  • util,intel: move probably_float to common code

  • agx: use util_is_probably_float

  • mailmap: update my personal email

  • panfrost: drop email from ancient copyright lines

  • asahi: fix some copyright headers

  • nir: optimize u2u32(unpack_32_2x16_split_*)

  • nir/lower_subgroups: generalize vote lowering

  • agx: use common code vote lowering

  • nir/lower_io: remove incorrect Intel _block cases

  • brw: move brw_nir_pack_vs_input to brw_nir.c

  • brw: move brw_can_coherent_fb_fetch to a C header

  • nir/lower_io: handle Intel URB intrinsics

  • nir/lower_subgroups: fix boolean clustered reductions

  • nir/opt_sink: sink Intel UBO loads

  • nir/opt_sink: sink pack_64_2x32_split

  • agx: drop NIR continue handling

  • util/sparse_bitset: add u_sparse_bitset_clear_all

  • brw: explicitly pad tgl_swsb

  • brw/eu_emit: relax assertion to allow ARF NULL

  • brw/nir_lower_fs_load_output: optimize pixel coord

  • brw: wire up MACL

  • brw: Move intel_nir_opt_peephole_imul32x16 later in compilation

  • brw: scalarize even 64-bit scratch access

  • brw: lower 16-bit mulh

  • brw: lower mem access sizes even for UBOs

  • brw: chop up unaligned access

  • brw: round up block components

  • nir: add frag_coord_w_rcp intrinsic

  • nir: add Intel RT write intrinsic

  • nir: add shuffle_intel

  • nir: add pixel_coord_intel

  • brw: subgroup lowering for jay

  • brw: disable hw generate local ID for jay

  • brw: disable nir_opt_uniform_atomics for Jay

  • brw: add Jay-specific SIMD selection rule

  • brw: lower ifind_msb for Jay

  • intel: add Jay

  • iris: wire up jay

  • anv: wire up jay

  • jay: fix W-entry calcs

  • jay: rematerialize address regs

  • jay: drop GRF reg stats

  • jay: roundrobin RA

  • jay: marginally improve send splitting heuristic

  • jay: tweak roundrobin

  • jay: generalize alignment heuristic

  • jay: improve vector affinities

  • jay: fix SEND scoreboarding

  • jay: fix simd split swsb bugs

  • jay: fix instr counts

  • jay: move deswizzle hack outside of swsb

  • jay: split up jay_from_nir.c

  • jay: load_simd_width_intel

  • jay: fix a bunch of opcode properties

  • jay: fix bfn cmod

  • jay: allow cmod on cvt

  • intel: fuse off Jay in Mesa 26.1

  • nir/opt_reassociate: fix exactness bug

Anders Roxell (7):

  • teflon: Add support for symmetric per-channel quantization

  • ethosu: Add support for per-channel quantization

  • ethosu: Handle per-channel zero_points

  • ethosu: fix RESIZE upscale mode

  • ethosu: clean up ADD elementwise scaling

  • teflon/tests: add micronet_large anomaly detection model

  • ethosu: fix blockdep to check for data dependencies

Andy Nguyen (1):

  • amd/addrlib: Add more GFX1013 GPUs

Anna Maniscalco (2):

  • freedreno/common: set has_astc_hdr true for a7xx targets

  • zink: don’t care about generated gs output primitive

Ansari, Muhammad (2):

  • amd/vpelib: Fix potential overflow calculation

  • amd/vpelib: Adding new wrapper for register profiling

Arjob Mukherjee (4):

  • doc: Added documentation for imagination tree

  • pvr: Fixup for deqp-vk.api 2d.optimal.* conformance

  • pvr: Fixup for deqp-vk.api 2d.optimal.* conformance

  • pvr: increase value of maxPerStageDescriptorStorageBuffers

Arkady Shlykov (2):

  • brw: Implement divergent atomics fusion optimization (single message approach)

  • anv: Add control over divergent atomics fusion opt via driconf

Assadian, Navid (1):

  • amd/vpelib: Reorder function pointers

Bas Nieuwenhuizen (1):

  • ac/llvm: Fix build with LLVM 23.

Benjamin Cheng (16):

  • radv/video: Use a more reliable way of computing tile sizes

  • radv/video: Use ac_video_dec for decode

  • radv/video: Split cdf buffer and encode ctx

  • ac/parse_ib: Fix VCN address parsing

  • radv/video: Disable qp map for h265 on vcn1

  • radeonsi/vcn: Use full pitch for pre-encode input

  • radv: Disable video features for some DRM modifiers

  • frontends/va: Assert that slices come in order

  • ac: Fix naming of hevc encode params IB

  • radv/video_enc: Use variable slice mode when possible

  • radeonsi/vcn: Reorder get_slice_ctrl_param

  • ac: Update FW required for variable slice mode

  • radv/video: Add low-latency flags

  • ac/surface: Filter swizzle modes for VCN

  • radv: Relax linear requirement to VCN1 and prior

  • radv/wsi: Re-use transfer queue if it exists

Benjamin Otte (1):

  • lavapipe: Fix features for nonsubsampled ycbcr formats

Bernd Kuhls (1):

  • blake3: add blake3_neon.c only for little endian archs

Boris Brezillon (2):

  • pan/kmod: Allow mmap() on foreign buffers

  • pan/format: Advertise support for AFBC(16x16,sparse,split)

Boyuan Zhang (1):

  • ac/vcn_dec: add addr_mode for VCN 5.0.1

Brian Paul (2):

  • gallivm: fix undefined CALLOC_STRUCT build error

  • util,loader: silence asprintf() unused result warnings

Caio Oliveira (49):

  • brw/scoreboard: Use std::vector when applicable

  • brw/scoreboard: Add tests showing implicit unordered dependencies in SWSB

  • brw: Provide ~ and &= operators for tgl_sbid_mode

  • brw/scoreboard: Support local implicit out-of-order dependencies

  • brw: Create a struct to hold parser state

  • brw: Move brw_last_inst macro to assembler

  • brw: Move the brw_codegen inside brw_asm_parser

  • brw: Remove global variables from brw_asm parser

  • brw: Remove tabs from brw_cfg.cpp

  • brw: Remove foreach_block_safe / reverse_safe

  • brw: Remove block_list in favor of blocks array

  • brw: Don’t increment block loads addresses unless needed

  • intel/compiler: Use SPDX annotations

  • spirv: Check Capability for identifying SPV_NV_mesh_shader

  • intel/mda: add difflog command

  • brw: Include backend NIR passes in mda files

  • brw: Use the “early break” loop macros when possible

  • intel/mda: Change the matching logic

  • intel/mda: Use -W for color words diff and -U for regular unified diff

  • brw: Remove outdated comment about remove_dead_variables

  • brw: Fix “GRF registers” stats output

  • brw: Print “GRF registers” in INTEL_DEBUG=shaders output

  • brw/print: Don’t print extra space at the end

  • anv: Don’t enumerate cooperative matrix configurations if disabled

  • anv: Simplify cooperative matrix feature advertising

  • brw: Fix cooperative matrix constant sources other than src0

  • brw: Make brw_builder::uniform() ignore previous group

  • brw: Explicitly set group=0 in generator for SYNC used in workaround

  • nir: Handle nir_instr_type_cmat_call in more places

  • brw/scoreboard: Don’t track dependencies for UNDEFs

  • brw: Add lowering for nir_cmat_call_op_per_element_op

  • anv: Enable cooperativeMatrixPerElementOperations

  • anv: Set PIPELINE_SELECT systolic mode based on shader usage

  • spirv: Refactor ALU opcode translation to take bit sizes

  • spirv: Pull constant source fixup to the existing loop

  • spirv: Fix spec constant to handle Select for non-native floats

  • spirv: Remove conversions from vtn_nir_alu_op_for_spirv_opcode()

  • nir: Fix constant folding for iadd_sat

  • anv: Add vkGetPhysicalDeviceCooperativeMatrixFlexibleDimensionsPropertiesNV

  • spirv: Use SPDX annotations

  • spirv: Remove dead code in subgroup instruction handling

  • nir: Add intrinsics for ShuffleUpINTEL and ShuffleDownINTEL

  • intel/compiler: Handle shuffle_*_intel intrinsics in bit size lowering

  • spirv: Lower ShuffleUpINTEL and ShuffleDownINTEL to intrinsics

  • brw: Always use split send in generator

  • anv: Fix assert in anv_nir_compute_push_layout

  • brw: In “Clear Accumulator” workaround, never set predicate_inverse

  • anv: Lower any remaining globals when cmat_calls are inlined

  • brw: Fix max_dispatch_width collection for CS with variable size

Caius-Moldovan-img (3):

  • pco: Move DITR and DITRP fencing from translation to legalization

  • pco: Add pseudo instruction fencing for DITR and DITRP

  • pco: Move part of legalization after register allocation

Calder Young (4):

  • Revert “anv,brw: Allow multiple ray queries without spilling to a shadow stack”

  • anv: Avoid dumping BVH before command buffer is submitted

  • anv: Fix address bit masking for indirect SBTs

  • anv: Fix support for indirect SBTs on Xe3+

Caleb Callaway (2):

  • driconf: LTO disable

  • anv/driconf: Disable shader LTO for MHW

Casey Bowman (3):

  • anv: Fix shaders-lineno implementation for eu stall sampling

  • intel/tools: Add xe3p format for intel_monitor

  • intel/ds: Modify rejection threshold to scale with requested sample period

Caterina Shablia (20):

  • panvk: fix sparse image non-opaque binds

  • panvk: let the mod handler handle DRM_FORMAT_MOD_ARM_16X16_BLOCK_U_INTERLEAVED

  • panvk: merge vm_bind ops in some cases

  • drm-uapi: update drm_fourcc.h

  • pan/genxml: add interleaved 64k clump ordering and block format

  • pan/lib: introduce standard_sparse_mapping_granularity

  • panvk: implement sparse in terms of interleaved 64k

  • pan/lib: remove deadcode

  • panvk: implement sparseResidencyImage3D

  • spirv: plumb spirv-dis –offsets

  • pan/bi: remove trailing space

  • panvk: propagate debug info through NIR when BIFROST_MESA_DEBUG=debuginfo is specified

  • pan/bi: propagate debug info

  • pan/bi: print shaders with debug info when BIFROST_MESA_DEBUG=shaders,debuginfo is specified

  • panvk: never report identicalMemoryLayout

  • pan/lib: use interleaved 64k in more cases

  • panvk: leave a TODO for U-interleaved copies

  • pan/lib: use tiled AFBC

  • radv: skip retiling if transitioning away from ZERO_INITIALIZED

  • radv: move all image checks into radv_image_need_retile

Chan, Roy (2):

  • amd/vpelib: fix uninitialized variable

  • amd/vpelib: add a new cap to differentiate scaler coeff handling

Chang, Tomson (1):

  • amd/vpelib: avoid using reg_update for multi-thread

Christian Gmeiner (62):

  • compiler/mesa/st: Move gl_advanced_blend_mode to util/blend.h

  • util/blend: Add advanced blend modes

  • util/blend: Add pipe_blend_overlap_mode enum

  • vulkan/runtime: Add helper to convert VkBlendOp to pipe_advanced_blend_mode

  • vulkan/runtime: Add helper to convert VkBlendOverlapEXT to pipe_blend_overlap_mode

  • vulkan/runtime: Add VK_EXT_blend_operation_advanced state tracking

  • nir: Extract blend equation helpers to shared header

  • nir/lower_blend: Add advanced blending support

  • asahi/hk: Implement VK_EXT_blend_operation_advanced

  • gallium: Add pipe cap for masked clears and support stencil masking

  • etnaviv: hwdb: Add BLT_64BPP_MASKED_CLEAR_FIX cap

  • etnaviv: blt: Enable masked clear for color and stencil

  • etnaviv: Emit alpha_to_coverage dither table only when needed

  • lavapipe: Implement VK_EXT_blend_operation_advanced

  • meson: Restore .clang-format for ninja clang-format target

  • pan/format: Disable storage image support for compressed formats

  • panvk: Support VK_EXT_zero_initialize_device_memory

  • pan/compiler: Fix progress reporting in pan_nir_lower_store_component

  • panvk: Support VK_EXT_astc_decode_mode

  • etnaviv: Validate MSAA sample count for depth/stencil formats

  • etnaviv: blt: Fix clear_bits overflow for 32-bit formats

  • panvk: Advertise VK_EXT_color_write_enable

  • panvk: implement VK_EXT_image_view_min_lod

  • etnaviv: blt: Use img->swizzle for CONFIG SWIZ fields

  • etnaviv: Add translate_pe_internal_format helper

  • etnaviv: Use BGRA-internal texture format with BLT/RS R/B swizzle

  • etnaviv: Compute blend color registers directly in etna_set_blend_color(..)

  • panvk: Support VK_EXT_depth_clamp_control

  • panvk: Support VK_EXT_legacy_dithering

  • u_blitter: Add single-triangle draw mode for NEAREST blit consistency

  • etnaviv: Enable single-triangle blitter mode

  • etnaviv: hwdb: Add WIDELINE_TRIANGLE_EMU cap

  • etnaviv: Limit max line width to 1.0 on GPUs needing wide line emulation

  • etnaviv: Mark TS config dirty after BLT blit

  • etnaviv: Implement stencil-only blit using util_blitter fallback

  • etnaviv: Add S8_UINT texture format support for stencil texturing

  • panvk: Use per-queue shader core count for CSF group creation

  • panvk: Advertise VK_ARM_scheduling_controls on CSF

  • isaspec: Use %g format for float display to ensure round-trip fidelity

  • panvk: advertise VK_EXT_nested_command_buffer on v10+

  • etnaviv: isa: Restrict COND field to conditional instructions

  • etnaviv: isa: Split texkill into concrete bitset variants

  • panvk: Advertise VK_VALVE_mutable_descriptor_type

  • etnaviv: isa: Add unary texkill variant

  • panvk: Advertise VK_EXT_shader_stencil_export

  • docs/features: VK_VALVE_mutable_descriptor_type: Add missing version info

  • panvk: Implement VK_EXT_memory_budget support

  • pan/kmod: Simplify pan_kmod_bo_mmap() to always map the whole BO

  • panvk: Implement VK_EXT_map_memory_placed

  • panvk: Add VK_EXT_conditional_rendering state and commands

  • panvk: Wrap draws and dispatches with conditional rendering

  • panvk: Support inherited conditional rendering in secondaries

  • panvk: Disable conditional rendering during meta operations

  • panvk: Advertise VK_EXT_conditional_rendering

  • panvk: Advertise VK_EXT_shader_atomic_float

  • panvk: Lower memcpy derefs

  • panvk: Advertise VK_KHR_shader_untyped_pointers on v9+

  • panvk: Advertise VK_EXT_attachment_feedback_loop_layout

  • panvk: Advertise VK_EXT_attachment_feedback_loop_dynamic_state

  • radv: Don’t advertise any features for R10X6G10X6B10X6A10X6_UNORM_4PACK16

  • util/format, vulkan: Add PIPE_FORMAT_X6R10X6G10X6B10X6A10_UNORM

  • panvk: Advertise VK_EXT_rgba10x6_formats

Christoph Pillmayer (20):

  • pan: Add some missing ForEachMacros to .clang-format

  • pan/bi: Fix spill limit value order

  • pan/bi: Reindex SSA before NIR->BIR

  • pan/bi: Copy agx_repair_ssa.c

  • pan/bi: Copy reindex_ssa.c from agx

  • pan/bi: Fixup bi_reindex_ssa.c for bi

  • pan/bi: Fixup bi_repair_ssa.c for bi

  • pan/bi: Repair SSA after spilling

  • pan/bi: Teach bi_print_instr about memory vars

  • pan/bi: Pull out size recording

  • pan/bi: Abstract away spills/fills when spilling

  • pan/bi: Don’t allocate lut space for temps

  • pan/bi: Account for MEMMOV in bi_record_sizes

  • pan/bi: Make SSA spilling vector aware

  • panvk: Don’t create MS2SS views for internal views

  • pan/bi: Fix coupling spill placement

  • pan/bi: Move FAUs to memory for memory phis

  • CODEOWNERS: Update panfrost

  • pan/bi: Fix MEMMOV size calculation

  • pan/bi: Fix source swizzle in bi_repair_ssa

Chuanyu Tseng (1):

  • Revert “amd/vpelib: Adding new wrapper for register profiling”

Collabora’s Gfx CI Team (13):

  • Uprev ANGLE to 63d1dd7c2dfccf6acbd92af224b48aa6ada45f1c

  • Uprev VVL to snapshot-2026wk06

  • Uprev Piglit to 0d79fb4a59c7d213ff144afa4c73e3b32ebe6500

  • Uprev VVL to snapshot-2026wk07

  • Uprev Piglit to 8e2c8bc0018f42b107d470a2de1bf7f53e8fb012

  • Uprev ANGLE to b90b9ee1a4f901e6ba9e649d8f6cf9098a944f50

  • Uprev Vulkan Validation Layers

  • Uprev Piglit to d0a16eee4f7b24abe7e3aab6ee77db8f82e5ad49

  • Uprev ANGLE to 599125448d7ad53b2868a7b5d2e3e8d3bfbc1717

  • Uprev VVL to adfdda5b665f59aae31acf5c12c73e64a8f89553

  • Uprev VVL to cb2acdf7f49053406770ae73cbb315229a9131eb

  • Uprev Piglit to 11ce9eb56edb00e6a7702d13168cc827ce5e0cbd

  • Uprev ANGLE to 5e591d03650dd427001e355f4884b857cadab113

Connor Abbott (65):

  • nir: Move is is_compact() out of unlower_io_to_vars

  • nir: Fix recompute_io_bases with compact i/o arrays

  • nir/recompute_io_bases: Fix handling of dual-source blending

  • nir/recompute_io_bases: Fix num_inputs with dual-slot VS inputs

  • nir: Allow lower_clip_fs with lowered IO

  • nir/lower_clip: Correctly handle driver_location in VS lowering

  • st/mesa: Call nir_recompute_io_bases after lowering FP variants

  • nir, ir3: Make ir3 GS varyings use a proper index

  • ir3, freedreno, tu: Move nir_lower_io out of post_finalize()

  • ir3: Fix ir3_nir_lower_layer_id indentation

  • ir3: Stop relying on variables in ir3_nir_lower_layer_id

  • ir3: Stop using variables when translating from NIR

  • ir3: Don’t use variables for passthrough TCS

  • ir3: Remove variables after i/o lowering

  • freedreno, turnip, ir3: Always gather streamout info from NIR

  • ir3: Stop asserting tess levels are scalar

  • ir3: Don’t support indirect inputs in FS

  • ir3, freedreno, turnip: Lower io earlier

  • freedreno, ir3: Fix branchstack register definitions on a5xx+

  • ir3: Split out max_branchstack and branchstack_size

  • ir3: Fix branchstack max_waves calculation

  • tu: Fix FDM texel size calculation

  • tu: Handle FDM-per-layer in CmdClearAttachments paths

  • tu: Use a patchpoint for subpass clears with FDM

  • tu: Implement VK_QCOM_multiview_per_view_viewports

  • tu/autotune: Take render pass layers into account

  • tu: Support VK_QCOM_multiview_per_view_render_areas

  • tu: Remove fdm argument from tu6_emit_tile_select

  • tu: Implement bin merging for views

  • tu: Implement bin skipping for zero-density regions

  • ir3: Fix barrier error case calculation

  • tu: Fix condition for re-emitting FDM-related state

  • tu: Use HW offset 0 in 3d loads/clears with FDM

  • ir3: Fix constlen trimming when more than one stage is trimmed

  • tu: Store tile the tile was merged with

  • tu: Refactor FDM sampling and bin merging

  • tu: Move FDM tile configuration to a new file

  • tu: Always call tu_emit_renderpass_begin()

  • tu: Pass through tile_config to FDM patchpoints

  • tu: Move immutable sampler handling above descriptor size calc

  • vulkan: Store a few more fields in vk_sampler

  • tu: Set polygon mode when blitting

  • tu: Fix setting will_be_resolved with MSRTSS

  • tu: Track which views an attachment is used as a resolve attachment

  • tu: Refactor immutable sampler handling with descriptor update templates

  • tu: Multiply bin size by GMEM extent

  • tu: Implement subsampled images

  • freedreno/afuc: Update cread/cwrite syntax in README

  • freedreno: Rename afuc to QRisc

  • vtn: Fix vtn_mediump_downconvert_value() for transposed matrices

  • vtn: Fix vtn_mediump_upconvert_value() with transposed matrices

  • nir: Use better calculation for alpha-to-coverage mask

  • tu, ir3, nir: Plumb through driver param for alpha-to-coverage

  • tu: Enable alpha-to-coverage emulation

  • freedreno: Name GS/DS ViewID register fields

  • ir3: Implement ViewIndex for GS

  • ir3: Support multiview in GS lowering

  • tu: Adjust multiview lowering for GS

  • tu: Fill GS/DS ViewID register fields

  • tu: Lower maxMultiviewViewCount to 6

  • tu: Enable multiviewGeometryShader

  • ir3: Don’t reset immediate count to 0 after lowering

  • ir3: Use correct immediate size for constlen calculation

  • tu: Fix LRZ+FDM offset+secondaries

  • tu: Disable LRZ when resuming if the GPU doesn’t support tracking

Daivik Bhatia (9):

  • broadcom/compiler: Update comment clarifying OpTerminate implementation

  • v3dv: use vk_graphics_pipeline_state for pipeline creation

  • v3d/v3dv: drop manual log2_tile_width/height asserts. Move the log2_tile_width/height asserts to pack header functions.

  • v3dv: parse V3DV_ENABLE_PIPELINE_CACHE with parse_debug_string

  • v3d/v3dv: drop unused UIF XOR disable plumbing

  • nir: Handle format swizzles for OOB image loads

  • v3dv: Implement robust_image_access_2 flag

  • broadcom/compiler: lower txf LOD for robustImageAccess2 on V3D 4.2

  • v3dv: Enable VK_KHR_robustness2

Daniel Lang (3):

  • etnaviv: hwdb: Import gc_feature_database from Amlogic

  • etnaviv: hwdb: Import gc_feature_database from D-Robotics

  • etanviv: hwdb: Import gc_feature_database from eYs3D

Daniel Schürmann (59):

  • nir/lower_non_uniform_access: flag IF as always divergent taken

  • panfrost/clc: call nir_opt_remove_phis after nir_opt_loop

  • asahi/clc: call nir_opt_remove_phis after nir_opt_loop

  • nir/opt_loop: Relax restrictions on opt_loop_peel_initial_break() for more loops

  • nir/opt_load_store_vectorize: use linear allocator instead of ralloc

  • nir/opt_load_store_vectorize: create add_entry_to_hash_table() helper

  • nir/opt_load_store_vectorize: Allow to vectorize at most one entry of each type across blocks

  • radv: vectorize UBO, SSBO and shared across blocks

  • radeonsi: vectorize UBO, SSBO and shared across blocks

  • nir/opt_load_store_vectorize: Vectorize speculatable instructions across blocks

  • nir/opt_load_store_vectorize: don’t use shared2 vectorization across blocks

  • nir/opt_loop_unroll: Always unroll loops with a known trip-count of 0

  • nir/loop_analyze: also set force_unroll if the array_size is larger than max_trip_count

  • nir/clone: Fix cloning indirect call instructions

  • aco/builder: Fix v_add_co_u32 carry-out to VCC if post_ra

  • aco/isel: Do IF-simplification if that didn’t happen during NIR optimizations

  • aco: don’t emit p_logical_start / p_logical_end after divergent branches

  • aco/isel: Don’t emit ELSE side of uniform branches which jump

  • aco/isel: Don’t emit ELSE side of divergent branches which jump

  • aco/lower_branches: Consider branch target of nested conditional branches

  • aco/print_asm: Sort block markers by block offset

  • aco: introduce notion of block_kind_loop_latch

  • aco/assembler: emit block_kind_loop_latch before the loop header

  • aco/insert_delay_alu: handle loop latch block before loop body

  • aco/lower_branches: Add try_rotate_latch_block() optimization

  • glsl_to_nir: emit loop continue construct

  • nir/divergence: rename divergent_loop_cf to divergent_cf

  • nir/divergence: Fix nir_block::divergent in presence of divergent breaks

  • nir/divergence: Ignore divergent_loop_{continue|break} for nir_block::divergent

  • nir/opt_remove_phis: recursively check loop header phis for triviality

  • nir/lower_continue_constructs: Simplify loops before lowering continue constructs

  • nir/lower_continue_constructs: Remove unnecessary handling of multiple continue statements

  • radv/rt: add and lower loop continue construct in traversal shaders

  • radv/dgc: add and lower loop continue construct

  • tgsi_to_nir: Add and lower loop continue constructs

  • dxil/nir: Remove nir_jump_continue from lower_subgroup_scan()

  • dozen: add and lower loop continue construct

  • nir/lower_goto_ifs: Add and lower loop continue constructs

  • ac: add and lower loop continue construct for streamout buffer info loop

  • tu/rt: add and lower loop continue construct in traversal shaders

  • lavapipe/rt: add and lower loop continue construct in traversal shaders

  • aco/tests: add and lower loop continue constructs in all tests which use continues

  • nir/tests: change opt_loop_peel_initial_break test to not use nir_jump_continue

  • nir: ensure that loop continue statements always link to continue constructs

  • nir: validate that loop continue statements always link to continue constructs

  • radv: call nir_opt_if late again

  • radv: increase limit for peephole_select in radv_optimize_nir_algebraic_early()

  • nir/opt_if: allow load_const instructions on ELSE-side for if-simplifaction

  • nir/opt_if: allow undef instructions on ELSE side for if-simplification

  • aco/isel: Remove loop_context* parameter from begin_loop() / end_loop() helper functions

  • aco/isel: Remove if_context* parameter from begin_if() / end_if() helper functions

  • aco/lower_branches: Don’t remove branches which jump over loops

  • aco/lower_branches: Fix try_rotate_latch_block()

  • aco/isel: remove handling of nir_jump_continue

  • aco/insert_exec_mask: remove handling of loop continues

  • aco/lower_phis: remove handling of block_kind_continue

  • aco/opt_value_numbering: remove handling of block_kind_continue

  • aco/lower_branches: remove handling of block_kind_continue

  • aco: remove remaining occurences of block_kind_continue

Daniel Stone (6):

  • panvk: Support VK_KHR_get_surface_capabilities2

  • panvk: Support VK_KHR_get_display_properties2

  • panvk: Support VK_EXT_acquire_drm_display

  • panvk: Support VK_KHR_present_id and present_wait

  • ci/panvk: Skip DRM WSI tests for v10/ASan

  • vulkan/wsi/wayland: Correctly map 24bpp format types

Danylo Piliaiev (23):

  • tu: Fix typo in min bounds calculation of FDM scissors

  • tu: Avoid disabling LRZ when possible for suspend/resume+depth-only draws

  • util/u_trace: Fix usage of variable-sized strings in non-queued case

  • docs/envvars: Document TU_DEBUG and IR3_SHADER_DEBUG

  • docs: Add documentation on how to debug GPU crashes and misrenderings

  • tu/a7xx: Fix D/S corruption when loading them via load_3d_blit

  • freedreno/rddecompiler: Fix shader editing when REG_BUNCH is used

  • ir3: Align TCS per-patch output to 64 bytes to prevent stale reads

  • tu: Fix double emission of PC_DS_CNTL due to missing break

  • tu: Add lrzWriteDisableReason to render_pass tracepoint

  • tu: vk_dont_care_as_load should not affect internal DONT_CARE cases

  • tu: Store gmem attachments after custom resolve in dyn RP

  • tu: Fix stomping of D/S test for custom resolve with D/S

  • tu: Don’t read .patch_input_gmem of unused attachment

  • tu/kgsl: Better detection of sparse support

  • tu: Fix imported memory not being affected by DEVICE_ADDRESS_CAPTURE_REPLAY

  • tu: u_trace usage fixes before u_trace refactoring

  • tu/autotune: Improve RP hash

  • tu: Support EXT_shader_image_atomic_int64

  • tu: Support transfer commands for R64 formats

  • tu: Add tracepoints for clear/copy/blit/lrz ops

  • tu: Fix CP_CCHE_INVALIDATE not being applied at the right point

  • tu/u_trace: Fix explicit toggle_name not being used

Dave Airlie (12):

  • lavapipe: add NV_cooperative_matrix2 flexible dimensions support

  • lavapipe: add NV_cooperative_matrix2 conversions support

  • nir: add cmat call to propogate invariants

  • lavapipe: add NV_cooperative_matrix2 reductions support

  • lavapipe: add support for NV_cooperative_matrix2 per element operations

  • gallivm: handle u16 correct on const loads.

  • st/mesh: handle mesh shader point size

  • nvk: don’t set sector promotion on texture headers

  • nouveau: drop sector promotion.

  • gallivm: handle llvm 22 coroutine end change

  • gallivm: handle llvm 22 scatter/gather intrinsic changes.

  • lavapipe: treat NULL pColorAttachmentLocations as no handles

David Headrick (2):

  • dozen: Add support for VK_EXT_memory_budget

  • dozen: Resolve Seg Fault in dzn_physical_device_create

David Rosca (40):

  • radeonsi/vcn: Clean up decode flags

  • radeonsi/vcn: Add low latency decode debug option

  • radv/video: Use coded size from session params instead of codedExtent

  • radv/video: Fix maxActiveReferencePictures for H265 decode

  • radv/video: Support AV1 encode frame size override

  • ac: Add video decode interface

  • ac: Add VCN ac_video_dec implementation

  • ac: Add VCN JPEG ac_video_dec implementation

  • ac: Add UVD ac_video_dec implementation

  • radeonsi: Don’t assert when using src texture in si_compute_clear_copy_buffer

  • radeonsi/video: Add video decoder using ac_video_dec

  • radeonsi/video: Remove old VCN and UVD decode implementation

  • radeonsi: Rename si_uvd_create_decoder to si_video_codec_create

  • radeonsi: Rename si_uvd.c to si_video.c

  • radeonsi: Move functions from radeon_video.c to si_video.c

  • radeonsi/video: Drop offsets parameter for si_vid_resize_buffer

  • radv/video: Remove old VCN and UVD decode implementation

  • radv: Drop videoarraypath debug option

  • ac/vcn_dec: Make the helper functions static

  • radv/video: Support UVD decode on hawaii and older

  • ac/vcn_dec: Fix tier2 dpb array size

  • vl: Also disable MPEG2 Main profile when mpeg12 decode is disabled

  • ac: Add variable slice mode interface

  • radeonsi/vcn: Use variable slice mode when possible

  • radv/video: Fix AV1 encode min tile size

  • radv/video: Fix coding pic_parameter_set_id in H264 slice header

  • frontends/va: Fix leaking H264/5 PPS/SPS objects when decoder wasn’t created

  • frontends/va: Fix leaks when create_video_codec fails

  • radv/video: Use quality level for encode preset instead of tuning mode

  • radeonsi: Set multi plane format also for imported textures

  • radeonsi/video: Fix setting decode surface format for single plane formats

  • radv/video: Remove unused function radv_vcn_sq_start

  • radeonsi/vcn: Remove encode op_preset overrides

  • radeonsi/vcn: Don’t force balance encode preset with sao on VCN5

  • d3d12: Use HEVC RefPicSet order from frontend

  • ac/parse_ib: Fix printing enc recon VAs on VCN5

  • radv/video: Fix initializing rc structs with default rate control

  • frontends/va: Fix finding LTRs from POCs in HEVC decode

  • frontends/va: Fix setting output color properties from color standard

  • frontends/va: Add missing NULL check for additional output surface

Derek Lesho (1):

  • zink: Guard bo map/unmap on map_count.

Dhruv Mark Collins (24):

  • freedreno,u_trace: Fix various UBSAN errors

  • tu: Increase clang-format ColumnLimit to 120

  • tu: Move tu_autotune_end_renderpass as late as possible

  • tu: Rewrite autotune in C++

  • util/rand_xor: Add extern C for C++ compatibility

  • tu/autotune: Add “Profiled” algorithm

  • util/math: Add ROUND_DOWN_TO_NPOT

  • tu/autotune: Prefer SYSMEM when only SW binning is possible

  • tu/autotune: Disable autotuning for small renderpasses by default

  • tu/autotune: Add “Preempt Optimize” mode

  • tu/autotune: Add prefer SYSMEM/GMEM mode

  • tu+util: Allow setting autotune mode from driconf

  • tu+util: Prefer SYSMEM for DXVK/VKD3D

  • tu/autotune: Add render mode locking to PROFILED algorithm

  • tu/autotune: Allow 99% max probability in profiled mode

  • tu/autotune: Only lock RPs sustain certain mode for 30s

  • freedreno/fdperf: Detect when counter values are invalid

  • zink+turnip/ci: Add failures uncovered by new autotune

  • tu: Disable features using performance counter for KGSL

  • tu: Only emit preempt optimization ambles when active

  • tu/autotune: Fail gracefully when CP counters are unavailable

  • fd/pps: Allocate performance counters from high-to-low

  • tu/autotune: Allocate performance counters from low-to-high

  • tu/query_pool: Avoid CP counter conflict with autotune

Dmitry Baryshkov (1):

  • freedreno/ci: update nightly expectations

Dmitry Osipenko (7):

  • intel: Check for userptr UAPI presence

  • intel: Add virtio-gpu native context

  • iris: Open-code drm prime ioctls

  • iris: Support virtio-gpu native context

  • anv: Support virtio-gpu native context

  • crocus: Use intel_ioctl() consistently

  • crocus: Support virtio-gpu native context

Duncan Brawley (4):

  • pvr: add basic support for shader statistics framework

  • pvr: Add support for VK_KHR_pipeline_executable_properties

  • pco: Use vertex input registers in register allocation

  • pco: Fix pco_last_igrp returning the first element instead of the last

Dylan Baker (15):

  • bin/pick: When the main widget is replaced, trigger a redraw

  • docs: add release notes for 25.3.4

  • docs: Add SHA sums for 25.3.4

  • docs: update calendar for 25.3.4

  • docs/release-calendar: Update calendar for 1 week bump

  • docs: add release notes for 25.3.5

  • docs: Add 25.3.5 SHA sums

  • docs: update calendar for 25.3.5

  • vulkan/runtime: Tie vulkan log printing to debug option rather than buildtype

  • docs: add release notes for 25.3.6

  • docs: Add SHA sums for 25.3.6

  • docs: Fix unescaped `*` in 25.3.6 release notes

  • docs: update calendar for 25.3.6

  • intel/tools: Don’t allocate in noop_drm_shim until after error checking

  • anv: assert we haven’t gone over the maximum number of push_buffers

Ella Stanforth (13):

  • vulkan: add plane aspect format helper

  • vulkan/runtime: use nir_shader_tex_pass for ycbcr lowering

  • pvr: fix transfer double stride

  • pvr/csbgen: fix packing multiple addresses

  • pvr: add multiplanar format support

  • pvr: handle packing texstate for ycbcr images

  • pvr: handle ycbcr swizzle

  • pvr: handle plane addresses for ycbcr images.

  • pvr: setup csc tables

  • pvr: implement chroma swap

  • pvr: workaround hardware clamping for YCBCR_IDENTITY conversion

  • pvr: add ycbcr formats

  • pvr: enable sampler ycbcr conversion

Emma Anholt (74):

  • nir: Fix C UB in imad24_ir3 evaluation.

  • nir/opt_algebraic_tests: Allow testing imad24_ir3.

  • nir/opcodes: Define the mul/mad_relaxed opcodes to return poison for OOB.

  • nir/opt_algebraic_tests: Allow testing mul/mad_relaxed opcodes.

  • nir/opcodes: Define udiv_aligned_4 to return poison for not-aligned-4.

  • nir/opt_algebraic_tests: Allow testing udiv_aligned_4.

  • nir/opt_algebraic_tests: Allow testing of fdot*_replicated opcodes.

  • nir/opt_algebraic_tests: Add support for expression swizzles.

  • nir/opt_algebraic_tests: Remove unnecessary input_count.

  • nir/opt_algebraic_tests: Move more of the base class code to be methods.

  • nir/opt_algebraic_tests: Rename and use the enum result type more.

  • nir/opt_algebraic_tests: Make sure we test the same inputs on BE as LE.

  • nir/opt_algebraic_tests: Test !nir_fp_preserve_signed_zero behavior.

  • nir/opt_algebraic_tests: Fix fuzzing levels for multi-component inputs.

  • nir/opt_algebraic: Fix a bit of imad24_ir3’s optimization.

  • nir/opt_algebraic_tests: Fix leak of the variable conds ht.

  • nir/opt_algebraic_tests: Fix annotating uint values.

  • nir/opt_algebraic_tests: Initialize an obvious dummy value for all defs.

  • ci: Skip dEQP-VK.wsi.direct_drm.

  • vulkan/wsi/display: Rename XCB RandR functions to mention “randr”

  • vulkan/wsi/display: Add some super useful debug messaging.

  • wsi/display: Fix up the swapchain init error paths.

  • vulkan/wsi/display: Avoid holding drm master for the device’s fd.

  • isaspec: Print a useful error for an assert I hit.

  • isaspec: Improve debug info for extractor_fallback().

  • isaspec: Print the bitset we’re processing when missing a field.

  • isaspec: Print the bit number when just a single bit is undefined.

  • ir3/tests: Print a helpful bit number on re-assembly failures.

  • nir,spirv: Add support for SPV_QCOM_image_processing.

  • ir3: Refactor bindless tex src info collection.

  • ir3: Add support for VK_QCOM_image_processing opcodes.

  • tu: Implement VK_QCOM_image_processing.

  • vulkan/wsi: Add some comments about how the vblank/flip sequencing happens.

  • wsi/display: Delete dead vblank_handler path.

  • vulkan/wsi: Delete ancient libdrm support for the page flip handler.

  • nir: Bump test timeouts.

  • tu: Add support for VK_FORMAT_E5B9G9R9_UFLOAT_PACK32 color attachments

  • lima/ci: Remove erroneous skips.

  • ci/freedreno: Clean up some expectations for the nightlies.

  • ci/intel: Clean up some expectations for the nightlies.

  • ci/tu: Skip more subgroup.clustered vector tests pre-merge.

  • ci/tu: Clear stale xfails from the nightlies.

  • ci: Add some flakes that I tripped over when test groups got reshuffled.

  • ci/deqp-runner: Drop silly catting of flakes/skips files together.

  • ci/deqp-runner: Bump to 0.23.2 for single-threaded and vkd3d support.

  • ci/deqp-runner: Enable a common single-threaded test list.

  • ci/vulkan: Enable dEQP-VK.wsi.direct_drm testing.

  • ci/rpi4: Move OOM-causing test skips to the single-thread list.

  • ci/tu: Move vkd3d-proton testing from nightly to pre-merge.

  • ci/zink: Skip ext-no-config-context for now, due to taking out the X server.

  • freedreno/a5xx: Convert a bunch of LO/HI regs to 64-bit regs.

  • docs/xlib: Drop docs about long-dead X features.

  • docs/xlib: Stop calling the fake GLX xlib frontend the most mature of sw.

  • docs/helpwanted: Rewrite this page.

  • docs/debugging: Drop this page.

  • docs/perf: Drop a bunch of old hints on getting faster GL1 swrast.

  • docs/shading.rst: Drop a bunch of old docs about the GLSL compiler.

  • docs: Rename the “shading” page to “GLSL” since it’s just that.

  • docs/index: Move the Xlib software GL driver with the other drivers.

  • docs: Conformance is done through SFC now, not SPI.

  • docs/systems: Minor touch-ups from reading the page.

  • ci/vulkan: Single-thread non-drm VK_KHR_display testing as well.

  • ci/radv: Add some flakes I hit while testing WSI.

  • ci/radv: Enable WSI testing.

  • ir3/ra: Clean up the ra_ctx_dump() output a bit.

  • ir3/ra: Fix DOUBLE_ONLY limit pressure computation.

  • ir3/ra: Assert that our calculated pressures aren’t bigger than the reg file.

  • freedreno/crashdec: Print an error instead of crashing on fopen() fail.

  • nir: Do NIR_DEBUG=print under a lock.

  • vulkan/wsi/display: Don’t re-probe connectors in between hotplugs.

  • vulkan/wsi/display: Check with an atomic commit if the swapchain fails.

  • ir3: Fix shared IMAD24 lowering.

  • tu: Add capture/replay for sparse buffers and descriptor buffer.

  • screenshot-layer: Fix leftover VK queues in the map at DeviceDestroy.

Emre Cecanpunar (1):

  • aco: drop optimizer peephole TODO comment

Eric Engestrom (124):

  • VERSION: bump to 26.1

  • docs: reset new_features.txt

  • docs/releasing: s/pull request/merge request/

  • docs/releasing: rephrase sentence about not letting the mr label script run

  • docs/releasing: skip ci when creating the branchpoint

  • docs: update calendar for 26.0.0-rc1

  • pick-ui: update for python 3.14 support

  • pvr/ci: document fixed tests

  • pvr/ci: sort expectations

  • pvr/ci: document last night’s flakes

  • docs/release-calendar: add 26.1 branchpoint and dates

  • hk: enable VK_EXT_present_timing

  • nir/meson: fix cpp_args of nir_opt_algebraic_pattern_tests

  • docs: update calendar for 26.0.0-rc2

  • Revert “meson: static link spirv-tools for darwin”

  • docs: update calendar for 26.0.0-rc3

  • pvr/drm-shim: avoid trying a random bvnc by default

  • pick-ui: add `Backport-to: *` as a synonym to `Cc: mesa-stable`

  • pvr/ci: rename deqp suite to a less generic name

  • pvr/ci: simplify the renderer string check

  • ci: split get_job_seconds() computation out of get_current_minsec() formatting

  • ci: sync time domains

  • mr-label-maker: label wsi files that have a label

  • Revert “add VK CTS validation report for a0 interpolation fix”

  • bin/gen_release_notes: fix support for python 3.14

  • docs: add release calendar for the 26.0.x cycle

  • docs: add release notes for 26.0.0

  • docs: add sha sum for 26.0.0

  • ci/deqp-runner: also limit the number of test log and caselist files

  • mr-label-maker: add label CI to bin/ci/*

  • docs/precompiled: modernize distro documentation

  • docs/precompiled: document debug packages

  • ci: close individual build sections by default

  • ci/vkd3d: drop duplicate submodule update

  • ci/vkd3d: fetch only the desired commit

  • ci/vkd3d: drop separate build directory

  • ci/vkd3d: drop no-longer-needed file

  • ci/vkd3d: ensure test-runner.sh has the right mode

  • ci/vkd3d: keep tests/ directory structure

  • ci/vkd3d: only build and install the files we actually want

  • ci/vkd3d: update tags

  • ci-tron: add job template for the x86_64 video test image

  • radv/ci: add vulkan fluster job on navi48

  • marge/pipeline_message: print job status for jobs still running or waiting for manual action

  • marge/pipeline_message: consider any job that hasn’t succeded as problematic

  • marge/pipeline_message: print details for any pipeline that hasn’t succeeded

  • ci: drop redundant MESA_IMAGE

  • nvk+zink/ci: update fails & flakes for nightly jobs

  • ci/build: include rusticl in debian-build-x86_64

  • nvk+zink/ci: add rusticl testing

  • docs: update calendar for 26.0.1

  • docs: add release notes for 26.0.1

  • docs: add sha sum for 26.0.1

  • docs/linkcheck: ignore a few more websites that don’t allow linkcheck

  • Revert “ci/gitlab_gql: disable schema fetch”

  • etnaviv/ci: update expectations

  • i915/ci: update expectations

  • r300/ci: update expectations

  • nvk/ci: update expectations

  • freedreno/ci: update expectations

  • anv/ci: update expectations

  • venus/ci: update expectations

  • radv/ci: document recent flakes

  • radeonsi/ci: document recent flakes

  • broadcom/ci: document recent flakes

  • etnaviv/ci: document recent flakes

  • freedreno/ci: document recent flakes

  • anv/ci: document recent flakes

  • nvk/ci: document recent flakes

  • llvmpipe/ci: document recent flakes

  • lavapipe/ci: document recent flakes

  • docs/linkcheck: ignore one more website that doesn’t allow linkcheck

  • docs: update link to ubuntu’s debug symbols documentation

  • etnaviv/ci: fix expectation

  • anv/ci: document more flakes

  • etnaviv/ci: fix expectation

  • freedreno: fix a few missed afuc -> qrisc renames

  • docs: update calendar for 26.0.2

  • docs: add release notes for 26.0.2

  • docs: add sha sum for 26.0.2

  • ci: let shader-db run on regular runners

  • ci: yaml-toml-shell-test runs on generic runners, not hw farm runners

  • ci: drop workaround for jobs not being created in fork pipelines

  • ci: changing .gitlab-ci.yml itself also means the container jobs must exist

  • docs: update calendar for 26.0.3

  • docs: add release notes for 26.0.3

  • docs: add sha sum for 26.0.3

  • ci: fix rebase mistake

  • ci: fix scheduled pipelines

  • docs: fix release calendar

  • meson: flip python check to avoid nesting conditions in the next commits

  • meson: move mako lib check inside python check

  • meson: move yaml lib check inside python check

  • Revert “meson: Fix build break on f43, gentoo, etc”

  • docs: fix various pep8 issues

  • docs: replace html redirects with http redirects

  • docs: delete now-unused html_redirects extension

  • freedreno/ci: document regressions

  • llvmpipe/ci: document regressions

  • nvk/ci: give more time to nightly job nvk-ga106-vkcts-valve

  • radv/ci: document recent flakes

  • radeonsi/ci: document recent flakes

  • vc4,v3d/ci: document recent flakes

  • turnip/ci: document recent flakes

  • nvk/ci: document recent flakes

  • zink+radv/ci: document recent flakes

  • zink+lvp/ci: document recent flakes

  • ci: vmware farm is offline, stop using it

  • ci: abort init-stage2.sh early if install dir is missing

  • ci/init-stage2: symlink install dir between both CI_PROJECT_DIR paths

  • ci: drop redundant existance check before `rm -rf`

  • ci: always make sure the results dir is created, not just when changing its path

  • ci: only clean the artifacts folder if gitlab hasn’t already done it

  • ci-tron: ensure the test jobs start with a clean job folder

  • docs: update calendar for 26.0.4

  • docs: add release notes for 26.0.4

  • docs: add sha sum for 26.0.4

  • Revert “ci-tron: ensure the test jobs start with a clean job folder”

  • VERSION: bump for 26.1.0-rc1

  • .pick_status.json: Update to 806fcc6193e305c22366baa17ccf88c8e1da1bda

  • VERSION: bump for 26.1.0-rc2

  • .pick_status.json: Update to d4d7055aee547f452689f8165e21ca100869e6fe

  • VERSION: bump for 26.1.0-rc3

  • .pick_status.json: Update to 2b9e491b6789f60a7993cc9b74fe5ac7fa60c9c5

Eric Guo (2):

  • panfrost: Fix NULL pointer dereference in panfrost_emit_images

  • panfrost: disable round_to_nearest_even for NEAREST samplers

Eric R. Smith (15):

  • mesa: do not unbind general point when different indexed points are deleted

  • pan: add some missing formats to pan_nir_lower_framebuffer

  • panfrost: optimize blending with DST_ALPHA when there is no alpha

  • panfrost: remove I8_UNORM from the blendable format table

  • panfrost: adjust format in blend shaders

  • panfrost: blending fixes for Midgard

  • pco: fix a typo in the check for optimization looping

  • panfrost: fix texel buffer calculations

  • panfrost: fix typos in architecture detection

  • panfrost: add sysval for number of samples

  • panvk: store number of samples in unused bits in the attribute descriptor

  • pan: change image2DMSArray lowering to use Z instead of Y

  • panvk: remove a redundant conditional

  • panfrost: make sure INDEX_OFFSET is cleared

  • panfrost: add helper function for checking for active queries

Erico Nunes (3):

  • Revert “ci: lima farm maintenance”

  • lima: add support for srgb framebuffers

  • lima: add support for srgb textures

Erik Faye-Lund (56):

  • pvr: drop needless include

  • pvr: avoid needless dispatches in powervr winsys

  • pvr/srv: query full pvr_device_info on winsys init

  • pvr/srv: define per-arch winsys-ops

  • pvr: prepare pvr_winsys_render_ctx_create_info for multi-arch

  • pvr: prepare pvr_winsys_compute_ctx_create_info for multi-arch

  • gallium/aux: do not hard-code linear interpolation

  • gallium: make needless linear interpolation optional

  • panfrost: expose the prefer_persp cap

  • v3d: move a failure to a flake

  • pan/ci: mark new xfails

  • pan/ci: skip a few more slow tests

  • pan/ci: clean up t720 expectations

  • panfrost: group image load/store flags a bit

  • panfrost: also check for PAN_BIND_STORAGE_IMAGE

  • panfrost: expose EXT_shader_image_load_store

  • pvr: add basic volcanic hw-definitions

  • pan/ci: add missing xfails from nightly run

  • pan/ci: update list of DRM-related skips

  • pan/ci: add missing t720-flakes

  • pan/ci: g720 and t720 isn’t the same

  • pan/ci: add some more flakes

  • pan/ci: correct syntax for flakes

  • pan/ci: update missed expectation

  • pan/ci: add a slow test to the skip-list

  • gallium/dri: set LIBVA_DRIVERS_PATH in devenv

  • gallium/va: set up symlinks in build-dir

  • panvk: fixup v7 check

  • pan/lib: kill compiler-warning

  • pan/lib: harmonize default-case handling

  • pan/ci: update traces result

  • util: add common ycbcr coefficient math code

  • compiler/nir: use common ycbcr math

  • vulkan: use common ycbcr code

  • gallium/vl: rename scale/bias variables

  • gallium/vl: do not adjust matrix twice

  • gallium/vl: use common ycbcr helpers

  • pan/genxml: remove non-existent YUV Enable for AFRC

  • pan/lib: do not try to use stencil-aspect of color attachment

  • pan/lib: set srgb-flag for afrc render-targets

  • pan/lib: divide extent by tile-extend, not itself

  • pan/lib: drop redundant assign

  • panvk: fix incorrect sorting

  • panvk: advertise wsi maintenance extensions

  • pan/ci: move flake from fails to flakes file

  • panvk: remove unused flag

  • docs/panfrost: fix heading-levels

  • pan/ci: update expectations

  • panvk: drop out-of-date TODO

  • pan/lib: fix up afbc and linear layout

  • pan/lib: emit high bits of buffer-size

  • nouveau: do not report unsupported feature

  • radeonsi: remove old, unsupported cap

  • panvk: do not enable extension without required feature

  • panvk: do not enable extension without required feature

  • dri: deprecate post-processing dri-confs

Erik Kurzinger (2):

  • wsi/display: retrieve monitor name from EDID

  • wsi/display: retrieve monitor size from EDID

Faith Ekstrand (195):

  • nvk: Enable ZPASS_PIXEL_COUNT in draw_state_init()

  • nak: Make OpF2F take a F16v2 source

  • nak: Use .xx swizzles for f2f.32.16

  • nir/lower_bool_to_bit_size: Use the correct num_components for conversions

  • nir/lower_bool_to_bit_size: Make smarter canonicalization choices

  • pan/bi: Run lower_alu_width after opt_algebraic_late

  • pan/bi: Add support for unpack_32_4x8

  • pan/bi: Add support for unpack_64_2x32

  • pan/bi: Stop calling nir_lower_pack()

  • pan/bi: Don’t attempt to fuse AND(ICMP, ICMP) if the AND is swizzled

  • pan/bi: Vectorize comparisons

  • pan/bi: Set lower_vector_cmp

  • pan/bi: Allow vector booleans

  • panvk: Don’t emit storage descriptors for compressed views

  • pan/texture: ASTC is not allowed for storage

  • pan/clear: Stop packing undefined bits in colors

  • pan: Add a new framebuffer abstraction

  • nir/print: Add panfrost blend intrinsics

  • nir/gather_info: Add support for panfrost tile load/store intrinsics

  • pan/fb: Add a common FB load shader builder

  • pan/fb: Add a mapping to the old FB info

  • panvk/csf: Use a panvk_rendering_state temp variable

  • panvk: Add and use a new pan_ptr_offset() helper

  • panvk: Return frame shader DCDs and modes from cmd_fb_preload()

  • panvk: Stop doing the pre/post DCD offsetting in common code

  • panvk: Memset fb state to zero

  • panvk: Drop all the crc_valid stuff

  • panvk/jm: Emit FRAGMENT_JOB ourselves

  • util/format: Place PAN_FORMAT_FOO_START after the first format

  • panvk: Refactor resolveMode handling

  • pan/desc: Set clean_tile.write_zs for interleaved Z/S stencil clears

  • panvk: Use thew new pan_fb_layout for setting up attachments

  • panvk: Use the new FB code for tile size selection

  • panvk: Use the new structs to re-populate fbinfo in force_fb_preload()

  • panvk/csf: Use the new structs in prepare_incremental_rendering_fbinfos()

  • panvk: Add a version of fb_preload which takes the new structs

  • panvk/csf: Use the new pan_fb code for emitting FBDs

  • panvk/jm: Re-generate FB info in cmd_preload_fb_after_batch_split()

  • panvk/jm: Use the new pan_fb code for emitting fragment jobs

  • panvk: Create both Z/S descriptors, even for separate Z/S

  • panvk/preload: Stop assuming 32 registers

  • panvk: Switch to the new preload shader framework

  • panvk/preload: Populate our own texture descriptors

  • panvk/trace: Trace using pan_fb_layout instead of info

  • panvk: Use pan_fb_load to decide when we have a clear

  • panvk: Drop pan_fb_info

  • panvk/jm: Refactor BeginRendering()

  • panvk/jm: Rework load/store/spill

  • panvk: Insert a pipeline barrier if we have any FB loads

  • panvk: Also load output attachments with LOAD_OP_NONE+STORE_OP_NONE

  • panvk: Use partial FB preloads to deal with alignments

  • pan/desc: Pass image views directly to attachment helpers

  • pan/desc: Drop cbuf_offset from emit_*_color_attachment()

  • pan/desc: Set Z/S MSAA averaging mode in common code

  • pan/genxml: Make sections more typesafe

  • pan/fb: Fill out our own descriptors

  • pan/fb: Refactor load shader building

  • pan/fb: Add support for more MSAA modes in shaders

  • pan/fb: Separate MSAA ops into in_bounds and border

  • pan/fb: Add new shader ops for copying from RTs

  • pan/fb: Add an option to only write sample0 of a render target

  • pan/fb: Add a concept of resolve ops and resolve shaders

  • panvk/meta: Set color_attachment_count based on bound attachments

  • panvk: Always set pan_fb_layout.rt_count to at least 1

  • panvk: Move cmd_fb_preload to cmd_frame_shaders

  • panvk/frame_shaders: Set no_shader_depth/stencil_read

  • panvk/frame_shaders: Set modes in cmd_preload_*_attachments()

  • panvk/frame_shaders: Only allow preload shaders to be killed

  • panvk/frame_shaders: Add support for resolve shaders

  • panvk: Respect storeOp for color attachments

  • panvk: Use resolve shaders for color resolves

  • panvk: Also use resolve shaders for Z/S

  • pan: Add a pan_format_supports_hw_blend() helper

  • panvk: Optimize resolves if possible

  • pan/ci: Mark couple of WSI crashes as flakes

  • util/format: Add a util_format_get_depth_bits() helper

  • pan/fb: Improve depth format asserts

  • pan/desc: Pass emit_*_attachments args through a struct

  • pan/fb: Set reverse_issue_order when needed

  • panvk: Avoid direct MSAA resolves to AFBC on v6 and earlier

  • pan/fb: Figure out clean tile enables up-front

  • pan/fb: Set clean_tile_write for mismatched superblock/tile sizes

  • pan/fb: Force pre-frame shaders to ALWAYS for clean tiles

  • panvk: Relax ms2ss afbc disablement

  • etnaviv: Call lower_bool_to_int32 not to_bitsize

  • nir/lower_bool_to_bitsize: Make all bN_csel sources match

  • nir,panfrost: Move lower_bool_to_bitsize to panfrost

  • pan/bi: Be more careful about bit sizes in b2f lowering

  • pan/bi: Delete the b32csel special case and assert sizes match

  • pan/compiler: Handle store_per_view_output in collect_varyings()

  • pan/compiler: Add a pan_varying_layout struct

  • pan/bi: Stop pretending to support f16 gl_Position

  • nir: Allow 8-bit vertex output stores

  • pan: Add a pass to resize I/O load/stores as needed by the varying layout

  • pan/bi: Allow 8-bit varying direct stores

  • pan/bi: Use the pan_varying_layout for Vallhall+ direct varing load/store

  • pan/bi: Stop lowering point size to float16 early

  • Revert “nir: Add a type parameter to nir_lower_point_size()”

  • pan/nir: Improve collect_noperspective_varyings_fs()

  • panvk: Scrape noperspective varyings out of the FS first

  • panvk: Compile shaders in pipeline stage order

  • panvk: Build the VS varying layout early

  • panvk/csf: Emit varying descriptors based on the VS varying layout

  • pan/bi: Handle varying layout mismatches in emit_load_vary()

  • panvk: Pass the varying layout from the VS to the FS compile

  • pan: Add a helper for generating more compact varying layouts

  • panvk: Use a new, more compact varying layout

  • vulkan/render_pass: Always use separate depth/stencil layouts

  • nak: Report progress from nak_nir_rematerialize_load_const()

  • pan/bi: Add new FS input load intrinsics

  • pan/bi: Lower FS input loads in NIR

  • pan,nir: Rework converted_mem_pan intrinsics

  • pan/bi: Lower VS outputs in NIR

  • pan/bi: Drop lower_sample_mask_writes

  • pan/bi: Drop bifrost_nir_lower_blend_components()

  • nir: Consider if uses in nir_def_all_uses_*

  • nir: Support primitive_id in lower_sysvals_to_varyings

  • treewide: Enable lowering of primitive ID in a bunch of Vulkan drivers

  • nak: Move lowering of load_*_id to lower_vtg_io.c

  • nak: Add support for load_primitive_id

  • vtn: Use a system value for primitive ID in fragment shaders

  • panvk: Drop lower_load_fs_input

  • pan/bi/ra: Dump verbose debug logging to stderr

  • pan/bi: v2x16 conversions don’t replicate

  • pan/buffer: Add the offset to the size for buffer textures

  • pan/buffer: Drop pan_buffer_view::offset

  • panvk: Reduce minTexelBufferOffsetAlignment

  • panvk: Rework setting dyn_buf_offsets

  • panvk: Track which dynamic buffers are SSBOs

  • panvk: Increase robust buffer access alignments

  • panvk: Set min_ubo/ssbo_alignment in spirv_options

  • pan/bi: Always vectorize UBO access

  • panvk: Replace robust2_modes with robust_modes

  • pan/bi: Vectorize SSBOs when not robust

  • pan/bi: Allow 64-bit vectors in bi_make_vec_to()

  • pan/bi: Handle 64-bit sources in bi_alu_src_index()

  • pan/bi: Properly handle large 8-bit vectors in bi_alu_src_index()

  • pan/bi: Move nir_op_mov handling to the top

  • pan/bi: Handle pack_*_split with vecN

  • pan/bi: Unify handling of pack_*

  • pan/bi: Unify handling of unpack_*

  • pan/bi: Simplify unpack_64_2x32_split_*

  • pan/bi: Rework mem_vectorize_cb

  • nir: Add sampler and resource heap system values

  • nir: Add intrinsics for descriptor heaps

  • nir: Add tex sources for descriptor heaps

  • spirv: Improve the error message for invalid SPIR-V sections

  • spirv: Add new SPV_KHR_descriptor_heap Builtins

  • spirv: Handle OpTypeBufferKHR

  • spirv,vulkan: Implement OpConstantSizeOfKHR

  • spirv: Handle ArrayStrideIdKHR and OffsetIdKHR decorations

  • spirv: Handle OpBufferPointerKHR

  • spirv: Mark DescriptorHeapKHR as implemented

  • vulkan: Rename some VK_EXT_descriptor_buffer properties

  • vulkan: Add a lowering pass for descriptor heap mappings

  • vulkan: Support descriptor heaps in vk_nir_convert_ycbcr()

  • vulkan/pipeline: Allow compiling compute/rt pipelines with a NULL layout

  • vulkan/shader: Call vk_nir_lower_descriptor_heaps()

  • vulkan: Add a vk_hash_descriptor_heap_mappings() helper

  • vulkan/pipeline: Reorder vk_pipeline_precomp_shader_deserialize()

  • vulkan/pipeline: Call vk_nir_lower_descriptor_heaps()

  • vulkan: Add a common implementation of GetPhysicalDeviceDescriptorSizeKHR

  • vulkan: Add a no-op implementation of [Un]RegisterCustomBorderColor()

  • pan/bi: Support more swizzle aliases in the bifrost pack code

  • pan/bi: Delete a few instruction encodings

  • pan/bi/ra: Allow offsets on tied sources

  • pan/bi: Add a bi_swizzle_from_half() helper

  • pan/bi: Compose swizzles in bi_half() and bi_byte()

  • pan/bi: Use bi_half() for texture MS indices

  • pan/bi: Return void from bi_swizzle_to_byte_channels()

  • pan/bi: Add a bi_swizzle_from_byte_channels() helper

  • pan/bi: Add a bi_try_compose_swizzles() helper

  • pan/bi: Add a bi_op_supports_swizzle() helper

  • pan/bi: Add a lowering pass for MKVEC and SWZ

  • pan/bi: Always use SWZ.v4i8 in bi_lower_swizzle()

  • pan/bi: Stop lowering swizzles on mkvec and swz

  • pan/bi: Emit MKVEC directly

  • pan/bi: Add bytewise copy propagation

  • pan/bi: Pack 8-bit vec2s

  • pan/bi: Vectorize 8-bit ops up to v4i8

  • pan/bi: Delete BI_SWIZZLE_1123

  • pan/bi: Add BI_SWIZZLE_NONE

  • pan/bi: Support all the swizzles in the packer

  • nir: Add a couple is_zero() helpers

  • pan/bi: Use nir_src_is_zero()

  • pan/bi: Handle arbitrary size constants

  • pan/nir: Stop doing manual optimization after resize_varying_io

  • pan/nir: Stop being so conservative about phi scalarizing

  • pan/nir: Use minimum-width constants instead of scalar

  • pan/bi: Simplify extract_i8 handling

  • nir: Add a nir_alu_src_comp_as_uint() helper

  • pan/bi: Handle vector 16-bit extract_[ui]8

  • pan/bi: Vectorize more conversions

  • panvk/csf: Emit INDEX_BUFFER[_SIZE] even for non-indexed draws

  • zink: Assert if we try to use a dedicated allocation with offset > 0

Felix DeGrood (4):

  • intel/tools: intel_measure.py correctly parse cmdbuf-only data

  • intel/tools: intel_measure.py avoid early exit on corrupted data

  • anv: report correct format for depth/stencil blorps in utrace

  • intel/decoder: update warning message when buildtype=release

Francisco Jerez (17):

  • intel/isl: Define ISL_AUX_STATE_COMPRESSED_HIER_DEPTH aux state.

  • iris/gfx12.5: Allocate indirect color state for depth surfaces.

  • iris/gfx12.5+: Keep HIZ_CCS aux usage while sampling from resolved depth surfaces.

  • anv/gfx12.5: Allocate indirect color state for depth surfaces.

  • anv: Use actual layout in anv_fast_clear_depth_stencil() instead of ANV_IMAGE_LAYOUT_EXPLICIT_AUX.

  • anv/gfx12.5: Can’t fast clear multisampled Z/S with HIZ CCS WT aux usage.

  • anv/gfx12.5: Resolve depth during layout transitions from ISL_AUX_STATE_COMPRESSED_HIER_DEPTH.

  • anv/gfx12.5: Infer ISL_AUX_STATE_COMPRESSED_HIER_DEPTH from anv_layout_to_aux_state().

  • anv/gfx12.5+: Keep HIZ_CCS aux usage while sampling from depth surfaces.

  • intel/measure: Define snapshot type for HiZ partial resolves.

  • intel/blorp: Add support for partial resolves of HiZ-CCS surfaces.

  • intel/isl: Teach ISL about HIZ CCS partial resolves.

  • anv/gfx12.5: Take advantage of partial resolves in depth layout transitions.

  • anv/gfx12.5: Apply HIZ-CCS resolve TC flush on full resolves for all gfx12.5.

  • iris/gfx12.5: Apply HIZ-CCS resolve DC flush after full resolves for all gfx12.5.

  • intel/isl: Add unit tests for ISL_AUX_STATE_COMPRESSED_HIER_DEPTH.

  • iris: Rework iris_sample_with_depth_aux() into helper that returns aux usage.

Francois Coulombe (1):

  • vulkan/wsi/headless: add sRGB swapchain format support

Frank Binns (14):

  • docs/pvr: fix some typos and wording

  • docs/pvr: some minor improvements

  • pvr/ci: document some recent flakes

  • pvr: remove asserts in pvr_get_image_subresource_layout()

  • pvr/ci: update fails to remove two tests that have started passing

  • pvr/ci: move some timing out tests from fails to skips

  • pvr: Fix alloc callbacks usage when freeing frame buffers

  • zink: add renderonly scanouts handling

  • zink: add a winsys library exposing renderonly screen creation

  • kmsro: wire Zink up as a fallback

  • pvr: re-enable fullDrawIndexUint32

  • pvr: re-enable multiDrawIndirect

  • pvr: re-enable depthBiasClamp

  • pvr: re-enable wideLines

GKraats (1):

  • crocus: Fix shader precompilation on Gen6 and higher

Georg Lehmann (297):

  • radv/gfx11: add a RADV_PERFTEST flag to expose bfloat16 cmat

  • aco/gfx12: use 64bit add/sub to swap sgprs

  • nir/opt_algebraic: optimize f2f16_rtz(b2f(a))

  • nir/opt_algebraic: optimize f2f16_rtz(min/max)

  • nir/opt_algebraic: remove f2f16 roundtrip conversions

  • nir/opt_algebraic: optimize f2f16_rtz of bcsel with constants

  • nir/opt_algebraic: optimize bcsel to hi 16bits with undef lo

  • aco/isel: optimize pack_32_2x16_split(undef, const)

  • aco/optimizer: fix parsing salu p_insert as shift

  • aco/optimizer: optimize pack(undef, f2f16_rtz(a)) for gfx6/7

  • aco/optimizer: optimize pack(undef, f2f16_rtz(a)) for gfx8+

  • aco/optimizer: optimize pack(undef, f2f16_rtz(a)) for salu

  • ac/nir/lower_ps_late: CSE partial packed exports

  • ac/nir/lower_ps_late: emit scalar f2f16_rtz for when one half of a packed export is undef

  • aco: fix demote in header of single iteration loop

  • aco: add a helper function for non supported DPP opcodes

  • aco/optimizer: use opcode_supports_dpp

  • aco: disable DPP for rev integer subs and shifts

  • aco/validate: allow dpp with scalar src1 on gfx11.5+

  • aco: undo operand swap if applying DPP fails

  • aco: don’t convert VOP3P to VOP3 when applying DPP

  • aco/ra: don’t move sgpr into v_fmac_f32_dpp src0

  • aco: apply DPP with scalar src1 on gfx11.5+

  • aco/optimizer: allow DPP with scalar src1 in alu_opt_info_is_valid

  • aco/optimizer: rework how dpp is applied

  • aco: only apply DPP with 3 or less uses

  • aco: allow v_cmpx with DPP

  • spirv: move NoContraction handling into vtn_handle_fp_fast_math

  • spirv: handle fast_math for opencl opcodes

  • spirv: use base type instead of bit size to determine fp_math_ctrl

  • spirv: consider both source and dest type for fast math

  • spirv: remove vtn_builder::exact

  • spirv: assert fp_math_ctrl was reset after use

  • nir/opt_algebraic: use correct syntax to create exact fsat

  • nir/algebraic: terminate opcode regex

  • nir/algebraic: remove manual opcode validation

  • nir/opt_algebraic: use contract instead of inexact for more patterns

  • nir/opt_algebaric: improve a < 0.0 ? 0.0 : sqrt(a) pattern

  • nir/opt_algebraic: optimize pack_half_2x16_rtz of float converted from 16bit

  • nir/search: gather union of all fp_math_ctrl

  • nir/search: preserve nan/inf/sz if any alu in a replaced expression did

  • nir/opt_algebraic: rework ignore_exact to work like other internal conditions

  • nir/algebraic: remove ability to create Value from Expression

  • nir/algebraic: make subexpression inexact on creation

  • nir/opt_algebraic: optimize unpack_32_2x16 of extract

  • nir/lower_alu_width: emit f2f32 for unpack_half_2x16

  • nir/lower_tex: use f2f32 instead of unpack_half

  • nir/opt_16bit_tex_image: remove unpack_half support

  • nir/format_convert: use f2f32 instead of unpack_half

  • nir/opt_algebraic: remove unpack_half_2x16_split

  • aco: remove unpack_half support

  • ac/llvm: remove unpack_half support

  • nak: remove unpack_half support

  • kk/compiler: remove unpack_half support

  • broadcom/compiler: use f2f32 when lowering image load

  • broadcom/compiler: remove unpack_half support

  • asahi/compiler: remove unpack_half support

  • brw/lower_storage_image: use f2f32 instead of unpack_half

  • brw: remove unpack_half support

  • elk/lower_storage_image: use f2f32 instead of unpack_half

  • elk: remove unpack_half support

  • microsoft/compiler: switch to a backend specific unpack half opcode

  • nouveau/codegen: remove support for unpack_half

  • panfrost/compiler/bi: remove unpack_half support

  • r600/sfn: implement minimal 16bit f2f32 support

  • r600/sfn: lower unpack_half to f2f32

  • r600/sfn: remove unpack_half support

  • nir: remove split unpack_half opcodes

  • aco: clean up emit_extract_vector a bit

  • aco/optimizer: repeat vector of split opt

  • aco/optimizer: don’t remove label_extract for splits

  • aco: improve emit_extract_vector for vector of vecs

  • aco/isel: split vector into dwords/words first

  • aco/isel: avoid extracts for continuous alu src components

  • aco/optimizer: apply further extracts to v_cvt_f32_ubyte

  • aco/optimizer: apply byte p_split_vector as extract

  • aco/optimizer: add second copy prop for pseudo instructions

  • aco/optimizer: only copy propagate p_split_vector if it can be eliminated

  • aco/optimzer: apply extract with any uses

  • aco/optimizer: use nan preserve flag to prevent incorrect med3

  • spirv: use nan/inf preserve instead of exact for fp compare

  • spirv: use nan/inf preserve for glsl.std.450 min/max instead of exact

  • mesa/prog_to_nir: use nan/inf preserve instead of exact for kill’s flt

  • gallium/ttn: use nan/inf preserve instead of exact for kill’s flt

  • radv/nir/rt: preserve inf/nan for emulated RT intersect

  • nir/format_convert: use nan/inf preserve flag for fmax instead of exact

  • nir/lower_double_ops: don’t create more exact ops than the input requires

  • nir/lower_uniform_subgroup: use nan/inf preserve instead of exact for feq

  • glsl: preserve inf/nan for precise/invariant

  • glsl: make fp (not) equal always nan/inf preserving

  • glsl: make fmin/fmax/fsat nan/inf preserving

  • nir/search: add option to set nan/inf/sz preserve on replacement patterns

  • nir/opt_algebraic: use nan/inf/sz preserve flags instead of exact for cmp/min/max replacement

  • nir/opt_algebraic: mark newly created fmulz nan/inf preserving

  • nir: remove special fp_math_ctrl rules

  • nir/opt_algebraic: remove inexact a * 0.0 patterns

  • nir/opt_algebraic: add a - a with nnan

  • nir/opt_algebraic: mark fmulz(finite, finite) -> fmul pattern as nsz

  • nir/opt_algebraic: remove inexact from floor->trunc pattern

  • nir/opt_algebraic: make pattern pushing fmul into bcsel exact

  • nir/opt_algebraic: make 1.0 - fsat(a) -> fsat(1.0 - a) pattern exact using nnan

  • nir/opt_algebraic: remove redundant patterns with fcmp(fneg(…), #c)

  • nir/opt_algebraic: use better float control for some fcmp patterns

  • nir/opt_algebraic: optimize flt/fge(#c, fadd(a, #b))

  • nir/opt_algebraic: update flt -> fneu patterns

  • nir: make alu fp_math_ctrl helpers const

  • nir/opt_algebraic: make bcsel(fcmp(b, a), b, a) -> fmin/fmax patterns exact

  • nir/opt_algebraic: make some fmin/fmax/fsat patterns exact using nsz/nnan

  • nir/opt_algebraic: make some more fcmp patterns exact using nnan

  • nir/opt_algebraic: make ffract(is_integral) exact using nnan

  • nir/opt_algebraic: make a < 0.0 ? -a : a exact using search helpers

  • nir/opt_algebraic: make fcmp(a+b, 0.0) -> fcmp(a, -b) exact using ninf

  • brw/nir_lower_fsign: try to fix NaN correctness

  • nir/algebraic: allow inexact optimizations with sz/inf/nan preserve

  • aco/optimizer: stop checking precise for med3

  • ci: skip invalid float_control2 tests

  • ci: update trace checksums

  • anv/ci: add cross signed zero expected fails

  • aco/lower_branches: consider jump target of conditional branches based on vcc

  • aco: handle all SALU that modifies PC in needs_exec_mask

  • aco/opt_postRA: don’t optimize across calls

  • aco: remove redundant can_use_DPP declaration

  • nir/opt_algebraic: remove few uses of integer nir_analyze_range

  • nir: remove non float nir_analyse_range support

  • nir: rename nir_analyze_range because it’s float only

  • nir: let nir_analyze_fp_range take a nir_def

  • nir/opt_algebraic: optimize 1.0 - b2f(a) to b2f(inot(a))

  • ci: disable debian-ppc64el and debian-s390x

  • zink: do not check type when emitting fp_fast_math_mode

  • nir/serialize: omit serializing fp_math_ctrl if it has to be 0

  • nir/opcodes: remove valid_fp_math_ctrl bits from some opcodes

  • nir/opt_algebraic: preserve signed zero when creating new b2f

  • nir/opt_algebraic: create more b2f if sign of zero doesn’t matter

  • vulkan,spirv: update headers

  • nir: add mixed float dot opcodes

  • spirv: implement SPV_VALVE_mixed_float_dot_product

  • aco: mixed float dot product opcodes

  • aco/ra: create v_dot2c_f32_f16

  • aco: allow modifiers for fp16 dot

  • aco: allow dpp for fp8/bf8 dot4

  • ac/llvm: implement mixed float dot

  • radv: expose VK_VALVE_shader_mixed_float_dot_product on supported hardware

  • nir/lower_subgroups: lower shuffles and bitwise reduce to 32bit before scalarizing

  • aco/insert_fp_mode: don’t skip setting round for fract

  • nir: print all fp_math_ctrl bits

  • nir/opt_algebraic: optimize b2f(a) - 1.0 to -b2f(a)

  • nir/opt_algebraic: optimize d3d9 iand(a, inot(b))

  • nir/opt_algebraic: generalize late fcmp(fneg(a), const) patterns

  • nir/opt_algebraic: remove fcmp+fneg patterns that are cleaned up earlier

  • nir/opt_algebraic: optimize all comparisons of b2f/b2i with constants

  • nir_opt_algebraic: remove more specific cmp+bcsel opts

  • nir/opt_algebraic: remove loops for b2f/b2i equality handling

  • nir_opt_algebraic: reassociate fadd into ffma where one factor is a constant

  • nir_opt_algebraic: remove unneeded is_not_const

  • nir/opt_algebraic: remove is_used_once on outer instruction

  • ci: update expectations

  • nir/opt_algebraic: optimize b2i(a) * b to bcsel

  • nir/lower_bool_to_float: assert that vector comparisons were lowered

  • nir: remove fall_equal/fany_nequal opcodes

  • ntt: lower vector comparisons using nir_lower_alu_to_scalar

  • i915,nv30,softpipe,svga,mesa/st: remove lower_vector_cmp for tgsi backends

  • kk,llvmpipe,nouveau: remove lower_vector_cmp from scalar backends

  • zink: use nir_lower_alu_to_scalar to lower vector compare

  • bifrost: use nir_lower_alu_width to lower vector comparisons

  • etnaviv: use nir_lower_alu_width to lower vector compare

  • lima: use nir_lower_alu_width to lower vector compare

  • freedreno/ir2: use nir_lower_alu_width to lower vector compare

  • r300: use nir_lower_alu_width to lower vector compare

  • nir: remove lower_vector_cmp

  • nir/opt_algebraic: remove pattern that skips iabs with range analysis

  • nir/opt_algebraic: fix frsq clamp pattern

  • nir/opt_algebraic: add is_not_uint_zero for b2i16(uge) pattern

  • nir: rewrite fp range analysis as a fp class analysis

  • nir: add fp class analysis for sin/cos

  • nir: add fp class analysis for flog2

  • nir: add fp class analysis for some intrinsics

  • nir: add fp class analysis for shadow compare

  • nir: add fp class analysis for fsub

  • nir: add fp class analysis for fround_even

  • aco/isel: skip min/max for SALU fsat if possible

  • nir/gather_tcs_info: use nir_analyze_fp_class directly

  • nir/search_helpers: switch to fp class analysis

  • nir: remove nir_analyze_fp_range

  • nir/search_helpers: use fp class analysis more

  • nir: remove more fsat using range analysis

  • nir: add fp class analysis tests

  • panvk/ci: document new crashes on bifrost

  • nir: add a pass to optimize fp_math_ctrl

  • radv: use nir_opt_fp_math_ctrl

  • nir: create more fsat using range analysis

  • nir/opt_algebraic: remove min(a, >= 1.0) before fsat

  • nir/opt_algebraic: skip more redundant alignment iand

  • nir: don’t assume indicies are always 32bit when accessing them as raw data

  • nir: support intrinsic indicies larger than 32 bits

  • nir: merge xfb and xfb2 into one 64bit intrinsic index

  • nir: add free bits in nir_io_semantics for future use

  • aco/scheld_vopd: make VOPDInfo more flexible by adding a swizzle

  • aco/sched_vopd: convert fma with inline constants to fmamk/fmaak

  • aco/opt_postRA: remove try_convert_fma_to_vop2

  • aco/sched_vopd: create dot2acc from VOP3P dot2

  • aco/ra: try to allocate registers for dot2 to allow VOPD

  • aco/ra: don’t tie definition when the operand is in a preserved reg

  • nir: fix nir_intrinsic_copy_const_indices for large indices

  • nir: add no_signed_zero flag to io semantics

  • nir/opt_fp_math_ctrl: handle input/output no_signed_zero flag

  • radv: set no_signed_zero for FS store_output when format doesn’t care

  • nir/opt_load_store_vectorize: use nir_intrinsic_has_align_mul

  • nir/gather_info: use nir_intrinsic_has_io_semantics

  • nir/lower_frexp: preserve fp_math_ctrl

  • ac/nir/cull: make fisfinite nan/inf preserving

  • nir/opt_algebraic: lower ninf fisfinite correctly

  • glsl: reset fp_math_ctrl when changing it per alu

  • glsl: make flt/fge/fabs/fneg inf preserving

  • nir/opt_algebraic: be more strict when optimizing fcmp(a + #b, #c)

  • ntt: set precise correctly for comparisons/min/max

  • nir/tests: test algebraic patterns with maximum fp_math_ctrl

  • nir/search_helpers: assume float sources without preserve flag can’t be inf/nan

  • nir/opt_algebraic: take advantage of range helpers including nnan

  • nir/opt_algebraic: turn fabs(a) into fneg(a) if a is not positive

  • nir/opt_algebraic: remove manual pattern that removes fmax(…, 0.0)

  • nir/opt_algebraic: remove manual patterns that optimizes flt([0.0, 1.0], 0.0)

  • nir/opt_algebraic: move some fsat patterns next to the other fsat patterns

  • nir/algebraic/tests: invert all excluded fp_math_ctrl flags

  • ir3: set progress for nir_opt_large_constants

  • nir/opt_large_constants: don’t add constants implemented with ALU to the constant data

  • nir/opt_large_constants: set fp_math_ctrl for bit exact results

  • nir/opt_large_constants: enable small constant optimization for non trivial strides

  • nir/opt_large_constants: optimize small vector constant arrays

  • nir/opt_large_constants: support negative small constants

  • nir/opt_large_constants: handle floating point power of two fractions

  • nir/opt_large_constants: only use 16bit float alu when supported

  • radv/ci: update restricted trace checksum

  • nir: replace lower_ldexp with has_ldexp

  • nir/opt_algebraic: create ldexp from exp2

  • nir/opt_algebraic: optimize more near useless bcsel

  • nir: rework nir_alu_src_is_trivial_ssa to take an alu src

  • nir/search: never insert movs for alu uses

  • nir/opt_algebraic: optimize more fmulz(1.0, a) remains

  • nir/opt_algebraic: optimize b2f(a) * b

  • broadcom/ci: skip rpi4 timeout

  • aco: skip fract for sin/cos on gfx6-8 if the src is already in range

  • gallivm: don’t optimize fadd(a, 0.0) with signed zero preserve

  • gallivm: remove dead code in lp_build_add

  • aco/optimizer: apply dpp to v_dot before RA for gfx10.3

  • nir/opt_fp_math_ctrl: ignore ffract input sign of zero

  • nir: add fp_math_ctrl as intrinsic index

  • nir: add fp_math_ctrl to ddx/ddy

  • nir/opt_uniform_subgroup: use ddx/ddy fp_math_ctrl

  • nir/opt_fp_math_ctrl: use ddx/ddy fp_math_ctrl

  • nir: add fp_math_ctrl to cmat alu ops

  • spirv: set fp_math_ctrl for cmat alu

  • radv: preserve fp_math_ctrl when lowering cmat alu ops

  • nak: preserve fp_math_ctrl when lowering cmat

  • brw: preserve fp_math_ctrl when lowering cmat alu

  • lavapipe: preserve fp_math_ctrl when lowering cmat alu

  • nir: add fp_math_ctrl to convert_alu_types

  • aco: spill VGPRs to LDS if it doesn’t further limit occupancy

  • aco: allow spilling to LDS in RT shaders without stack pointer

  • nir/lower_non_uniform_access: fix fusing loops for same index but different array variable

  • nir/lower_tex: fix lowering 16bit textureGatherOffsets

  • nir,radv: lower shadow compare gather to 16bit

  • aco/isel: use s_bitcmp1 for 1bit ubfe

  • nir/opt_algebraic: remove a few non 1bit bool patterns

  • nir/search: remove matching variable type

  • nir/opt_large_constants: optimize constant arrays with just two different values

  • nir/opt_load_skip_helpers: don’t skip helpers for store_scratch data

  • nir/opt_algebraic: update open coded flerp(…, b2f(c)) to bcsel patterns

  • nir/opt_algebraic: move some lower_lerp patterns

  • aco/isel: optimize 16/64bit non constant valu bit test

  • nir/opt_algebraic: create more 64bit bit test

  • aco/optimizer: do not try to create 3 byte constant operands

  • aco/spill: fix mixed lds+scratch spill/reload

  • nir: split exact bit into no_contract/reassoc/transform

  • spirv: map float control2 to fine grained nir flags instead of exact

  • nir/algebraic: actually seperate contract and inexact

  • nir/opt_reassociate: use nir_fp_no_reassoc instead of exact

  • intel/peephole_fma: use nir_fp_no_contract instead of exact

  • zink/ntv: seperate float control2 exact bits

  • aco/tests: fix med3 NaN tests

  • aco: use no contract/reassoc instead of exact

  • radv: clarify that copy prop is required for correctness after D16 opt

  • radv: remove point size in lowered io

  • radv: do not remove point size variable

  • radv: immediately remove phis after loop unrolling

  • radv: call radv_optimize_nir after lowering io

  • radv: remove radv_remove_color_exports

  • radv: lower lowered io to scalar

  • radv: do not vectorize fs out variables

  • radv: do not vectorize io variables

  • radv: do not remove dead variables

  • radv: remove some unneeded passes from radv_nir_lower_io_vars_to_scalar

  • radv: do not shrink vectors when lowering IO vars to scalar

  • radv: don’t lower io vars to scalar

  • radv: remove lower array vars to elem

  • radv: remove radv_nir_lower_viewport_to_zero

  • nir: disable fp class analysis for 64bit transcendentals

  • intel/nir_opt_peephole_ffma: fix fp_math_ctlr for modifiers

Gurchetan Singh (12):

  • gallium: fix sometimes-uninitialized warning

  • gfxstream: fixes related to -Wmissing-prototypes

  • gfxstream: fix build after vk.xml update

  • lavapipe: fix uninitialized variable warning

  • virtio/kumquat: add safety comments

  • gfxstream: explicitly assign INVALID_DESCRIPTOR

  • gfxstream: meson: remove duplicate includes

  • gfxstream: add vulkan_gfxstream_structure_type.h to codegen output

  • gfxstream: fixes to get Fuchsia headless to compile

  • gfxstream: simple compile fix

  • gfxstream: cereal: fix ‘None’ in gfxstream codegen

  • gfxstream: additional Goldfish logic for Android builds

Haixiang Tang (1):

  • zink/kopper: Allow surface creation for Pixmaps (non-window surfaces)

Hans-Kristian Arntzen (15):

  • vulkan/wsi: Add common infrastructure for EXT_present_timing.

  • vulkan/runtime: Expose PRESENT_STAGE_LOCAL as calibrateable domain.

  • anv: Add PRESENT_STAGE_LOCAL_EXT path for calibration.

  • vulkan/wsi: Add no-op present timing support to most backends.

  • wsi/wayland: Implement EXT_present_timing on Wayland.

  • radv: Enable EXT_present_timing.

  • turnip: Enable EXT_present_timing.

  • anv: Enable VK_EXT_present_timing.

  • nvk: Enable EXT_present_timing.

  • panvk: Enable EXT_present_timing.

  • vulkan/wsi: Implement QUEUE_OPERATIONS_END present timing query.

  • wsi/wayland: Fix some locking quirks around present ID update.

  • wsi/display: Implement present timing on KHR_display.

  • wsi/common: Allow timestampValidBits < 64 for present timing.

  • docs: Add VK_EXT_present_timing to new features.

Hoe Hao Cheng (1):

  • zink/codegen: do not enable extensions that are fully core-promoted

Hsieh, Mike (1):

  • amd/vpelib: Move feature skip after buffer size return

Hyunjun Ko (9):

  • anv/video: fix a typo in Vulkan AV1 decoding.

  • anv/video: Compute AV1 tile positions internally

  • anv: Add dummy workload for AV1 decode on affected platforms (Wa_1508208842)

  • anv/video: disable encoder on untested platforms

  • anv/video: set transform skip numbers according to qp

  • anv/video: set Qp passed from apps for h265 encoder

  • anv/video: Handle GPB(Generalized P and B frames) properly for H265 enc.

  • anv/video: set Sad Qp Lambda values properly for H265 encoder.

  • anv/video: remove unsupported feautres for encoders

Iago Toral Quiroga (5):

  • broadcom/compiler: drop unnecessary MOV

  • broadcom/compiler: don’t always clear undefined bits from sub-32 integers

  • broadcom/compiler: optimize alu(shr(x, 16).l) to alu(x.h)

  • broadcom/compiler: inform NIR scheduler about 0 cost ALU instructions

  • nir/opt_vectorize_load_store: allow sizes unaligned with high offset for loads

Ian Douglas Scott (1):

  • wsi/wayland: Use `wl_fixes` to destroy `wl_registry`

Ian Forbes (7):

  • svga: Implement GL_ARB_derivative_control

  • svga: Increase max_combined_shader_output_resources and SSBO limit to 16

  • svga: Implement GL_ARB_conditional_render_inverted

  • svga: Always emit VGPU10_OPCODE_DCL_GLOBAL_FLAGS for VGPU10

  • svga: Enable GL_ARB_vertex_type_10f_11f_11f_rev

  • svga: Make svga_screen::hud members atomic

  • svga: Implement GL_ARB_pipeline_statistics_query

Ian Romanick (29):

  • spirv: Use STACK_ARRAY instead of NIR_VLA

  • nir: Use STACK_ARRAY instead of NIR_VLA

  • elk: Use F16TO32 for nir_op_f2f32 of float16 source

  • brw: Call nir_opt_algebraic_late in brw_nir_create_raygen_trampoline

  • brw: Call nir_opt_algebraic_late later in brw_postprocess_nir_opts

  • elk: Call nir_opt_algebraic_late in elk_postprocess_nir

  • brw/cmod: Don’t propagate from CMP to ADD if there is a write between

  • elk/cmod: Don’t propagate from CMP to possible Inf + (-Inf)

  • elk/cmod: Don’t propagate from CMP to ADD if there is a write between

  • brw: Don’t mark_invalid in update_for_reads for non-VGRF destination

  • brw: Use brw_reg_is_arf in update_for_reads

  • brw: Also check for ADDRESS file in update_for_reads

  • brw/algebraic: Don’t optimize SEL.L.SAT or SEL.G.SAT

  • elk/algebraic: Don’t optimize SEL.L.SAT or SEL.G.SAT

  • brw/validate: Eliminate duplicate integer multiply validation

  • brw/validate: Implicit read of accumulator cannot also have explicit read

  • brw/validate: Perform more 3-src validation in brw_validate instead of brw_eu_emit

  • brw/emit: Src1 can be accumulator on Gfx12.5 and newer

  • brw: Fix encoding of accumulator sources of 3-source instructions

  • brw/asm: Don’t drop accumulator number in the assembler

  • anv: Use different logic to isolate lowest flag in anv_foreach_vk_stage

  • anv: Use u_foreach_bit

  • util: Use same method to clear bits in u_foreach_bit as util_bitcount

  • brw/algebraic: Allow mixed types in saturate constant folding

  • brw: Handle scalars and swizzles correctly in is_const_zero

  • brw/lower_regioning: Allow integer conversions in SEL

  • brw: Change the flags written by some CMP

  • brw/const: Don’t allow type changes when accumulators are involved

  • brw: brw_reg::nr for an accumulator is not part of the offset

Icenowy Zheng (39):

  • pvr: preliminary EXT_image_drm_format_modifier support

  • util/cpu: add a number of RISC-V extensions

  • util/cpu: support detecting RISC-V FD/C/V/Zb[abs] with riscv_hwprobe

  • pvr: only specially handle gfx subcmd for BeginQuery

  • pvr: suppress VkDescriptorSetLayoutBindingFlagsCreateInfo ignored warn

  • mailmap: map all mailboxes for Icenowy Zheng

  • gallium/frontends/dri: only reserve a few bind flags for MSAA cbufs

  • glsl: support adding point size to io_lowered shaders

  • pipe-loader: make get_driver_descriptor() return NULL for unknown driver

  • pipe-loader: fallback to zink instead of kmsro for render nodes

  • vulkan/wsi/headless: properly use CPU images for CPU devices

  • zink: skip all post-process when importing and resource_create fails

  • pvr: re-indent pseudocode for DDMADT behavior

  • glsl: initialize PSIZ variable to NULL when adding pointsize

  • pco: fix encoding of fred’s s0abs bit

  • pvr: drop master for the display FD if it’s not needed

  • pvr: Align width for PBE write when creating linear image

  • pvr: fix “obb” typo in oob_buffer_size when building vertex pds data

  • pvr: save vertex attribute size for DMA checking

  • pvr: move PVR_BUFFER_MEMORY_PADDING_SIZE definition to pvr_buffer.h

  • pvr: consider the size of DMA request when setting msize of DDMADT

  • vulkan/wsi/headless: properly cleanup swapchain init failure

  • vulkan/wsi/headless: implement wait_for_present for swapchain

  • pvr: support VK_EXT_non_seamless_cube_map

  • pvr: fix dirty tracking for stencil ops

  • docs: add missing bits for pvr’s VK_EXT_non_seamless_cube_map

  • pvr: fix pvr_clear_vdm_state_get_size_in_dw() inverted feature condition

  • pvr: set has_usc_alu_roundingmode_rne for all B-series Rogue cores

  • pvr: finalize query_indices array after ending last sub_cmd

  • pvr: fix the code copying query_indices to sub_query_indices

  • pvr: propagate get_vis_results flag from secondary cmdbuf gfx jobs

  • pvr: follow other drivers’ practice for copying build ID

  • pvr: skip emitting query program when copy result / reset with 0 queries

  • pvr: wait for graphics jobs in CopyQueryPoolResults

  • pvr: increase maxPerStageResources for new maxPerStageDescriptorStorageBuffers

  • pvr: do not setup deferred RTA clear for active render targets

  • pvr: properly handle deferred RTA clears for 2D array view of 3D image

  • pvr: add deferred RTA clear command to list after checking it’s not NULL

  • pvr: record deferred RTA clears for secondary cmdbuf subcmds

Iván Briano (5):

  • brw: fix local_invocation_index with quad derivaties on mesh/task shaders

  • anv, hasvk: handle MSAA resolving to a 3D slice

  • anv: don’t try to fast clear D/S with multiview

  • anv: fix anv_is_dual_src_blend_equation

  • brw: do not omit RT writes if dual_src_blend is on

Jakob Sinclair (6):

  • pan: improve debug printing of multiple registers

  • pan: move discard/kill_ssa flag after index for debug prints

  • pan: add sigil to SSA values for debug printing

  • pan/compiler: Do not assume split 64-bit registers in va_mark_last

  • pan/compiler: Fix style formatting in lower_split_src

  • pan/compiler: Use SHADDX instruction for i64 add

Jan Alexander Steffens (heftig) (1):

  • kk: Fix debug printf specifier

Janne Grunau (9):

  • asahi: ci: Rename asahi-g13g Vulkan CTS suite to asahi-agx2

  • asahi: Add OpenGL / EGL CTS CI expectations

  • asahi: Use GPU for buffer copies in resource_copy_region()

  • asahi: Implement clear_buffer using libagx_fill*

  • hk: Use aligned vector fill in hk_CmdFillBuffer if possible

  • hk: Increase maxFragmentCombinedOutputResources to HK_MAX_DESCRIPTORS

  • hk: ci: Skip timing out wsi xlib tests

  • hk: ci: Handle more spurious wsi CTS timeouts/fails

  • nir/gather_info: clear interpolation qualifiers only in fragment stage

Jarred Davies (2):

  • pvr: Fix allocating the required scratch buffer space for tile buffers

  • pvr: Add missing support for tile buffers to SPM EOT programs

Jason Macnak (4):

  • gfxstream: enable VK_EXT_primitives_generated_query

  • gfxstream: Fix StagingInfo destruction ordering

  • gfxstream: fix submit to not hold lock when calling into encoder

  • gfxstream: fix goldfish guards on fence functions

Jesse Natalie (11):

  • wgl: Delete stw_pfd_flag

  • wgl: Support PFD_SWAP_COPY pixel formats

  • wgl: Add a driconf option to force pixel formats with GDI support

  • driconf: Add a driconf entry for 文香白板 (Wenxiang whiteboard)

  • d3d12: Set packed_uniforms cap

  • meson: Include DirectX-Headers dependency for all VK Windows builds

  • d3d12: Fix importing external resources

  • wgl: Flush and wait when unbinding a context that references a swapchain

  • mesa/st: Pass the context to fence_finish as part of flush+wait

  • d3d12: Don’t allow CPU storage for huge buffers

  • wgl: Use an hwnd xor hdc for framebuffers

Jianxun Zhang (2):

  • anv: Limit modifier disabling workaround to specific GTK versions

  • driconf: Refactor CCS modifier disabling entry

Job Noorman (43):

  • ir3/isa: attach (sat) to dst

  • ir3/isa: fix shift/reduce conflict for mova.r

  • ir3/parser: make bison fail on warnings

  • tu,ir3: lower multiview indirect stores to register indirects

  • ir3: add block_can_be_predicated helper

  • ir3: don’t use predication for large blocks

  • ir3: update context builder after ir3_get_predicate

  • ir3: don’t predicate vote_all/vote_any

  • ir3/legalize: don’t drop sync flags on removed predt/predf

  • nir/opt_varyings_bulk: add data parameter to optimize callback

  • nir/opt_varyings: fix alu def cloning

  • nir/gather_info: gather per_view info

  • nir/gather_info: clear interpolation qualifiers before gathering

  • nir/recompute_io_bases: fix num_slots for per_view outputs

  • ir3: fix handle_partial_const with vectorized src

  • ir3: call nir_lower_io_vars_to_temporaries for GS outputs

  • ir3: call nir_io_add_intrinsic_xfb_info after IO lowering

  • tu: extract NIR lowering to a separate function

  • tu: use nir_opt_varyings_bulk for linking

  • nir/opt_uniform_subgroup: fix ballot_bit_count components

  • nir/lower_atomics: add support for bindless_image_atomic

  • ir3: allow imm src0 (IBO) on bindless atomics

  • ir3: support isam with less than 4 components

  • ir3: add support for r64u?i image loads/stores

  • ir3: add support for 64-bit image atomics

  • ir3/analyze_ubo_ranges: add const_align_vec4 helper

  • ir3/analyze_ubo_ranges: don’t over-align consts when loaded via preamble

  • ir3: simplify constlen calculation

  • ir3: remove unused ir3_context::has_relative_load_const_ir3

  • ir3/collect_info: remove max_const calculation

  • ir3/postsched: update legalize state for terminators

  • ir3: set cat6.dst_offset for ldc

  • ir3/legalize: track need_ss/sy_for_const per const reg

  • ir3/parser: set constlen when adding const regs

  • ir3/parser: add @constlen header

  • ir3: allow shared address src for ldg.k

  • ir3: add support for the ldg.k a1.x addressing mode

  • ir3/isa: fix load size encoding for ldg.k

  • ir3: use ldg.k load size

  • ir3/shared_ra: fix live-out reload after src reload

  • ir3/cf: fix rewriting uses with different dst types

  • ir3/shared_ra: use ir3_cursor instead of instr in reload helpers

  • ir3/shared_ra: insert reloads before tied dst pcopies

Jon Turney (1):

  • ddebug: Fix use of alloca() without #include “c99_alloca.h”

Jordan Justen (19):

  • intel/decoder: Use array of filenames in get_embedded_xml_data_by_name()

  • intel/genxml: Rename Xe2 genxml to xe2.xml and xe2_rt.xml

  • intel/genxml: Rename Xe3 genxml to xe3.xml and xe3_rt.xml

  • intel/genxml: Start Xe3P (GFX_VERx10 == 350) support (xe3p.xml, xe3p_rt.xml)

  • intel/genxml: Update README notes on hardware version numbers

  • intel/genxml: Fix Xe3P import filenames in intel_genxml.py

  • intel/genxml: Add gen125_rt.xml to default_imports in intel_genxml.py

  • intel/isl: Build for Xe3P (GFX_VERx10 == 350)

  • intel/shaders: Build for Xe3P (GFX_VERx10 == 350)

  • iris: Build for Xe3P (GFX_VERx10 == 350)

  • intel/l3: Add Xe3P (GFX_VERx10==350)

  • anv: Add Xe3P (GFX_VERx10==350)

  • intel/dev: Add INTEL_PLATFORM_NVL_P platform enum

  • intel/dev: Split out Xe3 threads and URBs macros

  • intel/dev: Add XE3P devinfo macros

  • intel/tools/intel_dev_info: Verify stage_names size in print_base_devinfo()

  • intel/dev: Handle Xe3P in intel_device_info_init_common() (for build tests)

  • intel/dev: Add NVL-P device info

  • intel/dev: Add NVL-P PCI IDs (with FORCE_PROBE required)

Jose Maria Casanova Crespo (14):

  • v3dv: disable blending when logicOpEnable is set

  • v3d: flush write jobs before BO replacement in DISCARD_WHOLE path

  • vc4: flush write jobs before BO replacement in DISCARD_WHOLE path

  • v3d: reject fast TLB blit when RT formats don’t match

  • v3d: simplify fast TLB blit format check

  • broadcom/ci: broaden glx-copy-sub-buffer flake entry on RPi5

  • broadcom/common: fix V3D 7.1 TFU ICFG IFORMAT values

  • broadcom/common: add tile alloc block size macros and sizing helper

  • v3d: use shared v3d_tile_alloc_sizes() and 128B initial blocks

  • v3dv: use shared v3d_tile_alloc_sizes() and 128B initial blocks

  • v3dv: defer tile_alloc creation in meta TLB ops

  • broadcom/compiler: really enable branch in delay slots validation

  • broadcom/compiler: MULTOP in branch delay slots doesn’t generate RTOP hazard

  • broadcom/compiler: move nir_lower_undef_to_zero out of optimization loop

José Expósito (2):

  • winsys/amdgpu: Fix userq job info log on PPC

  • venus: Fix error log on PPC

José Roberto de Souza (22):

  • intel/dev: Remove INTEL_DEVICE_INFO_MMAP_MODE_XD

  • intel/dev: Remove INTEL_DEVICE_INFO_MMAP_MODE_UC

  • intel/dev: Improve PAT entries comment

  • anv: Move anv_bo_get_mmap_mode() to i915 backend

  • intel/dev: Add INTEL_DEVICE_INFO_MMAP_MODE_INVALID

  • intel/isl/gfx12.5: Alow hierarchial depth buffer write through for multi sampled surfaces

  • intel/brw: Add BRW_DEPENDENCY_INSTRUCTIONS invalidation when instructions are added or removed in brw_opt_split_virtual_grfs()

  • iris: Fix invalid reads when uploading blend state

  • intel/brw: Use computed push constants size in brw_assign_urb_setup()

  • intel/brw: Add and call brw_lsc_supports_base_offset() in places that checks for support of this feature

  • intel/perf: Add HSW verx10 to intel_perf_query_result_write_mdapi()

  • intel/dev: Add URB min/max entries for Mesh and Task

  • intel/dev/xe3p: Add min URB entries for task and mesh shaders

  • anv: Fix CmdResetEvent2() with RESOURCE_BARRIER::Wait stage == none

  • anv: Remove asserts() added in resource_barrier_wait_stage()

  • anv: Always have a valid Resource barrier::Wait stage set

  • anv: Fix invalid resource barrier signal stage

  • anv: Fix placed address mmap with slab bo

  • anv: Rename and share get_scratch_surf() with other files

  • anv: Make use of anv_shader_get_scratch_surf() in genX_cmd_compute.c

  • anv: Use helper to get anv_address in emit_simple_shader_dispatch()

  • intel/brw: Remove unsed functions to get data port message type

Juan A. Suarez Romero (34):

  • broadcom/ci: remove asan failures from rpi3 and rpi4

  • broadcom/ci: re-evaluate timeout tests

  • broadcom/ci: re-adjust fractions

  • broadcom/ci: rename rusticl job

  • broadcom/ci: re-evaluate all the flakes

  • broadcom/ci: remove duplicate entries

  • broadcom/cle: bump up gen version for v3d

  • broadcom/cle: ensure zlib inflate assign memory

  • broadcom: don’t hardcode pagesize

  • broadcom/ci: update expected results

  • v3dv: serialize all the tests causing OOM

  • broadcom/ci: fetch custom packaged kernel in CI-Tron

  • broadcom/ci: update available devices

  • broadcom/ci: update expected results

  • v3d: fix leak in blit fast

  • v3d,v3dv: emit always set point size

  • st/pbo_compute: remove unused variables

  • broadcom/ci: update expected results

  • vc4/ci: update expected results

  • v3d: add support for GL_ARB_sample_shading

  • broadcom/ci: update expected results

  • broadcom/ci: update expected results

  • broadcom/ci: update expected results

  • v3dv: fix mutable resolve attachment format mismatch

  • vc4: fix unwanted buffer release on uploader

  • v3d/ci: add new OpenCL failure

  • v3dv/ci: add link to failing CTS test

  • vc4: add dot on static QPU unpack strings

  • vc4: make some dump functions return strings instead of printf

  • vc4: use Mesa logging functions

  • broadcom/compiler: make some dump functions return strings instead of printf

  • broadcom: use Mesa logging functions

  • broadcom/cle: parse once the XML spec

  • broadcom/ci: update expected results

Julia Zhang (2):

  • vulkan: return pQueue with matching flags

  • radv/amdgpu: handle DISCARDABLE flag in get_flags_from_fd

Juston Li (1):

  • anv: set missing protected bit for protected depth/stencil surfaces

Karmjit Mahil (14):

  • tu: Allocate cmd_buffer from its pool

  • tu: Set tu_ignore_frag_depth_direction driconf for Creed

  • zink: Fix incorrect assert checking for linear state format

  • freedreno/registers: Add some missing include in fd6_hw.h

  • freedreno/a6xx: Add missing include to fd6_pack.h

  • freedreno: Add fd{2,3,4,5}_hw.h and fd_hw_common.h

  • freedreno: Add check_xml_includes test

  • freedreno: Add check_xml_includes to meson setup

  • tu: Use “nir/” for the nir includes

  • tu: Undef before redefining MESA_LOG_TAG

  • tu: Update .clang-format include categories

  • tu: Reorder includes

  • tu: Cleanup some includes

  • tu: Remove unecessary forward declaration

Karol Herbst (76):

  • nvk: reorder exposed coop matrix types

  • clc: reorder headers to fix compilation errors due to UNUSED

  • clc: support some atomic and generic address space features

  • clc: enable generic address space and seq_cst and device scope atomic features

  • nir: fix nir_fixup_is_exported for LLVM-22

  • clc: fix compile compatability with LLVM-22

  • khronos-update: synchronize OpenCL header file list

  • khronos-update: add Intel’s OpenCL header

  • include: synchronize OpenCL headers

  • rusticl/platform: add rusticl_warn_once macro

  • rusticl/program: accept and ignore Intel’s 4G memory flags

  • nir: add nvidia IO intrinsics

  • nir: add BASE to nvidia memory intrinsics

  • nak: convert memory load/stores to nv variants

  • nir/opt_offsets: support negative offsets and 64 bit sources

  • nir/opt_offsets: support nvidias intrinsics

  • nak: replace get_io_addr_offset with nir_opt_offsets

  • rusticl/mesa: only use resource_from_user_memory if the cap is advertised

  • vtn/opencl: flush denorms for cbrt()

  • vtn: set default fp_math_ctrl values for kernels

  • nir: add nvidias shared memory non unform address shift

  • nak: add LDS/STS/ATOM address shift encoding

  • nak: Fold constant ishl into shared ld/st/atoms

  • zink: handle drivers with multiple subgroup sizes correctly

  • ac/llvm: handle int8 inside ac_build_optimization_barrier

  • zink: implement subgroup rotate

  • rusticl: support more subgroup extensions

  • asahi: support subgroup_rotate

  • nir: fix nir_alu_type_range_contains_type_range for fp16 to int

  • nir: fix nir_round_int_to_float for fp16

  • nouveau/drm-shim: implement get_zcull_info

  • nvk: run nir_opt_large_constants before nir_lower_load_const_to_scalar

  • nak: invalidate loop analysis with nak_nir_lower_load_store

  • nak: replace legalize_ext_instr with explicit lowering

  • nak: add input predicate to load_global_nv and OpLd

  • nak: use ldg input predicate in nak_nir_lower_non_uniform_ldcx

  • nak: support has_load_global_bounded on turing and newer

  • nvk: skip lowering load_global_constant_bounded on turing inside lower_load_intrinsic

  • nak: enable vectorize_vec2_16bit

  • nak: allow vector sources for f2f16 conversions

  • nak: vectorize f2f16

  • nak: vectorize f2f16 even more

  • nak: make nak_mem_vectorize_cb create only aligned and supported vectors

  • nir: rename fsin_amd and fcos_amd to a more generic name

  • nir: unvendor ac_nir_lower_sin_cos

  • nak: run nir_normalize_sin_cos on Volta+

  • ci: add api@clgetmemobjectinfo to fails

  • nak: rework swizzling on scalar FP16 ops

  • nak: remove OpF2F::dst_high

  • nak: support MUFU.F16

  • nak: add hw_test for MUFU.F16

  • nak: enable MUFU.F16 on Turing and newer

  • nak: add algebraic patterns to improve MUFU.F16

  • radeonsi: set valid_buffer_range for CL buffers

  • docs: clarify the use of autonomously acting tooling

  • docs: add AI disclosure requirements

  • radeonsi: properly report unified memory on APUs

  • rusticl/kernel: implement CL_KERNEL_GLOBAL_WORK_SIZE for custom devices

  • rusticl/device: Fix reporting of global memory on mixed memory devices

  • nak/copy_prop: allow modified F16v2 and F16 sources

  • nak: properly copy prop neg/abs float sources for flushed values

  • nak: add scalar tex encoding support

  • nak/nvdisasm_tests: test .SCR flag in TEX, TLD and TLD4

  • nak: scalarize tex, tld and tld4 on SM70+

  • nak/nvdisasm_tests: fix offset stride for gens older than Turing

  • nak: add ugpr latency classes for memory instructions

  • nak: add is_gpr_reg and is_ugpr_reg helpers

  • nak: uregs are 6 bits before Hopper, so enforce that

  • nak: the MS location comes last in TLD, same spot as depth compare in TEX

  • mesa/st: do not advertise CL subgroup features on the GL side

  • ci: install libstdc++-static on fedora

  • rusticl: link the C++ runtime statically

  • softfloat: make sign bit an unsigned int

  • nir: add fmul_rtz

  • nak: handle nir_op_fmul_rtz

  • nak: use fmul_rtz for NAK_INTERP_MODE_PERSPECTIVE

Kenneth Graunke (78):

  • nir: Add memory modes to URB load intrinsics

  • nir: Teach opt_load_store_vectorize how to handle Intel URB intrinsics

  • nir: Add load/store vectorizer option for rounding up masked stores

  • nir: Add a round_up_components callback to load/store vectorization

  • brw: Assert that urb_vec4_intel stores only have 4/8 components

  • brw: Skip vec8 store_urb_vec4_intel noop writemasks as well

  • brw: Avoid using URB global offset with per-slot offsets on <= Icelake

  • brw: Vectorize URB intrinsics using nir_opt_load_store_vectorize

  • brw: Use nir_imul_imm in per-vertex/per-primitive offset calculation

  • brw: Fix geometry shaders with non-constant vertex indices

  • iris: Switch to SPDX headers

  • brw: Drop urb_handle parameter from store_urb()

  • brw: Implement load_urb_output_handle_intel for VS/GS stages

  • brw: Move TES VUE map calculation before lowering outputs

  • brw: Set a valid varying_to_slot for VUE header fields other than PSIZ

  • brw: Add VUE header varyings to io_component()

  • brw: Split EOT handling out of emit_urb_writes()

  • brw: Convert VS/TES/GS outputs to URB intrinsics.

  • intel: Rename intel_msaa_flags to intel_fs_config

  • intel: Rename wm_prog_data to fs_prog_data

  • intel: Rename wm_prog_key to fs_prog_key

  • brw: Delete wm_prog_data::urb_setup_channel[]

  • elk: Delete mesh shader remnants

  • iris: Fix scratch shift after recent bindless changes

  • intel/elk, hasvk: Drop indirect_ubos_use_sampler option and DP code

  • brw: Make indirect_ubos_use_sampler a static inline bool taking devinfo

  • brw: Make use_tcs_multi_patch a static inline taking devinfo

  • iris: Move recompile debugging to work on iris program keys

  • anv: Make a devinfo local in anv_shader_lower_nir

  • anv: Pass devinfo to anv_shader_compute_fragment_rts, not compiler

  • anv: Drop an outdated comment about indirect descriptors

  • anv, brw: Consolidate ex_bso bits to a static devinfo inline

  • brw: Delete use_bindless_sampler_offset flag

  • brw: Pass devinfo to lower_bit_size, not compiler

  • brw: Make a devinfo temporary in lower_mem_access_bitsizes

  • brw: Drop brw_compiler option from brw_no_indirect_mask()

  • iris: Drop SBE component overrides for layer/viewport varyings

  • iris: Use the first FS input’s value for all undefined FS inputs

  • iris: Drop sprite coord checks from SBE_SWIZ setup

  • iris: Drop use of BFC(n) when it exists but COL(n) is unwritten

  • iris: Defeature native two-sided color support

  • brw: Use NUM_TOTAL_VARYING_SLOTS instead of VARYING_SLOT_TESS_MAX

  • brw: Drop BRW_VARYING_SLOT_PAD and brw_varying_slot enum

  • brw: Drop VUE header values and position from wm_prog_data->inputs

  • brw: Use memset for initializing varying/slot maps

  • brw: Only lower system values for passthrough TCS

  • brw: Drop extra validation from TCS passthrough creation

  • iris: Move passthrough TCS generation out of brw and into iris

  • iris: Create our own enums for system values

  • iris: Move ALT mode handling from brw to iris

  • nir: Fix divergence of Intel URB input/output handle intrinsics

  • brw: Purge source_depth_to_render_target

  • brw: Simplify GS load_invocation_id handling

  • brw: Combine brw_assign_*_urb_setup() into one function

  • brw: Fix single patch thread dispatch masks in NIR

  • brw: Lower TCS single patch invocation ID calculations in NIR

  • brw: Don’t emit HALT_TARGET for VS/TCS/TES/GS

  • brw: Simplify mark_last_urb_write_with_eot()

  • nir: Add is_sparse flag to texture builders

  • intel/nir: Use txf builder in intel_nir_lower_sparse

  • intel/nir: Set new image intrinsic parameters via builder helpers

  • intel/nir: Generalize lower_tex_compare to split_tex_residency

  • intel/nir: Replace tg4 with txl/txb/tex when splitting texture residency

  • nir: Add nir_texop_resinfo_intel

  • brw: Use nir_texop_resinfo_intel for query_levels and txs

  • nir: Increase tex opcode bits from 5 to 6 in nir_instr_set

  • anv: Use nir_lower_memory_model

  • intel/compiler: Use nir_static_workgroup_size helper

  • brw: Support load_simd_width_intel for fragment shaders

  • brw: Set nir->info.{min,max}_subgroup_size in brw_nir_apply_key

  • brw: Have brw_nir_apply_key call brw_nir_lower_simd for all stages

  • nir, brw: lower scratch in NIR

  • nir: Add nir_texop_sparse_residency[_txf]_intel operations

  • intel: add INTEL_JAY environment variable

  • intel/nir: Make intel_nir_lower_sparse work for either brw or jay

  • jay: Assert that source is not null in jay_copy_strided

  • jay: Make lower_immediates bail if there are no sources

  • jay: Clear default group for quad swizzles

Khem Raj (1):

  • glx: fix const qualifier warnings found with C23 glibc support

Kitlith (2):

  • panvk: Free drm device in can_present_on_device

  • pvr: Free drm device in can_present_on_device

Konstantin Seurer (33):

  • vulkan: Handle inactive primitives with LBVH builds

  • vulkan: Avoid NAN in the IR BVH

  • vulkan: Limit the number of LBVH invocations

  • radv/rra: Fix nullptr dereference

  • vulkan: Make sure no NaNs end up in the BVH

  • radv/bvh: Make sure internal nodes are collapsed when possible

  • radv: Use stderr for shader printf

  • radv: Stop saving descriptors before acceleration structure OPs

  • radv: Fix setting the viewport for depth stencil FS resolves

  • util/ralloc: Allow creating a linear context without parent context

  • vulkan/cmd_queue: Use a linear allocator

  • lavapipe: Extend vk_cmd_queue_entry_base for internal commands

  • vulkan: Remove vk_cmd_queue_entry::driver_data

  • vulkan/cmd_queue: Remove get_array_member_copy

  • vulkan/cmd_queue: Fixup stride for multi draws

  • vulkan/cmd_queue: Do not zero initialize vk_cmd_queue_entry

  • radv/meta: Add and use helpers for setting state

  • radv/meta: Rework saving/restoring state

  • vulkan/cmd_queue: Rework copy codegen

  • vulkan/cmd_queue: Don’t explicitly set struct members to NULL

  • nir/tests: Test nir_opt_large_constants

  • spirv,nir: Preserve more information about the descriptor type

  • vulkan: Add helper for dispatching BVH build stages

  • vulkan: Request less scratch space for lbvh

  • vulkan: move internal_node_count to vk_acceleration_structure_build_state

  • vulkan: Remove bvh_state

  • vulkan: Init all update scratch at once

  • radv/bvh: Prefer selecting quads as the first pair of a HW node

  • radv: Add an option for dumping BVH stats

  • radv: Add RT prolog information to hang reports

  • radv: Refactor declaring shader args

  • radv: Set debug info in radv_shader_create_uncached

  • radv: Gather debug info about shader args

Kovac, Krunoslav (1):

  • amd/vpelib: Apply external CSC

Krzysztof Sobiecki (3):

  • gallium/dril: Don’t use gbm if there is no gbm configured

  • os: Don’t use PATH_MAX as it’s not portable.

  • os: Add support for GNU/HURD compilation and use of dri swrast/llvmpipe.

Lakshman Chandu Kondreddy (1):

  • freedreno/layout, tu: Fix UBWC block sizes for PIPE_FORMAT_R8_G8B8_420_UNORM

Lars-Ivar Hesselberg Simonsen (25):

  • panfrost/bi: Fix unbound texel buffers

  • pan/decode/jm: IDVS decode improvements

  • panvk: Fix dcd_flags1 dirty bit

  • pan/genxml/v13: Fix HSR Prepass typo

  • pan/genxml/v13: Add HSR operation enums

  • pan/compiler: Add pass to collect HSR info

  • panvk/v13: Set HSR flags

  • panvk/v13: Support HSR Prepass

  • pan: Drop ASTC support for v5 texel buffers building

  • pan: Move buffer functions to pan_buffer

  • pan/va: Remove non-existent unused CLPERs

  • pan/va: Clean up unused/removed instructions

  • pan/va: Add opcode modifier to ISA.xml

  • pan/va: XMLify opcode2

  • pan/va/disasm: Move instr print to function

  • pan/va: Generalize opcode/opcode2

  • pan/va/disasm: Clean up hardcoded values

  • pan/va/disasm: Move src discard marker behind reg

  • pan: Centralize preload registers

  • pan/model: Redo gpu_prod_id in the model

  • pan: Add support for 64 bit gpu_id

  • pan/va/isa: Src for X16_TO* takes lane, not swizzle

  • pan/model: Expose prod_id and rev functions

  • panfrost: Add support for 64 bit gpu_id

  • panvk: Fix debug flag overlap

Leder, Brendan Steve (Brendan) (1):

  • amd/vpelib: Add missing JFIF enum

Leon Perianu (6):

  • pvr: fix logic for when to reset fill blit

  • pvr: fix format table properties duplicate

  • pvr: remove hardcoded buffer alignment and image alignment

  • pvr: refactor image initialization with helper functions

  • pvr: use align64 for large buffer memory requirements

  • pvr: enable VK_KHR_maintenance4

Lin, Ricky (1):

  • amd/vpelib: Augment swizzling modes

Link Mauve (1):

  • docs/panfrost: fix outdated path to complete GPUs list

Lionel Landwerlin (111):

  • iris: fix incorrect intrinsic usage on ELK

  • anv/iris: add drirc to enable sampler state & compute surface state prefetch

  • vulkan/wsi/direct: remove VkDisplay created from GetDrmDisplayEXT on ReleaseDisplayEXT

  • vulkan/pipeline: don’t consider capture-replay flag for shader hashing

  • anv: fix shader heap replay addr

  • anv/xe: move special WaitIdle optimization to submission path

  • anv: implement VK_KHR_internally_synchronized_queues

  • anv: flush render caches on first pipeline select

  • anv: fix nested command buffer relocations

  • anv: add missing constant cache invalidation for descriptor buffers

  • isl: fix 32bit math with 4GB buffer size

  • brw: make PULL_CONSTANT opcodes more like MEMORY opcodes

  • brw: enable ex_bso for LSC_SS

  • anv: rename/document a layout helper

  • anv: rework descriptor set indexing in NIR

  • anv: remove set index for descriptor buffers

  • anv: add a couple of surfaces to read descriptors

  • brw: handle non-GRF aligned pushed UBO masking

  • anv: delay BRW prog_data filling

  • anv: drop unused argument for compute_push_layout

  • anv: use internal surface state on Gfx12.5+ to access descriptor buffers

  • anv: remove unused arguments

  • brw/iris: move ubo range analysis pass to iris

  • intel/tools: print out GRF size in intel_dev_info

  • anv: enable compute bti prefetch

  • anv: apply the same ccs disabling for Xe3 than Xe2

  • anv: disable ccs modifier reporting when ccs modifiers are disabled

  • anv: move ALU registers used for mi commands

  • anv: enable sharing binding table pool programming

  • anv: predicate BTP emissions

  • anv: add a drirc to control binding table block size

  • anv: dirty descriptors after blorp operations

  • anv: program HW to gather push constants at 3DSTATE_CONSTANT parsing time on Gfx9

  • anv: specialize code for Wa_18019110168

  • anv: remove snprintf for aux op transition

  • anv: remove old comment related to pre softpin platforms

  • anv: add a debug printout for dirty descriptors

  • anv: make a helper for push constant allocation

  • anv: optimize binding table flushing

  • anv: track descriptor buffers used & promoted to push constants

  • anv: avoid filling PC reason for timestamp u_trace captures

  • anv: pack bind_map further

  • anv: delay num-workgroups usage until push remapping

  • anv: merge 2 push constants fields

  • anv: add a shrinking push constant loading pass

  • brw/nir: handle inline_data_intel more like push_data_intel

  • anv: implement inline parameter promotion from push constants

  • anv: fix dynamic buffes again

  • anv: add missing handling for attachment locations in secondaries

  • docs/anv: add some debug tips

  • anv: dirty all push constant stages in simple shader

  • anv: remove unused bind_map field

  • anv: add an analysis pass to detect compute shaders clearing data

  • anv: add drirc option to workaround missing application barriers on typed/untyped data

  • blorp: add mda support

  • brw: use scalar build for immediate offsets

  • brw/nir: enable constant offsets for global_constant_uniform_block_intel

  • brw/nir: add new intrinsics to load data from the indirect address

  • blorp: switch to new load_indirect_address_intel intrinsic

  • anv/brw: remove push constant load emulation from the backend compiler

  • anv: fix dEQP-VK.memory.address_binding_report*

  • anv: only go into buffer relocs after we’ve looked at all batches

  • anv: fix pulling constant data in compute/mesh/task shaders

  • brw/nir: improve shader_indirect_data_intel handling

  • anv: fix internal compute shader constant data pull

  • anv: use companion RCS for hiz ops on compute queue

  • anv: reduce optimalBufferCopy(Offset|RowPitch)Alignment

  • vulkan/runtime: add new helper for vertex strides

  • vulkan/runtime: build (address|copy)_flags for vk_buffer

  • vulkan/runtime: add implementation of older entrypoints using KHR_DAC

  • anv: enable compression control on Android VP17

  • vulkan/runtime: break view mask from renderpass information

  • anv: don’t queue pipe control reasons wihout a trace

  • anv: limit aux disabling on concurrent images to pre-Xe2

  • vulkan/runtime: fix missing copy image layout

  • vulkan/runtime: fix incorrect entrypoint call for CmdCopyImageToBuffer2

  • anv: deal with Wa 14024015672 on the blorp path

  • anv: move depth/stencil BeginRendering handling prior to color

  • anv: rename variables in CmdBeginRendering

  • anv: batch rendering initialization commands

  • anv: merge null surface state packing with previous attachments

  • anv: document more stalling

  • anv: rework color_aux operation tracking

  • drm-uapi: Sync xe_drm.h

  • intel/dev: add state cache perf fix support xe detection

  • brw: fence SLM writes between workgroups

  • nir: improve deref_instr_get_variable

  • isl: speedup buffer fills by dropping swizzle programming

  • nir/lower_image_atomics_to_global: add heap support

  • nir/lower_non_uniform: add heap support

  • nir/lower_robust_access: add heap/bindless support

  • vulkan/runtime: convert descriptor heap pipeline flag to shader flag

  • nir/divergence: handle resource_intel like other intrinsics

  • nir: add heap variant of load_param_intel

  • nir/lower_io: add index support for load_param_intel

  • nir: divergence analysis support for image_heap_load_param_intel

  • brw: make the program key available on pass_tracker

  • anv: use arrays instead of vectors for descriptor set mapping

  • anv: bump MAX_SETS to 32

  • anv: don’t relocate memory from blob

  • brw: add support for < 32bit io values

  • anv: enable storageInputOutput16

  • brw: don’t support frontfacing ternary optimization on != 32bit

  • elk: don’t support frontfacing ternary optimization on != 32bit

  • anv: avoid C23

  • anv: fix compute push constant allocations on pre Gfx12.5 platforms

  • anv: fix invalid value for push block index

  • anv: fix debug printfs on hang

  • anv: fixup compute queue detection

  • anv: fix null pointer access

  • anv: fix arc artifacts on Farming simulator 2022

Liu, Mengyang (1):

  • aco: fix broken VGPRs reservation for 64-bit attributes in VS prologs

Liviu Prodea (2):

  • clc: Fix static link with clang>=22

  • util: Fix use of undeclared identifier ‘NULL’ in src/util/os_misc.h when compiling with clang

Lorenzo Rossi (49):

  • nvk,nak: Add nir_printf_fmt

  • nir/opt_varyings: Skip code-motion for upconversions

  • pan/compiler: Remove mediump from collect_varyings

  • panvk: Constant fold location in panvk_lower_nir_io

  • pan/compiler: Add formats to varyings info

  • panvk/jm: Emit varying descriptors based on the VS varying layout

  • panvk: Cleanup shaders linking

  • panfrost: Build explicit varying layout

  • pan/compiler: Dont build ABI automatically

  • panfrost: Refine names in pan_cmdstream descriptor functions

  • panfrost/bi: Emit varying descriptors based on the VS varying layout

  • panfrost/val: Emit varying descriptors based on the VS varying layout

  • pan/compiler: Remove collect_varyings

  • pan/compiler: Remove unused descriptor info

  • pan/compiler: Mostly remove auto32 varting store

  • pan/compiler: Remove auto32 quirk

  • panfrost/docs: Document varyings

  • panfrost: Switch to compact varyings layout

  • pan: Remove dead code for sso_abi builder and fixed_varyings

  • people: Update my email

  • mailmap: Update my email

  • pan/bi: Add is_psiz_store flag in bi_instr

  • pan/bi,nir: Divide memory_access from segments

  • Revert “pan/bi: Model pos/vary segments in STORE instructions”

  • panfrost: Lower indirect derefs before lower_io

  • pan/bi: Resize varyings IO early

  • pan/compiler: Remove dead ABI function

  • panvk,panfrost: Always emit ld_var_buf when possible

  • panfrost/docs: Fix v9+ varyings description

  • pan/compiler: Remove unused hack in varyings stores

  • panfrost/bi: Separate va_shader_output from bitmasks

  • pan: Add PAN_MAX_MULTIVIEW_VIEW_COUNT

  • pan/compiler: Refactor va_shader_output_from_ in common code

  • pan/compiler: Split lower_varyings_io into fs_inputs and vs_outputs

  • pan/compiler: Group outputs in lower_vs_outputs

  • pan/compiler: Make lower_vs_outputs write needs_extended_fifo

  • pan/compiler: Add bound assert on emit_split_i32

  • pan/compiler: Replace frag_coord_zw_pan with var_special_pan

  • pan/compiler: Replace bi_lower_ldexp16 with algebraic pass

  • pan/compiler: Split bi_debug.c from bifrost_compile.c

  • pan/compiler: Split bifrost_nir.c from bifrost_compile.c

  • pan/compiler: Don’t crash nopersp if pos is undefined

  • pan/compiler: Fix noperspective int varyings

  • panfrost,panvk: Move postprocess near shader_compile

  • panfrost: Move lower_res_indices before postproc

  • panfrost,panvk: Move lower_texture_late inside postproc

  • panfrost,panvk: Move lower_texture_early inside preproc

  • pan/compiler: Document compilation pipeline expectations

  • panvk/jm: Fix tls_size overwrite in indirect draws

Louis Montagne (1):

  • zink: relax build-id length assertion for Mach-O

Loïc Molinari (17):

  • util/perf: Replace tabs with spaces

  • util/perf: Reorder ending CPU scope traces to match beginning order

  • util/perf: Add support for conditional CPU scope traces

  • pan/trace: Add wrappers for Mesa CPU scope traces

  • panfrost: Port CPU scope traces to PAN_TRACE_*()

  • panfrost: Add new CPU scope traces

  • panfrost: Fix clean_pixel_write_enable forced check for AFBC

  • pan/desc: Only set clean_pixel_write_enable on clear (v4)

  • pan/desc: Emit common RGB render target config in pan_emit_rt()

  • pan/desc: Force pan_merge() ending semicolon

  • pan/desc: Move funcs closer to callers

  • pan/desc: Cache clean tile state

  • pan/desc: Issue TSIX-2033 only affects pre-frame shaders

  • panfrost: Update clean_pixel_write_enable flag name for v6+

  • panfrost: Fix -Wunused-variable warning on arch > 7

  • panfrost: Fix -Wunused-variable warnings using ASSERTED

  • panfrost: Fix -Wunused-but-set-variable warnings using ASSERTED

Lucas Fryzek (8):

  • drisw: Properly mark shmid as -1 when alloc fails

  • x11: Add helper util to check for xshm support

  • egl/dri: Check that xshm can be attached

  • glx: Check that xshm can be attached

  • vulkan/wsi: Check that xshm can be attached

  • lvp: Mark opaque FD and dmabuf as compatible is supported

  • lvp: Export -1 as sync fd

  • tu: fix reg size for a8xx_gen1

Lucas Stach (1):

  • etnaviv: idle the pipe before flushing texture caches

Luigi Santivetti (14):

  • pvr: fix logic for setting DSMERGE and PICKD

  • pvr: fix src/dst image formats for DS resolve ops

  • pvr: fix ds merge logic for blit image to image

  • pvr: fixup for wrong conflict resolution in rebase

  • pvr: allow pvr_get_copy_format to handle ycbcr formats

  • pvr: drop redundant check on VK_FORMAT_X8_D24_UNORM_PACK32

  • pvr: stop using samples to determine what src and dst formats

  • Revert “pvr: Fixup for deqp-vk.api 2d.optimal.* conformance”

  • pvr/ci: update bxs flakes to add one ycbcr test

  • zink: fix format conversion logic for the alpha emulation case

  • pco: fix Mesa-CI regression in pco texture packed formats

  • pvr: expose partial usc mrt init routine

  • pvr: keep compiler resources in sync with attachments

  • pvr: add missing multi-arch support for pipeline exec and stats

Maaz Mombasawala (7):

  • ci: Update vmware farm admins.

  • svga: Update ci failure expectations.

  • svga: Update CI expectations

  • svga: Update CI expectations.

  • Revert “ci: disable vmware farm”

  • svga: Use gfx-ci kernel in CI

  • Revert “ci: vmware farm is offline, stop using it”

Marc Alcala Prieto (1):

  • pan/cs: Fix cs_run_fragment() calls with swapped arguments

Marek Olšák (122):

  • ac/nir/meta: tune image clear & copy performance for gfx12

  • ac/nir/meta: tune 12B clear buffer performance for gfx12

  • ac,radeonsi: set optimal COMPUTE_DISPATCH_INTERLEAVE for buffer clears/copies

  • radeonsi: don’t use pipe_surface_size in clears

  • radeonsi: add faster color clear for gfx12

  • radeonsi: test bigger sizes for AMD_TEST=dmaperf

  • radeonsi: disable 2D compute dispatch SE interleaving

  • nir/print: fix a crash due to unhandled GLSL_SAMPLER_DIM_EXTERNAL

  • nir: add ACCESS_SPARSE

  • nir: add dest_type to load_buffer_amd

  • nir/opt_16bit_tex_image: lower dst of load_buffer_amd

  • radeonsi: unify tex descriptor loads

  • ac,radeonsi: add AC_NIR_TEX_BACKEND_FLAG_IS_IMAGE

  • aco: handle ACCESS_SPARSE and ACCESS_SKIP_HELPERS for load_buffer_amd

  • ac: set missing dest_type for image_deref_load

  • aco,ac/llvm: force IDXEN=1 for buffer format opcodes on GFX9

  • ac/llvm: handle ACCESS_SPARSE in load_buffer_amd

  • ac/nir: rename ac_nir_lower_tex -> ac_nir_lower_image_tex

  • ac/nir: lower buffer txf to load_buffer_amd in NIR

  • ac/nir: lower buffer image_load to load_buffer_amd in NIR

  • ac: remove txf buffer code from ACO & LLVM

  • ac: remove image_load buffer code from ACO & LLVM

  • ac/llvm: fix buffer_load_format with TFE by replacing inline asm with LLVM code

  • ac/llvm: remove scalarizing emit_intrin_1f_param_scalar

  • ac/llvm: simplify emitting FP intrinsics

  • ac/llvm: pass soffset to buffer_load/store_format

  • radeonsi: move VB descriptor emission and upload into separate functions

  • radeonsi: don’t allocate a VB descriptor upload buffer if it’s not needed

  • nir: reassociate a $op (b ? #c : #d) for div, mod, rem

  • ac: unify DCC clear code definitions

  • radv: use DCC definitions more

  • ac: unify and demystify CMASK clear codes

  • ac: unify HTILE codes and encoding

  • ac: add FMASK codes

  • ac: lower load_workgroup_ids for ACO in NIR

  • ac: lower load_subgroup_id for ACO in NIR

  • ac/nir: add ac_nir_lower_intrinsics_to_args_options structure

  • ac: lower load_num_workgroups in NIR

  • ac/llvm: remove unused functions

  • nir: handle get_ubo_size as a resource query in nir_shader_gather_info

  • nir: add ACCESS to get_ubo_size

  • nir: allow get_ssbo_size to return a 64-bit result

  • nir/lower_non_uniform_access: add an option not to lower tex & image queries

  • nir/opt_non_uniform_access: use new query flags

  • radeonsi: remove CB_RESOLVE

  • radeonsi: don’t fail a CB_RESOLVE assertion on gfx11

  • ac/nir/meta: don’t scalarize sparse loads if the address is aligned to load size

  • ac/nir/meta: use the clear/copy compute shader if CP DMA doesn’t support sparse

  • ac/nir/meta: properly align sparse buffer clears with 12-byte clear values

  • radeonsi: remove the CP DMA workaround for sparse for GFX6-9

  • radeonsi: replace null descriptors with memset

  • st/mesa: optimize glCopyImageSubData for 3D and array textures

  • amd: add meson variable idep_amd_generated_headers for all generated headers

  • amd: add gfx11 and gfx12 CP packet definitions

  • ac/gpu_info: handle more VRAM types

  • ac/llvm: encode LLVM cache flags manually

  • ac: tidy up ac_hw_cache_flags

  • radeonsi: don’t set any EXPCLEAR state on gfx12+

  • ac: set the correct number of Z planes for ALLOW_EXPCLEAR

  • ac: add ac_cu_info::has_fmask, adjust FMASK checks

  • radv: make radv_postprocess_binary_config more correct and more readable

  • radv,radeonsi: PA_SC_BINNER changes for gfx12

  • radeonsi: rename si_shader_gs/vs -> si_shader_gs/vs_legacy

  • radeonsi: don’t fail si_compute_blit for compressed/subsampled formats properly

  • radeonsi: add debug options forcing fast clear, gfx and compute blits

  • radeonsi: remove AMD_TEST=blitperf

  • amd/packets: remove non-existent CLEAR_STATE from gfx12 definitions

  • amd: generate a packet parser/printer automatically from packet definitions

  • ac: enable the new auto-generated CP packet parser

  • ac: replace some packet field definitions in sid.h by generated ones

  • meson.build: require python 3.10, try python3.12

  • nir/inline_uniforms: rename num_offsets -> num_uniforms

  • nir/inline_uniforms: rename new_num -> new_num_uniforms

  • nir/inline_uniforms: update comments

  • nir/inline_uniforms: track visited state per component

  • nir: change export_amd intrinsics to use enabled_channels instead of write_mask

  • nir: change export_amd intrinsics to use target instead of base

  • Inline SHA1_DIGEST_LENGTH

  • Inline SHA1_DIGEST_STRING_LENGTH

  • Inline mesa_sha1, SHA1_CTX

  • Inline SHA1* functions, remove sha1.h

  • Inline _mesa_sha1_init/update/final functions

  • Remove redundant BLAKE3_KEY_LEN32

  • Inline _mesa_sha1_compute/format, remove the other unused ones

  • Remove mesa-sha1.h

  • util: rename the sha1 test to blake3 test

  • Rename SHA1_* names to BLAKE3_*

  • Rename sha1_* and sha_* names to blake3_*

  • Rename sha words to blake3

  • Rename more sha and sha1 names to blake3

  • Rename SHA1 words to BLAKE3

  • Rename *_sha1 names to *_blake3

  • Final rename of sha1 names to blake3

  • Change remaining SHA-1 occurences to BLAKE3

  • driconf: unbreak profiles for “runner” by merging them and ignoring sha1s

  • driconf: rename sha1 option to blake3

  • radeonsi: recompute IO bases after optimizations

  • radeonsi/meson: don’t use llvm variables when LLVM is disabled

  • ac/llvm: remove almost duplicated ac_build_varying_gather_values

  • ac/llvm: inline ac_build_gather_values_extended

  • radeonsi: remove unnecessary ac_to_integer in si_llvm_ps_build_end

  • radeonsi: fix compiler selection for fixed-func TCS

  • radeonsi: fix an assertion failure for sampler descriptor loads with LLVM

  • ac/nir/meta_cs_blit: use uint16 for coordinates to fix 64K blits

  • gallium/u_blitter: allow using the single triangle for scaled blits too

  • radeonsi: fix blits via util_blitter_draw_rectangle

  • radeonsi: disable streamout queries for u_blitter

  • radeonsi: add 64K texture support to gfx blits

  • radeonsi: remove always-set SI_SAVE_FRAGMENT_STATE

  • radeonsi: sink si_get_pipe_constant_buffer in si_blitter_begin

  • radeonsi: draw using a single triangle in u_blitter

  • nir: return a failure value from nir_system_value_from_intrinsic

  • nir: factor out nir_system_value_from_instr from nir_opt_varyings

  • nir/opt_varyings: move expressions with view_index into preceding shaders

  • nir/tests: test nir_opt_varyings with sysvals

  • ac,radv: use AC_TRACKED_DB_PA_SC_VRS_OVERRIDE_CNTL for PA_SC_VRS_OVERRIDE_CNTL

  • radv,radeonsi: don’t set PA_SC_HIS_INFO

  • ac,radv: remove AC_TRACKED_DB_VRS_OVERRIDE_CNTL as well

  • amd/packets: fix the size of 1-bit bitfields

  • amd/packets: remove the underscore between opcode number and word index, use %x

  • amd/packets: add disable_wr_confirm alias to dis_wc

  • amd: switch to new packet definitions for all packets

Mario Kleiner (3):

  • v3dv: Enable VK_KHR_present_id and VK_KHR_present_wait

  • v3dv: Enable VK_EXT_hdr_metadata.

  • dri: Fix “cosmetic” undefined behaviour warning for RGB[A]16_UNORM formats.

Martin Roukala (né Peres) (3):

  • ci: disable the valve-kws farm

  • Revert “ci: disable the valve-kws farm”

  • zink/ci: mark the unvanquished trace on vangogh as flake

Mary Guillemard (40):

  • nvk/nvkmd: Do not limit exec_push count in nvkmd_nouveau_exec_ctx

  • mr-label-maker: Mark CI files for NVK with the NVK label

  • nvk: Reenable compression support with nouveau 1.4.2

  • nvk: Report NIR shader in pipeline executable properties

  • nvk: Reorder view_mask checks in nvk_mme_clear

  • nvk: Rename DRAW_BEGIN scratch to DRAW_TOPOLOGY

  • nvk: Use DRAW_CONTROL_A on Turing+

  • nvk: Early return in draw commands when no draw will be performed

  • hk: Fix crash in hk_handle_passthrough_gs

  • vulkan: Do not override the shader_flags in case of no task shader

  • nir: Add isbewr_nv intrinsic and extends isberd_nv

  • nak: Legalize ISBERD

  • nak: Implement ISBEWR and extend ISBERD implementation

  • nak/nvdisasm_tests: Test ISBERD and ISBEWR

  • nir, nvk, nak: Add base to isbewr_nv and isberd_nv

  • docs/nvk: Fix link for subchannel switches

  • nvk/mme: Add missing nullcheck in nvk_mme_test_state_state

  • nvk: Put nvk_mme in the nouveau test suite

  • nvk/mme: Enable testing for Kepler

  • nvk: Validate push constant offset in nvk_root_descriptor_table

  • nvk: Move viewport and scissor emit to their own function

  • nvk: Broacast viewport0 and scissor0 in case of FSR on Turing

  • nir/dead_cf: Add missing load_ssbo_ir3 handling

  • nir/dead_cf: Add missing load_global_bounded handling

  • nir/dead_cf: Add missing load_global_nv handling

  • nak: Do not allow load_helper_invocation reordering

  • agx: Fix alpha-to-coverage bit size

  • nvk: Use SET_PRIMITIVE_TOPOLOGY instead of MME scratch

  • nvk: Move shader size and offset calculations to nvk_shader_get_shader_size

  • nvk: Wire up shader program prefetch method

  • nvk: Ensure that shader I-cache prefetch is enabled on Ada+

  • nvk: Do not fill cb0 at queue creation

  • nvk: Do not use SET_L1_CONFIGURATION on 3D state init

  • nvk: Set VAF eviction policy to nornmal

  • nvk: adjust reduce color thresholds default values

  • nvk: Remove old comments from draw state init

  • bin: Add Tested-by in rb.py

  • nvk: Adjust maxFragmentCombinedOutputResources to match max descriptors limit

  • hk: Add HK_MAX_RTS to maxFragmentCombinedOutputResources

  • nak: Allows predicate in legalize_ext_instr

Matt Arsenault (2):

  • ac/llvm: Remove -promote-alloca workaround

  • ac/llvm: Use new denormal_fpenv attribute for llvm >= 23

Matt Coster (1):

  • ci,cirnm: Fix program name in usage example

Matt Turner (6):

  • brw/cse: fix `operands_match` corrupting non-IMM register data

  • brw/cse: use copies in `operands_match` instead of in-place modification

  • elk/cse: fix `operands_match` corrupting non-IMM register data

  • elk/cse: use copies in `operands_match` instead of in-place modification

  • intel/elk: Remove dead TXL_LZ/TXF_LZ opcodes

  • radv: fix UB in radv_format_pack_clear_color for snorm formats

Mauro Rossi (12):

  • vulkan/runtime: Fix gnu-empty-initializer error in vk_pipeline.c

  • lavapipe: Fix gnu-empty-initializer error in NV_cooperative_matrix2 conversions

  • lavapipe: Fix gnu-empty-initializer error in NV_cooperative_matrix2 reductions

  • vulkan/runtime: Fix gnu-empty-initializer error in vk_shader.c

  • radv: Fix gnu-empty-initializer error in radv_pipeline_graphics.c

  • radv: Fix gnu-empty-initializer error in radv_pipeline_rt.c

  • radv: Fix gnu-empty-initializer error in radv_pipeline_compute.c

  • radv: Fix gnu-empty-initializer error in radv_shader_object.c

  • radv: Fix gnu-empty-initializer error in prolog_stage

  • intel/jay: fix static_assert expression

  • radv: Fix gnu-empty-initializer errors in 480a94fb

  • radv: Fix gnu-empty-initializer errors in 8c10eab1

Maíra Canal (11):

  • broadcom/ci: skip tests that causes GPU resets/hangs in RPi 3

  • broadcom/compiler: Don’t lower to LCSSA before calling nir_divergence_analysis()

  • vc4: drop redundant shader->failed reassignment

  • nir: add load_texture_scale intrinsic

  • vc4: fail VS compilation on divergent loops

  • broadcom/ci: don’t skip dynamic loop tests in RPi 3

  • v3d: increase BO allocation size when growing CLs

  • v3d: use the state uploader for the image view texture shader state

  • v3d: sub-allocate sampler view texture state from state uploader

  • v3d: Rename cle_buffer_min_size to page_size

  • v3d: use devinfo->page_size for state uploader default size

Mel Henning (66):

  • nvk: Use layout->vk.dynamic_descriptor_count

  • nvk: Use pipeline_layout.dynamic_descriptor_offset

  • hk: Use layout->vk.dynamic_descriptor_count

  • hk: Use pipeline_layout.dynamic_descriptor_offset

  • kk: Use layout->vk.dynamic_descriptor_count

  • kk: Use pipeline_layout.dynamic_descriptor_offset

  • nvk: Ignore meta ops in occlusion queries

  • nvk: Disable large pages for now

  • nvk: Add a NVK_MME_VAL_MASK macro

  • nvk: Use macros for nvk_mme_set_tess_params tests

  • nvk: Add CCW, POINT_MODE flags for set_tess_params

  • nvk: Compute tess prims in the MME macro

  • nvk: Remove prims from tess state

  • nvk: Move tess flags between other fields

  • nvk: Use some additional drf macros

  • nak: Split out TesselationCommonShaderInfo

  • nak: Handle unspecified tess spacing

  • nvk: Merge tese/tesc state in the MME

  • vulkan/wsi: Call wl_display_roundtrip on our queue

  • nvk: Initialize SET_ALPHA_TO_COVERAGE_OVERRIDE

  • nvk: Report additional host_image_copy layouts

  • zink: Emit float controls for preserve_denorms too

  • zink: Generalize spirv_builder_emit_exec_mode_id3

  • zink: Use float_controls2

  • zink: Use NMin/NMax for fmin/fmax if nan_preserve

  • nvk,nak: Store offsets in a const extern struct

  • nak: Remove some unused fs_key parameters

  • nvk: Don’t include u_math.h in generated headers

  • nouveau/headers: Don’t use 128-bit comparisons

  • nouveau/headers: Use UINT64_C in drf.h

  • libcl_vk: Add VkCopyMemoryIndirectCommandKHR

  • nouveau/headers: Add P_IMMD_WORD()

  • nvk: VK_KHR_copy_memory_indirect

  • drm-uapi: Sync nouveau_drm.h

  • nouveau/winsys: Fetch zcull_info on device create

  • nouveau/headers: Preserve _ before 0-9 in to_camel

  • nil: Add zcull support

  • nvk: Enable basic zcull support

  • nvk: Enable zcull for VK_ATTACHMENT_LOAD_OP_LOAD

  • nvk: Remove unused cmd.tls_space_needed

  • driconf: force_vk_vendor on No Man’s Sky + NVK

  • nvk: Use SET_GLOBAL_RENDER_ENABLE

  • nvk: Use the MME for cond rendering on Turing+

  • nvk: Expose VK_KHR_depth_clamp_zero_one

  • nvk: Disable descriptorBufferCaptureReplay for now

  • nir/lower_io: Add global_bounded to io_offset_src

  • nir/mem_access_bit_sizes: Handle global_bounded

  • nak: Fix mufu’s f16 bit on sm90+

  • nvk/lower_descriptors: Move load_root_table up

  • nvk/lower_descriptors: Use more load_root_table

  • nvk/lower_descriptors: .base in load_root_table

  • nvk/lower_descriptors: Add load_root_table_array()

  • nvk/lower_descriptors: Change ROOT_DESC addr space

  • nvk: Rename macro loop index from i to _index

  • nvk: Swizzle root_table.dynamic_buffers[]

  • nvk: Initialize NVC597_SET_ROOT_TABLE_VISIBILITY

  • nvk: Reorder nvk_root_descriptor_table

  • nvk: Factor out build_push_write_push_const

  • nak: Turn nak_const_offsets into a function.

  • nak: Add an is_graphics param to nak_const_offsets

  • nak: Add printf_cb to nak_constant_offset_info

  • nvk/cmd_indirect: Pass pdev into more functions

  • nvk: Move mme_set_anti_alias_tests to a check func

  • nvk: Wire up ROOT_TABLE

  • nvk: SET_ROOT_TABLE_PREFETCH

  • nvk: Disable zcull save/restore regions for now

Michael Cheng (15):

  • vulkan: add vk_shader_ops::replay_at vfunc stub

  • anv: Implement RT shader group handle capture/replay

  • anv: Rename instruction_state_pool to shader_heap

  • intel/blorp: add explicit clear op enums for stencil and linear paths

  • intel/blorp: Remove unused blorp_gfx8_hiz_clear_attachments

  • intel/blorp: use dedicated clear ops in clear paths

  • vulkan/runtime: allow drivers to enable vk_log output in release builds

  • anv: enable perf warning logging in release builds

  • hasvk: enable perf warning logging in release builds

  • intel/ds: report when OA metric access is blocked by kernel policy

  • intel/ds: report when OA metrics are unavailable

  • anv: log fast color clear fallback reasons in vkCmdClearAttachments

  • anv: log fast depth clear fallback reasons in vkCmdClearAttachments

  • anv: log aux disable and aux-skip reasons during image setup

  • anv: log aux disable reasons in image init and DRM modifier selection

Michal Krol (3):

  • gallium: add rasterization_stream to pipe_rasterizer_state

  • draw: fix per-stream vertex buffer leak in non-LLVM path

  • lavapipe: implement transformFeedbackRasterizationStreamSelect

Michel Dänzer (13):

  • Pass the destination buffer size minus one to strncpy

  • ci: Drop -Wno-error=vla-cxx-extension from debian-x86_64-msan job

  • ci: Drop remaining -Wno-error stanzas from debian-x86_64-asan/ubsan jobs

  • ci: Drop -Wno-error=stringop-overread from debian-release job

  • ci: Drop some -Wno-error stanzas from the debian-android job

  • ci: Drop -Wno-error stanzas from debian-no-libdrm job

  • ci: Drop half of -Wno-error stanzas from fedora-release job

  • ci: Drop most -Wno-error stanzas from debian-arm64 jobs

  • ci: Drop a couple of -Wno-error stanzas from alpine-build-testing job

  • vulkan/wsi/x11: Guard XCB_PRESENT_OPTION_SUBOPTIMAL by ignore_suboptimal

  • vulkan/wsi/x11: Don’t use modifiers when ignoring SUBOPTIMAL

  • winsys/amdgpu: Prefer render node FD for ac_drm_device_initialize

  • winsys/amdgpu: Use render node only as fallback

Mike Blumenkrantz (100):

  • zink: re-allow transient images during blitting

  • zink: break out ntv into separate meson dep

  • ntv: emit extra decorations for matrix members of structs

  • ntv: stop explicitly tracking variables for samplers/images

  • ntv: handle a couple trivial builtin loads

  • ntv: shore up shader_temp var handling

  • ntv: add push const variable to ctx->vars hash table

  • ntv: handle glsl texture types

  • ntv: handle bare sampler arrays

  • ntv: add basic vulkan support

  • ntv: emit demote extension/capability when emitting demote

  • ntv: add a simple pass to convert vulkan descriptor access to direct derefs

  • ntv: stop tracking ubo variables

  • ntv: avoid setting Block decoration repeatedly on bo struct types

  • ntv: improve setting Aliased decoration on bo emits

  • ntv: emit ViewIndex with flat for fragment stage

  • ntv: handle nir_intrinsic_load_first_vertex as basevertex

  • zink: fix broken compiler assert

  • zink: only do pre-sync transfer barrier after a renderpass

  • zink: only update the value of VkAttachmentFeedbackLoopInfoEXT, not the pNext

  • zink: use maintenance10 info for DRLR optimization

  • ci: add ASAN_OPTIONS=malloc_fill_byte=1 for asan jobs

  • ntv: also use base glsl type for non-zink array derefs

  • ntv: ignore stuff for get_ssbo_size() in vulkan mode

  • zink: add TRANSFER_WRITE -> HOST_READ sync to end of batch

  • st/bitmap: only release YUV samplerviews

  • ntv: run nir_cleanup_functions() in ntv_shader_prepare()

  • ntv: re-gather shader info after ntv_shader_prepare

  • ntv: run nir_remove_dead_variables during ntv_shader_prepare()

  • ntv: call nir_lower_variable_initializers() from ntv_shader_prepare

  • radv: fix multiview fast clears

  • ntv: do gl-style shared/task lowering for vulkan mode

  • ntv: run opt_algebraic late for prep optimization pass

  • vk/cmd_queue: use arrays to directly manage refcounting

  • vk/cmd_queue: handle descriptor layout refcounting

  • vk/cmd_queue: return cmd instead of error code

  • vk/cmd_queue: pass command to struct copying methods

  • vk/cmd_queue: generate CmdBindDescriptorSets

  • vk/cmd_queue: generate CmdPushDescriptorSet

  • vk/cmd_queue: move pipeline layout refs into builder

  • vk/cmd_queue: generate the rest of the descriptor functions

  • vk/cmd_queue: generate CmdPushConstants2

  • egl/device: fix the fix for explicit sw rejection in non-sw EGL_PLATFORM=device

  • zink: reapply zsbuf state after unordered blits

  • zink: allow renderpass termination for clears with ZINK_DEBUG=rp and GENERAL layouts

  • zink: run opt_combine_stores when optimizing

  • aux/trace: handle set_sample_locations

  • llvmpipe: enable GLSL 4.60

  • nir: fix nir_is_io_compact for mesh shaders

  • mesa/st: fix unlower_io_to_vars to work with mesh shaders

  • lavapipe/llvmpipe: make mesh draw params consistent

  • llvmpipe: support EXT_mesh_shader

  • mesa/st: make st_texture_get_current_sampler_view static

  • mesa/st/sampler_view: use a local variable for buffer sv format

  • mesa/st/sampler_view: use a local variable for texture sv format

  • mesa/st/sampler_view: eliminate st_sampler_view::srgb_skip_decode

  • mesa/st/samplerview: explicitly block releasing in-use samplerviews

  • zink: work around drivers with broken mesh shader properties

  • nir/print: print per_vertex for variables

  • llvmpipe: save mesh shader when calling u_blitter

  • llvmpipe: fix mesh cap exports

  • lavapipe: fix mesh property exports

  • llvmpipe: set prefer_real_buffer_in_constbuf0 and delete user buffer path

  • r300: import util_framebuffer_init

  • gallium/util: kill off util_framebuffer_init

  • util/cso: use the mesh_shader pipe cap for mesh support

  • ntv: always emit const coord components for fbfetch loads

  • mesa/renderbuffer: always add PIPE_BIND_SAMPLER_VIEW to rendering textures

  • llvmpipe: fix color fbfetch

  • softpipe: delete pipe_context::create_surface

  • svga: delete pipe_context surface hooks

  • llvmpipe: delete pipe_context surface hooks

  • svga: simplify some surface management

  • crocus: clean up surface management

  • tc: delete unused surface ref code

  • freedreno: clean up some surface management

  • nouveau: delete unused surface hook

  • tegra: delete pipe_context surface hooks

  • freedreno: delete pipe_context surface hooks

  • r300: clean up some surface management

  • r300: delete pipe_context surface hooks

  • gallium: add a destructor param to surface refcounting functions

  • gallium: delete pipe_context surface hooks

  • gallium: add a pipe_context param to pipe_surface_reference()

  • svga: move surface context member onto internal surface type

  • gallium: kill off pipe_surface::context

  • zink: use EXT_primitive_restart_index

  • lavapipe: update prim restart index on index buffer bind

  • lavapipe: VK_EXT_primitive_restart_index

  • radv: handle null pCounterBuffers with xfb binds

  • vulkan/runtime: handle null pCounterBuffers with xfb binds

  • llvmpipe: fix min_samples + A2C

  • lavapipe: fix indirect memory copies

  • lavapipe: fix pushconst data updating

  • util/format: support 256-bit formats in util_format_get_tilesize()

  • lavapipe: use the right type for DGC mesh draws

  • lavapipe: rework immutable samplers

  • lavapipe: allow fbfetch with shader objects

  • vk/cmd_queue: always ceil() param lens

  • llvmpipe: always set view_index for linear rasterizer

Mixie (6):

  • xlib: clear currentDpy when releasing the current context

  • xlib: use XMesaDestroyVisual when destroying display visuals

  • xlib: use XMesaDestroyVisual instead of manual free

  • xlib: fix skipping visuals in destroy_visuals_on_display

  • xlib: remove vishandle from XMesaVisual and fix XVisualInfo leak

  • xlib: clear currentDpy when switching current context

Mohamed Ahmed (7):

  • nil/modifiers: Clarify drm_format_mods_for_format rejecting modifiers for unsupported color formats

  • nvk: Calculate and stash the plane offset and alignment at create time

  • nvk: Extend tiled_shadow to be multiplanar

  • nvk: Defer tiled shadow plane memory allocation to draw time

  • nvk: Enable multiplanar YCbCr linear modifiers

  • nvk: Use the pre-calculated offsets for sparse binds

  • nvk: Remove nvk_image_plane_size_align_B()

Máté Pinczel (1):

  • nak: implement uror and urol using shf

Nanley Chery (63):

  • intel/isl: Use 1x ACM Tile64 swizzle on Xe2

  • intel/isl: Use 1x Ys/Yf swizzle for IMS layout

  • intel/isl: Set TileAddressMappingMode for CMS/UMS

  • intel/isl: Fix miptail selection for compressed textures

  • iris: Disable some 8bpp fast-clears within miptail

  • iris: Increase imported dmabuf alignment for 64K+ BOs

  • iris: Use PIPE_BIND_SHADER_IMAGE more

  • iris: Limit resolves for atomics to R32 formats

  • iris: Allow Yf and Ys tilings more often

  • intel/isl: Rework miptail restrictions with CCS

  • intel/isl: Reduce scope of Yf-disabling workaround

  • anv: Disable multisampled host transfer support

  • anv: Ensure host-transfer tilings are supported by ISL

  • blorp: Fix Tile64 clear redescription assertion

  • intel: Add and use ISL_SURF_USAGE_PREFER_4K_ALIGNMENT

  • intel/isl: Refactor tiling selection in isl_surf_init_s

  • intel/isl: Prefer the smallest suggested tiling

  • intel/isl: Drop HIZ/MCS checks in CCS support query

  • intel/isl: Prefer suggested tilings which use CCS

  • anv: Query the plane in anv_can_fast_clear_color()

  • anv,iris: Don’t fast-clear 3D + Ys on gfx12.0

  • intel: Enable CCS support for Yf and Ys

  • intel/isl: Fix QPitch of arrayed MCS

  • iris: Set missing flags on clear color changes

  • iris: Use the CLEAR state on Xe2+ for MCS

  • anv: Update predicated resolve documentation

  • anv: Fix the fast clear type for FCV writes

  • anv: Reset fast-clear type in transition_color_buffer()

  • anv: Support partial resolves on any level/layer

  • anv: Set compressed bit separately from fast-clear type

  • anv: Delete conversion of CCS_D partial resolve

  • anv: Inline the CCS/MCS predicated resolve functions

  • anv: Line wrap anv_CmdClearColorImage

  • anv: Don’t return the Xe2+ fast-clear type early

  • anv: Use variable default value for some images using CLEAR

  • anv: Support fast clears on more layers

  • anv: Don’t partial resolve LOD1+ for non-FCV CCS

  • intel/blorp: Avoid unused surface redescription calc

  • intel/blorp: Optimize non-zero-layer fast-clears

  • intel/blorp: Bump pitch when clearing unaligned bottom rows

  • anv: Fix clear state of WSI blit sources during presentation

  • anv: Treat non-WSI PRESENT_SRC as TRANSFER_SRC

  • anv: Don’t set the display flag on WSI blit sources

  • anv: Drop anv_image::wsi_blit_src

  • intel: Disable CCS_E support for YCRCB on gfx12

  • intel/isl: Add YCRCB CMF mappings for Xe2+

  • anv: Enable YCRCB CMFs on Xe2+

  • intel/blorp: Fix the redescribed fast-clear qpitch

  • intel/isl: Replace mc_format with aux_format

  • intel/isl: Increase 3D miptail workaround scope

  • intel/isl: Generalize and move some Yf/Ys miptail limits

  • intel/isl: Relax some alignments in get_image_surf()

  • intel/isl: Use a fixed alignment for single slices

  • intel/blorp: Lower bit-casting code in blorp_copy()

  • intel/blorp: Add blorp_surf::has_replicated_pixel

  • anv: Prepare for format width changes in blorp_copy()

  • anv: Add WaSamplerCacheFlushBetweenRedescribedSurfaceReads

  • intel/blorp: Make blorp_copy() format queries aux-dependent

  • intel/blorp: Use stencil hardware less for CPB copies

  • intel/blorp: Add blorp_surf_convert_to_single_level_tile()

  • intel/blorp: Redescribe surfaces for copies

  • isl: Apply VALIGN_8 fast-clear restriction on Xe3P+

  • intel/blorp: Fix width scaling for YCBCR copies

Natalie Vock (26):

  • aco: Fix parameter stack size calculation

  • radv/rt: Refactor shader group stack size calculation to include traversal stack

  • aco: Don’t exclude discardable parameters from register preservation

  • radv/rt: Fix some tail-call compatibility checks

  • radv/rt: Fix discardable attributes on chit and traversal shaders

  • meson: Identify LTO builds in the package version

  • mesa: Prevent building with LTO

  • radv/rt: Only use ds_bvh_stack_rtn if the stack base is possible to encode

  • radv: Initialize nir_lower_io_to_scalar progress variable

  • radv/nir: Correctly handle workgroup sizes not aligned to 32

  • radv/rt: Bump ray query stack base limit for GFX12

  • radv/rt: Fix shared ray query stack on top of application LDS

  • vulkan: Rename {encode,update}_bind_pipeline to {encode,update}_prepare

  • vulkan: Bump MAX_ENCODE_PASSES

  • radv/rt: Fix cases in which the bound BVH build pipeline gets clobbered

  • radv/rt: Remove RADV_OFFSET_UNUSED

  • radv/rt: Don’t enable midpoint sorting

  • radv/rt: Don’t combine config of unreachable shaders

  • radv: Disable RADV_DEBUG=llvm in release builds

  • aco/spill_preserved: Only compute preserved registers if in a callee

  • aco/isel: Fix stack_ptr synthesis

  • aco/lower_to_hw_instr: Run p_init_scratch if the program has a call

  • radv: Rewrite the RT prolog in NIR

  • aco: Nuke ACO-side prolog selection

  • nir/deref: Elide loads/stores from deref cast of undef

  • radv: Run nir_opt_deref after first optimization loop

Nataraj Deshpande (1):

  • anv: Fix ASTC emulation sync in CopyImage and CopyBufferToImage

Nick Hamilton (12):

  • pvr: Temporarily disable the buffer device address extension

  • pco: Fix for atomic operations on an image buffer

  • pvr: Fix the isp samples per tile calculation

  • pco: Fix multiview sampling of subpass input attachments

  • pvr: Fix incorrect subpass merging optimisation

  • pvr: Rename pvr_render_input_attachment

  • pvr: Add missing support for preserve attachments

  • pvr: Update CI fails list after render pass fixes

  • pvr: Add support for fragment pass through shader

  • pvr: Fix for multiple attachments being assigned to the same tile buffer.

  • pco: fix clamping the array index when shaderImageGatherExtended is enabled

  • pvr: Revert don’t csb emit multi-layer clear attachments without rta support

OPNA2608 (2):

  • vc4: Fix printing of get_tiling.modifier

  • rocket: Fix printing of rknpu_mem_create.dma_addr

Olivia Lee (13):

  • Revert “panvk: advertise VK_EXT_primitives_generated_query on v10+”

  • hk: fix hk_passthrough_gs_key size computation

  • hk: fix passthrough GS key invalidation

  • panvk/csf: use different resource registers for precomp vs user dispatch

  • pan/va: weaken barrier requirements for allow_merging_workgroups

  • pan/va: move allow_merging_workgroups decision to drivers

  • pan/va: don’t merge workgroups when subgroups are used

  • panvk/csf: take merged workgroups into account for divergence

  • panvk/csf: lower divergent values introduced by merged workgroups

  • panvk/csf: enable allow_merging_workgroups when possible

  • panfrost: don’t try to emit varying shader stats on v12+

  • panvk/csf: check printf buffer one last time when queue is lost

  • pan/bi: fix memory access alignment

Olle Lögdahl (3):

  • aco/isel: move if_context and loop_context to heap

  • aco/isel: use iterative visitor during traversal

  • aco/isel: added test-case for iterative cf visitor

Osama Abdelkader (1):

  • vulkan/wsi: Fix realloc error handling in wsi_get_modifiers_for_format

Patrick Lerda (13):

  • r600: fix cayman msaa shading behavior

  • r600: disable l8_srgb on r700 and older gpus

  • r600: fix rv770 dot4 operations

  • r600: make vertex r10g10b10a2_sscaled conformant on palm and beyond

  • r600: fix rv770 clamp to max_texel_buffer_elements

  • r600: enable GL_EXT_shader_realtime_clock

  • r600: update cubearray imagesize calculation

  • r600: improve vs_as_ls switch reliability

  • r600: fix cs atomic operations when the shader is called multiple times

  • r600: fix alpha-to-coverage and alpha-to-one used together

  • r600: fix atomic buffer offset

  • r600: update vertex emit_varying_pos

  • r600: fix atomic_counter_post_dec

Paulo Zanoni (13):

  • anv: anv_get_image_format_features2() can be static

  • anv: don’t pass vk_format to anv_get_image_format_features2()

  • anv: unify blit_cts_workaround handling

  • anv: extract anv_color_format_supports_drm_modifier_tiling()

  • anv: extract 2 subvariants of anv_get_image_format_features2()

  • anv: extract anv_get_color_format_features()

  • vtn_bindgen2: limit the nir_opt_peephole_select optimization

  • elk: don’t use instr->const_index[] directly

  • anv: avoid VK_STRUCTURE_TYPE_BIND_MEMORY_STATUS warnings

  • intel/blorp: remove always-true #if

  • intel/genxml: move the GPGPU_DISPATCHDIM* registers to genxml

  • intel/isl: fix assert when surf->size_B is > UINT_MAX

  • intel/isl: warn about excessive num_elements only once

Pavel Ondračka (36):

  • r300: split unaligned 3D texsubimage uploads by layer

  • r300: align macro-tiled stride-addressed textures in X

  • r300/ci: enable glx tests

  • i915/ci: update expectation missed in piglit uprev

  • mesa: implement FRAMEBUFFER_RENDERABLE internalformat query

  • r300/ci: update expectations

  • i915/ci: update expectations

  • r300: handle polygon-mode points in point sprite path

  • r300/ci: update expectations

  • r300: Z16 polygon offset fixes

  • r300: copy target when merging alpha output instruction

  • r300: disable HiZ for PIPE_FUNC_ALWAYS

  • r300/ci: enable HiZ in CI

  • r300: make occlusion queries work without a bound depth buffer

  • r300: pause and resume occlusion queries for blitter/meta paths

  • r300: return zero for finished occlusion queries with no emitted results

  • frontends/dri: fix NUM_PLANES for imported dma-buf images

  • r300: disable clip-discard watermark for triangles

  • r300: pad short vertex shaders to avoid R3xx hangs

  • r300/ci: update expectations

  • r300/ci: expectation update

  • r300: fix bias presubtract algebraic transformation

  • r300: lower Z16 polygon offset scale coefficient

  • r300: don’t apply odd macroblock rounding to 3D textures

  • r300: disable zmask clears for large surfaces

  • r300: add shared HyperZ pipe-count helper

  • r300: split large HiZ clears into multiple packets

  • st/bitmap: release the temporary bitmap sampler view

  • r300: keep all vertex atributes 32bit on big endian

  • r300: fix BE 32-bit CBZB clear values

  • r300: fix BE CBZB clears for swapped 8888 formats

  • gallium/u_blitter: remove unused CONST declaration when using IMM

  • r300/ci: rv380 piglit

  • r300/ci: update expectations

  • r300: fix MSAA resolve COLORPITCH tiling after pipe_surface de-pointerization

  • r300: dirty VS state when switching variants

Pierre-Eric Pelloux-Prayer (37):

  • radeonsi/sqtt: retrieve sqtt data after the flush ended

  • radeonsi/sqtt: use pipe_aligned_buffer_create to allocate bo

  • radeonsi/sqtt: use pipe_buffer_map instead of ws->buffer_map

  • radeonsi/sqtt: allocate BOs in VRAM

  • radeonsi/sqtt: use radeon_add_to_buffer_list

  • mesa/vbo: update NeedFlush before flushing

  • dri: prevent read_sbc from going backward

  • ac: keep a single instance of sid_table

  • radeonsi: move mediump code to a separate compilation unit

  • radeonsi: split shaders/draw code from si_debug to a new file

  • radeonsi: move buffer high-level functions to si_buffer.c

  • radeonsi: move si_ps_key_update_framebuffer to si_state.c

  • ac: add u_stub.h helper

  • meson: add with_gfx_compute property

  • radeonsi: use with_gfx_compute to disable parts of the driver

  • radeonsi: remove nir references when graphics is disabled

  • frontends/va: fix undefined ref error

  • mesa: don’t wraparound st_context::work_counter

  • radeonsi: move spi_shader_*_format to si_shader_variant_info

  • radeonsi: account for outputs_written when updating spi_shader_col_format

  • radeonsi/test: update failures

  • gallium/u_blitter: add a new fs_color_clear variant

  • drm-shim: fix shim on GLX

  • winsys/amdgpu: remove assert

  • ac: remove ac_null_device

  • ac/info: add ac_fill_tiling_info

  • ac/info: add ac_fill_memory_info

  • ac/info: add ac_fill_hw_ip_info

  • ac/info: add ac_identify_chip

  • ac/info: move more memory properties to ac_fill_memory_info

  • ac/info: remove has_bo_metadata

  • ac/info: add ac_fill_bug_info

  • ac/info: add ac_fill_feature_info

  • ac/info: add ac_fill_hw_info

  • ac/info: add ac_fill_tess_info

  • ac/info: constify ac_fill_compiler_info

  • ac/tests: use amdgpu shim devices

Pohsiang (John) Hsu (12):

  • mediafoundation: refactor update picture desc

  • mediafoundation: remove published codecapi

  • mediafoundation: update version to 1.09

  • mediafoundation: in slice generation mode, send METransformNeedInput once per frame.

  • d3d12: add workaround for max subregion number reported in slice auto mode

  • mediafoundation: add workaround for max subregion number reported in slice auto mode

  • mediafoundation: fix hevc vui time_scale

  • mediafoundation: set defualt unwrapped poc for h264 to true

  • mediafoundation: set reasonable number of reference frames if the user didn’t set CODECAPI_AVEncVideoMaxNumRefFrame

  • d3d12: ifdef the surfaces member from d3d12_batch under HAVE_GALLIUM_D3D12_GRAPHICS

  • mediafoundation: add support for GPU priority setting via IMFDXGIScheduler

  • mediafoundation: remove published codecapi

Priya Hosur (1):

  • ac/nir/ngg: re-enable use of known compile-time GS connectivity

Qiang Yu (11):

  • radeonsi: move sqtt draw code to shared function with mesh pipeline

  • radeonsi: mesh shader support sqtt

  • radeonsi: be able to record sqtt for frame 0|1 and no swap

  • radeonsi: not overlap ib print for multi context

  • radeonsi: fix mesh shader outputs kill

  • winsys/amdgpu: add timeline point support to fence lists

  • winsys/amdgpu: use timeline syncobj chunks in kernelq submission

  • radeonsi: add timeline semaphore support to fence operations

  • radeonsi: advertise GL_NV_timeline_semaphore

  • docs: add GL_NV_timeline_semaphore support for radeonsi

  • ac,radeonsi,radv: fix print IB assertion fail for reserved fields

Radu Costas (7):

  • pvr, pco: Commonize texture packing code

  • pco: Add hwinfo check for features in sampler code

  • pco: Commonize atomic sync operations

  • pvr,ci: Update expected fails list with new tests

  • pvr, ci: Update expected failures list

  • pco: Amend errant nir_move_option

  • pvr, ci: Remove tests from expected failure list

Raviraj Uppal (2):

  • ac/nir: Fixed OpenGL CTS transform feedback overflow detection test case The ordered atomic commits the post-add offset to memory, but overflow was computed using the pre-add offset, causing partial overflows to be missed and counters to become corrupted.

  • driconf: disable allow_rgb16_configs for SPECviewperf

Reilly Brogan (1):

  • amd,compiler: fix const errors found with C23 glibc support

Renato Pereyra (5):

  • pps: On data source register, report all counters as enabled by default

  • pps: Remove timestamps from counter descriptions

  • pps: Skip emitting repeated zero counter values

  • intel: Add pid and tid to Vulkan QueueSubmit events

  • intel: Include available counter descriptions in the perfetto counter spec

Rhys Perry (113):

  • aco/insert_fp_mode: remove incorrect assertion

  • radv: fix RADV_DEBUG=shaderstats with RT pipelines

  • aco: add lv1/lv2 as alias for v1/v2.as_linear()

  • aco: use lv1/lv2 instead of v1/v2.as_linear()

  • aco: use lv1.resize() pattern

  • radv: fix when incomplete rt pipeline libraries are loaded from cache

  • radv: improve skipping of creation of NIR for cached rt pipeline libraries

  • aco: use ABI::numClobbered() more

  • aco: use Program::stack_ptr instead of Program::static_scratch_rsrc

  • aco: add return address to call_clobbered_regs

  • aco: move return address to a clobbered register

  • aco/insert_waitcnt: improve s_setpc_b64/s_swappc_b64/end_with_regs a bit

  • radv: include ahit/isec shaders in radv_get_shader_from_executable_index

  • nir/search: remove creation of swizzle

  • nir/search: use memcmp/memcpy/memset

  • aco: consider 64-bit transcendental normal valu for s_delay_alu

  • radv: add ngg_wave_id_en to radv_shader_info

  • radv,aco/gfx11: preserve s2 when NGG_WAVE_ID_EN=1

  • aco: only consider cost of memory loads at waitcnt

  • aco/sched_ilp: improve scheduling with VMEM/DS->VALU WaW

  • aco/ra: track cost of moving variables

  • aco/ra: move variables from affinity register to avoid waitcnt

  • aco/ra: prefer phi operands which don’t create waitcnt

  • aco/ra: create vectors for affinities of split definitions

  • nir/opt_intrinsics: optimize inot(inverse_ballot(const))

  • ac/nir,radv,radeonsi: flip branches to avoid waitcnts

  • nir/load_store_vectorize: more carefully add entries from loop preheader

  • nir/load_store_vectorize: don’t update last_entry after a barrier

  • nir: fix fmin_agx/fmax_agx constant folding

  • nir: only set fp_math_ctrl if meaningful

  • nir/algebraic: remove ignore_exact

  • aco: fix gfx6-8 store_scratch() with function calls

  • aco/ra: don’t modify parallelcopies if get_reg_for_affinity fails

  • aco: fix VALUReadSGPRHazard with s_call_b64/s_swappc_b64

  • aco: reset all vgpr_used_by_vmem_ in resolve_all_gfx11

  • aco: resolve hazards before calls

  • radv: disable fast math for frag_coord.z adjust

  • radv: combine v_mov_dpp into fma in frag_coord.z adjust

  • radv: fix size of reserved in radv_physical_device_cache_key

  • radv: remove radv_physical_device::max_shared_size

  • ac/nir: remove gfx_level parameter from ac_nir_lower_indirect_derefs

  • ac/nir: remove ac_nir_lower_ps_late_options::family

  • ac/gpu_info: fix outdated comment

  • ac/gpu_info: remove padding from ac_cu_info

  • amd: move various flags to ac_cu_info

  • amd: add ac_cu_info::has_vrs_frag_pos_z_bug

  • amd: add ac_cu_info::local_invocation_ids_packed

  • ac/llvm: pass ac_cu_info to ac_llvm_context_init

  • radv: don’t cache esgs_ring_size/gsvs_ring_size

  • aco: perform dce for blocks skipped for process_block()

  • ac/gpu_info: move some NGG flags to ac_cu_info

  • ac/nir: use ac_nir_lower_ngg_options for ac_nir_lower_ngg_mesh

  • ac/nir: don’t pass radeon_info to NGG lowering

  • ac/nir: don’t pass radeon_info to ac_nir_set_options

  • ac/nir: pass ac_cu_info to ac_nir_compute_tess_wg_info

  • ac/nir/ngg: add ac_cu_info shortcut

  • amd: rename ac_cu_info to ac_compiler_info

  • ac/gpu_info: print most of ac_compiler_info

  • aco: fix PS epilog dual-source blending with only one color output

  • ac/nir: fix when both dual source outputs are unwritten

  • radeonsi: replace dual_src_blend_swizzle flag with dual_src_blend

  • radeonsi: fix PS epilog dual-source blending with only one color output

  • radeonsi: fix dual-source blending with only one output

  • radv: don’t mask PS epilog spi_shader_col_format with dual source blending

  • nir/range_analysis: set deleted key

  • nir: add nir_fp_analysis_state

  • nir/range_analysis: use SSA index for hash table keys

  • nir/range_analysis: use function pointers for lookup

  • nir/range_analysis: use sparse array for float analysis

  • zink: use hash_table_u64 instead of _mesa_hash_table_create_u32_keys

  • amd/common/virtio: use hash_table_u64

  • util: make UINT32_MAX a reserved key for _mesa_hash_table_create_u32_keys

  • nir/range_analysis: use uint16_t for sparse array elements

  • nir/range_analysis: use a dense array

  • nir/range_analysis: cache results of non-alu fp class queries

  • amd/drm-shim: enable conformant_trunc_coord for navi31

  • radv: improve lower_array_layer_round_even condition

  • aco/tests: fix assembler tests with LLVM 22

  • aco/tests: fix assembler/isel tests with LLVM 23

  • radv: don’t copy radv_vertex_input_state in CmdSetVertexInputEXT

  • radv: fix memory leak in radv_rt_nir_to_asm

  • radv: add radv_shader_debug_info

  • radv: simplify radv_shader_dump_debug_info

  • radv: add radv_parse_binary_debug_info

  • radv: add radv_shader_debug_info parameter to radv_shader_create

  • radv: move radv_shader_create out of radv_compute_pipeline_compile

  • radv: copy spirv in radv_graphics_shaders_nir_to_asm instead

  • radv: move radv_shader_create out of radv_graphics_shaders_nir_to_asm

  • radv: move radv_shader_create out of radv_graphics_shaders_compile

  • radv: move radv_shader_create out of radv_rt_nir_to_asm

  • radv: create radv_rt_spirv_to_nir

  • util: allow any key for hash tables

  • util: simplify hash_table_u64

  • util: fix UBSan error with _mesa_bfloat16_bits_to_float

  • nir/tests: fix NaN/inf checks in skip_test()

  • nir/algebraic: optimize exact f2u32(fmul(unpack_norm))

  • nir/propagate_invariant: include intrinsics

  • nir/propagate_invariant: set fp_math_ctrl for intrinsics

  • nir/propagate_invariant: include derefs

  • nir/propagate_invariant: be more conservative with NULL variables

  • nir/propagate_invariant: be more conservative with aliasing variables

  • nir/propagate_invariant: handle images

  • nir: add and use block predecessor helpers

  • nir: add nir_loop_has_back_edge helper

  • nir/cf: don’t remove block predecessors while iterating

  • nir: use a u_dynarray for block predecessors

  • aco: ignore copykill+latekill operands in get_temp_reg_changes

  • docs/aco: add live variable analysis documentation

  • ir3/array_to_ssa: skip remove_trivial_phi for non-array phis

  • ir3/array_to_ssa: initialize ir3_instruction::data

  • ir3/ra: fix copy-paste error

  • aco: support s_bitset

  • aco/ra: create s_bitset

Rob Clark (115):

  • ir3: Use fd_dev_info from ir3_compiler

  • ir3: Handle dual-wave reconvergence

  • freedreno/common: Fix gen8 EFU float control

  • freedreno: Force single wavesize if double threadsize is unsupported

  • tu: Drop HIC support for depth images

  • freedreno/lrz: Correct lrz fc layout for gen8

  • tu: Re-enable LRZ for gen8

  • freedreno/a6xx: Better program state size calc

  • freedreno/decode: Skip bindless dumps on pre-bindless hw

  • freedreno/decode: Enable –bindless for cffdump tests

  • freedreno/decode: Add multi-plane descriptor coverage

  • freedreno/registers: Drop a6xx descriptor chip use

  • freedreno/registers: Rename A6XX_TEX_MEMOBJ

  • freedreno/decode: Fix gen8 descriptor address

  • freedreno/decode: Extract out helper to set varset

  • freedreno/decode: Decode all descriptor variants

  • freedreno/registers: Descriptor variants

  • freedreno/decode: Add script support for enum types

  • freedreno/decode: Replace/remove __tonumber()

  • freedreno/decode: Allow direct access to domain bitfield

  • freedreno/decode: Add lua handler to filter descriptors

  • freedreno/decode: Allow dom[1] to be NULL

  • ir3: Rename cat6 UBO/UAV descriptor src

  • ir3: Disasm shader descriptor stats

  • freedreno/decode: Split out domain based decoding

  • freedreno/decode: fix domain decode for “structs”

  • freedreno/decode: add shader stats object

  • freedreno/decode: pass more info to descriptor handler

  • freedreno/decode: call show_descriptor() for UBO and SAMPLERs as well

  • freedreno/decode: filter unused descriptors in lua

  • freedreno/decode: Dump filtered bindless descriptors by default

  • freedreno/decode: Fix query bin vals

  • freedreno/decode: Expose gpu buffers to lua

  • freedreno/decode: Handle strips

  • freedreno/decode: Allow raw access to pm4 packets

  • freedreno/decode: Emulate CP_MEM_WRITE

  • freedreno/decode: Shorten query string

  • freedreno/decode: Split out endswith() helper

  • freedreno/decode: Filter redundent _HI regs

  • freedreno/decode: Keep intereactive for query mode

  • nir: Fill in missing conversion opts

  • freedreno/decode: Fix endswith()

  • freedreno/registers: Update CP_COND_WRITE

  • freedreno/registers: Update GRAS_BIN_FOVEAT

  • tu: Split out stomp_regs() helper

  • tu: Mark TU_CMD_DIRTY_COMPUTE_DESC_SETS after stomping

  • freedreno/registers: Move binning regs to “cmd”

  • freedreno/registers: Rename some unknown A2D regs

  • freedreno/registers: Split out “blit” usage

  • freedreno/registers: Split out “resolve” usage

  • freedreno/registers: Split out compute usage

  • freedreno/registers: Move remaining rp_blit to draw

  • freedreno/registers: Usage additions/corrections

  • freedreno/rnn: Track reg usage

  • freedreno/decode: Remove prefetch-test

  • freedreno/decode: Use reg usage for reg summary

  • freedreno: Move some draw regs into driver

  • gallium: Add PIPE_QUERY_TIMESTAMP_RAW

  • freedreno/a6xx: Implement PIPE_QUERY_TIMESTAMP_RAW

  • nir: Fix validation error after nir_round_int_to_float()

  • ir3: More COMPUTE vs KERNEL

  • freedreno+ir3: Implement CL isam mode

  • gallium: Switch TIMESTAMP_RAW back to callback

  • gallium: Add warning about PIPE_QUERY_x’s ABIness

  • rusticl: Add CL specific bind flag

  • freedreno/a6xx: Hide 10_10_10_2 for opencl

  • ir3: Lower 8b usub_sat

  • freedreno/fdl: Set layer_size in explicit_layout case

  • freedreno: Add missing cl_gl_sharing cap

  • freedreno: Fix stdout vs stderr logging

  • freedreno/a6xx: Barrier debug

  • freedreno: Flip logging to debug

  • freedreno: Block rusticl on older gens

  • freedreno/a6xx: Fix num_groups programming

  • freedreno: Attach fence to last batch

  • freedreno: Reuse last_fence when possible

  • freedreno/a6xx: Don’t emit epilogue per-tile

  • freedreno/a6xx: Rework flushing events

  • freedreno/decode: Add missing a6xx/a7xx reg decoding

  • freedreno: Use linear for 1d/1d_array

  • freedreno: Reduce advertised memory

  • freedreno/drm: bo cache logging vs tsan

  • freedreno: Initialize debug once

  • ir3: Initialize debug once

  • freedreno: Avoid shadow blits for compute contexts

  • freedreno/a6xx: Avoid touching long lived stateobj refcnt

  • rusticl: Let backend control convert_alu_types lowering

  • ir3: Handle (some) convert_alu_types in backend

  • freedreno: Rename a830

  • freedreno: Split up freedreno_devices.py

  • freedreno: Add –nvtop arg

  • freedreno/fdl: Use 4k alignment for tiled

  • ir3: Move shader upload under variants_lock

  • freedreno/drm: Fix bo_flush race

  • freedreno: Check for flushed batches

  • freedreno: Update pscreen->num_contexts

  • freedreno: Don’t re-bind global buffers

  • freedreno: Move pvtmem to screen

  • freedreno/a6xx: Fix sharable cs races

  • freedreno/drm: Shareable stateobjs

  • ir3: Lower ffma

  • ir3: Late lowering of fmul+fadd to ffma

  • freedreno/ci: Update trace expectations

  • ir3: Set max_workgroup options

  • freedreno/registers: Add a couple missing bitfields

  • freedreno/registers: Remove left-over comment

  • meson: Fix build break on f43, gentoo, etc

  • freedreno/a6xx: Move A2D reg write to ncrb

  • freedreno/common: Fix upstream a830 chip_id

  • freedreno/registers: Update gmu reg offsets

  • freedreno/a6xx: Fix supported-blit fmt check

  • freedreno/common: Drop gen8 0x78000 offset

  • freedreno: Add a829

  • freedreno/a6xx: Fix blit fmt check

  • tu/kgsl: Add UBWC_5 and UBWC_6 support

Rob Herring (Arm) (17):

  • ethosu: Fix padding calculation

  • teflon/tests: Add 16-bit output support

  • teflon: Add debug string for concatenation

  • test_teflon: Fix crash with read-only buffers

  • test_teflon: Fix missing UInt16/Int16 output size

  • test_teflon: Add 32-bit integer output comparison

  • teflon: Add support for setting the tensor type size

  • ethosu: Add support for 16-bit tensors

  • ethosu: Add scalar ADD support

  • ethosu: Handle reversing IFM and IFM2 operands

  • ethosu: Handle IFM2 H/W/D broadcast

  • teflon: Support ReLU activation for ADD ops

  • ethosu: Support ReLU activation for ADD ops

  • ethosu: Fix buffer overrun in stridedslice

  • ethosu: Fix U85 AvgPool for greater than 8x8 kernel sizes

  • ethosu: Drop 2nd allocation of IFM and OFM

  • ethosu: Move ethosu_allocate_feature_map() to ethosu_lower.c

Robert Mader (5):

  • lavapipe: enable dmabuf import for planar drm formats

  • lavapipe: Remove some dead code

  • llvmpipe: Stop aligning height to raster block size for unbacked handles

  • nir/lower_tex: Reinstate LSB to MSB shift

  • llvmpipe: Implement manual context resets

Rohan Garg (2):

  • anv: set a private binding when the image is not externally shared

  • anv: refactor add_aux_state_tracking_buffer for conciseness

Rohit Athavale (1):

  • mediafoundation: Test compile steps v/s step , and set build flag

Roland Scheidegger (3):

  • llvmpipe: get rid of unused code in float to small float code

  • llvmpipe: don’t rely on cpu denorms for float to smallfloat conversion

  • llvmpipe: disable denorms in compute shaders on x86/sse

Romaric Jodin (1):

  • pan/bi: lower phis to scalar early

Rouf, Farhan (1):

  • amd/vpelib: Embedded Buffer Size for 3DLUT FL

Rudi Heitbaum (1):

  • mesa: retain const qualifier from pointer

Ruijing Dong (1):

  • ac/vcn: correct a typo in av1 dec header

Ryan Houdek (1):

  • freedreno/fdl: Fix compiling with GCC and AVX2

Ryan Mckeever (2):

  • panvk: lower multisampled images before nir_lower_descriptors

  • panvk: enable fragmentStoresAndAtomics for Bifrost

Ryan Zhang (5):

  • panvk: guard against NULL pointers to avoid crash

  • panvk/csf: use DEFERRED_FLUSH for fragment job cache flush

  • panvk: trivial fix to remove repeated assignment

  • panvk/csf: rework IR descriptor handling for tiler OOM

  • panvk: add VK_IMAGE_LAYOUT_DEPTH_READ_ONLY_OPTIMAL to host copy layouts

Sagar Ghuge (14):

  • anv: Mark RootNodeOffset at 256B always

  • nir: Add nir_resource_intel_internal entry

  • anv: Set max outstanding ray queries to 1024

  • intel/blorp: drop unused BLORP_BATCH_COMPUTE_ENGINE flag

  • anv: Improve bvh_no_build option

  • anv: Fix Wa_14021821874, Wa_14018813551, Wa_14026600921

  • anv: Write IR header using shader instead of CS

  • anv/rt: Drop header update using blorp code path

  • intel/genxml: Add new State Cache Perf Fix Disabled field

  • anv: enable BTP+BTI RCC keying for some workloads

  • anv/bvh: Drop atomic on instance_count

  • intel/compiler: Handle TerminateOnFirstHit in ray query execution

  • intel/compiler: Remove unused brw_nir_memclear_global helper

  • anv: Fix Wa_14021821874, Wa_14018813551, Wa_14026600921

Samuel Pitoiset (358):

  • vulkan: fix missing begin debug marker for HPLOC

  • spirv: Update the JSON and headers

  • vulkan: update spec to 1.4.340

  • radv: move emitting SQTT markers closer to the draw/dispatch packets

  • radv: use the SQTT enable bit for PKT3_DRAW_{INDEX}_INDIRECT_MULTI

  • radv: use the SQTT enable bit for PKT3_DISPATCH_MESH_INDIRECT_MULTI

  • radv: use the SQTT enable bit for PKT3_DISPATCH_TASKMESH_INDIRECT_MULTI_ACE

  • radv: fix applying radv_ssbo_non_uniform=true for Crysis 2/3 remastered

  • radv: add a workaround for a synchronization bug in Strange Brigade Vulkan

  • radv/meta: fix 3D color resolves with compute when base slice isn’t zero

  • radv/meta: return the flush bits from radv_clear_hiz()

  • radv: optimize barriers when clearing HiZ on GFX12

  • radv/sqtt: rework acquiring GPU timestamps

  • radv/sqtt: rework acquiring timed cmdbufs

  • radv/sqtt: reduce the number of timed cmdbufs

  • radv: rework app workarounds implemented using internal layers

  • vulkan: add support for VK_KHR_internally_synchronized_queues

  • radv: advertise VK_KHR_internally_synchronized_queues

  • radv: zero-initialize image view objects

  • radv: fix tracking of pipelines used in secondaries

  • radv/meta: remove declared but unused radv_decompress_resolve_rendering_src()

  • radv/amdgpu: remove radv_dummy_winsys_create()

  • radv/meta: remove unused emit_depth_stencil_resolve()

  • radv/amdgpu: bypass GL2 for command buffer BOs

  • radv: disable unordered submits when SQTT queue events are enabled

  • ac,radv,radeonsi: shorten some emit macro names

  • radv: emit pending flushes after late decompressions with fbfetch

  • radv: stop delaying decompression passes for feedback loops with DRLR

  • radv: emit the VRS surface as part of the framebuffer state on GFX11+

  • radv: track redundant PA_SC_VRS_OVERRIDE_CNTL register writes

  • radv: remove occurrences of VK_IMAGE_LAYOUT_SHARED_PRESENT_KHR

  • radv/meta: decompress resolve src outside of depth/stencil resolves

  • radv/meta: make radv_decompress_resolve_src() static

  • radv/meta: stop saving/restoring rendering state for FS/HW resolves

  • radv/meta: fix the key for DCC decompress on compute

  • radv/meta: remove useless check in radv_CmdClearAttachments()

  • radv/meta: remove dead code for VK_FORMAT_R4G4_UNORM_PACK8

  • radv/meta: remove dead DCC clear code about E5B9B9R9_UFLOAT_PACK32

  • radv: fix late decompressions for fbfetch with more corner cases

  • radv: only pass custom sample locations when relevant

  • radv: emit late decompressions for fbfetch slightly earlier

  • radv/meta: stop saving/restoring rendering state for color/depth decompressions

  • radv/meta: remove unused saving/restoring rendering state logic

  • zink/ci: update checksum of one trace running on VANGOGH

  • radv/meta: remove dead code in the gfx depth/stencil clear path

  • radv: move color/depth-stencil init surface helpers to radv_image_view.c/h

  • radv: remove declared but unused radv_get_dcc_max_uncompressed_block_size()

  • radv: move {depth,stencil}_compress_disable to the image view extra info

  • radv: add a new dirty bit for the GFX12 HiZ workaround

  • radv: emit the framebuffer state when rendering begins

  • radv/meta: fix CmdCopyBufferToImage2() on compute queue with compressed HTILE

  • radv/meta: add a function to fixup HTILE metadata for copies on compute queue

  • radv/meta: stop fixing up HTILE after a partial copy

  • radv: set COMPRESSION_EN=1 for depth or stencil storage images when supported

  • radv/meta: do not disable compression for depth/stencil expand on compute

  • radv/meta: skip some HTILE operations when it’s decompressed on image stores

  • radv/meta: remove an useless barrier when fixing up HTILE for copies on compute

  • radv/meta: stop using custom sample locations for color resolves

  • radv: pass VkSampleLocationsInfoEXT for depth/stencil expand

  • radv: clear rendering state before performing resolves

  • radv: make sure rendering isn’t already active in CmdBeginRendering()

  • radv: do not resolve when rendering is suspended

  • radv: do not set the resume rendering flag for custom resolves

  • radv/meta: stop trying to reduce the number of format variants

  • radv/meta: use R32G32 formats for R64 slow color clears

  • radv: enable trimming FS color exports for internal shaders

  • radv/meta: stop fixing up DCC after a partial resolve using compute

  • radv/meta: remove an useless barrier when fixing up DCC for compute resolves

  • radv/meta: add a function to fixup DCC metadata for compute resolves

  • radv: rename radv_image_use_dcc_image_stores()

  • radv/meta: fix partial depth/stencil resolves with compute

  • radv: cleanup barriers after a depth/stencil expand

  • radv/meta: stop fixing up HTILE after a partial resolve using compute

  • radv/meta: add HTILE support to radv_fixup_resolve_dst_metadata()

  • radv/meta: optimize a barrier with depth/stencil compute resolves

  • radv/meta: move the barrier for depth/stencil compute resolves outside

  • ac,radv,radeonsi: use correct swizzle/pitch for depth-only images with SDMA

  • radv/meta: remove useless DCC decompressions for image<->buffer

  • radv/ci: mark more WSI flakes for NAVI21

  • radv/ci: mark more WSI tests as flakes on NAVI21

  • radv: remove a redundant check in radv_image_is_renderable()

  • radv/meta: rename some variables for btoi 96-bit shader

  • radv/meta: rename r32g32b32 to 96bit

  • radv/meta: rework get_image_stride_for_96bit() and make it non-static

  • vulkan/runtime: add a separate function to build ETC2 decode core shader

  • vulkan/runtime: add support for ETC2 emulation with copy_memory_indirect

  • radv: simplify buffer-to-image and image-to-image operations for 96-bit formats

  • radv: fix potential corruption after FMASK decompression on GFX6-8

  • radv: skip some operations when the image is already zero-initialized

  • radv/meta: fix depth/stencil resolves with different regions

  • radv: reserve CS space for the HiZ WA on GFX12

  • radv: skip some redundant operations when rendering is resumed

  • radv: handle the cache flush workaround for mips before emitting the fb

  • radv: suspend/resume dynamic rendering completely

  • radv: fix independent sets with dynamic buffers and GPL

  • ac/nir: fix writemask for dual source blending on GFX11+

  • radv: fix potential GPU hangs with secondaries on transfer queue

  • radv/nir: use radv_physical_cache_key::no_rt more

  • radv/nir: use radv_physical_cache_key::emulate_rt more

  • radv/nir: use radv_physical_cache_key::bvh8 more

  • radv: use radv_physical_cache_key::disable_trunc_coord more

  • radv: use radv_physical_cache_key::disable_aniso_single_level more

  • radv: use radv_physical_cache_key::disable_shrink_image_store more

  • radv: use radv_physical_cache_key::clear_lds more

  • ac/nir: stop passing radeon_info for addr->coord helpers

  • radv/meta: stop using pdev for shaders that use addr<->coord helpers

  • radv/meta: stop using pdev for building the resolve meta shaders

  • radv/meta: stop using pdev for some query resolve shaders

  • radv: remove an useless check for VK_IMAGE_LAYOUT_PREINITIALIZED

  • radv/meta: replace radv_meta_resolve_type by VkImageAspectFlags

  • radv/meta: add depth/stencil support to the core resolve shader

  • radv/meta: use the same shader for color/depth&stencil compute resolves

  • radv/meta: inline one function in the compute resolve path

  • radv/meta: pass a resolve mode for compute color resolves

  • radv/meta: simplify creating pipelines for color/depth&stencil compute resolves

  • radv/meta: add a single function for color/depth&stencil compute resolves

  • radv/meta: fix the shader stage for push constants in the fragment resolve path

  • radv/meta: move the barrier for color fragment resolves outside

  • radv/meta: use the same shader for color/depth&stencil fragment resolves

  • radv/meta: inline one function in the fragment resolve path

  • radv/meta: pass a resolve mode for fragment color resolves

  • radv/meta: simplify creating pipelines for color/depth&stencil fragment resolves

  • radv/meta: add a single function for color/depth&stencil fragment resolves

  • ac/cmdbuf: only set DCC_WRITE_COMPRESS for DCC on SDMA 5+

  • radv: only enable DCC/HTILE if it’s compressed with SDMA

  • ac/sdma: fix pitch assertion for SDMA7

  • radv: fix copying images with different swizzle modes on SDMA7

  • radv: remove redundant radv_sdma_surf::micro_tile_mode

  • radv: tidy up determining 3D alignment for SDMA

  • ac/parse_ib: fix parsing some packets on SDMA7

  • ac,radv,radeonsi: merge tiled/linear surfaces into one struct

  • radv: mark linear images for SDMA as potentially compressed on GFX12

  • ac/sdma: rework and fix metadata for SDMA7

  • radv: fix computing pitch/slice_pitch for compressed block formats

  • radv/meta: remove redundant barriers in vkCmdResolveImage2()

  • radv/meta: remove an useless barrier after color resolves with graphics

  • radv/meta: remove an useless barrier before color resolves with graphics

  • radv/meta: optimize synchronization with compute resolves

  • radv/meta: rework the barrier for depth/stencil resolves

  • radv/meta: set the depth/stencil resolve region depth to 1

  • radv: disable STORAGE for depth-only formats

  • radv: remove radv_disable_depth_storage drirc

  • radv: always fast-clear non-MSAA color image with comp-to-single on GFX10-10.3

  • ac/surface: remove remaining occurrences of HiS on GFX12

  • radv: add radv_image_has_hiz()

  • radv: add radv_image_has_hiz_metadata()

  • radv: remove unused radv_gfx12_get_hiz_clear_value()

  • radv: initialize HiZ also for depth-only images

  • radv: always enable DISABLE_CONSERVATIVE_ZPASS_COUNTS on GFX11

  • radv: fix a GPU hang with PS epilogs and secondary command buffers

  • radv: remove redundant radv_sdma_surf::is_3d

  • radv: remove redundant radv_sdma_surf::is_linear

  • radv: simplify 96-bit copies with SDMA

  • radv: use vk_image_buffer_copy_layout() for SDMA buf layout

  • radv: simplify computing offset/extent of SDMA surfaces

  • radv: remove unnecessary radv_sdma_surf::{blk_w,blk_h}

  • radv: simplify getting bpe for SDMA surfaces

  • radv: tidy up radv_sdma_surf

  • radv: replace radv_sdma_surf by ac_sdma_surf

  • radv/meta: fix HTILE fixup after copying depth/stencil image copies

  • radv: fix local invocation index for mesh/task and quad derivatives on GFX12

  • radv: apply the 1D workgroup optimization for mesh/task shaders too

  • radv: dump the PS epilog in the GPU hang report

  • amd/drm-shim: add rembrandt

  • amd/drm-shim: add phoenix

  • amd/drm-shim: bump version_minor to 52

  • ac,radeonsi: pre-compute some raster config in ac_gpu_info

  • ac,radeonsi: move guardband computations to common code

  • radv: use common guardband computations

  • radv: optimize clipping performance with PA_SU_HARDWARE_SCREEN_OFFSET

  • zink/ci: update the lists for CEZANNE and VANGOGH

  • zink/ci: update traces expectations for VANGOGH/GFX1201

  • radv: initialize HiZ for UNDEFINED transitions on transfer queue

  • ac,radv,radeonsi: add has_db_force_stencil_valid_bug

  • radv: set {color,ds}_samples for inherited rendering state

  • radv: only emit FORCE_S_VALID(1) for MSAA depth/stencil images

  • radv: fix missing L2 cache invalidation with streamout on GFX12

  • radv: rewrite a comment explaining why PFP waits for ME with streamout

  • ci: uprev vkd3d

  • radv/meta: make some functions non-static

  • radv: implement VK_KHR_copy_memory_indirect

  • radv: advertise VK_KHR_copy_memory_indirect on GFX8+

  • radv: handle FRAGMENT_SHADING_RATE_ATTACHMENT_READ properly on GFX10.3

  • radv/meta: remove redundant cache flushes when copying VRS rates to HTILE

  • radv: tidy up radv_postprocess_nir()

  • radv: remove empty gather_shader_info_cs()

  • radv: remove unused radv_device parameter to few functions

  • radv: use radv_physical_device_cache_key::use_ngg_culling more

  • radv: remove radv_nir_compiler_options::info

  • radv: remove unnecessary radv_device parameter to few functions

  • radv: remove radv_use_llvm_for_stage()

  • ac/rtld: remove radeon_info

  • radv/meta: use the fragment resolve path by default

  • radv/meta: remove CB_RESOLVE

  • vulkan: adjust MESA_VK_PIPELINE_RAY_TRACING_FLAGS with beta extensions disabled

  • vulkan: update spec to 1.4.346

  • vulkan: use vk_object_zalloc() for acceleration structs

  • vulkan: add support for vkCreateAccelerationStructure2KHR()

  • vulkan: add helpers for device address range

  • vulkan: add vk_image_memory_copy_layout()

  • vulkan: add helpers for depth/stencil only layouts

  • radv: only set the relevant image views for custom depth/stencil resolves

  • radv: stop checking whether HTILE is compressed with the UNDEFINED layout

  • radv/meta: add separate ds layouts support to the HTILE expand pass

  • radv/meta: stop setting the other depth/stencil attachments when unused

  • radv: handle separate depth/stencil layouts correctly for fbfetch decompressions

  • radv: handle separate depth/stencil layouts correctly for barriers

  • radv: fix missing HTILE decompression with separate depth/stencil layouts

  • radv: always use separate depth/stencil layouts for rendering

  • radv: cleanup valid image layouts in radv_layout_is_htile_compressed()

  • vulkan: do not pass vk_instance for debug report messages

  • aco,radv,radeonsi: remove debug report support in ACO

  • radv: stop passing radv_device for SPIR-V debug reports

  • radv/meta: fix missing sync for compute resolves

  • radv/amdgpu: free the VA range in case the BO allocation failed

  • radv/amdgpu: remove dead code in radv_amdgpu_winsys_bo_create()

  • vulkan: stop passing vk_device to vk_set_subgroup_size()

  • radv: remove unnecessary radv_device parameter to few functions

  • radv: move radv_printf_data to radv_debug_nir

  • radv: move valid VA debug info to radv_valid_va data

  • radv: stop associating NIR with device for debugging tools

  • radv: move setting NIR options for meta shaders

  • radv: stop passing radv_device for creating NIR meta shaders

  • radv/meta: cleanup determining the resolve method

  • radv/meta: decompress source resolve images slighly earlier

  • radv: use nir_intrinsic_printf in radv_build_printf()

  • radv: use nir_shader::uses_printf for lowering printf

  • ac/gpu_info: remove a TODO about LOAD_CONTEXT_REG on GFX6-7

  • ac/cmdbuf: add an assertion for COPY_DATA+PFP with registers

  • radv: use LOAD_CONTEXT_REG_INDEX when supported for streamout

  • radv: remove useless PFP_SYNC_ME when loading color/ds metadata on GFX6-7

  • radv: update color/ds clear metadata in ME

  • radv: emit PFP_SYNC_ME right after STRMOUT_BUFFER_UPDATE is emitted

  • radv: stop allowing users to disable the global BO list

  • radv: remove radv_device::use_global_bo_list

  • radv: cleanup functions that writes descriptors

  • radv: stop allocating an array of BO for descriptors

  • radv/amdgpu: remove the virtual BOs tracking logic

  • radv: remove adding a BO to the per-cmdbuf list when unnecessary

  • radv: ignore the GFX12 HiZ WA for internal blits

  • radv: only consider HiZ as valid after clears with the depth aspect

  • radv: fix a perf issue when clearing depth/stencil images on GFX12

  • radv: cleanup redundant radv_image_has_hiz_metadata() checks

  • vulkan: fix memory leak in vkQueueBeginDebugUtilsLabelEXT()

  • radv/ci: stop skipping ASTC tests with ANGLE+STONEY

  • radv/ci: stop skipping some memory related tests on POLARIS10

  • radv/ci: remove skipping mesh shader tests for NAVI31

  • Revert “radv: remove adding a BO to the per-cmdbuf list when unnecessary”

  • radv/ci: move slow tests to radv-slow-skips.txt

  • radv/ci: add a new dEQP test suite for nightly jobs

  • radv/ci: add new jobs that run full VKCTS on NAVI21/NAVI31/GFX1201

  • radv/amdgpu: always return VK_ERROR_INVALID_EXTERNAL_HANDLE for host ptr imports

  • radv/ci: fix radv-slow-skips.txt path

  • radv/ci: fix a typo in radv-navi10-vkcts-full

  • radv: replace radv_copy_flags by VkAddressCopyFlagsKHR

  • radv: implement VK_KHR_device_address_commands

  • radv: advertise VK_KHR_device_address_commands

  • nir: make nir_variable::descriptor_set a 32-bit variable

  • vulkan/runtime: handle custom border color index with samplers

  • docs: add missing description of RADV_PERFTEST=rtcps

  • radv: introduce RADV_EXPERIMENTAL envvar for experimental features

  • radv: do not try to resize the SPM BO for per-submit captures

  • radv: improve dumping RGP captures

  • radv: add RADV_DEBUG=fullsync

  • treewide: cleanup non-existent descriptor types from nir_intrinsic_desc_type()

  • nir: introduce nir_descriptor_type for Vulkan like descriptors

  • amd/drm-shim: bump version_minor to 54

  • amd: bump required DRM version to 3.54 (Linux kernel 6.6+)

  • nir,treewide: add nir_image_intrinsic_type

  • vulkan: add DGC support with descriptor heap

  • nir: add texture_heap_offset/sampler_heap_offset to nir_build_tex()

  • nir/lower_mediump: add heap support

  • nir/opt_shrink_vectors: add heap support

  • nir/opt_sink: add heap support

  • nir/opt_move_discards_to_top: add heap support

  • nir/lower_image: add heap support

  • nir/divergence_analysis: add missing nir_intrinsic_image_heap_texel_address

  • nir/opt_intrinsics: add heap support

  • nir/opt_uniform_atomics: add heap support

  • nir/opt_access: add heap support

  • nir/opt_group_loads: add heap support

  • nir/opt_preamble: add heap support

  • nir/validate: add heap support

  • nir/gather_info: add heap support

  • nir/lower_helper_writes: add heap support

  • nir/opt_shrink_stores: add heap support

  • radv: fix a typo when determining if a VS needs a prolog

  • radv: emit BOP events after every draw to workaround a VRS bug on GFX12

  • vulkan: rename VK_EXT_device_fault features

  • vulkan,spirv: update headers

  • spirv: fix OpUntypedVariableKHR with optional data type parameter

  • spirv: handle untyped pointer storage class with descriptor heap

  • vulkan: remove unused parameters in vk_build_descriptor_heap_address()

  • vulkan: fix determining the heap ptr

  • vulkan: update spec to 1.4.348

  • ac/nir: adjust lowering of query size for descriptor heap

  • ac/nir: add descriptor heap support to ac_nir_lower_image_tex()

  • ac/nir: add descriptor heap support to opt_flip_if_for_mem_loads()

  • nir: add new variable modes for the resource/sampler heap pointers

  • spirv: change the resource/sampler builtins variable mode

  • spirv: set the image format for image intrinsics

  • spirv: emit nir_intrinsic_image_heap when resource/sampler ptrs are used

  • spirv: implement SpvOpUntypedImageTexelPointerEXT

  • vulkan: adjust lowering of descriptor heaps

  • nir: remove resource/sampler heap ptrs sysvals

  • spirv: mark all resources as non-uniform by default with descriptor heap

  • vulkan: stop emitting global_addr_to_descriptor

  • nir: remove nir_intrinsic_global_addr_to_descriptor

  • radv/meta: fix computing extent for image->image with both compressed formats

  • nir: allow heap image intrinsics in nir_rewrite_image_intrinsic()

  • radv: pre-compute the primitive restart index

  • radv: implement VK_EXT_primitive_restart_index

  • radv: advertise VK_EXT_primitive_restart_index

  • radv/meta: remove an outdated comment in vkCmdClearAttachments()

  • radv: replace remaining occurrences of VK_ACCESS_xxx

  • vulkan: mark RP attachments as invalid when no rendering create info

  • nir: add new system values for descriptor heap RT traversal inputs

  • radv: zero-allocate graphics shader stages

  • radv: add a new helper to make a sampler descriptor

  • radv: add support for custom border colors with descriptor heap

  • radv: make radv_make_sampler_descriptor() non-static

  • radv: use 32-bit memory types for descriptor heap buffers

  • radv: add shader info about whether descriptor heap is used

  • radv: declare shader arguments for resource/sampler heaps

  • radv/rt: declare shader arguments for resource/sampler heaps

  • radv: keep track of descriptor heap mapping in the shader layout

  • radv: call vk_nir_lower_descriptor_heaps()

  • radv/nir: adjust lowering of ycbcr tex instructions for descriptor heap

  • radv/nir: adjust lowering of immediate samplers for descriptor heap

  • radv/nir: rename radv_nir_apply_pipeline_layout

  • radv/nir: lower descriptor heap in radv_nir_lower_descriptors

  • radv: set descriptor heap sizes/alignments for VTN

  • radv: allow to create pipelines with a NULL pipeline layout

  • radv: copy mapping info for graphics pipeline libraries

  • radv: implement vkWrite{Resource,Sampler}DescriptorsEXT()

  • radv: implement vkCmdBind{Resource,Sampler}HeapEXT()

  • radv: add support for emitting descriptor heaps

  • radv: implement vkCmdPushDataEXT()

  • radv: implement vkGetPhysicalDeviceDescriptorSizeEXT()

  • radv: add support for capture&replay with descriptor heap

  • radv: add support for inherited descriptor heap for secondaries

  • radv: flush caches with descriptor heap access flags

  • radv: add support for DGC with descriptor heap

  • radv: advertise VK_EXT_descriptor_heap with RADV_EXPERIMENTAL=heap

  • radv/ci: set RADV_EXPERIMENTAL=heap

  • ci: uprev VKCTS main to 634a3fc62d82c34de68c3b1add25e6b7f5777524

  • radv/ci: remove a hack for the number of deqp instances with RENOIR

  • radv/ci: update flakes of VKCTS jobs

  • radv/ci: fix setting RADV_EXPERIMENTAL=heap

  • radv/ci: document a descriptor heap failure

  • vulkan: add an option to lower SHADER_RECORD_INDEX to non-uniform

  • radv: lower SHADER_RECORD_INDEX to non-uniform

  • radv: fix GPU hangs with PS epilogs and secondaries properly

  • radv: re-introduce DGC+multiview support and enable it for vkd3d-proton only

  • vulkan: add missing VkMemoryRangeBarriersInfoKHR support

  • radv: add missing VkMemoryRangeBarriersInfoKHR from DAC

  • radv/meta: fix expanding HTILE on compute with multisampling

  • radv: fix determining needed dynamic states when rasterization is disabled

Sergi Blanch Torne (7):

  • ci: disable Collabora’s farm due to maintenance

  • Revert “ci: disable Collabora’s farm due to maintenance”

  • ci: disable Collabora’s farm due to maintenance

  • Revert “ci: disable Collabora’s farm due to maintenance”

  • ci: fix envvar default value

  • ci: nightly run xfiles for gc2000 and a618 piglit jobs

  • ci: nightly run xfiles for a618 angle job

Shih, Jude (1):

  • amd/vpelib: Gate assertion on debug flag

Silvio Vilerino (20):

  • d3d12: Add missing using Microsoft::WRL:ComPtr in d3d12_context_common

  • d3d12: Add HAVE_GALLIUM_D3D12_VIDEO guards for d3d12_video_encoder_set_max_async_queue_depth/d3d12_video_encoder_get_last_slice_completion_fence

  • pipe: Add PIPE_VIDEO_CAP_SLICE_STRUCTURE_AUTO for PIPE_VIDEO_SLICE_MODE_AUTO

  • d3d12: Implement PIPE_VIDEO_CAP_SLICE_STRUCTURE_AUTO reporting

  • mediafoundation: Query PIPE_VIDEO_CAP_SLICE_STRUCTURE_AUTO

  • ci: Bump DirectX-Headers and Agility SDK dependencies to v1.619.1

  • d3d12: Implement trim notification residency eviction

  • d3d12: Truncate move_rects_support.bits.max_motion_hints 16 bit var to 65535, not 65536

  • d3d12: d3d12_video_encode_support_caps was assigning a stack variable address to capEncoderSupportData in/out arg

  • d3d12: Fix video fence leak and double assign

  • d3d12: Video sliced encode: Use same ID3D12Fence/different per slice values as optimization

  • pipe: Add fence_get_win32_event since fence_get_fd return int type is smaller than HANDLE/void*

  • d3d12: Implement pipe_screen::fence_get_win32_event

  • mediafoundation: Use WaitForMultipleObjects for AUTO slices wait in sliced encode mode

  • mediafoundation: Prefetch the slice fence handles before the waits

  • mediafoundation: Pre-create all MFSamples to avoid per slice COM allocation in the hot loop

  • mediafoundation: Remove unnecessary staging variable in ProcessSliceBitstreamZeroCopy

  • d3d12: Check queues are registered before unregistering in unregister_work_queue

  • mediafoundation: MFTRegisterWorkQueue/MFTUnregisterWorkQueue to validate null param instead of crash

  • Revert “d3d12: Video sliced encode: Use same ID3D12Fence/different per slice values as optimization”

Simon Perretta (18):

  • pco: update formatless skip check

  • pvr: remove drm device config table

  • pvr: add initial yuv tex/smp state words

  • pvr: rename PVR_HAS_ERN to PVR_HAS_ENHANCEMENT

  • pvr: add support for drm-shim

  • docs/pvr: add drm-shim documentation

  • pvr: drop pvr_assert macro

  • pvr: handle SRC_SMRG_D32S8_D32S8 in tq shader

  • pvr: set transfer flags based on derived formats

  • pvr: allow primary drm node to be optional

  • pvr: report nir shader in pipeline executable properties

  • pvr: downgrade vs out/fs in mismatch assertion to a warning

  • pco: add encodings and mappings for smp integer and array flags

  • pco: use vm/icm for tile buffer store coverage mask

  • pco: add native u{add,sub}{carry,borrow,sat} ops

  • pvr: advertise VK_KHR_shader_integer_dot_product

  • pco: reserve additional outputs for trilinear sampled coeffs

  • pco: amend tg4 lowering

Stéphane Cerveau (1):

  • anv/ci: add vulkan fluster job on adl

Sushma Venkatesh Reddy (1):

  • brw: Use lookup tables for Gfx12+ 3src type encoding/decoding

Tanner Van De Walle (1):

  • d3d12: Fix d3d12_surface_destroy() to match pipe_surface_destroy_func()

Tapani Pälli (34):

  • intel/dev: update mesa_defs.json from workaround database

  • anv: add handling for Wa_14026600921

  • intel/genxml: bring some missing fields to gen125.xml

  • drirc/anv: implement steps to disable RHWO for Wa_14024015672

  • iris: implement steps to disable RHWO for Wa_14024015672

  • blorp: fix asserts hit with msaa blorp blits on xe3

  • anv: route clear operations on compute to companion

  • intel/dev: update mesa_defs.json from workaround database

  • anv: use workaround framework for Wa_1508208842

  • intel/genxml: add CHICKEN_RASTER_2 with required bit for Xe3

  • anv: set DisableAnyMCTRresponsefix to zero on init

  • iris: set DisableAnyMCTRresponsefix to zero on init

  • anv: skip compressed flag for bo if not supported by modifier

  • util: bring back fix to avoid strict aliasing bugs in xxhash

  • intel/dev: implement urb handle limits for Wa_16025326720

  • anv: update btp address after CmdExecuteCommands

  • anv: dirty descriptors in set_dirty_for_bind_map if sha changed

  • intel/dev: add NVL_U, NVL_P platforms to gen_wa_helpers.py

  • intel/dev: update mesa_defs.json from workaround database

  • intel/compiler: move validation assert after brw_shader_debug_log

  • anv: remove barrier special handling for RT_BTI_CHANGE

  • anv: add required barrier for Wa_14026570320

  • genxml/mi: add additional bit to FF_MODE and autostrip helper

  • anv: use mi_set_autostrip_state for autostrip control

  • iris: use mi_set_autostrip_state for autostrip control

  • intel/compiler: expose inferred_exec_pipe from scoreboarding

  • intel/compiler: implement dummy mov for Wa_18035690555

  • intel/dev: update mesa_defs.json from workaround database

  • anv: fix Wa_14024015672 interaction in blorp

  • intel/compiler: implement macl part of Wa_18035690555

  • drirc/anv: add flag to disable VK_EXT_subgroup_size_control

  • drirc: set anv_disable_subgroup_size_control for bg3

  • drirc: use anv_disable_drm_ccs_modifiers for any GTK version

  • anv: do not use resource barrier with split barriers

Thomas H.P. Andersen (2):

  • nvk: prepare for driver internal layers

  • nvk: add app workaround layer

Thong Thai (1):

  • radeonsi: remove radeonsi prefix from si_pipe.h includes

Tim Van Patten (1):

  • anv: Enable Vulkan 1.4 for SDK 37+

Timothy Arceri (18):

  • nir: make nir_collect_src_uniforms() private

  • nir: make nir_add_inlinable_uniforms() private

  • nir: update asserts in inline uniforms

  • nir: speedup nir_find_inlinable_uniforms()

  • mesa: add _mesa_lookup_state_param_idx() helper

  • st/glsl_to_nir: make sure the variant has the correct locations set

  • mesa/st: use same path for setting state ref locations

  • st/glsl_to_nir: update state var locations earlier

  • glsl: relax precision matching on unused uniforms ES

  • glsl: add workaround for MDK2 HD

  • glx: guard glx_screen frontend_screen member

  • mesa: add force_explicit_uniform_loc_zero workaround

  • util/driconf: add workarounds for Lethis - Path Of Progress

  • nir: remove is_only_uniform_src() restriction

  • nir: test loop analyze sets exact trip flags correctly

  • radeonsi: add Gun Godz workaround

  • glcpp: fix paste within macro function expansion

  • amd/radeonsi: dont clamp packed user varyings

Tomeu Vizoso (43):

  • dril: don’t build a rocket_dri.so

  • teflon/tests: Fail tests with unsupported output types

  • teflon/tests: Add EfficientDet model

  • teflon/tests: Add InceptionV1 model

  • teflon/tests: Add MobileNetV2 model

  • teflon/tests: Add SSD MobileNetV2 model

  • teflon/tests: Add MoveNet Lighning and Thunder models

  • ethosu: Update tests baseline for new models

  • ethosu: Update test expectations

  • teflon: Fix leak of tensor structs

  • ethosu: Add U85 fields, these are compatible with the U65

  • ethosu: Let maxblockdeps be arch-specific

  • ethosu: Make the UBlock sizes arch-specific

  • ethosu: Compute is_partkernel during scheduling

  • ethosu: Switch to the weight encoder from Regor

  • ethosu: Invert lowering order of concatenation suboperations

  • ethosu: Add debug option for forcing U85 generation

  • ethosu: Don’t emit redundant state changes

  • ethosu: Add a separate scheduler for the U85

  • ethosu: Improve parallelism by detecting overlaps for BLOCKDEP

  • ethosu: Expand pooling to U85

  • ethosu: Refactor ethosu_allocate_feature_map to return the new offset

  • ethosu: Emission changes for U85

  • ethosu: Implement simplified scaling for U85

  • ethosu: Fix ublock selection for 8-bit depthwise/pooling on U85-256

  • ethosu: map BOs at creation time and unmap at destruction

  • ethosu: Fix scalar ADD on U85

  • ethosu: Properly emit IFM_BROADCAST and IFM2_BROADCAST on U85

  • ethosu: Set test baseline for the Corstone 1000 (U85)

  • etnaviv/ml: Skip all synthetic tests as we now have several real models

  • rocket: Skip all synthetic tests as we now have several real models

  • gallium: pipe_tensor.resource → pipe_tensor.data

  • gallium: replace padding_same with per-side padding

  • gallium: add pipe_ml_device, pipe_screen::get_ml_device()

  • ethosu: move hardware description from ethosu_screen to ethosu_ml_device

  • ethosu: add U85-256 support to ethosu_ml_device_create()

  • ethosu: parse optional SRAM size from device spec string

  • ethosu: Specifying SRAM size in pipe_ml_device ID

  • gallium: add pipe_context::ml_subgraph_deserialize()

  • ethosu: implement ml_subgraph_deserialize()

  • ethosu: handle NULL bias tensor in convolution

  • gallium: add ml_device_destroy callback to pipe_ml_device

  • ethosu: implement ml_device_destroy for standalone ML device

Trigger Huang (1):

  • vulkan/queue: pass protected submit info to driver

Urja Rannikko (1):

  • hash_table: fix use-after-free by reorganization of destruct callbacks

Utku Iseri (19):

  • zink: add a variable splitter for component-xfb + unlowering

  • zink: manually ignore centroid with sample shading

  • zink: set mediump is 32 bits

  • zink: set flat interpolation for integer FS inputs

  • zink: switch to using unlower_io_to_vars

  • zink: remove rework_io and revectorization

  • zink: add an rpstores debug option

  • zink: track which stages a sampler gets bound to

  • zink: add arrayness I/O matching

  • zink: emulate clip distance

  • zink: add arm and panvk to invalid<->linear

  • zink: ignore msrtss support on panvk

  • pan/genxml: make pandecode comparisons return -1,1

  • panvk: pass heap explicitly to as_alloc/free

  • panvk: increase mappable VA range to 48 bits

  • panvk: expose swapchain_mutable_format support

  • panvk: use AUTO_VA with v9

  • panvk: add max supported va to physical device

  • panvk: BDA capture/replay support on v10+

Val Packett (1):

  • util/rust: Add memory map access mode detection to descriptor API

Valentine Burley (84):

  • tu: Assign stable unique_id to buffer objects

  • tu: Implement VK_EXT_device_memory_report

  • zink/ci: Fix a few job timeouts

  • zink/ci: Re-enable optimal_keys for zink-tu-a618

  • zink/ci: Enable optimal_keys for zink-tu-a750

  • tu: Handle VkDrmFormatModifierPropertiesList2EXT

  • lavapipe/ci: Update Android CTS expectations

  • ci/deqp: Retry GitHub API lookup for main-branch check

  • tu: Fix memory leak of patchpoints_ctx in dynamic rendering

  • tu/ci: Document a618-vk-asan failure

  • tu: Free cmd_buffer from its pool

  • tu: Simplify cmd_buffer allocation

  • docs: Update features.txt for Turnip

  • tu: Enable VK_KHR_compute_shader_derivatives for a6xx

  • tu: Advertise VK_EXT_shader_uniform_buffer_unsized_array

  • tu: Advertise VK_KHR_depth_clamp_zero_one

  • docs/features: Remove VK_KHR_android_surface

  • venus/ci: Increase android-angle-venus-anv-cml-cts timeout

  • venus/ci: Skip invalid SkQP test on ANV

  • ci: Update kernel to pull in new build for freedreno

  • turnip/ci: Move a618-vk job to new sc7180 runner

  • turnip/ci: Promote a618-vk-asan to pre-merge

  • turnip/ci: Skip more slow tests

  • turnip/ci: Remove a618-vk-full job

  • ci: Disable Collabora’s farm due to network issues

  • Revert “ci: Disable Collabora’s farm due to network issues”

  • zink/ci: Drop fixed VU from VVL filters

  • zink: Enable optimal keys for GPL on Turnip

  • radeonsi/ci: Skip subgroups.arithmetic tests on Mendocino

  • ci/lava: Uprev lava-job-submitter

  • ci: Update kernel to Linux 6.19.6

  • freedreno/ci: Switch sm8650 to gfx-ci/linux kernel

  • etnaviv/ci: Switch CI-tron to gfx-ci/linux kernel

  • intel/ci: Document recent Intel flakes

  • venus/ci: Remove hanging timeout override for ADL and TGL jobs

  • ci: Strip qemu from rootfs

  • ci/android: Disable wifi for Cuttlefish

  • ci/android: Update Cuttlefish build

  • ci/container: Generalize debian/x86_64_test-android container

  • ci/container: Prepare test-android for multi-arch support

  • ci: Add test-android container for arm64

  • venus/ci: Add an Android Venus on Turnip job on a618

  • ci: Update kernel to pick up new network adapter

  • tu: Add support for VK_EXT_depth_clamp_control

  • ci: Enable legacy-wayland=bind-wayland-display for debian-arm32 and debian-arm64

  • zink/ci: Enable mesh shader tests on lavapipe

  • zink/ci: Run zink-lavapipe on regular runner

  • ci/android: Update Cuttlefish build

  • ci: Uprev GL & GLES CTS

  • ci: Capture weston logs

  • tu/drm/virtio: Add missing lock to virtio_bo_init_dmabuf

  • tu/drm/virtio: Move set_iova into success path of virtio_bo_init_dmabuf

  • tu/drm/virtio: Avoid freeing zombified tu_sparse_vma

  • tu/drm/virtio: Do not free iova from heap for lazy BOs

  • tu/drm/virtio: Fix GEM handle leak in tu_bo_init error path

  • tu/drm/virtio: Fix GEM handle leak on failed dmabuf res_id lookup

  • ci: Avoid mixing libwayland versions in build and test containers

  • ci: Drop legacy-wayland option for debian-arm32

  • ci: Drop duplicate Intel shader-db run

  • ci: Run Intel shader-db on Lunar Lake and Panther Lake

  • ci: Enable ZSTD support for ZRAM in the kernel

  • zink/ci: Move zink-tu-a618 to sc7180-trogdor-kingoftown

  • ci/venus: Skip crashing Android CTS test on ANV

  • util: Add more libdrm stubs

  • egl: Use util/libdrm.h instead of xf86drm.h

  • meson: Add support for buidling zink + Turnip/KGSL

  • meson: Fix Turnip libdrm-linking check

  • ci: Enable EGL and GLX in debian-no-libdrm

  • meson: Update freedreno-kmds comment

  • lavapipe/ci: Skip flaky Android CTS test

  • venus/ci: Move android-angle-venus-tu-a618 to sc7180-trogdor-kingoftown

  • ci/android: Add 5-minute timeout to Cuttlefish launch

  • ci/android: Refactor replacing Vulkan drivers

  • ci/android: Enable virtio freedreno KMD support

  • ci/android: Update Cuttlefish build

  • turnip/ci: Add Android job with ANGLE on a618

  • pan/ci: Document recent flakes and timeouts

  • ci/freedreno: Move remaining lazor a618 jobs, retire device type

  • ci: Disable Collabora’s farm due to network issues

  • Revert “ci: Disable Collabora’s farm due to network issues”

  • anv/ci: Add full VKCTS pre-merge job on Raptor Lake

  • zink/ci: Remove Cezanne job

  • tu/drm/virtio: Fix tu_wait_fence timeout handling

  • freedreno/drm/virtio: Fix wait_fence ret ordering

Vignesh Raman (1):

  • ci/gitlab_gql: disable schema fetch

Vinson Lee (9):

  • compiler/clc: Fix const correctness in libclc_add_generic_variants

  • freedreno/decode: Fix const correctness in get_tex_count

  • freedreno/decode: replace lua_pushunsigned with lua_pushinteger

  • llvmpipe: fix build on macOS due to st_mtim

  • nil: Fix Rust test link failure under Coverity due to missing -lm

  • d3d12: Fix MinGW cross-build error in resource_state_if_promoted

  • zink: remove unused variable in zink_instance.py

  • st/mesa: fix implicit conversion warning in st_atom_framebuffer

  • vulkan/screenshot-layer: initialize info to NULL

Wang Ruitang (1):

  • amd/common/virtio: use device fd to init sync provider

Wei Hao (1):

  • radeonsi: fix threaded shader compilation finishing after context is destroyed

Wei Zhao (1):

  • vulkan/wsi/wayland: use roundtrip instead of flush on swapchain free

Wenfeng Gao (2):

  • mediafoundation: Support externally provided motion hints

  • mediafoundation: Fix the frame number validation logic for motion hint

Wujian Sun (1):

  • mesa: Fix inconsistent multisampled CopyTexImage checks

Xianzhong Li (1):

  • panfrost: Fix GEM handle refcount leak in panfrost_bo_import

Yiwei Zhang (77):

  • venus: track prime blit dst buffer memory in the wsi image

  • venus: track dedicated image during mem alloc

  • venus: add vn_renderer_bo_export_sync_file helper

  • venus: refactor vn_AcquireNextImage2KHR

  • venus: properly handle wsi implicit in-fence

  • venus: refactor Android ANB tracking to avoid confusions with WSI

  • venus: remove obsolete asserts for ANB image creation

  • ci/android: revive some previously skipped tests

  • pan/kmod: drop pan_kmod_bo_check_import_flags validation

  • pan/kmod: clean up unused flags from bo import paths

  • venus: fix a prime blit assert…again

  • venus: sync latest protocol for VK_EXT_descriptor_heap support

  • venus: implement all descriptor heap commands

  • venus: cmd inheritance info fix to consider descriptor heap

  • venus: pipeline layout is now optional

  • venus: skip image cache for VkOpaqueCaptureDataCreateInfoEXT

  • venus: add vn_descriptor.h to be shared between different desc systems

  • venus: rename format_update_mutex for general purpose

  • venus: cache descriptor size query

  • venus: ensure descriptor writes invariance

  • venus: take care of combined image sampler descriptor for ycbcr

  • venus: fill descriptor heap feats and props

  • venus: expose VK_EXT_descriptor_heap behind a debug option

  • venus: workaround a gcc-15 dead store elimination (DSE) bug

  • venus: sync latest protocol for VK_KHR_shader_fma

  • vulkan/wsi/drm: force prime buffer blit for WSI_DEBUG_BUFFER

  • venus: sync protocol for strict aliasing compliance

  • venus: the GCC DSE workaround is no longer needed

  • venus: amend to mark descriptor size cache initialized

  • venus: RegisterCustomBorderColorEXT can be async when index is requested

  • venus: expose VK_EXT_descriptor_heap by default

  • pan/fb: fix return type for mali_to_glsl_dim

  • ci/venus: skip broken drm display tests

  • util: convert tabs to spaces for ralloc.c

  • pan: fix to not clear out of bitset range

  • lvp: avoid advertising dmabuf support for kms_swrast

  • lvp: hide import-only dmabuf support from zink

  • ci/venus: update expectation based on nightly job runs

  • virgl: set DRM_RDWR for exported dma-bufs (non-blob)

  • venus: force prime blit on Nvidia GPU

  • vulkan/android: add new helpers for aliased ANB support (spec v8+)

  • lvp: add lvp_image_init helper

  • lvp: support VK_ANDROID_native_buffer v8+

  • llvmpipe: drop unused dt_format

  • lvp: import_memory_fd returns a boolean

  • lvp: properly initialize AHB image layout

  • lvp: fix dedicated allocation requirements for AHB images

  • ci/lvp: update android cts expectations

  • lvp: drop redundant lvp_image::offset

  • lvp: rename lvp_image_plane::plane_offset to offset

  • lvp: fix multi-planar image memory binding with explicit layout

  • llvmpipe: follow winsys handle attributes when imported with explicit layout

  • lvp: raise LVP_MAX_PLANE_COUNT to 3 and update ci expectations

  • lvp: follow winsys handle size when imported with explicit layout

  • lvp: refactor image plane initialization

  • venus: fix to relax the KHR_external_memory_fd requirement

  • vulkan/anv: use vk_device_get_timestamp and drop vk_clock_gettime

  • util/list: fix formatting

  • panvk: hide swapchainMaintenance1 behind WSI guard

  • ci/panvk: update expectations with new flakes

  • docs/venus: update instructions around Intel pat issue

  • docs/venus: adjust driver support list and drop obsolete descriptions

  • docs/venus: add QEMU instructions

  • docs/venus: add Android Cuttlefish instructions

  • vulkan/wsi/win32: add wsi_win32_find_idle_image helper

  • vulkan/wsi/win32: respect acquire timeout for sw wsi

  • venus: add vn_get_query_pool_results for non-qfb

  • venus: relocate vn_query_feedback_wait_ready into qfb query

  • venus: add vn_relax_warn to check if at warn order

  • venus: ensure qfb can catch device lost

  • venus: add vn_get_semaphore_counter_value that takes vn_relax_state

  • venus: ensure sfb can catch device lost

  • venus: add vn_get_fence_status that takes vn_relax_state

  • venus: ensure ffb can catch device lost

  • ci/venus: update expectation for an expected fail

  • docs/vulkan: fix the order of KHR and EXT extensions

  • docs/vulkan: fix the order of platform and vendor extensions

Yogesh Mohan Marimuthu (2):

  • winsys/amdgpu: pointers to be NULL if num 0 for kernel ioctl

  • winsys/amdgpu: call userq wait ioctl only once

Yonggang Luo (2):

  • vulkan/anv:Remove unused anv_clock_gettime

  • pvr: Remove two unused function

You, Min-Hsuan (1):

  • amd/vpelib: refactor minor change

Yuxuan Shui (4):

  • vulkan/wsi/x11: Make sure error is returned if create_swapchain fails

  • wsi/display: add connectors to connectors list during allocation

  • wsi/display: initialize Xlib display connector property IDs in all cases

  • wsi/display: move set atomic cap out of wsi_display_get_connector

Zan Dobersek (15):

  • tu: handle DS_DEPTH_BOUNDS_TEST_BOUNDS state under TU_DYNAMIC_STATE_RB_DEPTH_CNTL

  • tu: avoid incorrect pipeline draw state for disabled depth/stencil attachments

  • tu: allocate transient attachments used for LRZ

  • tu/kgsl: wait-only submit handling should not ignore sparse bind commands

  • freedreno/common: make a8xx magic regs common between all such devices

  • freedreno/common: set up a830 properties

  • tu/a8xx: fix tu_desc_set_ubwc() to avoid unwanted bitfield override

  • tu: use pkt_field macros in tu_desc_{get,set}_addr()

  • fd: make RD dump output path configurable through FD_RD_DUMP_PATH

  • tu/a8xx: add missing register state in tu_clear_sysmem_attachments()

  • fd: support a8xx in rddecompiler

  • fd/replay: kgsl context should use no-fault tolerance, report reset state

  • tu/kgsl: bump msm_kgsl.h header

  • tu: only support userspace-managed perfcounters on a7xx and earlier

  • tu/a8xx: remove enforced TU_DEBUG_FLUSHALL

Zeyang Lyu (1):

  • radv: Fix incorrect misaligned_mask_invalid for VK_EXT_vertex_input_dynamic_state

Zhao, Jiali (1):

  • amd/vpelib: Re-enable new feature support check

aerith (1):

  • zink: fix codegen for extensions with non-standard struct names

anonymix007 (1):

  • vulkan/runtime: Implement VK_TIME_DOMAIN_QUERY_PERFORMANCE_COUNTER_KHR

emre (1):

  • nvk: fix barrier cache invalidation

irql-notlessorequal (8):

  • hasvk: Allow NULL index buffers

  • hasvk: Remove no longer valid assert

  • hasvk: Handle VkBindMemoryStatusKHR on buffer/image memory bind

  • hasvk: Add support for Cmd*DescriptorSet*2KHR

  • hasvk: Advertise VK_KHR_maintenance6

  • docs/features: Mark VK_KHR_maintenance6 complete for hasvk

  • Revert “hasvk: Remove no longer valid assert”

  • hasvk: Stop advertising blockTexelViewCompatibleMultipleLayers

jaap aarts (1):

  • radv/sqtt: Prevent concurrent submit when sqtt is enabled

jiajia Qian (1):

  • nir/opt_phi_precision: Fix bit size mismatch when moving widening conversions

juntak0916 (1):

  • nvk: fix BindImageMemory2 per-bind status result

kingstom.chen (1):

  • radv/rt: only run move_rt_instructions() for CPS shaders

osy (2):

  • vulkan: external sync for vk_sync_binary

  • kk: enable VK_KHR_external_{fence,semaphore}_fd

rdh (1):

  • mesa: allow MAX_TRANSFORM_FEEDBACK_BUFFERS in GL40+ contexts

scavenger (1):

  • add VK CTS validation report for a0 interpolation fix

utzcoz (1):

  • gfxstream: Fix vkSetDebugUtilsObjectNameEXT crash for unwrapped objects