4611 lines
216 KiB
ReStructuredText
4611 lines
216 KiB
ReStructuredText
Mesa 25.0.0 Release Notes / 2025-02-19
|
||
======================================
|
||
|
||
Mesa 25.0.0 is a new development release. People who are concerned
|
||
with stability and reliability should stick with a previous release or
|
||
wait for Mesa 25.0.1.
|
||
|
||
Mesa 25.0.0 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 25.0.0 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: 96a53501fd59679654273258c6c6a1055a20e352ee1429f0b123516c7190e5b0 mesa-25.0.0.tar.xz
|
||
SHA512: 7f5b6674c40b6c8dcab7934512ff754b40a6a8a466422c90236f614d322033d4d465307ddcd983f9f3afb1310e132ec3186a085d261c95493a0c460b2ec59ce8 mesa-25.0.0.tar.xz
|
||
|
||
|
||
New features
|
||
------------
|
||
|
||
- cl_khr_depth_images in rusticl
|
||
- Vulkan 1.4 on radv/gfx8+
|
||
- VK_KHR_dedicated_allocation on panvk
|
||
- VK_KHR_global_priority on panvk
|
||
- VK_KHR_index_type_uint8 on panvk
|
||
- VK_KHR_map_memory2 on panvk
|
||
- VK_KHR_multiview on panvk/v10+
|
||
- VK_KHR_shader_non_semantic_info on panvk
|
||
- VK_KHR_shader_relaxed_extended_instruction on panvk
|
||
- VK_KHR_vertex_attribute_divisor on panvk
|
||
- VK_KHR_zero_initialize_workgroup_memory on panvk
|
||
- VK_KHR_shader_draw_parameters on panvk
|
||
- VK_KHR_shader_float16_int8 on panvk
|
||
- VK_KHR_8bit_storage on panvk
|
||
- VK_EXT_4444_formats on panvk
|
||
- VK_EXT_global_priority on panvk
|
||
- VK_EXT_global_priority_query on panvk
|
||
- VK_EXT_host_query_reset on panvk
|
||
- VK_EXT_image_robustness on panvk
|
||
- VK_EXT_pipeline_robustness on panvk
|
||
- VK_EXT_provoking_vertex on panvk
|
||
- VK_EXT_queue_family_foreign on panvk
|
||
- VK_EXT_sampler_filter_minmax on panvk
|
||
- VK_EXT_scalar_block_layout on panvk
|
||
- VK_EXT_tooling_info on panvk
|
||
- depthClamp on panvk
|
||
- depthBiasClamp on panvk
|
||
- drawIndirectFirstInstance on panvk
|
||
- fragmentStoresAndAtomics on panvk/v10+
|
||
- sampleRateShading on panvk
|
||
- occlusionQueryPrecise on panvk
|
||
- shaderInt16 on panvk
|
||
- shaderInt64 on panvk
|
||
- imageCubeArray on panvk
|
||
- VK_KHR_depth_clamp_zero_one on RADV
|
||
- VK_KHR_maintenance8 on radv
|
||
- VK_KHR_shader_subgroup_rotate on panvk/v10+
|
||
- Vulkan 1.1 on panvk/v10+
|
||
- VK_EXT_subgroup_size_control on panvk/v10+
|
||
- initial GFX12 (RDNA4) support on RADV
|
||
|
||
|
||
Bug fixes
|
||
---------
|
||
|
||
- radeonsi: regression with running DaVinci Resolve under rusticl since 666a6eb871d5dec79362bdc5d16f15915eb52f96
|
||
- [ANV][LNL] - Black Myth: Wukong (2358720) - Corruption is visible near the edge of water.
|
||
- [ANV][LNL] - Hogwarts Legacy (990080) - Pixelated corruption is visible when looking out at the water.
|
||
- radv/video/h265: pps.flags.transform_skip_enabled_flag = 1 randomly hangs GPU
|
||
- [ANV][LNL] - Steel Rats (619700) - Game crashes after opening logos play before reaching main menu
|
||
- nvk: Implement host-only descriptors
|
||
- Gnome-shell Wayland fails to start with segfault at modifier-less driver
|
||
- [ANV][LNL] - DYNASTY WARRIORS: ORIGINS (2384580) - Dithered transparency has vertical bands.
|
||
- AMD Radeon R9 270 randomly causes video playback applications to crash with "amdgpu: The CS has been rejected"
|
||
- Rendering issues on GravityMark with RadeonSI ACO
|
||
- i915: multiple tests assert with tgsi_ureg.h:893: ureg_swizzle: Assertion \`reg.File != TGSI_FILE_NULL' failed.
|
||
- shaders/closed/steam/deus-ex-mankind-divided/260.shader_test fails NIR validation
|
||
- shaders/closed/steam/deus-ex-mankind-divided/260.shader_test fails NIR validation
|
||
- panvk : vk_pipeline_cache_object_deserialize: Assertion \`reader.current == reader.end && !reader.overrun' failed.
|
||
- 46a8d5e7ef61735416d0c54886a7a9930621ae2c causes a permission denied spam
|
||
- [BUILD] Build Failure: Implicit Function Declaration 'timespec_sub_saturate' (loader_wayland_helper.c)
|
||
- intel genX_acceleration_structure: missing dependency to bvh/header.spv.h
|
||
- KHR_subgroup glsl parsing broken
|
||
- intel: add config options to disable ELK compiler bits
|
||
- a618: godot-tps-gles3-high trace reproducible flakes
|
||
- radv: mesh shader depth-only rendering is broken
|
||
- anv: Enable VK_FORMAT_A4R4G4B4_UNORM_PACK16_EXT for Android 15
|
||
- Using a buffer allocated on a rx 6800XT for scanout on a Ryzen 7950X results in glitches
|
||
- Systemfreeze from mesa version 1:24.3.0-1-x86_64 and above with Chromium and derivatives [and more or less all other graphic related things]
|
||
- msm_kgsl.h:560:21: error: expected ‘:’, ‘,’, ‘;’, ‘}’ or ‘__attribute__’ before ‘*’ token
|
||
- [radeonsi] VC1 hardware decoding over vaapi outputs green screen
|
||
- consecutive glDrawPixels do not reflect a changed pixel mapping
|
||
- Crashing while Processing Shaders in Marvel Rivals on Mesa 24.3.2 & Mesa 24.3.3
|
||
- Assertion \`nir_cf_node_get_function(&block->cf_node)->structured' failed
|
||
- r300: Conditional jump or move depends on uninitialised value in Xnine.mova test
|
||
- anv: Mesh shaders with two OpSetMeshOutputsEXT instructions are not supported
|
||
- hasvk: apps crash since "intel/compiler: Remove usage of variable length arrays"
|
||
- nir_validate should check metadata
|
||
- anv: vkcube(pp) segfault in multi-GPU config, apparent vkCreateSwapchainKHR failure
|
||
- anv,regression: Black square artifacts in Fenyx Rising on BMG
|
||
- [anv] Cyberpunk visual corruption on BMG
|
||
- [ANV][LNL] - Cyberpunk 2077 (1091500) - Flickering mesh during benchmark.
|
||
- Intel Arc A770: Crosshair in THE FINALS renders too large
|
||
- 3d render issues in Chromium after 1:24.3.1-3 update over 1:24.2.7-1 of mesa package
|
||
- intel/compiler: Out of bounds read in brw_eu_compact.c
|
||
- intel/compiler: Out of bounds read in brw_eu_compact.c
|
||
- egl,dri2: Segfault when running wayland clients on non-default GPU
|
||
- anv,regression: Visual glitches in Ghost of Tsushima on BMG
|
||
- anv, regression: Resident Evil 2 d3d12 freezes in main menu on a Arc b580
|
||
- radeonsi: fails to build with libc++
|
||
- Random mesa crashes in kwin_wayland on a 6600XT
|
||
- enc->enc_pic.enc_pic_order_cnt_type always zero even if pic->pic_order_cnt_type non-zero that application set
|
||
- [anv] Visual corruption in Cyberpunk on LNL and BMG
|
||
- [anv] Borderlands 3 visual corruption on BMG
|
||
- [ANV] LNL triangle corruption on clothing in HogwartsLegacy-trace-dx12-1080p-ultra
|
||
- Intel: Dark graphical glitches on cars and characters on Disney Speedstorm
|
||
- Regression in VA-API decoding
|
||
- freedreno: fails to build with Android NDK 27c
|
||
- hk_cmd_draw.c:3471:32: error: expression in static assertion is not constant
|
||
- anv/gfx12: Enable non-zero fast clears for non-FCV CCS_E
|
||
- gen12: 5% regression in factorio
|
||
- 32-bit: error: format ‘%lx’ expects argument of type
|
||
- regression;bisected;FTBFS: commit b13e2a495e9e3da56add7d852ca01b2cd7eef52d breaks x86_32 mesa build
|
||
- glxext.c: error: 'struct glx_screen' has no member named 'frontend_screen'
|
||
- regression;bisected;FTBFS: commit ae76a6a04596bfdbd37bab165bc5f2a5ff60d389 breaks x86 mesa build
|
||
- Can't allocate dpb buffer on firefox
|
||
- Segmentation fault resetting a query pool used to get BLAS properties
|
||
- libvulkan_lvp link fails if glslangValidator is not installed
|
||
- lvp acceleration structure broken on \`main` but not on \`staging/24.x`
|
||
- radv: warning that "radv is not a conformant Vulkan implementation" on Navi 32
|
||
- [anv][UHD630] DXVK 2.5 - 2.5.2 with DXVK_HUD=compiler or DXVK_HUD=fps freezes the game or the entire system (Works without compiler/fps HUD, DXVK 2.4.1 works fine)
|
||
- Licenses seems incomplete/misleading
|
||
- anv: Symbol clash in intel_batch_decoder build when expat not available
|
||
- glcts failures on LNL/BMG
|
||
- Lavapipe vulkan 1.4 support?
|
||
- d3d12 vaapi: thread safety issues
|
||
- anv: Missing textures and glitches in It Takes Two (game)
|
||
- [anv][bisected] GravityMark segfault when enabling u-trace on RT workload
|
||
- features.txt does not have a Vulkan 1.4 section despite some drivers already supporting the new version
|
||
- Black screen bug that only affects AMD
|
||
- Failure to correctly decode H.264, possibly specific to use of array output view
|
||
- X1-85: Portal 2: Bottom of portal gun disappears
|
||
- X-Plane 12: Prop disc rendering regression
|
||
- Errors when enumerating devices create incorrect expecations
|
||
- Resident evil 3 remake hanging - f8b584d6 regression
|
||
- R6700XT: QP value doesn't affect output when using CQP rate control w/ H264/H265 VAAPI encoders
|
||
- Bug in Mesa headers: \`error: redefinition of typedef 'GLsync'`
|
||
- nak: Crash when starting The First Descendant
|
||
- [r300] Regression in f424ef18010 breaks wayland on RS480M
|
||
- anv: Missing text in Age of Mythology Retold on a Arc b580
|
||
- RustiCL: and Clover broken with 9b7ea720c93 (!32713 (merged))
|
||
- nvk: Artifact Classic crash at loading screen
|
||
- radeonsi VAAPI - vc-1 interlaced decoding garbled on Polaris
|
||
- VDPAU AV1 hardware decoding broken for Mesa 25.0.0-devel
|
||
- mesa: st_glsl_to_nir call to nir_opt_fragdepth might not be valid with MSAA
|
||
- rusticl: warning: pointers cannot be transmuted to integers during const eval
|
||
- rusticl: warning: pointers cannot be transmuted to integers during const eval
|
||
- X1-85: Half Life 2 water rendering artifacts
|
||
- crash on video playback
|
||
- anv: Allow buffer compression for vkd3d by default?
|
||
- anv: bellwright needs force_vk_vendor=-1 %command% to launch
|
||
- [anv] Possible regression from !31269
|
||
- Up to 60% perf drop in SynMark DrvRes benchmark
|
||
- Memory leak on closing and re-opening X11 windows
|
||
- SIVPE errors on GPU-based screen recording (Radeon 890M)
|
||
- d3d12: va-api: build failure regression since 24.3.0-rc1 with MinGW GCC and clang
|
||
- anv: Marvel Rivals XeSS crash, game needs force_vk_vendor=-1 env variable
|
||
- anv: \`MESA: warning: INTEL_HWCONFIG_MIN_GS_URB_ENTRIES (2) != devinfo->urb.min_entries[MESA_SHADER_GEOMETRY] (0)`
|
||
- aco: two nir_shader_clock are miss optimized to one for GFX12
|
||
- aco: opengl buffer blit test fail when using aco on GFX12
|
||
- aco: nir_ddx/ddy v_interp optimization does not work on GFX12
|
||
- VAAPI b_depth 2 causes "manage_dpb_before_encode UVD - Failed to find ref0" error
|
||
- regression;bisected;FTBFS: commits 37d47913437e2e9f72283ea8bffce00efc40fce2 and e67e44522f4f5de4fcde53ad0fb75e396ef31f52 breaks x86 mesa build
|
||
- anv: Enable storage image compression on TGL
|
||
- zink: zink_create_quads_emulation_gs doesn't write primitive ID
|
||
- DZN/DXIL doesn't validate GTK shaders
|
||
- black screen and "Failed to add framebuffer" error in wayland compositors when not filtering dmabuf formats with ccs modifiers on intel graphics when upgrading to mesa 24.3.0
|
||
- nir: nir_opt_if_merge_test fails validation with NIR_DEBUG=validate_ssa_dominance
|
||
- radv: Vulkan AV1 video decode glitches
|
||
- radv: support RGP captures for purely compute pipelines
|
||
- regression;bisected: c49a71c03c9166b0814db92420eadac74cbc4b11 leads to artifacts if on top of launched game (in full screen mode) show list running apps (Hold Alt + Tab)
|
||
- !32067 broke piglit "spec\@egl_khr_create_context\@no-error context gl"
|
||
- Intel: Re-enable bo cache in iris driver (Xe2)
|
||
- [amdgpu][regression] GPU Hang/Reset Triggered by Several Applications
|
||
- ANV: X4 Foundations crashes with vkAllocateDescriptorSets -12
|
||
- About twenty vulkan-samples cases will crash caused by the same error while running on PanVK
|
||
- Firestorm crashes on startup with Mesa 24.3
|
||
- anv: Use-after-free detected by AddressSanitizer while running dEQP-VK
|
||
- GPU process crash via WebGPU shader - UAF in mesa gcm_schedule_early_instr at src/compiler/nir/nir_opt_gcm.c:477
|
||
- radv: DCC causes glitches in Red Dead Redemption 2
|
||
- A5xx rendering issues with firefox
|
||
- [ANV][Regression] Broken rendering in Flycast + Per-Pixel Alpha Sorting
|
||
- [TGL][anv] Performance regression in Dota 2 replay
|
||
- vtn: OpTypeStruct in kernel parameters trigger assertion in glsl_types.h
|
||
- anv: Assertion failure in \`dEQP-VK.image.extended_usage_bit_compatibility.image_format_list.s8_uint_optimal_transfer_src_bit`
|
||
- radv: Resident Evil 6 Benchmark Tool has artifacts on 7900 XTX when DCC is enabled, game launched on 4K monitor without scaling and with FullHD settings
|
||
- [AMD RX 6700 XT] Artifacts while upscaling games in fullscreen mode
|
||
- Distorted pixelated graphics with Radeon RX 7900 XT with some games
|
||
- Total War Warhammer 2 Graphical Glitch
|
||
- Glitching artifacts in tile shaped patterns on 6700 XT, when using upscaled fullscreen game on labwc
|
||
- anv: Page fault when using MTL simulator in dEQP-VK.ray_tracing_pipeline.data_spill.report_intersection.float32
|
||
- mesa_cache_db.c:316:33: error: call to undeclared function 'mremap'
|
||
- [trunk] shaders fail hard in openmw after cbfc225e2bda2c8627a4580fa3a9b63bfb7133e0
|
||
- u_perfetto.h:33:9: error: unknown type name 'clockid_t'; did you mean 'clock_t'?
|
||
- brw_fs_opt_copy_propagation incorrectly handles size changes of uniforms
|
||
- RADV Command buffer reuse doesn't reinitialize is_secondary
|
||
- Virgl:Qcom sa8155 GL_MAX_FRAGMENT_SHADER_STORAGE_BLOCKS/GL_MAX_VERTEX_SHADER_STORAGE_BLOCKS is too small to run antutu benchmark apk
|
||
- nouveau paraview msaa corruption 23.1 bisected regression
|
||
- mesa fails to build due to missing SPV_ENV_UNIVERSAL_1_6 symbol
|
||
|
||
|
||
Changes
|
||
-------
|
||
|
||
Aaron Ruby (6):
|
||
|
||
- meson: Remove experimental from gfxstream driver build
|
||
- gfxstream: Some cleanup in manual entrypoints
|
||
- gfxstream: Remove VK_HOST_CONNECTION macro
|
||
- gfxstream: Fix unused variable warnings in ResourceTracker.cpp
|
||
- vulkan/util: Add c99_compat.h inclusion for cpp 'restrict' compatibility
|
||
- gfxstream: Remove internal vk_util.h and vk_struct_id.h entirely
|
||
|
||
Adam Jackson (2):
|
||
|
||
- docs/envvars: Remove mention of IRIS_ENABLE_CLOVER
|
||
- docs/envvars: Combine WGL sections
|
||
|
||
Alejandro Piñeiro (1):
|
||
|
||
- docs/features: mark VK_EXT_scalar_block_layout as supported for vc7+
|
||
|
||
Aleksi Sapon (9):
|
||
|
||
- draw: primitive ID is per-patch
|
||
- llvmpipe: spec\@arb_tessellation_shader\@execution\@gs-primitiveid-instanced is fixed
|
||
- zink: spec\@arb_tessellation_shader\@execution\@gs-primitiveid-instanced is fixed
|
||
- draw: front-face injection must check geometry shader primitive type
|
||
- llvmpipe: PointCoord is offset when multisampling is enabled
|
||
- meson: fix finding Python on Windows
|
||
- llvmpipe: fix lp_test_arit on Windows
|
||
- llvmpipe: LLVM v2f32 trunc/floor/ceil/nearbyint generates optimal x86 code since at least version 8
|
||
- llvmpipe: disable anisotropic filtering for non-2D textures
|
||
|
||
Alyssa Rosenzweig (206):
|
||
|
||
- nir/opt_algebraic: optimize patterns from Skia
|
||
- nir/opt_algebraic: add more 64-bit patterns
|
||
- nir/opt_algebraic: add another 64-bit pattern
|
||
- nir: add amul flag
|
||
- nir: add late_lower_int64 option
|
||
- nir: add ilea_agx/ulea_agx opcodes
|
||
- nir/builder: use amul over ishl on agx
|
||
- nir/opt_algebraic: don't lower amul if requested
|
||
- nir/lower_uniforms_to_ubo: use amul
|
||
- rusticl: respect late_lower_int64
|
||
- agx: vectorize SSBOs
|
||
- agx: model IC dispatch
|
||
- agx: fix bfeil timing
|
||
- hk: reduce max SSBO size
|
||
- libagx: promote math to use AGX address mode
|
||
- agx: rewrite address mode lowering
|
||
- agx: change int conversion test
|
||
- agx: add pseudo for signext
|
||
- agx: optimize signext+iadd
|
||
- agx: fold zext into int sources
|
||
- agx: add tests for sign/zero-extend propagate
|
||
- agx: fix atomics in tess count shaders
|
||
- hk: don't advertise impossible modifiers
|
||
- agx: optimize signext imad
|
||
- agx: fuse iadd+large shift into imad
|
||
- agx: make imad+ishl rules actually work
|
||
- hk: drop assert
|
||
- hk: fix meta shader name
|
||
- libagx: fix cl warning
|
||
- libagx: drop branch
|
||
- libagx: drop dead code
|
||
- libagx: vectorize triangle def'n
|
||
- libagx: drop Clockwise
|
||
- libagx: simplify index patch expression
|
||
- libagx: don't key unroll to index size
|
||
- libagx: fix unroll kernel constant qualifier
|
||
- libagx: drop silliness in restart kernel
|
||
- agx: fuse also 8-bit address math
|
||
- asahi: extract agx_get_num_cores
|
||
- asahi: correct core count, max freq
|
||
- asahi: fix a2c with sample shading, harder
|
||
- asahi: assert/cse resource valid
|
||
- asahi: don't take compiled_shader in agx_build_internal_usc
|
||
- asahi: drop dead param
|
||
- asahi: factor out more compiled shader
|
||
- asahi: move agx_gather_device_key
|
||
- util: add u_tristate data structure
|
||
- panfrost: switch to u_tristate
|
||
- agx: make needs_g13x_coherency a tri-state
|
||
- nir/lower_convert_alu_types: use intrinsics_pass
|
||
- nir/conversion_builder: avoid redundant uint->uint clamp
|
||
- nir/opt_algebraic: optimize convert_uint_sat(ulong)
|
||
- nir: add names to function parameters
|
||
- nir/print: print function signature
|
||
- nir/print: annotate entrypoints
|
||
- nir/print: print parameter names in calls
|
||
- vtn: gather function parameter names
|
||
- vtn: use rzalloc in bindgen
|
||
- vtn: use named parameters in bindgen
|
||
- vtn: preserve name, is_return in bindings
|
||
- nir: split off some definitions for OpenCL
|
||
- compiler: make glsl_sampler_dim available to CL
|
||
- nir/lower_system_values: add ID to 32-bit lowering
|
||
- nir: add nir_fixup_is_exported pass
|
||
- vtn: introduce vtn_bindgen tool
|
||
- libagx: switch to vtn_bindgen
|
||
- libagx: move out of lib/
|
||
- libagx: DCE
|
||
- asahi: drop dead ACCESS
|
||
- asahi,agx: move texture lowering into the compiler
|
||
- asahi: drop desc align alloc
|
||
- asahi/decode: disasm 3D helper progs
|
||
- asahi/clc: drop getopt
|
||
- agx: vectorize scratch access
|
||
- agx: gather workgroup size
|
||
- asahi,hk: reenable rgb32 buffer textures
|
||
- hk: generalize internal launch
|
||
- hk: expose missing eds3 feature
|
||
- hk: handle mismatching colour vs z/s dimensions
|
||
- hk: implement EXT_depth_bias_control
|
||
- hk: be robust against invalid MSAA inputs
|
||
- hk: do not increment GS queries for passthru GS
|
||
- hk: use common wg size
|
||
- hk: add cmd buffer to hk_cs
|
||
- hk: dce
|
||
- libagx: fix return type
|
||
- libagx: don't export vertex_id_for_top
|
||
- asahi/genxml: fix 0 encoding for groups
|
||
- asahi/genxml: fix 128-bit in CL path
|
||
- asahi/genxml: optimize out masking with shr
|
||
- asahi/genxml: define missing macros
|
||
- asahi: add XML for cdm stream link with return
|
||
- asahi: refmt
|
||
- vtn: ignore SpvFunctionParameterAttributeSret
|
||
- nir/pack_bits: handle 8-bit vec8 -> 64-bit
|
||
- nir: add nir_lower_calls_to_builtins pass
|
||
- asahi/clc: switch to nir_lower_calls_to_builtins
|
||
- nir: add nir_foreach_entrypoint macros
|
||
- nir: add workgroup size to functions
|
||
- vtn: plumb through OpEntryPoint
|
||
- vtn: gather workgroup size in libraries
|
||
- nir: add nir_function::pass_flags
|
||
- nir: add nir_remove_entrypoints helper
|
||
- nir: add nir_lower_constant_to_temp helper
|
||
- nir: add helpers for precompiled shaders
|
||
- asahi,vtn: precompile kernels
|
||
- libagx: increase wg size for query copy
|
||
- asahi: crash on fault
|
||
- hk: fix incorrect index size translate
|
||
- hk: fix z bias perf regression
|
||
- hk: implement hack for layered no attachments
|
||
- hk: clarify bounds check calculations
|
||
- agx: disable bounds check optimization
|
||
- agx: reduce preamble/main alignment
|
||
- asahi: drop dead pool stuff
|
||
- asahi: don't leak rodata
|
||
- hk,asahi,libagx: unify a bit of code
|
||
- asahi: drop dead
|
||
- asahi: fix page size alignment
|
||
- asahi: fix u_blitter related leaks
|
||
- asahi: label individual pools
|
||
- asahi,hk: mmap BO on first use
|
||
- asahi: add more asserts around bo add
|
||
- asahi: fix agx_batch_add_bo
|
||
- asahi: add =bodump debug help
|
||
- asahi: fix agxdecode memory mapping
|
||
- hk: implement timestamps
|
||
- hk: claim 1.4
|
||
- zink: fix gl_PrimitiveID reads with quads
|
||
- nir/search_helpers: handle bcsel in is_only_used_as_float
|
||
- nir/opt_algebraic: optimize sign bit manipulation
|
||
- nir/opt_load_store_vectorize: match amul like imul
|
||
- nir,asahi: make argument alignment configurable
|
||
- mesa_clc: add depfile support
|
||
- libagx: switch to depfile support
|
||
- libagx: remove redundant source files
|
||
- vulkan: rename depth bias graphics states
|
||
- vulkan: bump layer api versions
|
||
- nir: add printf_abort intrinsic
|
||
- nir/lower_printf: allow fixed address
|
||
- nir/lower_printf: lower aborts
|
||
- nir/lower_printf: use unsigned math
|
||
- nir/lower_printf: use 64-bit math
|
||
- util/printf: be robust against truncated buffers
|
||
- util/printf: add context-ful helpers
|
||
- vulkan: add vk_check_printf_status helper
|
||
- nir/lower_point_size: skip non-var derefs
|
||
- clc: plumb cl_khr_subgroup_ballot
|
||
- libcl: add a common header for CPU/GPU stuff
|
||
- libcl: add VkDraw(Indexed)IndirectCommand definitions
|
||
- util/bitpack_helpers: make partially CL safe
|
||
- asahi: allow c23 extensions
|
||
- asahi/clc: remap __FILE__
|
||
- asahi,hk: wire up printf, abort
|
||
- agx: implement halts
|
||
- libagx: drop pointless helper
|
||
- libagx: port to common libcl.h
|
||
- compiler: use libcl.h for CL
|
||
- compiler: add mesa_prim_has_adjacency helper
|
||
- asahi: use mesa_prim_has_adjacency
|
||
- nir: add lower_scratch_to_var pass
|
||
- compiler/glsl_types: add glsl_get_word_size_align_bytes
|
||
- agx: optimize scratch access
|
||
- radeonsi: use mesa_prim_has_adjacency
|
||
- asahi: fix mmap'ing imported BOs
|
||
- hk,libagx: move hk_draw to the gpu
|
||
- asahi: use common draw
|
||
- libagx: add missing agx_vdm_return
|
||
- agx: add more 8-bit address fusing rules
|
||
- asahi: reformat
|
||
- agx: match another address pattern
|
||
- libagx: move index size helpers to the gpu
|
||
- libagx: refactor index buffer code
|
||
- libagx: factor out load/store_index
|
||
- hk: use index buffer overflow check
|
||
- hk: factor out hk_draw_as_indexed_indirect
|
||
- hk,libagx: accelerate index buffer robustness
|
||
- hk,libagx: handle adjacency without a GS
|
||
- libagx,hk: handle pipeline stats queries without a GS
|
||
- libagx: use designated initializers
|
||
- hk: avoid compiling unneeded VS->GS variants
|
||
- hk: fix primitive restart dirty tracking
|
||
- glsl: fix glsl_get_word_size_align_bytes
|
||
- nir: pass a callback to nir_lower_robust_access
|
||
- nir/lower_robust_access: fix robustness with atomic swap
|
||
- libagx: add agx_barrier enum
|
||
- nir,asahi,hk: add barrier argument to MESA_DISPATCH_PRECOMP
|
||
- intel: set max_buffer_size to nir_lower_printf
|
||
- nir/lower_printf: drop null check
|
||
- nir/lower_printf: drop default max buffer size
|
||
- nir,util: move printf serializing into util
|
||
- util: add u_printf_hash helper
|
||
- util/u_printf: add singleton implementation
|
||
- util/u_printf: allow printing from singleton
|
||
- nir/lower_printf: add option to hash format strings
|
||
- nir/lower_printf: support dynamic buffer size
|
||
- nir: add nir_lower_printf_buffer pass
|
||
- agx: defer printf address lowering
|
||
- nir/lower_printf: drop static buffer addr lowering
|
||
- util,vulkan,asahi,hk: hash format strings
|
||
- nir/lower_robust_access: do not preserve control flow
|
||
- nir: fix O(N^2) behaviour in nir_remove_dead_variables
|
||
- meson: project-wide fs = import('fs')
|
||
- clc,libagx: drop --in for mesa_clc
|
||
- clc,libagx: automatically set lang version
|
||
- nir/serialize: strip function names names
|
||
|
||
Antonino Maniscalco (1):
|
||
|
||
- nir,zink,asahi: support passing through gl_PrimitiveID
|
||
|
||
Antonio Ospite (53):
|
||
|
||
- ci/deqp: replace local android patches with upstream solution
|
||
- docs/android: update docs/android.rst after libgallium_dri updates
|
||
- docs/android: improve documentation about building llvmpipe for Android
|
||
- docs: remove leftover mention of meson dri3 option
|
||
- ci/android: unset compiler env vars in debian/android_build.sh
|
||
- ci/android: add a script to build LLVM libraries for Android
|
||
- ci/container: remove S3_JWT_FILE when container_job_trampoline.sh exits
|
||
- ci: set GIT_COMMITTER_DATE in a locale-agnostic format
|
||
- ci/deqp: refresh some patches to apply on top of recent VK-GL-CTS
|
||
- ci/deqp: cherry-pick fixes for building GL and GLES deqp on Android
|
||
- ci/deqp: enable building testlog tools on Android too
|
||
- ci/deqp: collect the mustpass lists also for the android target
|
||
- ci/android: fix problem with deqp version file when building for Android
|
||
- ci/android: build deqp for DEQP_API=VK
|
||
- ci/android: build llvmpipe driver for Android by forcing llvm fallback
|
||
- ci/android: don't copy the DRI drivers which are not needed anymore
|
||
- ci/android: restart all services after copying the new mesa libraries
|
||
- ci/android: handle premature exit of .gitlab-ci/cuttlefish-runner.sh
|
||
- ci/android: update version of cuttlefish host tools
|
||
- ci/android: add sudo to EPHEMERAL deps for debian/x86_64_test-android.sh
|
||
- ci/android: get custom cuttlefish images from the S3
|
||
- ci/android: make cuttlefish-runner.sh more robust against different Android images
|
||
- ci/android: better separate host and guest mesa artifacts
|
||
- ci/android: use a custom kernel when launching cuttlefish
|
||
- ci/android: fix warning when using chown
|
||
- ci/android: fix result dir for Android guest execution of deqp-runner
|
||
- ci/android: don't call cuttlefish-host-resources script
|
||
- ci/android: reorder PATH and LD_LIBRARY_PATH values to clarify priority
|
||
- ci/android: also copy mesa vulkan libraries to the Android guest
|
||
- ci/android: update list of deqp files pushed to the guest system
|
||
- ci/android: use a native adb connection
|
||
- ci/android: set XDG_CACHE_HOME and pass --shader-cache-dir to deqp-runner
|
||
- ci/android: use a /data/deqp subdirectory on guest to store dEQP files
|
||
- ci/android: set VK_DRIVER_FILES before launching cuttlefish
|
||
- ci/android: add ci rules to test llvmpipe on Android
|
||
- ci/android: add ci rules to test venus on Android
|
||
- ci/android: upgrade DEBIAN_TEST_ANDROID_TAG
|
||
- ci/android: fix meson C++ cross-compiler argument detection
|
||
- ci/android: update ANDROID_NDK and ANDROID_SDK_VERSION
|
||
- ci/android: use ANDROID_SDK_VERSION when building deqp components
|
||
- ci/android: use ANDROID_SDK_VERSION for debian-android job too
|
||
- ci/android: rename variable ANDROID_NDK to ANDROID_NDK_VERSION
|
||
- docs/android: bump suggested platform-sdk-version to 34
|
||
- freedreno/meson: remove C++ cross-build arguments HACKs
|
||
- freedreno/meson: sort list of options passed to get_supported_arguments()
|
||
- ci/android: update CUTTLEFISH_BUILD_NUMBER
|
||
- ci/android: define an INSTALL var for the source of mesa artifacts
|
||
- ci/android: improve handling of expectation files
|
||
- ci/android: fix pulling results from Android device
|
||
- ci/android: post-process testlog XML and create a junit.xml
|
||
- ci/android: pass --max-fails to deqp-runner in cuttlefish-runner.sh
|
||
- ci/android: pass --allow-downgrades when installing cuttlefish host tools
|
||
- ci/android: stop pushing libglapi.so since it's not available anymore
|
||
|
||
Arseny Kapoulkine (1):
|
||
|
||
- radv: On GFX11, use box sorting heuristic based on ray flags
|
||
|
||
Arvind Yadav (1):
|
||
|
||
- amd: Add amdgpu userqueue IOCTL functions
|
||
|
||
Asahi Lina (16):
|
||
|
||
- asahi: Add pipe bind flags to resource debug
|
||
- asahi: Add PIPE_BIND_SHARED to imported resources
|
||
- asahi: Extract agx_decompress_inplace()
|
||
- asahi: Introduce batch->feedback to disable compression in PBE
|
||
- asahi: In-place decompress shared resources for feedback loops
|
||
- hk: Add virtio implicit sync support
|
||
- hk: Fix DRM modifier selection for compressed surfaces
|
||
- hk: Enable missing swapchainMaintenance1 support
|
||
- asahi: Use 64bit size fields
|
||
- hk: Bump up max buffer size
|
||
- asahi: UAPI update to add GET_TIME & cleanup
|
||
- asahi: Fix agx_gpu_time_to_ns & implement DRM_ASAHI_GET_TIME
|
||
- asahi: UAPI update to add support for user timestamp buffers
|
||
- asahi: Add timestamp buffer ops
|
||
- asahi: Virt UABI update
|
||
- asahi: hk: Enable timestamps for virt
|
||
|
||
Autumn Ashton (1):
|
||
|
||
- radv/video: Fix bitstreamStartOffset including dstBufferOffset
|
||
|
||
Bas Nieuwenhuizen (1):
|
||
|
||
- util/perf: Fix some warnings.
|
||
|
||
Benjamin Cheng (4):
|
||
|
||
- ac/vcn: allow sq signature package to be skipped
|
||
- radv/video: support event for pre-VCN4 encode queues
|
||
- radv/video: support event for pre-VCN4 decode queues
|
||
- radv/video: enable by default on vcn2/3 with latest fw
|
||
|
||
Benjamin Lee (36):
|
||
|
||
- panvk: inherit sample count in secondary cmdbufs
|
||
- nir: clamp small W in nir_lower_viewport_transform
|
||
- nir: document order requirement for nir_lower_viewport_transform
|
||
- panvk: refactor fbinfo into a temp var in get_tiler_desc
|
||
- panvk: treat provoking vertex as dynamic state
|
||
- panvk: set provoking vertex in fbinfo
|
||
- panvk: advertise VK_EXT_provoking_vertex
|
||
- nir: handle arbitrary per-view outputs in nir_lower_multiview
|
||
- nir: document index semantics in nir_lower_multiview
|
||
- nir: treat per-view outputs as arrayed IO
|
||
- nir: add option to use compact view indices
|
||
- panvk: implement multiview support
|
||
- panvk: only clear enabled views
|
||
- panvk: disable position fifo optimization when multiview enabled
|
||
- panvk: advertise multiview support on v10+
|
||
- panvk: add note about pan_lower_store_component requirements
|
||
- nir: update docs for nir_get_io_arrayed_index_src
|
||
- panvk: set uses_sample_shading NIR flag when sample shading is forced
|
||
- panvk: fix sample position when sample shading is disabled
|
||
- panvk/csf: fix alpha-to-coverage
|
||
- panfrost: add intrinsic to load frag coord at a barycentric
|
||
- panfrost: add nir pass to lower noperspective varyings
|
||
- panfrost: collect noperspective varyings in shader info
|
||
- panvk: pass noperspective_varyings sysval as a push constant
|
||
- panfrost: add pass to lower noperspective varyings to a constant
|
||
- panvk: use static noperspective when statically linking VS and FS
|
||
- panfrost: factor FS shader key into a helper function
|
||
- panfrost: specialize VS on FS interpolation qualifiers
|
||
- panvk: handle sample mask writes on 1-sample targets
|
||
- panvk: remove load_multisampled_pan sysval
|
||
- panfrost/va: add FLUSH instruction
|
||
- panfrost/va: implement fquantizetf16 ftz
|
||
- panvk: disable round_to_nearest_even for NEAREST-filtered samplers
|
||
- panfrost: remove incorrect usage of MALI_PIXEL_KILL_STRONG_EARLY
|
||
- panfrost: fix hang by using MALI_PIXEL_KILL_WEAK_EARLY in color preload
|
||
- panfrost: remove is_blit flag
|
||
|
||
Benjamin Otte (1):
|
||
|
||
- vulkan/wsi: Support alpha swapchains on win32
|
||
|
||
Benjamin ROBIN (1):
|
||
|
||
- util/disk_cache: Do not try to delete old cache if cache is disabled
|
||
|
||
Bo Hu (5):
|
||
|
||
- gfxstream: snapshot: avoid double boxing dispatchable handle
|
||
- gfxstream: snapshot: DescriptorSet allocate and update
|
||
- gfxstream-guest: update offset to correct value
|
||
- update decoder.py to clean up un-used ApiCallInfo
|
||
- remove the mReconstructionMutex in load
|
||
|
||
Boris Brezillon (103):
|
||
|
||
- panvk: Enable CI on G610
|
||
- pan/ci: Move g610-vk jobs to post-merge CI
|
||
- panvk: Change the prototype of panvk_select_tiler_hierarchy_mask()
|
||
- panvk: Kill unused fields in panvk_cmd_graphics_state
|
||
- panvk: Move the panvk_cmd_graphics_state definition to panvk_cmd_draw.h
|
||
- panvk: Move panvk_cmd_compute_state to a common place
|
||
- panvk: Move is_dirty() to panvk_cmd_draw.h and rename it
|
||
- panvk: Don't link the VS and FS shaders on v10
|
||
- panvk: Sanitize the driver-internal dirty state tracking
|
||
- panvk: Move common gfx bits to a new source file in the common dir
|
||
- panvk: Cache the fs_required() result
|
||
- panvk/csf: Fix a wait-LS operation in finish_cs()
|
||
- panvk/cs: Poison cmdbuf registers when PANVK_DEBUG=cs is set
|
||
- panvk/ci: Update CI expectations to have a green CI
|
||
- panfrost: Increase AFBC body alignment requirement on v6+
|
||
- panfrost: Add a helper to expose the maximum effective tile size
|
||
- panfrost: Add the concept of render block
|
||
- panfrost: Add support for AFBC(split)
|
||
- panfrost: Advertise support for AFBC(32x8,sparse,split)
|
||
- pan/decode: Flush the dump file before crashing
|
||
- panvk/csf: Keep a cache of the CS reg file at the panvk_queue level
|
||
- panvk/csf: Fix cross command buffer render pass suspend/resume
|
||
- panvk/csf: Explain why the tiler is set to 0xdeadbeefdeadbeef
|
||
- panvk: Fix panvk_plane_index() for D32_SFLOAT_S8_UINT
|
||
- pan/cs: Add cs_exception_handler_ctx
|
||
- pan/cs: Align exception handlers with NOPs
|
||
- pan/cs: Add dynamic save_reg to exception handler
|
||
- pan/cs: Add block macro for exception handler
|
||
- panvk/csf: Fix register overlap in issue_fragment_jobs()
|
||
- pan/cs: Return the dump region size when an exception handler is defined
|
||
- pan/cs: Return exception handler size/address
|
||
- panfrost: Add cs_exception_handler_def() to the ForEachMacros list
|
||
- panvk/csf: Use the information returned by cs_exception_handler_def()
|
||
- panfrost: Use the handler size returned by cs_exception_handler_def()
|
||
- panvk: Filter out input-attachment usage on non renderable formats
|
||
- pan/decode: Untangle CS disassembling and interpretation
|
||
- pan/decode: s/interpret_ceu/interpret_cs/
|
||
- pan/decode: Rename pandecode_cs() into pandecode_interpret_cs()
|
||
- pan/decode: Add a helper to print CS binaries without interpreting them
|
||
- pan/decode: Provide a helper to print messages outside of the decoding path
|
||
- pan/cs: Add a LOAD_IP pseudo instruction
|
||
- pan/cs: Add an event-based tracing mechanism
|
||
- panvk/csf: Use event-based CS tracing
|
||
- panvk/csf: Don't disable SIMULTANEOUS_USE when tracing is enabled
|
||
- panvk: Add a flag to force SIMULTANEOUS_USE
|
||
- pan/texture: Move the plane info retrieval logic to a helper function
|
||
- pan/texture: Stop passing the view format around
|
||
- pan/texture: s/index/plane_index/ in panfrost_emit_plane()
|
||
- pan/texture: Stop passing a layout to panfrost_emit_plane()
|
||
- pan/texture: Pass pan_image_section_info around
|
||
- nir: Let nir_lower_texcoord_replace_late() report progress
|
||
- panfrost: s/NIR_PASS_V/NIR_PASS/
|
||
- panfrost: Use nir_shader_intrinsics_pass() for the line_smooth lowering pass
|
||
- panvk: s/NIR_PASS_V/NIR_PASS/
|
||
- pan: s/NIR_PASS_V/NIR_PASS/
|
||
- panvk: Move the descriptors preparation out of CreateImageView()
|
||
- vk/meta: Pass depth/stencil attachments only when a clear is requested
|
||
- panvk: Ignore the view aspects when dealing with depth/stencil attachments
|
||
- pan/cs: Fix cs_builder allocation failure robustness
|
||
- panvk: Wrap our descriptor lowering passes in NIR_PASS()
|
||
- panvk: Stop using magic values for the sysval push constant offset/range
|
||
- panvk: Automate sysval access from NIR shaders
|
||
- panvk: Lower dynamic push_constant loads in desc_copy logic
|
||
- panvk: Lower load_push_constant with dynamic offset to global loads
|
||
- pan/bi: Get rid of bi_lower_load_push_const_with_dyn_offset()
|
||
- panvk: Don't define push_constant range/base when we don't have to
|
||
- pan/indirect: Don't use .base to pass the push_constant offset
|
||
- pan/mi: Don't pretend we support push constants
|
||
- pan/bi: Disallow non-zero .{range,base} on load_push_constant instructions
|
||
- pan/bi: Fix mem_access_size_align_cb() for push constants
|
||
- panvk: Don't lower load_base_vertex
|
||
- panvk: Fix first_vertex/base_instance types
|
||
- pan: Don't pretend we support load_{vertex_id_zero_base,first_vertex}
|
||
- panvk: Don't lower load_blend_const_color_rgba
|
||
- panvk: Factor-out the sysvals initialization logic
|
||
- panvk: Pass a cmdbuf to blend_emit_descs()
|
||
- panvk: Pack push constants
|
||
- panfrost: Kill the mali_ptr typedef
|
||
- panfrost: Kill the uXX typedefs
|
||
- panfrost: Move MALI_EXTRACT_INDEX to pan_format.h
|
||
- panfrost: Move MAX_{MIP_LEVELS,IMAGE_PLANES} to pan_texture.h
|
||
- panfrost: Kill panfrost-job.h
|
||
- panvk: Don't invalidate the viewport on cull mode updates
|
||
- panvk/jm: Fix depth clipping with small viewport depth range
|
||
- panvk: Fix an alignment issue on x86
|
||
- panvk: Fix panvk_priv_mem_bo() on 32-bit platforms
|
||
- panfrost/ci: Add panvk and panfrost to the debian-x86_32 job
|
||
- pan/genxml: s/PAN_PAN_HELPERS_H/PAN_PACK_HELPERS_H/
|
||
- pan/genxml: Include pan_pack_helpers.h instead of copying it
|
||
- pan/genxml: Generate MALI_XXX_PACKED_T macros
|
||
- panfrost: Fix instanced draws when attributes have a non-zero divisor
|
||
- pan/cs: Fix the tracepoint register dump loops
|
||
- pan/cs: Allow undefined value if condition=always in cs_branch_label()
|
||
- pan/cs: cs_{break,continue} are not for_each macros
|
||
- panvk/csf: Make all sync operations on the CSG scope
|
||
- panvk/csf: Use cs_sr_reg64() instead of cs_reg64() when setting the OQ pointer
|
||
- panvk/csf: Rework the occlusion query logic to avoid draw flushes
|
||
- panvk/csf: Fix add_memory_dependency() for input attachment access
|
||
- panvk/csf: Add a knob to force texture cache invalidation on RUN_FRAGMENT
|
||
- panvk: Don't clobber registers if the render pass was suspended
|
||
- pan/decode: Fix the blend_count mask
|
||
- panvk/csf: Don't free the resources twice when init_render_desc_ringbuf() fails
|
||
- panvk: Initialize device virtual address space after the VM creation
|
||
|
||
Brad Smith (1):
|
||
|
||
- util: Support elf_aux_info() on OpenBSD arm and ppc
|
||
|
||
Brian Paul (2):
|
||
|
||
- svga: add svga_resource_create_with_modifiers() function
|
||
- svga: fix printing 64-bit value for 32-bit build
|
||
|
||
Caio Oliveira (90):
|
||
|
||
- intel/executor: Fix exec_size in \@read macro for Xe2
|
||
- intel/brw: Add test for combining SWSB dependencies in SENDs
|
||
- intel/brw: Allow extra SWSB encodings for Xe2
|
||
- intel/common: Properly dispose resources in mi_builder tests
|
||
- intel/common: Prepare mi_builder tests to support Xe KMD
|
||
- intel/common: Implement Xe KMD in mi_builder tests
|
||
- intel/common: Enable mi_builder test for PTL
|
||
- intel/brw: Add SHADER_OPCODE_BALLOT
|
||
- intel/brw: Add SHADER_OPCODE_QUAD_SWAP
|
||
- intel/brw: Omit type and region in payload sources when printing IR
|
||
- intel/brw: Use <V,W,H> notation for FIXED_GRF and ARF source when printing IR
|
||
- intel/executor: Enable PTL
|
||
- intel/brw: Fix decoding of cond_modifier and saturate in EU validation
|
||
- intel/brw: Fix SWSB output when printing IR
|
||
- intel/brw: Dump IR after lower scoreboard pass
|
||
- util/ra: Remove unimplemented function declaration
|
||
- intel/brw: Add is_control_source for the new subgroup ops
|
||
- mr-label-maker: Rules for intel/executor
|
||
- intel/brw: Enable EU validation and compaction tests for PTL
|
||
- intel/brw: Dump errors when brw_assemble() fails EU validation
|
||
- intel/compiler: Use #pragma once instead of header guards
|
||
- intel/brw: Remove overloads for brw_print_instruction/s functions
|
||
- intel/brw: Consider if SEND is gather variant when setting ex_desc
|
||
- intel/brw: Add TGL_PIPE_SCALAR value
|
||
- intel/brw: Add assembly support for ARF scalar register
|
||
- intel/brw: Add validation for ARF scalar register
|
||
- intel/executor: Add example using scalar register and send gather
|
||
- intel/brw: Skip some regioning EU validation for Vx1 and VxH modes
|
||
- intel/brw: Extract format enum in EU validation code
|
||
- intel/brw: Add validation for some Xe2 register regioning restrictions
|
||
- intel/brw: Add some tests for new Xe2 register regioning restrictions
|
||
- intel/brw: Add SHADER_OPCODE_READ_FROM_CHANNEL and LIVE_CHANNEL
|
||
- intel/brw: Disallow cmod in some cases of ARF scalar as destination
|
||
- intel/brw: Use variable instead of manually count the passes
|
||
- intel/brw: Rename brw_inst.h to brw_eu_inst.h
|
||
- intel/brw: Rename brw_inst to brw_eu_inst
|
||
- intel/brw: Rename brw_compact_inst to brw_eu_compact_inst
|
||
- intel/brw: Rename brw_inst_bits/set_bits to brw_eu_inst_bits/set_bits
|
||
- intel/brw: Rename brw_inst_* helpers to brw_eu_inst_*
|
||
- intel/brw: Rename brw_compact_inst_* helpers to brw_eu_compact_inst_*
|
||
- intel/brw: Gather brw_reg related implementations in brw_reg.cpp
|
||
- intel/brw: Add missing call to invalidate analysis
|
||
- intel/brw: Move two NIR passes to brw_nir.c
|
||
- gallium/meson: Ensure all needed sym_config are set.
|
||
- intel/brw: Remove 'fs' prefix from passes filenames
|
||
- intel/brw: Remove 'fs' prefix from passes and related functions
|
||
- intel/brw: Add missing bits in 3-src SWSB encoding for Xe2+
|
||
- intel/brw/xe2+: Do not use $.dst or $.src SWSB annotations in SENDs
|
||
- intel/compiler: Use INFINITY spill cost to represent no_spill
|
||
- util: Add operator new[] to linear context helper declarations
|
||
- intel/compiler: Use linear allocator for ACP trees in copy-prop
|
||
- intel/brw: Remove uses of VLAs
|
||
- intel/elk: Add ELK_MAX_MRF_ALL for static allocating arrays
|
||
- intel/elk: Remove uses of VLAs
|
||
- intel/elk: Fix typo in assertion
|
||
- util/ra: Move less used data out of ra_node
|
||
- util/ra: Don't store a pointer to graph per ra_node
|
||
- util/ra: Bump the initial size of adjacency lists
|
||
- util/ra: Don't store a pointer to a ra_regs per ra_reg
|
||
- intel/brw: Rename brw_fs_validate to brw_validate
|
||
- docs: Update syntax on Performance tips page
|
||
- intel/brw: Rename brw_fs_generator.cpp to brw_generator.cpp
|
||
- intel/brw: Add brw_generator.h header
|
||
- intel/brw: Rename fs_generator to brw_generator
|
||
- intel/brw: Add missing cases to flags_written()
|
||
- intel/brw: Remove extra wrapping around fs_visitor in tests
|
||
- intel/brw: Rename brw_fs_builder.h to brw_builder.h
|
||
- intel/brw: Rename fs_builder to brw_builder
|
||
- intel/brw: Stop using namespace for brw_builder
|
||
- intel/brw: Move a few builder helpers to brw_builder.h/cpp
|
||
- intel/brw: Move shuffle_from_32bit_read implementation to brw_builder
|
||
- intel/brw: Apply conventions to lower_src_modifiers helper
|
||
- intel/brw: Rename brw_fs_reg_allocate.cpp to brw_reg_allocate.cpp
|
||
- intel/brw: Remove 'fs' prefix from reg alloc code
|
||
- intel/brw: Rely on existing helper for dispatch width of geometry stages
|
||
- intel/elk: Fix wrong destination to memset
|
||
- intel/brw: Use brw prefix for some schedule instructions identifiers
|
||
- intel/brw: Use brw prefix instead of namespace in dynamic_msaa_flags()
|
||
- intel/brw: Remove unused enum
|
||
- intel/executor: Fix typo when copying result into Lua table
|
||
- intel/tools: Use idep_libintel_common in meson
|
||
- intel/tools: Add helpers for decoder_init/disasm
|
||
- intel/tools: Merge libaub into libintel_tools
|
||
- intel: Add meson option -Dintel-elk
|
||
- intel/brw: Add scoreboard support for scalar register
|
||
- intel/brw: Plumb through generator whether SEND is gather variant
|
||
- intel/brw: Add SHADER_OPCODE_SEND_GATHER
|
||
- intel/brw: Add lowering for SHADER_OPCODE_SEND_GATHER
|
||
- intel/brw: Use SHADER_OPCODE_SEND_GATHER in Xe3
|
||
- intel/brw: Fallback to SEND from SEND_GATHER if possible
|
||
|
||
Caleb Callaway (2):
|
||
|
||
- docs: Intel GPU performance tips
|
||
- docs: clarify ASPM performance tips
|
||
|
||
Casey Bowman (1):
|
||
|
||
- vulkan/screenshot-layer: Add region command option
|
||
|
||
Caterina Shablia (9):
|
||
|
||
- pan/bi: fix a typo
|
||
- pan/va: fix WMASK packing
|
||
- pan/bi: handle read_invocation
|
||
- pan/bi: handle ballot, ballot_relaxed and as_uniform
|
||
- pan/bi: lower some subgroup intrinsics
|
||
- pan/bi: lower the rest of subgroup ops using nir_lower_subgroups
|
||
- pan/bi: add a MEMORY_BARRIER pseudo-instruction
|
||
- pan/bi: handle barriers with SUBGROUP scope
|
||
- panvk: enable subgroupSizeControl
|
||
|
||
Chen, Phoebe (1):
|
||
|
||
- amd/vpelib: Refactor YUV format check
|
||
|
||
Chia-I Wu (69):
|
||
|
||
- panvk: ensure res table is restored after meta
|
||
- panvk: add memory mmap/munmap helpers
|
||
- panvk: do not leak mapped memory
|
||
- panvk: update CI expectations
|
||
- panvk: add get_subqueue_stages
|
||
- panvk: rework collect_cache_flush_info
|
||
- panvk: rework collect_cs_deps
|
||
- panvk: always skip frag->tiler subqueue wait
|
||
- panvk: skip frag subqueue self-wait within a render pass
|
||
- panvk: skip tiler subqueue self-wait within a render pass
|
||
- panvk: improve should_split_render_pass
|
||
- panvk: fix a missing cache invalidation
|
||
- panvk: update expectations for G610
|
||
- vulkan: include host write in expanded dst access flags
|
||
- panvk: add normalize_dependency
|
||
- panvk: improve VK_QUEUE_FAMILY_EXTERNAL support
|
||
- panvk: add support for VK_EXT_queue_family_foreign
|
||
- panvk: fix base_workgroup_id sysval
|
||
- ci: update the comment on MESA_VK_ABORT_ON_DEVICE_LOSS
|
||
- panvk: report queue lost timely when PANVK_DEBUG=sync
|
||
- panvk: implement check_status on v10+
|
||
- panvk: no need to map IB internally on valhall
|
||
- panvk: clang-format issue_fragment_jobs
|
||
- panvk: fix frag_completed for layered rendering
|
||
- panvk: minor clean up to prepare_blend
|
||
- panvk: fix dirty check for prepare_blend
|
||
- panvk: expand top-of-pipe and bottom-of-pipe
|
||
- panvk: use u_foreach_bit to loop over mask bits
|
||
- panvk: fix vs image support
|
||
- panvk: add panvk_queue_submit_init
|
||
- panvk: add panvk_queue_submit_init_storage
|
||
- panvk: add panvk_queue_submit_init_waits
|
||
- panvk: add panvk_queue_submit_init_cmdbufs
|
||
- panvk: add panvk_queue_submit_init_signals
|
||
- panvk: add panvk_queue_submit_ioctl
|
||
- panvk: add panvk_queue_submit_process_signals
|
||
- panvk: add panvk_queue_submit_process_debug
|
||
- panvk: clean up panvk_queue_submit
|
||
- panvk: move pandecode_next_frame a bit earlier
|
||
- panvk/csf: fix SIMULTANEOUS_USE gpu faults
|
||
- panvk/csf: fix subqueue ctx memory pool
|
||
- panvk: use cs_tracing_ctx::enabled for exception handler
|
||
- panvk: add u_trace_context to panvk_device
|
||
- panvk: define cmdbuf begin/end tracepoints
|
||
- panvk/csf: add CS_REG_SCRATCH_COUNT
|
||
- panvk/csf: add u_trace to panvk_cmd_buffer
|
||
- panvk/csf: add vk_sync to panvk_queue
|
||
- panvk/csf: flush and process trace events for one-time cmdbufs
|
||
- panvk/csf: flush and process trace events for all cmdbufs
|
||
- panvk: improve C++ compat for perfetto
|
||
- panvk: add u_trace perfetto support
|
||
- panvk: silence a perfetto init warning
|
||
- vulkan: add vk_device_get_timestamp
|
||
- vulkan: add common GetPhysicalDeviceCalibrateableTimeDomainsKHR
|
||
- vulkan: add common GetCalibratedTimestampsKHR
|
||
- anv: use common calibrated timestamp support partially
|
||
- hasvk: use common calibrated timestamp support
|
||
- radv: use common calibrated timestamp support
|
||
- tu: use common calibrated timestamp support
|
||
- nvk: use common calibrated timestamp support
|
||
- hk: remove calibrated timestamp support
|
||
- panvk: no need to zero availability on query create
|
||
- panvk: no need to check query count on query create
|
||
- panvk: no need to zero results on query reset
|
||
- panvk/csf: no need to sb wait on query begin
|
||
- panvk/csf: no need to sb wait on query end
|
||
- panvk/csf: no need to sb wait on query copy
|
||
- panvk/csf: no need to flush caches after query copy
|
||
- panvk/csf: add a comment on query synchronization
|
||
|
||
Christian Gmeiner (20):
|
||
|
||
- broadcom/common: Make v3d_device_info.h usable for C++
|
||
- v3d: Move v3d_ioctl(..) to src/broadcom/common
|
||
- v3dv: Switch to v3d_ioctl(..)
|
||
- v3d: Move v3d_X(..) to src/broadcom/common
|
||
- v3dv: Switch to v3d_X(..)
|
||
- broadcom: Add perfcount library
|
||
- v3d: Switch to use libbroadcom_perfcntrs
|
||
- v3dv: Switch to use libbroadcom_perfcntr
|
||
- etnaviv: blt: Add DBG(..) why blt usage was not possible
|
||
- etnaviv: rs: Add DBG(..) why blt usage was not possible
|
||
- v3d: Sync v3d_drm.h with drm-misc-next
|
||
- broadcom: Add perfetto data source
|
||
- pps: Add support for v3d ds
|
||
- perfetto: Add v3d data sources to system.cfg
|
||
- perfetto: Add v3d data sources to gpu.cfg
|
||
- docs: Update perfetto with the latest status
|
||
- etnaviv: isa: Support src2 for texld
|
||
- etnaviv: isa: Support src2 for texldb and texldl
|
||
- egl/meson: Specify which symbols to export
|
||
- v3dv: Add some CPU tracepoints
|
||
|
||
Christopher Michael (5):
|
||
|
||
- v3d: Add check to see if v3d supports cpu_queue
|
||
- v3d: Add check to see if v3d supports multisync
|
||
- v3d: Add support for timestamp queries
|
||
- v3d: Add support for time elapsed queries
|
||
- v3d: Add support for PIPE_QUERY_TIMESTAMP_DISJOINT
|
||
|
||
Collabora's Gfx CI Team (5):
|
||
|
||
- Uprev Piglit to eebe1b555f51dbb702f696d08ad5ae8153bcdcdd
|
||
- Uprev Piglit to d04d6fff00849a2a8e29ef3251c6ca04a2f68dc7
|
||
- Uprev Piglit to 468221c722481c470e6a23760b914c33143c2af6
|
||
- Uprev Piglit to 4c0fd15fd956ec70c5509bedee219d602b334464
|
||
- Uprev Piglit to 631b72944f56e688f56a08d26c8a9f3988801a08
|
||
|
||
Connor Abbott (55):
|
||
|
||
- vulkan/runtime: Add driver callbacks for BVH building
|
||
- vulkan/runtime,radv: Add shared BVH building framework
|
||
- vulkan/runtime,radv: Add shared BVH building framework
|
||
- ir3: Fix reload_live_out() in shared RA
|
||
- tu: Add Vulkan 1.4 features and properties
|
||
- tu: Expose Vulkan 1.4 on a7xx
|
||
- tu: Move queue-related code to a new file
|
||
- tu: Refactor the submit path
|
||
- tu/kgsl: Make wait_timestamp_safe() return VkResult
|
||
- tu/knl: Move u_trace fence handling to generic code
|
||
- tu: Rename bo_list to submit_bo_list
|
||
- util/dynarray: Add macro for appending an array
|
||
- tu: Make userspace RD dump generic
|
||
- freedreno/fdl: Make tiled r8g8 images have 4k alignment
|
||
- tu: Re-enable tiled non-ubwc R8G8 images
|
||
- freedreno/fdl: Fix 3d mipmapping height alignment
|
||
- freedreno/fdl, tu: Make mutable part of the image layout
|
||
- freedreno/fdl: Don't enable r8g8 special case for mutable images
|
||
- freedreno/fdl, tu: Allow swaps with mutable tiled images
|
||
- tu: Allow UBWC with images with swapped formats.
|
||
- vk/bvh: Fix clang build error with turnip
|
||
- ir3: Allow collect sources to be undef
|
||
- ir3: Support assembling/disassembling ray_intersection and resbase
|
||
- ir3: Plumb through two-dimensional UAV loads
|
||
- ir3: Plumb through ray_intersection intrinsic
|
||
- tu: Implement cmd_fill_buffer_addr internal function
|
||
- tu: Implement buffer_write_cp
|
||
- freedreno: CP_SCRATCH_WRITE exists on a7xx too
|
||
- freedreno: Add new a7xx CP_REG_RMW and CP_REG_TO_SCRATCH fields
|
||
- freedreno/a7xx: Document partial workgroup register
|
||
- tu: Stop emitting HLSQ_CS_KERNEL_GROUP_*
|
||
- tu/a7xx: Emit HLSQ_CS_LAST_LOCAL_SIZE dynamically
|
||
- tu: Implement unaligned dispatches
|
||
- tu: Add common define for maxTexelBufferElements
|
||
- tu: Create meta device
|
||
- freedreno: Introduce ray tracing features
|
||
- tu/kgsl: Bump uapi header
|
||
- tu: Plumb through raytracing fuse
|
||
- tu: Move fd_dev_info() before name generation
|
||
- tu: Display when raytracing is disabled in device string
|
||
- tu: Support VK_KHR_acceleration_structure
|
||
- tu: Support VK_KHR_ray_query
|
||
- tu: Expose VK_KHR_ray_tracing_maintenance1
|
||
- tu, ir3: Implement a750 RT workaround
|
||
- ir3: Use nir_split_struct_vars for temporaries
|
||
- vk/bvh: Add default stubs for unsupported entrypoints
|
||
- anv: Delete acceleration structure stubs
|
||
- radv: Delete acceleration structure stubs
|
||
- tu: Use image view format for sysmem resolves
|
||
- tu: Handle non-identity GMEM swaps when resolving
|
||
- tu: Handle non-identity GMEM swaps for input attachments
|
||
- tu, freedreno: Write PC_DGEN_SU_CONSERVATIVE_RAS_CNTL
|
||
- tu: Stop setting binning fields on a7xx
|
||
- tu: Support VK_EXT_conservative_rasterization on a7xx
|
||
- tu: Add missing assignment to shared_viewport
|
||
|
||
Constantine Shablia (23):
|
||
|
||
- panvk: move samplerAnisotropy in the order it appears in struct definition
|
||
- panvk: enable shaderInt64
|
||
- panvk: elaborate the comment on the maxMemoryAllocationCount limit
|
||
- panvk: adjust maxSamplerAllocationCount limit
|
||
- nir: introduce instance_index system value
|
||
- nir: lower INSTANCE_{ID,INDEX} to an offset load_instance_{index,id} respectively
|
||
- Revert "nir: lower INSTANCE_{ID,INDEX} to an offset load_instance_{index,id} respectively"
|
||
- Revert "nir: introduce instance_index system value"
|
||
- panvk: replace vkGetBufferMemoryRequirements2 with vkGetDeviceBufferMemoryRequirements
|
||
- panvk: never prefer or require dedicated allocation for buffers
|
||
- panvk: never require dedicated allocation for images
|
||
- panvk: add panvk_image_init helper
|
||
- panvk: implement vkGetDeviceImageMemoryRequirements
|
||
- panvk: enable shaderInt8, VK_KHR_8bit_storage and VK_KHR_shader_float16_int8
|
||
- pan/util: sort files in meson.build
|
||
- panvk: order KHR extension enables alphabetically
|
||
- panvk/csf: use gfx_state_set_dirty instead of touching state directly
|
||
- pan,nir: introduce load_attribute_pan
|
||
- pan/bi: handle load_attribute_pan
|
||
- panvk: Fix base_{instance,vertex} handling
|
||
- panvk: lower drawid to zero
|
||
- panvk: enable shaderDrawParameters
|
||
- panvk: enable drawIndirectFirstInstance
|
||
|
||
Corentin Noël (6):
|
||
|
||
- virgl: Propagate the GL_MAX_stage_SHADER_STORAGE_BLOCKS for each stage
|
||
- virgl: Simply loop over the resources to figure-out if it is already added
|
||
- virgl: Update virgl_hw.h from virglrenderer
|
||
- virgl: Use MAX_SAMPLERS instead of MAX_SHADER_SAMPLER_VIEWS
|
||
- virgl/ci: Remove screen size arguments
|
||
- virgl/ci: Re-enable virgl-traces
|
||
|
||
Daniel Schürmann (49):
|
||
|
||
- aco/ra: set Pseudo_instruction::scratch_sgpr to SCC if it doesn't need to be preserved
|
||
- aco/ra: use bitset for sgpr_operands_alias_defs
|
||
- aco/ra: explicitly assign scratch SGPR for linear phis
|
||
- aco: remove Pseudo_instruction::tmp_in_scc
|
||
- aco/insert_NOPs: implement vector-based RegCounterMap as replacement for VGPRCounterMap
|
||
- aco/insert_NOPs: use RegCounterMap as replacement for the CounterMap implementation
|
||
- aco/insert_NOPs: add early exit to handle_valu_partial_forwarding_hazard_instr
|
||
- aco/print_asm: allow for empty blocks with arbitrary offsets
|
||
- aco/assembler: constify assembly functions
|
||
- aco/assembler: Actually insert s_inst_prefetch instructions when aligning blocks for loops
|
||
- aco/assembler: change ctx.loop_header to uint32_t instead of Block*
|
||
- aco/assembler: chain branches instead of emitting long jumps
|
||
- aco: remove definition from SOPP branch instructions
|
||
- aco: remove definition from Pseudo branch instructions
|
||
- aco/assembler: Don't emit target basic block index when chaining branches
|
||
- aco/print_ir: don't print disconnected empty blocks
|
||
- aco/optimizer_postRA: set branch()->never_taken if exec is constant non-zero
|
||
- aco: move try_optimize_branching_sequence() to postRA optimizations
|
||
- aco/jump_threading: remove branch sequence optimization
|
||
- aco: move branch lowering optimization into separate file 'aco_lower_branches.cpp'
|
||
- aco/lower_branches: remove edges between blocks if there is no direct branch
|
||
- ac/lower_ngg: Fix collecting buffer offsets from 4 lanes on gfx12
|
||
- ac/lower_ngg: move break blocks after loop in streamout code generation for gfx12/ACO
|
||
- ac/lower_ngg: move readlane into break blocks in streamout code generation for gfx12/ACO
|
||
- nir/divergence: change nir_has_divergent_loop() to return true only for divergent breaks
|
||
- aco/jump_threading: don't remove loop preheaders
|
||
- aco/assembler: Find loop exits using the successor's loop nest depth
|
||
- aco: consider s_cbranch_exec* instructions in needs_exec_mask()
|
||
- aco/lower_branches: do eliminate_useless_exec_writes_in_block() during branch lowering.
|
||
- aco/lower_branches: implement try_remove_simple_block() in lower_branches()
|
||
- aco: move try_merge_break_with_continue() to lower_branches()
|
||
- aco/lower_branches: allow for non-fallthrough loop exits in try_merge_break_with_continue()
|
||
- aco: delete aco_jump_threading.cpp
|
||
- aco/lower_branches: stitch linear blocks if there is exactly one successor with one predecessor
|
||
- nir/from_ssa: only consider divergence if requested
|
||
- Revert "nir: add nir_clear_divergence_info, use it in nir_opt_varyings"
|
||
- aco/insert_NOPs: refactor VALUReadSGPRHazard detection
|
||
- aco/insert_NOPs: implement VALU -> VALU case for VALUReadSGPRHazard on GFX12
|
||
- nir/loop_analyze: only iterate loop header phis in compute_induction_information()
|
||
- nir/loop_analyze: remove nir_loop_variable::in_if_branch and nir_loop_variable::in_nested_loop
|
||
- nir/loop_analyze: remove nir_loop_variable::in_loop
|
||
- nir/loop_analyze: directly record induction variables into nir_loop_info
|
||
- nir/loop_analyze: don't initialize nir_loop_variable separately
|
||
- nir/loop_analyze: replace nir_loop_variable array with hash table
|
||
- nir/loop_analyze: insert only induction vars into hash map
|
||
- nir/loop_analyze: ignore terminating induction variable in guess_loop_limit()
|
||
- nir/loop_analyze: re-use the same nir_loop_variable struct before and after the increment
|
||
- nir/loop_analyze: store nir_loop_induction_variable hash table in loop_info
|
||
- nir/loop_analyze: stack-allocate loop_info_state
|
||
|
||
Daniel Stone (22):
|
||
|
||
- ci: Don't run Meson tests in critical-path jobs
|
||
- ci: Slash ASan and UBSan build coverage
|
||
- ci: Give much more time to ASan and UBSan jobs
|
||
- ci: Let rootfs builds run for 2 hours (!)
|
||
- pipe_loader: Fix pipe_i915 with the dynamic loader
|
||
- ci: Disable Werror on wrapped subprojects
|
||
- ci: Remove obsolete compiler-wrapper
|
||
- ci: Move build containers above test containers
|
||
- ci/fedora: Install which into build image
|
||
- ci: Define LLVM_VERSION as a container property
|
||
- ci: Require LLVM_VERSION to be set explicitly
|
||
- ci/debian: Upgrade Debian images to LLVM 19
|
||
- ci: Fix dependency on lint job
|
||
- ci: Fix kernel section nesting
|
||
- ci: Move dEQP message into section
|
||
- ci: Pass build targets to dEQP CMake
|
||
- ci: Don't build Vulkan for GL dEQP
|
||
- ci: Trim down VVL external builds
|
||
- ci: Capture Ninja log
|
||
- ci: Only build Perfetto in build-test jobs
|
||
- ci: Only build what we use for testing jobs
|
||
- ci: Move r300/nine/nvk builds out of critical path
|
||
|
||
Danylo Piliaiev (31):
|
||
|
||
- ir3/parser: Print the line where parsing error occurred
|
||
- nir/nir_opt_offsets: Do not fold load/store with const offset > max
|
||
- freedreno/registers: Define Fragment Shading Rate registers
|
||
- ir3,tu: Add support for Fragment Shading Rate and plumb it into Turnip
|
||
- tu/a7xx: Implement VK_KHR_fragment_shading_rate
|
||
- ir3/parser: Add fullnop and fullsync sections for debugging
|
||
- tu: Enable UBWC for 3D images without mipmaps
|
||
- freedreno/fdl: Pass fd_dev_info to fdl6_layout
|
||
- tu,freedreno: Enable linear mipmap tail for UBWC images
|
||
- tu: Disable fragmentShadingRateWithShaderSampleMask due to issues
|
||
- tu,ir3: Add workaround for reading shading rate on A7XX gen1,gen2
|
||
- tu: Handle cmdbuf and rp_blit flags of TU_DEBUG_STALE_REGS_FLAGS
|
||
- tu/perfetto: Always emit submission event and time it
|
||
- tu/perfetto: Add app and engine names to the command buffer tracepoint
|
||
- ir3: Make allocation of consts more generic and order independent
|
||
- ir3: Use generic consts alloc for driver params
|
||
- tu,ir3: Make push consts be able to start from higher than c0.x offsets
|
||
- ir3: Use generic const alloc for everything and call it once
|
||
- tu: Allocate consts for driver params as early as possible
|
||
- tu: Do not re-calculate static blend LRZ state
|
||
- freedreno/regs: Set correct shr for GRAS_LRZ_BUFFER_PITCH.ARRAY_PITCH
|
||
- tu: Fix LRZ for arrayed depth
|
||
- tu: Handle 8x MSAA for LRZ
|
||
- freedreno,tu: Unify LRZ layout calculations
|
||
- tu: Track at which draw call LRZ is disabled
|
||
- tu: Do not disable LRZ for whole RP if it is disabled in RP
|
||
- ir3: Consider const alloc alignment in free space size calcs
|
||
- tu: Fix stale A7XX_GRAS_LRZ_CNTL2 in 3d blits or !valid lrz case
|
||
- tu/a7xx: Always have depth/stencil in corresponding resolve groups
|
||
- tu: Get correct src view when storing gmem attachment
|
||
- tu: Handle mismatched mutability when resolving from GMEM
|
||
|
||
Dave Airlie (9):
|
||
|
||
- nir/functions: force inlining for barriers.
|
||
- v3dv: report correct error on failure to probe
|
||
- venus: handle device probing properly.
|
||
- vulkan: update to 302 headers for av1 encode
|
||
- lavapipe: fix beta build due to changes in AMDX ext
|
||
- radv/video: set max slice counts to 1 for h264/5 encode
|
||
- anv: add default av1 tables from media-driver
|
||
- genxml: add av1 fields
|
||
- anv: add initial support for AV1 decoding
|
||
|
||
David (Ming Qiang) Wu (3):
|
||
|
||
- frontends/va: adding PIPE_FORMAT_P012
|
||
- frontends/va: add PIPE_VIDEO_PROFILE_AV1_PROFILE2
|
||
- radeonsi/vcn: support 12bit YUV420 AV1 decoding
|
||
|
||
David Heidelberg (14):
|
||
|
||
- util: Drop 3Dnow optimisation leftovers
|
||
- util: Remove MMX/MMXext detection code
|
||
- util: Drop ancient Intel CPU detection
|
||
- util: drop XOP detection code
|
||
- llvmpipe: align with u_cpu_detect struct changes
|
||
- compiler/rust: drop duplicated bindgen check
|
||
- ci/freedreno: update Adreno 306 expectations
|
||
- ci/freedreno: increase Adreno 618 timeout to 1h
|
||
- docs: remove deprecated component list and licenses
|
||
- docs: Clarify project name and include Mesa3D
|
||
- docs: move license(s) to licenses directory
|
||
- c11: use SPDX-License-Identifier header
|
||
- licenses: add missing licenses
|
||
- drm-uapi: update licenses statement
|
||
|
||
David Rosca (148):
|
||
|
||
- radeonsi/vcn: Fix coding AV1 render size
|
||
- frontends/va: Add minus_1 to AV1 render_width/height
|
||
- gallium: Add PIPE_VIDEO_CAP_SKIP_CLEAR_SURFACE
|
||
- frontends/va: Support skip clear on surface creation
|
||
- frontends/vdpau: Support skip clear on surface creation
|
||
- radeonsi: Support PIPE_VIDEO_CAP_SKIP_CLEAR_SURFACE
|
||
- radeonsi/vcn: Stop clearing decode internal buffers
|
||
- radv/video: Fix H264 slice control
|
||
- radv/video: Fix HEVC slice control
|
||
- radv/video: Report correct encodeInputPictureGranularity
|
||
- radv/video: Avoid selecting rc layer over maximum
|
||
- radv/video: Use 64x16 alignment for HEVC encode
|
||
- radv/video: Override pic_init_qp_minus26 in PPS
|
||
- radeonsi/vcn: Use correct frame context buffer for preencode on VCN5
|
||
- radeonsi: Check all supported formats in si_vid_is_target_buffer_supported
|
||
- frontends/va: Create surfaces with correct fourcc for RT format
|
||
- frontends/va: Stop reallocating to prefered format in EndPicture
|
||
- frontends/va: Stop reallocating from progressive to interlaced in EndPicture
|
||
- frontends/va: Stop reallocating buffers for protected playback
|
||
- frontends/va: Stop reallocating according to JPEG sampling factor
|
||
- frontends/va: Check if target buffer is supported in EndPicture
|
||
- frontends/va: Stop reallocating buffers in EndPicture
|
||
- frontends/va: Use compositor blit with different number of planes
|
||
- frontends/va: Only use interlaced surfaces when progressive is not supported
|
||
- pipe: Remove video update_decoder_target
|
||
- radeonsi/vpe: Set correct surface swizzle mode
|
||
- radeonsi/vpe: Don't allow DCC surfaces
|
||
- frontends/va: Return correct pixel formats in surface attributes query
|
||
- frontends/va: Change default fourcc for RGB 10bit to X2R10G10B10
|
||
- gallium/vl: Implement rendering to 3-plane YUV formats
|
||
- gallium/vl: Don't support planar RGB as video format
|
||
- frontends/va: Enable 3-plane YUV formats as postproc output
|
||
- radeonsi/vcn: Support tiling for JPEG decode
|
||
- radv/video: Fix IB signature checksum
|
||
- radv/video: Always use setup reference slot when valid
|
||
- ac/surface: Add RADEON_SURF_VIDEO_REFERENCE
|
||
- radeonsi: Support PIPE_BIND_VIDEO_DECODE/ENCODE_DPB
|
||
- radeonsi/vcn: Create decode DPB surfaces with PIPE_BIND_VIDEO_DECODE_DPB
|
||
- radeonsi/vcn: Create encode DPB surfaces with PIPE_BIND_VIDEO_ENCODE_DPB
|
||
- frontends/va: Add support for VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME_3
|
||
- frontends/va: Store picture type for buffers in encode DPB
|
||
- radeonsi/vcn: Don't allow encoding H264 B-frame references
|
||
- frontends/va: Move mjpeg sampling_factor to pipe_mjpeg_picture_desc
|
||
- radeonsi/vcn: Remove code handling buffer_get_virtual_address failure
|
||
- radeonsi/vcn: Unmap bitstream buffer in radeon_dec_destroy
|
||
- radeonsi/vcn: Gracefully handle decode errors and report to frontend
|
||
- radeonsi/vcn: Make sure JPEG target buffer format matches sampling factor
|
||
- radeonsi/vcn: Cleanup JPEG supported formats
|
||
- radeonsi/vpe: Silence expected errors with unsupported output format
|
||
- gallium/vl: Add plane order for Y8_400 format
|
||
- gallium/vl: Fix plane order for IYUV format
|
||
- frontends/va: Stop converting formats in Put/GetImage
|
||
- radeonsi: Update minimum supported encode size for VCN5
|
||
- radeonsi/vcn: Align bitstream buffer to 128 when resizing
|
||
- radeonsi/uvd: Align bitstream buffer to 128 when resizing
|
||
- radeonsi/vcn: Enable write combine for decode
|
||
- radeonsi/vcn: Don't keep last fence
|
||
- radeonsi/vcn: Use local variable for destory fence
|
||
- pipe: Remove PIPE_DEFAULT_DECODER_FEEDBACK_TIMEOUT_NS
|
||
- frontends/va: Get AV1 decode subsampling_x/y
|
||
- radeonsi/vcn: Return error when decoding 12bit VP9 and 4:2:2/4:4:4 AV1
|
||
- frontends/va: Fix decoding VC1 interlaced video
|
||
- frontends/va: Don't allow Render/EndPicture without BeginPicture
|
||
- frontends/va: Don't allow EndPicture without calling driver begin_frame
|
||
- ac/parse_ib: Parse VCN IB_COMMON_OP_WRITEMEMORY
|
||
- radv/amdgpu: Set VCN version for ac_parse_ib
|
||
- frontends/va: Fix deinterlace filter
|
||
- radeonsi/vcn: Change required FW version for rc_per_pic_ex on VCN3
|
||
- radv/video: Fix DPB tier2 surface params
|
||
- radv/video: Use correct array index for decode target and DPB images
|
||
- radv/video: Remove dt_field_mode handling code
|
||
- radv: Fix sampling from image layers of video decode target
|
||
- ac/surface: Don't force linear for VIDEO_REFERENCE with emulated image opcodes
|
||
- frontends/va: Get buffer feedback with locked mutex in MapBuffer
|
||
- radeonsi/vcn: Use compute only context
|
||
- gallium/vl: Fix unbinding sampler views
|
||
- gallium/vl: Create sampler state also when gfx is not supported
|
||
- gallium/vl: Add rgba compute shader
|
||
- gallium/vl: Add param to create compute only vl_compositor
|
||
- gallium: Add param to create compute only multimedia context
|
||
- frontends/va: Use compute only context if driver prefers compute
|
||
- radeonsi/vcn: Fix crash when failing to allocate internal buffers
|
||
- frontends/va: Only report surface alignment when non-zero
|
||
- frontends/va: Allow creating DRM PRIME surfaces without surface descriptor
|
||
- frontends/va: Set csc matrix in PutSurface
|
||
- gallium/vl: Fix creating buffers with auxiliary planes
|
||
- radeonsi: Add radeon_bitstream and use it in radeon_vcn_enc
|
||
- radeonsi/vce: Remove support for FW 50 and older
|
||
- radeonsi/vce: Set more header params
|
||
- radeonsi/vce: Move dual pipe context to offset 0 of CPB
|
||
- radeonsi/vce: Use app DPB management
|
||
- radeonsi/vce: Support slice encoding
|
||
- radeonsi/vce: Support VBAQ
|
||
- radeonsi/vce: Support quality presets
|
||
- radeonsi/vce: Support min/max QP and max frame size
|
||
- radeonsi/vce: Support intra refresh
|
||
- radeonsi/vce: Support raw packed headers
|
||
- radeonsi/vce: Set input pic swizzle mode on GFX9
|
||
- radeonsi/vce: Cleanup
|
||
- radeonsi/uvd: Stop clearing decode internal buffers
|
||
- radeonsi/uvd: Optimize bitstream buffer resizing
|
||
- radeonsi/uvd: Set decode target swizzle mode on GFX9
|
||
- radeonsi/uvd_enc: Rework DPB allocation
|
||
- radeonsi/uvd_enc: Use app DPB management
|
||
- radeonsi/uvd_enc: Consider input surface size for padding
|
||
- radeonsi/uvd_enc: Support Pre-Encode
|
||
- radeonsi/uvd_enc: Support VBAQ
|
||
- radeonsi/uvd_enc: Support quality presets
|
||
- radeonsi/uvd_enc: Support slice encoding
|
||
- radeonsi/uvd_enc: Support intra refresh
|
||
- radeonsi/uvd_enc: Support temporal layer rate control
|
||
- radeonsi/uvd_enc: Support min/max QP and max frame size
|
||
- radeonsi/uvd_enc: Support dynamic rate control changes
|
||
- radeonsi/uvd_enc: Support raw packed headers
|
||
- radeonsi/uvd_enc: Set input pic swizzle mode on GFX9
|
||
- radeonsi: Enable implemented VCE/UVD encode features
|
||
- gallium/vl: Fix sampler view components for Y8_400 format
|
||
- gallium/vl: Add vl compositor layer mirror
|
||
- gallium/vl: Clear remaining planes in YUV conversion
|
||
- gallium/vl: Use matrix for scale and crop in cs compositor
|
||
- gallium/vl: Implement rotation and mirror in cs compositor
|
||
- frontends/va: Simplify format check in PutSurface
|
||
- frontends/va: Disable color conversion for luma-only source formats
|
||
- frontends/va: Stop using util_compute_blit
|
||
- frontends/va: Refactor vlVaPostProcCompositor to be usable outside processing
|
||
- frontends/va: Support rotation and mirror for processing
|
||
- frontends/va: Implement format conversions in PutImage/GetImage
|
||
- gallium/auxiliary: Remove util_compute_blit
|
||
- radeonsi: Fix reporting support for AV1 Profile2
|
||
- radeonsi/vcn: Fix AV1 coded size for VCN 5.0
|
||
- radeonsi: Report surface alignment for AV1 encode
|
||
- gallium/vl: Add compute shader deinterlace filter
|
||
- frontends/va: Stop using extra context for deinterlacing
|
||
- frontends/va: Implement QuerySurfaceStatus as SyncSurface with 0 timeout
|
||
- frontends/va: Don't flush before resource_get_handle
|
||
- frontends/va: Remove vlVaBuffer derived_image_buffer
|
||
- frontends/va: Add surface pipe_fence for vl_compositor rendering
|
||
- gallium/vl: Don't flush in vl_compositor yuv_deint and rgb_to_yuv
|
||
- frontends/va: Add context mutex
|
||
- frontends/va: Unlock driver mutex for SyncSurface/Buffer fence wait
|
||
- frontends/va: Fix decoding VC1 streams with multiple slices
|
||
- ac/vcn_dec: Fix AV1 film grain on VCN5
|
||
- radeonsi/video: Avoid stream handle duplicates in PID namespace
|
||
- frontends/vdpau: Set H264 chroma_format_idc
|
||
- radeonsi/vcn: Set correct chroma format for H264 decode
|
||
- radeonsi/uvd: Set correct chroma format for H264 decode
|
||
- radv/video: Fix setting balanced preset for HEVC encode with SAO enabled
|
||
- radv/video: Move IB header from begin/end to encode_video
|
||
|
||
David Tobolik (2):
|
||
|
||
- rusticl/style: use Arc::clone instead of .clone()
|
||
- rusticl/style: add util for conversion with err
|
||
|
||
Deborah Brouwer (36):
|
||
|
||
- freedreno/ci: add prefix for a630-vk-asan tests
|
||
- ci: Remove duplicate slash before $RESULTS_DIR
|
||
- ci/b2c: update RESULTS_DIR for .b2c-test jobs
|
||
- ci: add a tool to summarize a failed pipeline
|
||
- ci/pipeline_message: add unit tests for tool
|
||
- ci: move pipeline_summary tool to .marge/hooks
|
||
- ci: debian/x86_64_pyutils remove redundant rules
|
||
- ci: python-test rename artifacts
|
||
- ci: yaml-toml-shell-test: use pyutils container
|
||
- ci: separate python tests and artifacts
|
||
- ci: post gantt: use logging instead of print
|
||
- ci: add some static typing to the gantt scripts
|
||
- ci: make the gantt scripts available as modules
|
||
- ci: post gantt: add --marge-user-id option
|
||
- ci: post gantt: add --project-id option
|
||
- ci: post gantt: add pipeline-id to gantt filename
|
||
- ci: post gantt: ignore pipeline_summary message
|
||
- ci: gantt chart: include in-progress jobs
|
||
- ci: add --ci-timeout option for gantt scripts
|
||
- ci: add pytests for the gantt chart scripts
|
||
- ci: update token retrieval method for gantt charts
|
||
- ci: collapse yamllint and shellcheck sections
|
||
- ci: run-pytest.sh: allow script to run locally
|
||
- ci: add .flake8 linting to ci scripts and tests
|
||
- ci: update_traces_checksum: fix E501 line too long
|
||
- ci: update the pyutils container
|
||
- ci: stop using a venv for run-pytest.sh
|
||
- ci: set python version 3.11 for run-pytest.sh
|
||
- ci: pipeline_message: catch module loading errors
|
||
- ci: pipeline_message: improve job list formatting
|
||
- ci: pipeline_message: add test to parse error logs
|
||
- ci: pipeline_message: ignore \`error_type` errors
|
||
- ci: pipeline_message: ignore harmless build logs
|
||
- ci: pipeline_message: ignore \`generated` errors
|
||
- ci: pipeline_message: parse \`fatal` messages
|
||
- ci: pipeline_message: reset empty errors
|
||
|
||
Derek Foreman (3):
|
||
|
||
- vulkan/wsi/wayland: Fix time calculation
|
||
- vulkan/wsi/wayland: Avoid spurious discard event at startup
|
||
- vulkan/wsi/wayland: Move timing calculations to the swapchain
|
||
|
||
Detlev Casanova (3):
|
||
|
||
- ci/fluster/lava: Add fluster in LAVA rootfs
|
||
- ci/fluster: Add radeonsi-raven-vaapi-fluster jobs
|
||
- ci/deqp-runner: uprev from 0.20.2 to 0.20.3
|
||
|
||
Dylan Baker (25):
|
||
|
||
- VERSION: bump to 25.0
|
||
- docs: reset new_features.txt
|
||
- docs/release-calendar: update one more time for pushed back release
|
||
- docs: add release notes for 24.3.0
|
||
- docs/relnotes/24.3.0: Add SHA sums
|
||
- docs/release-calendar: remove 24.3 RC dates
|
||
- docs: Add calendar entries for 24.3 release.
|
||
- anv: advertise Vulkan 1.4
|
||
- anv: bump max number of push constants to 256
|
||
- anv: Add new Vulkan 1.4 features and properties
|
||
- anv: bump conformance version to 1.4
|
||
- maintainer-scripts: Bump Vulkan release version to 1.4
|
||
- docs: add release notes for 24.3.1
|
||
- docs: Add SHA sums for 24.3.1
|
||
- docs: update calendar for 24.3.1
|
||
- clc: Tell clang to track imported dependencies
|
||
- docs: add release notes for 24.3.2
|
||
- docs: Update checksums for 24.3.2
|
||
- docs: update calendar for 24.3.2
|
||
- docs/release-calendar: Move next release to January 2nd
|
||
- intel/tests: Fix coverity warning about possibly leaked memory
|
||
- intel/tests: Fix missing assignment of error condition
|
||
- docs: add release notes for 24.3.3
|
||
- docs: Add SHA sums to 24.3.3 release notes
|
||
- docs: update calendar for 24.3.3
|
||
|
||
Eric Engestrom (139):
|
||
|
||
- meson: bump spirv-tools version needed to v2022.1
|
||
- radeonsi/ci: add more flakes seen recently
|
||
- radv/ci: add more flakes seen recently
|
||
- broadcom/ci: add more flakes seen recently
|
||
- freedreno/ci: add more flakes seen recently
|
||
- ci: upgrade the fedora image from 38 to 41
|
||
- ci/build: drop "verify after bump to F39" as that did not help
|
||
- ci/build: add workaround for incorrect maybe-uninitialized error
|
||
- ci: move error handling functions at the end
|
||
- ci: use quiet alias for commands
|
||
- ci: make error handling quieter
|
||
- broadcom/ci: add flakes seen recently
|
||
- freedreno/ci: add flakes seen recently
|
||
- nvk+zink/ci: add flakes seen recently
|
||
- radv+zink/ci: add flakes seen recently
|
||
- ci: raise priority of release manager pipelines
|
||
- ci: reduce priority of nightly pipeline jobs from 50 to 45
|
||
- meson: move openmp block out of the middle of the x11 deps block
|
||
- meson: define only once the versions of the x11 deps
|
||
- radv/ci: document flakes seen recently
|
||
- broadcom/ci: document flakes seen recently
|
||
- nvk/ci: document flakes seen recently
|
||
- freedreno/ci: document flakes seen recently
|
||
- docs: update calendar for 24.2.7
|
||
- docs: add release notes for 24.2.7
|
||
- docs: add sha sum for 24.2.7
|
||
- turnip/ci: document regression
|
||
- ci/crosvm: remove noise inside deqp-runner output
|
||
- v3dv/ci: mark whole group as flaky
|
||
- docs: fix invalid expression in new pipe cap
|
||
- docs: fix invalid expression in teflon docs
|
||
- intel/ci: disable CML jobs because of networking issues
|
||
- intel/ci: add missing .intel-common-manual-rules to .{iris,crocus,i915g}-manual-rules
|
||
- ci/build: drop mold wrapper for \`ninja install`
|
||
- ci: drop override forcing ld to be gold (and forcing gold to be installed everywhere)
|
||
- ci: when installing mold, make its use automatic
|
||
- ci: bump image tags
|
||
- radeonsi/ci: drop two failures that are mysteriously fixed by using mold?
|
||
- ci/container: move deqp build section into the script itself
|
||
- ci/container: move apitrace build section into the script itself
|
||
- ci/container: move crosvm build section into the script itself
|
||
- ci/container: move deqp-runner build section into the script itself
|
||
- ci/container: move fossilize build section into the script itself
|
||
- ci/container: move gfxreconstruct build section into the script itself
|
||
- ci/container: move kdl build section into the script itself
|
||
- ci/container: move libclc build section into the script itself
|
||
- ci/container: move llvm-spirv build section into the script itself
|
||
- ci/container: move mold build section into the script itself
|
||
- ci/container: move ninetests build section into the script itself
|
||
- ci/container: move piglit build section into the script itself
|
||
- ci/container: move rust build section into the script itself
|
||
- ci/container: move vkd3d-proton build section into the script itself
|
||
- ci/container: move vulkan-validation build section into the script itself
|
||
- ci/container: move wayland build section into the script itself
|
||
- ci/container: add sections around the other build scripts
|
||
- ci/container: close debian_{setup,cleanup} sections
|
||
- ci/lava: add setup-test-env.sh to the rootfs
|
||
- ci/container: add section around strip-rootfs.sh
|
||
- ci: bump image tags
|
||
- zink+nvk/ci: fix deqp binary used for gles tests
|
||
- zink+radv/ci: fix deqp binary used for gles tests
|
||
- ci/deqp: move testlog-to-* tools to /deqp
|
||
- ci/deqp: only compress caselists when they exist
|
||
- ci/deqp: build testlog tools on android
|
||
- ci/deqp: fetch & checkout exactly the commit/tag/branch requested
|
||
- ci/deqp: avoid downloading 1.47 GiB multiple times
|
||
- ci/deqp: error out in case of invalid build API
|
||
- ci/deqp: build glcts in gles build, for gles*-khr tests
|
||
- ci/deqp: add build of \`main` branch
|
||
- ci/deqp: make sure the main commit is actually from the main branch
|
||
- ci/deqp: fully isolate deqp builds
|
||
- ci: bump image tags
|
||
- ci/container: setup sections in all image builds
|
||
- radv/ci: document regression of test_shader_sm66_is_helper_lane in 7469f99e...25b8f4f7
|
||
- meson: simplify logic a bit
|
||
- meson: drop unused variables
|
||
- meson: reuse variable
|
||
- meson/megadriver: s/_/-/ in an argument name to be consistent
|
||
- meson/megadriver: simplify setting common megadriver arguments
|
||
- meson/megadriver: support various lib suffixes
|
||
- ci/deqp: simplify paths since we are already in /deqp-$deqp_api/
|
||
- ci/deqp: fix the "is this a build on main?" check
|
||
- ci/deqp: support having commit backports and local patches for main too
|
||
- ci/deqp: simplify generating the version description file
|
||
- ci/deqp: mention the deqp api in the version string
|
||
- ci/deqp: only print the commit list header when the list is not empty
|
||
- ci/lava: turn the $BUILD_VK check into a proper if block
|
||
- ci/deqp: add a deqp-vk build on the \`main` branch
|
||
- ci: bump image tags
|
||
- radv/ci: use deqp-vk-main in radv jobs
|
||
- docs: update calendar for 24.2.8
|
||
- docs: add release notes for 24.2.8
|
||
- docs: add sha sum for 24.2.8
|
||
- ci/meson: make meson wrap fallback list more readable
|
||
- ci/meson: add FORCE_FALLBACK_FOR variable for build jobs to use
|
||
- docs/release-calendar: add 25.0 branchpoint and RCs schedule
|
||
- docs/release-calendar: fixup sed fail
|
||
- docs/release-calendar: push the 25.0 branchpoint back by 2 weeks
|
||
- docs: update calendar for 24.3.4
|
||
- docs: add release notes for 24.3.4
|
||
- docs: add sha sum for 24.3.4
|
||
- docs/release-calendar: push back the 24.3.x releases by one week
|
||
- docs: update url to vulkan features & extensions
|
||
- anv,gfxstream,panvk,zink: update urls to vulkan docs
|
||
- radv,lvp: fix url to VkAabbPositionsKHR docs
|
||
- ci: make linker warnings fatal
|
||
- VERSION: bump for 25.0.0-rc1
|
||
- [25.0-only] hk: comment out dead variable
|
||
- .pick_status.json: Update to 5b856a741d6dc18d409a0c06ad6492cc3ee9a6bd
|
||
- .pick_status.json: Mark 0ee5015da4c386c0ef8b6ff12fd2bb34022d86a6 as denominated
|
||
- .pick_status.json: Update to e49df902b4c1b98569921d8b858e6e3855bf10e0
|
||
- .pick_status.json: Update to e192d7d615dec9c9c04447c4b9ab0244d6380944
|
||
- .pick_status.json: Mark 39969409f6fb60b21aea36be4d5424718fcc26b8 as denominated
|
||
- VERSION: bump for 25.0.0-rc2
|
||
- .pick_status.json: Update to fdaf7c7b9647874e66e79653050f9d0999dc9134
|
||
- docs/android: drop libglapi.so now that it's gone
|
||
- .pick_status.json: Mark 5f54beb30728f6510ce50071ddaef5f9157b16ef as denominated
|
||
- gfxstream: fix signedness of shifts
|
||
- gfxstream: drop dead variables
|
||
- gfxstream: use \`range` variable for its intended purpose
|
||
- gfxstream: mark unused variables as such
|
||
- .pick_status.json: Update to ee9edd46254884ab7fe6c96518e23d421d5f5344
|
||
- llvmpipe/tests: include math.h for INFINITY
|
||
- ci: don't run on tag pipelines
|
||
- ci: only trigger the CI for release managers when pushing to staging branch
|
||
- .pick_status.json: Update to 18f0807408425da11cb1d8cd1d73de369317440d
|
||
- .pick_status.json: Update to 30a3d567c8b996fde86b07d2bad018013a54ff44
|
||
- ci: run containers builds on staging branches
|
||
- .pick_status.json: Mark 13e987669ccee373948753e113e9ce7e9bdbef55 as denominated
|
||
- VERSION: bump for 25.0.0-rc3
|
||
- .pick_status.json: Update to e41438275e005bbb20fc9c8115d7d29343c292d8
|
||
- ci: debian-testing-ubsan is used by tests
|
||
- ci/yaml-toml-shell-py-test: don't run on post-merge pipelines
|
||
- ci/yaml-toml-shell-py-test: run on direct push pipelines
|
||
- .pick_status.json: Update to a9b6a54a8cce0aab44c81ea4821ee564b939ea51
|
||
- .pick_status.json: Update to 06d8afff640c66e51517bf4bebd2a58abb2fa055
|
||
- .pick_status.json: Update to 2361ed27f34774f0a73324915a9ddb57f43e112a
|
||
- .pick_status.json: Update to 56aac9fdecad0f7d335f82653832927486f07d44
|
||
- .pick_status.json: Update to 6b20b0658489afe745a28b8f09c57067e45b47f3
|
||
|
||
Eric R. Smith (28):
|
||
|
||
- util: rename PIPE_FORMAT_Y8_U8V8_422_UNORM
|
||
- dri, mesa: fix NV16 texture format
|
||
- egl, mesa: add support for NV15 and NV20 textures
|
||
- dri: fix NV15 and NV20 definitions to make sure they will be used
|
||
- panfrost: add panfrost support for NV15, NV16 and NV20
|
||
- panvk: fix depth bias calculation
|
||
- panfrost: add a perf warning when resources need to be converted
|
||
- panfrost: convert resources before binding them to images
|
||
- panfrost: check afbc status in panfrost_query_compression_modifiers
|
||
- mesa: when blitting between formats clear any unused components
|
||
- aux: add support for dumping the swizzle in pipe_blit_info
|
||
- mesa: update more drivers to handle pipe_blit_info swizzle_enable
|
||
- format: Add R8_G8B8_422_UNORM format
|
||
- panvk: update feature support
|
||
- panvk: split device and instance version numbers
|
||
- panvk: advertise version 1.1 support
|
||
- panfrost: fix read/write resource confusion in afbc_pack
|
||
- panfrost: fix potential memory leak
|
||
- panvk: fix fs_required()
|
||
- panfrost: apply DEPTH_STENCIL flag consistently
|
||
- panfrost: Allow ATEST input to be a FAU index
|
||
- panfrost: ensure sample_mask is written before color
|
||
- panvk: re-enable fragmentStoresAndAtomics for v10
|
||
- drm-uapi: update drm_fourcc.h to latest version
|
||
- panfrost: support MTK 16L32S detiling
|
||
- panfrost: avoid potential divide by 0 calculating timer_resolution
|
||
- panfrost: fix YUV center information for 422
|
||
- panfrost: fix backward propagation of values in loops
|
||
|
||
Erico Nunes (2):
|
||
|
||
- ci/lima: update piglit ci expectations
|
||
- ci/lima: enable again
|
||
|
||
Erik Faye-Lund (134):
|
||
|
||
- panvk: drop unused include
|
||
- panfrost: use mesa_log infra instead of stdio
|
||
- glx: avoid null-deref
|
||
- panfrost: use 64-bits for layout calculations
|
||
- panvk: set correct max extents for images
|
||
- panvk: support binding swapchain memory
|
||
- panvk: wire up swapchain image creation
|
||
- panvk: remove duplicate property
|
||
- panvk: implement sampleRateShading
|
||
- panvk: check for maxResourceSize-overflow in vkCreateImage
|
||
- panvk: document reason for maxResourceSize-limit
|
||
- docs: mark GL_ARB_shader_subroutine as always supported
|
||
- docs: mark GL_ARB_get_program_binary as always supported
|
||
- docs: update GL_OES_shader_image_atomic support
|
||
- docs: update GL_ARB_multi_draw_indirect support
|
||
- docs: refer to panfrost by version
|
||
- docs: fixup a few mistakes with panfrost
|
||
- docs: add missing panfrost extensions
|
||
- lima: fixup typo
|
||
- lima: add assert to validate list-lenght
|
||
- lima: avoid memleak on error
|
||
- panfrost: sanity-check alignment
|
||
- panvk: correct signedness of timestamps
|
||
- panvk: widen type before multiplying
|
||
- mesa/main: properly check for EXT_memory_object
|
||
- mesa/main: properly check for EXT_memory_object_fd
|
||
- mesa/main: properly check for EXT_memory_object_win32
|
||
- mesa/main: properly check for EXT_semaphore
|
||
- mesa/main: properly check for EXT_semaphore_win32
|
||
- st/mesa: check requirements for MESA_texture_const_bandwidth
|
||
- mesa: error-check GL_TEXTURE_TILING_EXT params
|
||
- panvk: report minmax-support for sampled formats
|
||
- panvk: expose KHR_dedicated_allocation
|
||
- vulkan/meta: plug a couple of memory leaks
|
||
- panvk: free preload-shaders after compiling
|
||
- panvk, nvk: spell width correctly
|
||
- panvk/ci: correct name of skips-file
|
||
- panvk/ci: remove duplicate skips
|
||
- panvk/ci: add some missing skips
|
||
- panvk/ci: update ci results for g610
|
||
- panvk/ci: add a few flakes
|
||
- panvk/ci: add a full panvk job
|
||
- panfrost: match 4-bit format order
|
||
- panfrost: add missing 4-bit formats
|
||
- panvk: expose EXT_4444_formats
|
||
- panvk/ci: update g52 results
|
||
- panvk/ci: update g610 results
|
||
- panvk: expose scalarBlockLayout
|
||
- panvk/ci: remove duplicate skips
|
||
- panvk/ci: update g52 results
|
||
- panvk/ci: update g52-vk-full job
|
||
- panvk: do not expose subgroup support
|
||
- panvk: disable imageCubeArray on bifrost
|
||
- panvk: soften the language around opt-in
|
||
- panvk: do not require opt-in for panvk on v10
|
||
- panvk/ci: correct timeouts as crash
|
||
- panvk/ci: fixup g52 skip sorting
|
||
- panvk/ci: add a few more g52 skips
|
||
- panvk: fixup bad indent
|
||
- panvk: only validate the push-sets that we update
|
||
- panvk: back out of vk 1.1 support
|
||
- panvk: make vk-version helper internal to source
|
||
- docs: add new panvk features
|
||
- panvk: fix image size for cube-arrays on bifrost
|
||
- Revert "panvk: disable imageCubeArray on bifrost"
|
||
- st/mesa: document ARB_texture_float quirk
|
||
- pan/cs: fix broken allocation-failure check
|
||
- panfrost: clean up mmap-diagnostics
|
||
- panfrost: report errors from panfrost_bo_mmap
|
||
- panfrost: handle mmap failures
|
||
- panfrost: handle NULL-batches
|
||
- panfrost: propagate cs_builder error instead of asserting
|
||
- panfrost: handle pool-allocation errors
|
||
- panfrost: handle errors allocating csf oom-handler
|
||
- panfrost: try to survive start-up alloc fails
|
||
- pan/ci: update t860 ci xfails
|
||
- panvk: drop fragmentStoresAndAtomics support for now
|
||
- vulkan: add vk_descriptor_type_is_dynamic helper
|
||
- v3dv: use vk_descriptor_type_is_dynamic
|
||
- turnip: use vk_descriptor_type_is_dynamic
|
||
- dozen: use vk_descriptor_type_is_dynamic
|
||
- panvk: use vk_descriptor_type_is_dynamic
|
||
- radv: use vk_descriptor_type_is_dynamic
|
||
- asahi: use vk_descriptor_type_is_dynamic
|
||
- turnip: use vk_descriptor_type_is_dynamic
|
||
- pvr: use vk_descriptor_type_is_dynamic
|
||
- panvk: use vk_descriptor_type_is_dynamic
|
||
- lavapipe: use vk_descriptor_type_is_dynamic
|
||
- anv: use vk_descriptor_type_is_dynamic
|
||
- hasvk: use vk_descriptor_type_is_dynamic
|
||
- dozen: use vk_descriptor_type_is_dynamic
|
||
- nvk: use vk_descriptor_type_is_dynamic
|
||
- panvk/ci: update expected failures
|
||
- docs: fixup broken markup
|
||
- docs: fixup link in radv docs
|
||
- docs/ci: treat warnings as errors
|
||
- docs: update panvk status
|
||
- panvk/ci: drop needless envvar
|
||
- Revert "panfrost: Disable CRC by default"
|
||
- pan/ci: update t760 checksum
|
||
- pan/ci: update opencl expectations
|
||
- docs/panfrost: document vulkan support
|
||
- docs: update panvk status
|
||
- docs/features: fixup panvk KHR_shader_draw_parameters-support
|
||
- pan/va: fix base-level for nir_texop_lod
|
||
- pan/ci: add some occasional flakes
|
||
- docs/features: add a few missing extensions
|
||
- docs/features: mark panfrost as supporting GL_OES_texture_view
|
||
- pan/ci: drop empty trailing variables-list
|
||
- panfrost: reuse tiler hierarchy mask selection from panvk
|
||
- panfrost: limit maximum texture size
|
||
- panfrost: do not artificially limit texture-sizes
|
||
- pan/midgard: use macros for mir_prev_op / mir_next_op
|
||
- pan/midgard: constify pointers
|
||
- pan/compiler: don't pass midgard_instruction by value
|
||
- panvk: expose subgroup operations
|
||
- panvk: expose vk1.1 on v10 hardware
|
||
- pan/bi: bump iter_count to 2000
|
||
- panvk: do not expose EXT_subgroup_size_control on bifrost
|
||
- panvk/ci: update expected failures
|
||
- panfrost: mark helper as static
|
||
- panfrost: handle allocation errors when afbc-packing
|
||
- panfrost: unify emit_tls and emit_fbd
|
||
- panfrost: propagate allocation scratchpad allocation errors
|
||
- panfrost: propagate errors from panfrost_batch_create_bo
|
||
- panfrost: in-place map/unmap shouldn't grow
|
||
- gallium/aux: do not assert on map-failures
|
||
- meson: build panvk by default on arm
|
||
- panvk: fix line-rasterization of bifrost
|
||
- panvk/ci: add back incorrectly removed crash
|
||
- pan/ci: add flaky tests to the flake-list
|
||
- pan/ci: add fail from llvm 19 upgrade
|
||
- panvk: correct number of read bytes for dynamic buffers
|
||
- panvk: report passing the VK CTS
|
||
|
||
Ernst Persson (1):
|
||
|
||
- intel/vulkan: Add bvh build dependency
|
||
|
||
Evan (1):
|
||
|
||
- amd/vpelib: Shaper Refactor
|
||
|
||
Faith Ekstrand (27):
|
||
|
||
- vulkan: Allow the same item to show up twice in core version <requires>
|
||
- vulkan: Add Vulkan 1.4 feature aliases
|
||
- treewide: Stop putting enum in front of Vulkan enum types
|
||
- vulkan: Update XML and headers to 1.4.303
|
||
- nvk: Increase push constant space to 256B
|
||
- nvk: No-op implement VK_KHR_global_priority
|
||
- nvk: Add new Vulkan 1.4 features and properties
|
||
- nvk: Advertise Vulkan 1.4
|
||
- nvk: Only support Vulkan 1.4 on Turing+
|
||
- nvk: Move Vulkan 1.4 features to the 1.4 section
|
||
- nvk: Move Vulkan 1.4 properties to the 1.4 section
|
||
- nvk: Set a command buffer error if pushbuf alloc fails
|
||
- nvk: Call nir_opt_access
|
||
- nak: Use ldc.constant for load_global when CAN_REORDER is set
|
||
- nvk: Handle pCounterBuffers == NULL in Begin/EndTransformFeedback
|
||
- nvk: Fix scissor bounds
|
||
- nvk: Rename nvk_descriptor_set::mapped_ptr
|
||
- nvk: Respect VK_DESCRIPTOR_POOL_CREATE_HOST_ONLY_BIT_EXT
|
||
- nvk: Implement descriptorBufferPushDescriptors
|
||
- nvk: Pull shaders from the state command buffer in nvk_cmd_process_cmds()
|
||
- nvk: Handle shader==NULL in nvk_cmd_upload_qmd()
|
||
- nvk: Allow sparse loads on EDB buffers
|
||
- nak: Handle sparse texops with unused color destinations
|
||
- nvk: Use suld for EDB uniform texel buffers
|
||
- nvk: Align UBO/SSBO addresses down rather than up
|
||
- nak: Use suld.constant when ACCESS_CAN_REORDER is set
|
||
- nvk: Use suld.constant for EDB uniform texel buffers
|
||
|
||
Felix DeGrood (6):
|
||
|
||
- iris: Use vfg distribution mode = RR_STRICT for Xe2+
|
||
- anv: Use vfg distribution mode = RR_STRICT for Xe2+
|
||
- anv: allow compressed buffers types on vkd3d titles
|
||
- anv: remove unnecessary driconf entries for anv_enable_buffer_comp
|
||
- vk/overlay-layer: defer log creation to swapchain creation
|
||
- intel/perf: add new perf consts to support more metrics
|
||
|
||
Feng Jiang (2):
|
||
|
||
- virgl: Ensure that PIPE_SHADER_CAP_MAX_CONST_BUFFERS is less than PIPE_MAX_CONSTANT_BUFFERS
|
||
- radv/rt: Fix memleak in radv_init_header()
|
||
|
||
Francisco Jerez (27):
|
||
|
||
- intel/fs/xe2: Fix up subdword integer region restriction with strided byte src and packed byte dst.
|
||
- intel/brw/xe3+: Relax SEND EOT register assignment restrictions.
|
||
- intel/brw: Saturate shifted subgroup index to avoid reading past the end of register file.
|
||
- intel/brw: Use urb_read_length instead of nr_attribute_slots to calculate VS first_non_payload_grf.
|
||
- intel/brw/xe3+: Mask subgroup shuffle index to be within valid range to avoid VRT hangs.
|
||
- anv/gfx12.5: Request subgroup size 8 for RT trampoline shader.
|
||
- intel/brw: Allow specifying a required subgroup size for fragment shaders.
|
||
- intel/blorp: Specify a subgroup size requirement of 16 for fast clear or repclear shaders.
|
||
- intel/common/xe2+: Allow SIMD32 PS for all multisample cases.
|
||
- intel/brw/xe3: Define XE3_MAX_GRF.
|
||
- intel/brw/xe3: Extend regalloc sets to maximum Xe3 GRF size.
|
||
- intel/brw/xe3+: Bump number of SBID tokens for Xe3.
|
||
- intel/brw/xe3+: Disable round-robin allocation heuristic on Xe3+.
|
||
- intel/brw: Indent body of brw_compile_fs() not applicable to xe3+.
|
||
- intel/brw: Indent conditional block from brw_compile_fs() not applicable to Xe2+.
|
||
- intel/brw: Exit early from run_fs() if compilation failed before optimization loop.
|
||
- intel/brw/xe3+: brw_compile_fs() implementation for Xe3+.
|
||
- intel/brw/xe3+: Optimize CS/TASK/MESH compile time optimistically assuming SIMD32.
|
||
- intel/brw: Report number of GRF registers used in brw_stage_prog_data.
|
||
- intel/brw: Define ptl_register_blocks() helper.
|
||
- intel/genxml/xe3+: Update definitions for shader state setup.
|
||
- iris/xe3+: Set RegistersPerThread during shader state setup based on prog_data.
|
||
- intel/blorp/xe3+: Set RegistersPerThread during shader state setup based on prog_data.
|
||
- anv/xe3+: Set RegistersPerThread during shader state setup based on prog_data.
|
||
- anv/xe3+: Set RegistersPerThread for bindless shader dispatch.
|
||
- iris/xe3+: Enable VRT.
|
||
- anv/xe3+: Enable VRT.
|
||
|
||
Frank Binns (2):
|
||
|
||
- pvr: add TI j721s2 as a supported device
|
||
- pvr: add 36.53.104.796 (BXS-4-64) to the list of supported GPUs
|
||
|
||
Friedrich Vock (15):
|
||
|
||
- vulkan/rmv: Correctly set heap size
|
||
- vulkan/runtime/bvh: Set leaf_node_count for updates
|
||
- radv,driconf: Apply DOOM Eternal/idTech workarounds for Indiana Jones
|
||
- aco/lower_to_hw_instr: Check the right instruction's opcode
|
||
- radv/rt: Remove nir_intrinsic_execute_callable instrs in monolithic mode
|
||
- aco: Fix dead instruction/index handling for try_insert_saveexec_out_of_loop
|
||
- nir: Serialize all parameter attributes
|
||
- nir,vtn: Add return info to parameters
|
||
- nir: Add parameter divergence info
|
||
- vtn: Set parameter type in glsl_type_add_to_function_params
|
||
- nir: Add indirect calls
|
||
- nir: Apply passes to all functions
|
||
- nir: Add nir_instr_is_before helper
|
||
- nir: Free liveness info when invalidating metadata
|
||
- nir: Add indirect call optimizations
|
||
|
||
GKraats (1):
|
||
|
||
- i915g: fix glClearColor using a 1 byte color format
|
||
|
||
Georg Lehmann (79):
|
||
|
||
- radv: run copy prop before vectorizing
|
||
- nir/opt_16bit_tex_image: optimize extract half sources
|
||
- nir: add nir_def_all_uses_ignore_sign_bit
|
||
- pan/bi: use nir_def_all_uses_ignore_sign_bit
|
||
- aco: use nir_def_all_uses_ignore_sign_bit
|
||
- nir: handle fmul(a,a)/ffma(a,a,b) in nir_def_all_uses_ignore_sign_bit
|
||
- aco/gfx8: use ds_swizzle_b32 rotate mode
|
||
- nir: return def for debug info in nir_instr_def
|
||
- nir/instr_set: replace nir_instr_get_def_def with nir_instr_def
|
||
- nir/instr_set: support instrs with no def
|
||
- nir: cse terminate/demote
|
||
- nir/opt_undef: replace undef in a separate pass
|
||
- nir/opt_undef: use some nir helpers
|
||
- nir/opt_undef: keep undefs used by partial undef vectors
|
||
- nir/opt_undef: handle unpack/pack like mov/vec
|
||
- aco/isel: use undef Operands for p_create_vector created from nir vecs
|
||
- util: add BITSET_LAST_BIT_BEFORE
|
||
- nir/move_discards_to_top: single final iteration
|
||
- nir/move_discards_to_top: don't move across is_helper_invocation
|
||
- radv/ci: document test_shader_sm66_is_helper_lane as fixed
|
||
- freedreno/ci: update a630 KSP checksum
|
||
- nir/opt_intrinsic: rework sample mask opt with vector alu
|
||
- nir/opt_intrinsic: fix sample mask opt with demote
|
||
- radv: optimize sample mask comparisons
|
||
- aco/optimizer: label fcanonicalize like a copy if there is nothing to flush
|
||
- nir/opt_algebraic: optimize ffma(b2f, b2f, c)
|
||
- nir/opt_algebraic: optimize d3d9 ftrunc
|
||
- nir/opt_algebraic: optimize d3d9 ceil
|
||
- nir/opt_algebraic: mark a - ffract(a) as nan incorrect.
|
||
- radv: fix reporting mesh/task/rt as supported dgc indirect stages
|
||
- radv: rework vk_property initialization
|
||
- aco/gfx12: disable vinterp ddx/ddy optimization
|
||
- aco/gfx12+: do not use v_pack_b32_f16 to pack untyped data
|
||
- radeonsi/ci: add vangogh ubo fail
|
||
- zink: spec\@ext_framebuffer_multisample\@blit-mismatched-formats was fixed
|
||
- aco/gfx11+: use v_and_b32 to extract local id 0
|
||
- radv: track holes in the clip/cull masks
|
||
- nir: add constant clip/cull distance optimization
|
||
- radv: use nir_opt_clip_cull_const
|
||
- nir/uub: properly limit float support to 32bit
|
||
- nir: add unsigned upper bound support for f2i32
|
||
- nir: add unsigned upper bound support for fsat
|
||
- aco/gfx12: don't assume memory operations complete in order
|
||
- aco/ra: don't write to exec/ttmp with mulk/addk/cmovk
|
||
- aco/ra: disallow s_cmpk with scc operand
|
||
- aco/ra: don't write to scc/ttmp with s_fmac
|
||
- nir/opt_remove_phis: rematerialize equal alu
|
||
- nir/opt_algebraic: optimize min(max(a, b), a)
|
||
- nir: optimize unpacking 8bit values from a 64bit source
|
||
- aco/isel: skip and(exec) for top level demote_if/terminate_if
|
||
- aco: rename p_early_exit_if to if_not
|
||
- aco: allow p_exit_early_if_not with exec condition
|
||
- aco/insert_exec: exit shader using exec for top level discard
|
||
- aco: create v_cmpx with s_andn2(exec, v_cmp)
|
||
- nir: sink/move alu with two identical, non constant sources.
|
||
- amd: switch to FRONT_FACE_ALL_BITS(0)
|
||
- nir: add load_front_face_fsign
|
||
- amd: support load_front_face_fsign
|
||
- nir: add nir_alu_srcs_negative_equal_typed
|
||
- nir,amd: optimize front_face ? a : -a
|
||
- aco/optimizer: fix signed extract of sub dword temps with SDWA
|
||
- aco/insert_exec: reset top exec for p_discard_if
|
||
- radv: run peephole_select in optimize_nir_algebraic
|
||
- nir/peephole_select: allow load_vector/scalar_arg_amd
|
||
- aco: guard small_vector move/copy operator against self assignment
|
||
- aco: support less trivial component types in small_vec
|
||
- aco: implement some more std::vector functions for small_vec
|
||
- nir/opt_algebaric: convert fadd(a, a) to a * 2.0
|
||
- aco: update is_dual_issue_capable for gfx11.5+
|
||
- aco/sched_ilp: continue open clauses
|
||
- aco/sched_ilp: add dependencies of later clause instrs more aggressively
|
||
- aco/sched_ilp: only remove WaW/WaR for inter clause dependencies
|
||
- aco/sched_ilp: reorder VINTRP
|
||
- aco/sched_ilp: new latency heuristic
|
||
- aco/sched_ilp: rename priority to wait_cycles
|
||
- aco/sched_ilp: use more realistic memory latencies
|
||
- aco/sched_ilp: base latency and issue cycles on aco_statistics
|
||
- nir: fix range analysis for frcp
|
||
- nir: fix frsq range analysis
|
||
|
||
Gert Wollny (6):
|
||
|
||
- virgl/vtest: take handle from host when using protocol version >=3
|
||
- virgl/vtest: When trying to use protocol 3 check host feature
|
||
- virgl/vtest: change interface of virgl_vtest_submit_cmd
|
||
- virgl/vtest: Add support for creating blob resources
|
||
- ci: Upref virglrenderer version
|
||
- radeon/evergreen: ensure equal sizes for depth-stencil npot textures
|
||
|
||
Guilherme Gallo (9):
|
||
|
||
- ci/lava: Set default exit code to 1 for failed jobs
|
||
- ci/lava: Improve exception handling for job failures
|
||
- ci/lava: Uprev freezegun
|
||
- ci/intel: Set HWCI modules for puff DUT
|
||
- ci/iris: Force UART for puff boards
|
||
- ci/iris: Rebalance iris-cml-deqp jobs
|
||
- ci/iris: Fix iris-cml-traces expectations
|
||
- ci/iris: Update iris-cml-deqp CI expectations
|
||
- ci/container: set up S3_JWT_FILE also for container jobs
|
||
|
||
Gurchetan Singh (17):
|
||
|
||
- util: add c++ guards to u_mm.h
|
||
- gfxstream: move isHostVisible function
|
||
- gfxstream: nuke android::base::SubAllocator
|
||
- gfxstream: use vulkan_lite_runtime
|
||
- gfxstream: nuke EntityManager.h include
|
||
- gfxstream: aemu: vendor it
|
||
- gfxstream: modify libaemu for Mesa use case
|
||
- gfxstream: guest: use internal version of AEMU headers + impls
|
||
- gfxstream: use canonical Mesa dependencies
|
||
- gfxstream: conditionals for using gfxstream::aemu
|
||
- gfxstream: delete qemu_pipe target
|
||
- gfxstream: for Android, look for the autogenerated files
|
||
- gfxstream: change output location
|
||
- gfxstream: remove abort()
|
||
- gfxstream: fix issues with VK1.4 build
|
||
- gfxstream: remove references to Fuchsia Goldfish
|
||
- gfxstream: fix some integration bugs
|
||
|
||
Hans-Kristian Arntzen (11):
|
||
|
||
- vulkan/wsi/wayland: Use X11-style image count strategy when using FIFO.
|
||
- radv: Fix missing gang barriers for task shaders.
|
||
- radv/winsys: Report VA mappings in bo_log too.
|
||
- radv: Add sparse mappings to radv_check_va.py.
|
||
- wsi/x11: Do not use allocation callbacks on a thread.
|
||
- wsi/wayland: Only use commit timing protocol alongside present time.
|
||
- wsi/wayland: Don't fallback to broken legacy throttling with FIFO
|
||
- wsi/wayland: Handle FIFO -> MAILBOX transitions correctly
|
||
- wsi/wayland: Remove unused present_mode member.
|
||
- wsi/wayland: Add forward progress guarantee for present wait.
|
||
- radv: Add radv_invariant_geom=true for Indiana Jones.
|
||
|
||
Hsieh, Mike (1):
|
||
|
||
- amd/vpelib: Refactor 3D LUT parameters
|
||
|
||
Hyunjun Ko (10):
|
||
|
||
- anv: define ANV_VIDEO_H264_MAX_DPB_SLOTS
|
||
- anv: Enable remapping picture ID
|
||
- anv: handle negative value of slot index for h265 decoding.
|
||
- intel/genxml: define MEMORYADDRESSATTRIBUTES for Gen12.5 with TILEF
|
||
- anv/video: Fix to return supported video format correctly.
|
||
- anv: calculate global parmeters correctly for AV1 decoding
|
||
- anv: support in-loop super resolution for AV1 decoding
|
||
- anv: fix to set default cdf buf correctly.
|
||
- anv: change bool to VkResult
|
||
- anv: Fix to set CDEF flter flag correctly for AV1 decoding
|
||
|
||
Iago Toral Quiroga (15):
|
||
|
||
- v3d: add a V3D_DEBUG option to force synchronous execution of jobs
|
||
- broadcom: handle double buffer on V3D 7.1 tile size calculations
|
||
- v3d: group tile spec into a struct inside the job
|
||
- v3d: save a pointer to the TILE_BINNING_MODE_CFG packet in the CL
|
||
- v3d: do tile state BO allocation later
|
||
- v3d: only enable double-buffer for jobs where it might make sense
|
||
- v3dv: add missing support for double-buffer on V3D 7.x
|
||
- v3d: drop blank line
|
||
- v3d: store size of qpu program for compiled shaders
|
||
- broadcom: add helpers for double-buffer heuristic
|
||
- v3d: use heuristic to enable double-buffer mode
|
||
- v3dv: use the double buffer heuristic helpers
|
||
- broadcom: move double-buffer heuristic helpers to the compiler
|
||
- v3dv: fix missing access bit flag when checking for texel buffer reads
|
||
- v3dv: fix crash on 32-bit builds
|
||
|
||
Ian Romanick (57):
|
||
|
||
- brw/emit: Add correct 3-source instruction assertions for each platform
|
||
- brw/copy: Don't copy propagate through smaller entry dest size
|
||
- brw/cse: Don't eliminate instructions that write flags
|
||
- brw/lower: Don't emit spurious moves to or from NULL register
|
||
- brw/opt: Always do copy prop, DCE, and register coalesce after lower_regioning
|
||
- brw/opt: Always do both kinds of copy propagation before lower_load_payload
|
||
- brw/build: Add scalar_group() helper
|
||
- brw/lower: Lower invalid source conversion to better code
|
||
- Fix copy-and-paste bug in nir_lower_aapoint_impl
|
||
- brw/lower: Don't "fix" regioning of broadcast
|
||
- brw: Use resize_sources several more places
|
||
- brw/build: Use SIMD8 temporaries in emit_uniformize
|
||
- brw/copy: Allow copy prop into src1 of broadcast
|
||
- nir/algebraic: Optimize some trivial bfi
|
||
- brw/algebraic: Fix ADD constant folding
|
||
- brw/algebraic: Fix MUL constant folding
|
||
- brw/emit: Fix typo in recently added ADD3 assertion
|
||
- brw/algebraic: Partial constant folding of ADD3
|
||
- brw/const: Allow mixing signed and unsigned immediate sources
|
||
- brw/copy: Don't try to be clever about ADD3 constant propagation
|
||
- brw: Emit immediate value for MAD in canonical position
|
||
- brw/copy: Commute immediates for MAD multiplicands
|
||
- brw/algebraic: Constant fold multiplicands of MAD
|
||
- brw/algebraic: Don't restrict MAD(a, b, 1) optimization to float32
|
||
- brw/const: Refactor checking whether an immediate source is allowed
|
||
- brw/const: Allow constants in integer MAD
|
||
- brw/const: Allow HF constants in MAD on Gfx11
|
||
- brw/const: Remove TODO that isn't allowed by the hardware
|
||
- brw/algebraic: Pull brw_constant_fold_instruction out of the switch statement
|
||
- brw/emit: Fix BROADCAST when value is uniform and index is immediate
|
||
- brw: Add devinfo parameter to fs_inst::regs_read
|
||
- brw: Basic infrastructure to store convergent values as scalars
|
||
- brw/lower: Allow uniform and scalar sources to many kinds of SEND
|
||
- brw/nir: Fix up handling of sources that might be convergent vectors
|
||
- brw/lower: Adjust source stride on DF is_scalar sources to MAD on Gfx9
|
||
- brw/lower: Properly handle UNIFORM globals address in lower_trace_ray_logical_send
|
||
- brw/emit: Allow scalar sources to HF math instructions on Xe2
|
||
- brw/nir: Prepare try_rebuild_source for scalar values
|
||
- brw/build: Prepare BROADCAST for scalar values
|
||
- brw/nir: Treat load_const as convergent
|
||
- brw/nir: Treat some load_uniform as convergent
|
||
- brw/nir: Treat load_workgroup_id as convergent
|
||
- brw/nir: Treat some ALU results as convergent
|
||
- brw/nir: Treat some load_ubo as convergent
|
||
- brw/nir: Treat load_inline_data_intel as convergent
|
||
- brw/nir: Treat load_reloc_const_intel as convergent
|
||
- brw/nir: Treat load_btd_{global,local}_arg_addr_intel and load_btd_shader_type_intel as convergent
|
||
- brw/nir: Treat load_*_uniform_block_intel as convergent
|
||
- brw/nir: Treat some resource_intel as convergent
|
||
- brw/nir: Eliminate nir_to_brw_state::uniform_values
|
||
- brw/nir: Don't try optimize around emit_uniformize
|
||
- brw/nir: Simplify get_nir_image_intrinsic_image and get_nir_buffer_intrinsic_index
|
||
- brw/nir: Treat some ballot as convergent
|
||
- brw/nir: Don't generate scalar byte to float conversions on DG2+ in optimize_extract_to_float
|
||
- iris: Add missing nir_metadata_preserve in iris_lower_storage_image_derefs
|
||
- crocus: Add missing nir_metadata_preserve in crocus_lower_storage_image_derefs
|
||
- brw/copy: Fix handling of offset in extract_imm
|
||
|
||
Icenowy Zheng (4):
|
||
|
||
- zink: do not set transform feedback bits when not available
|
||
- meson: prefer 'python3' to 'python' when finding python3
|
||
- zink: emit consts as uint only on IMG proprietary drivers
|
||
- zink: use lazy descriptors for IMG proprietary drivers
|
||
|
||
Igor Torrente (2):
|
||
|
||
- Zink: Add NVK to the non \`driver_workarounds.implicit_sync` list
|
||
- NVK: Enable RW DMA-BUF export
|
||
|
||
Ivan Avdeev (1):
|
||
|
||
- radv: add a flag to indicate ray tracing support
|
||
|
||
Iván Briano (6):
|
||
|
||
- intel/rt: fix ray_query stack address calculation
|
||
- intel/decoder: fix INTEL_DEBUG=bat
|
||
- anv: remove unused/misleading/wrong parameters from the RT trampoline
|
||
- vulkan: calculate remaining layers of 2d view of 3d image correctly
|
||
- anv: disable logic op for float/srgb formats
|
||
- hasvk: disable logic op for float/srgb formats
|
||
|
||
James Hogan (3):
|
||
|
||
- glsl: Expose gl_ViewID_OVR back to GLSL 1.30
|
||
- mesa: Fix multiview attachment completeness check
|
||
- mesa: Fix FramebufferTextureMultiviewOVR num_views check
|
||
|
||
Janne Grunau (1):
|
||
|
||
- panvk: Silence warning on incompatible DRM render devices
|
||
|
||
Jason Macnak (3):
|
||
|
||
- Simplify ApiInfo
|
||
- Pass VkSnapshotApiCallInfo-s through VkDecoderGlobalState
|
||
- Update VkDecoderSnapshot locking
|
||
|
||
Jesse Natalie (4):
|
||
|
||
- microsoft/compiler: Put holes in driver_location based on I/O variable sizes
|
||
- microsoft/clc: Initialize printf buffer for tests
|
||
- microsoft/compiler: Skip POS for io compaction
|
||
- microsoft/compiler: Update clip/cull split pass to handle clip/cull getting merged
|
||
|
||
Jianxun Zhang (5):
|
||
|
||
- anv,hasvk,genxml: Rename genxml files using verx10
|
||
- isl: Refactor WA 22015614752
|
||
- iris: Allow compression on multi-sampled stencil (xe2)
|
||
- isl: Allow CCS in more cases (xe2)
|
||
- isl: Move a CCS restriction in GFX 12.x
|
||
|
||
Job Noorman (87):
|
||
|
||
- ir3/ra: prevent moving source intervals for shared collects
|
||
- ir3,tu: include ir3 debug flags in shader hash key
|
||
- ir3,tu: filter debug flags included in the hash key
|
||
- ir3: fold shared movs into other movs
|
||
- nir: add ir3-specific bitwise triop opcodes
|
||
- nir/search: make is_only_used_by_iadd reusable
|
||
- nir/search: add is_only_used_by_{iand,ior} helpers
|
||
- ir3: fix backend support for bitwise triops
|
||
- ir3: add codegen for bitwise triops
|
||
- ir3: add pass to select bitwise triops
|
||
- ir3/isa: allow rpt6/rpt7
|
||
- ir3: add workaround for predication hardware bug
|
||
- nir/lower_subgroups: support unknown subgroup size
|
||
- ir3: use generic lowering for 64b scan/reduce
|
||
- ir3: remove unused ir3_nir_lower_64b_subgroups
|
||
- nir: add read_getlast_ir3 intrinsic
|
||
- ir3: add codegen for read_getlast_ir3
|
||
- ir3: add helper to get the subgroup size
|
||
- ir3: rename cluster_size to brcst_cluster_size
|
||
- nir/lower_subgroups: add extra filter data to options
|
||
- nir/lower_subgroups: disable boolean reduce when not supported
|
||
- ir3: add support for clustered subgroup reductions
|
||
- tu: advertise VK_SUBGROUP_FEATURE_CLUSTERED_BIT
|
||
- nir/lower_subgroups: add option to only lower clustered rotates
|
||
- ir3: lower clustered rotates to shuffles
|
||
- tu: advertise VK_SUBGROUP_FEATURE_ROTATE_CLUSTERED_BIT_KHR
|
||
- ir3: don't update builder cursor for IR3_CURSOR_AFTER_BLOCK
|
||
- ir3: add ir3_after_instr_and_phis helper
|
||
- ir3: use generic INSTR0 implementation for ir3_NOP
|
||
- ir3: refactor builders to use ir3_builder API
|
||
- ir3: reformat after refactoring in previous commit
|
||
- ir3: add reformatting commits to .git-blame-ignore-revs
|
||
- ir3/isa: fix conflict between stib.b and stsc
|
||
- ir3/isa: fix cat3-alt immed src
|
||
- ir3/isa: fix isaspec for sad.s32
|
||
- ir3: teach backend about sad
|
||
- ir3: add codegen for sad
|
||
- ir3/cp: only mark mad srcs as swapped when swap succeeded
|
||
- ir3/cp: extract common src swapping code
|
||
- ir3/cp: make try_swap_mad_two_srcs more generic
|
||
- ir3/cp: add support for swapping srcs of sad
|
||
- ir3/validate: print file/line info
|
||
- ir3,freedreno: remove binning outputs after vs ucp lowering
|
||
- ir3/cp: swap back correct srcs when swap failed
|
||
- ir3: always set wrmask for movmsk
|
||
- ir3: emit uniform iadd3 as two adds
|
||
- ir3: output early-preamble stat as integer
|
||
- ir3/ra: fix non-trivial collect detection
|
||
- ir3/ra: allocate shared collects dst over its srcs when possible
|
||
- ir3/parser: fix parsing integer as float
|
||
- ir3/a7xx: properly handle alias scope and type
|
||
- ir3/a7xx: disasm halfness of alias dst
|
||
- ir3/a7xx: implement and document unknown alias field
|
||
- ir3/a7xx: handle alias.rt dst
|
||
- ir3/a7xx: document alias.rt
|
||
- ir3/print: add support for alias
|
||
- ir3: teach backend about alias
|
||
- ir3: introduce alias goups
|
||
- ir3: add validation for alias
|
||
- ir3: add ir3_compiler::has_alias
|
||
- ir3: add support for alias.tex
|
||
- ir3: optimize alias register allocation by reusing GPRs
|
||
- ir3/legalize: insert (ss) to read consts after stc
|
||
- ir3/legalize: insert (sy) to read consts after ldc.k
|
||
- ir3/dce: support partial writes from collects
|
||
- ir3: add some preamble helpers
|
||
- ir3: make find_end a global helper
|
||
- tu,ir3: inform ir3 of dynamically remapped FS slots
|
||
- ir3: make shader output struct non-anonymous
|
||
- ir3: reuse ir3_find_output in ir3_find_output_regid
|
||
- tu: add chip param to tu6_emit_fs_outputs
|
||
- tu: add support for aliased render target components
|
||
- freedreno: add chip param to emit_fs_output
|
||
- freedreno: add support for aliased render target components
|
||
- ir3: add support for alias.rt
|
||
- ir3: disable alias.rt pre-a750
|
||
- ir3: account for inserted nops in delay calculation
|
||
- freedreno: move ForEachMacros into freedreno
|
||
- freedreno: remove unused entries from ForEachMacros
|
||
- freedreno: add missing entries to ForEachMacros
|
||
- ir3: schedule alias.rt at the end of the preamble
|
||
- ir3: rematerialize preamble defs in block dominated by sources
|
||
- ir3: add helper to calculate src read delay
|
||
- ir3: make delay slots a compiler property
|
||
- ir3/a7xx: update delays slots
|
||
- ir3/a7xx: enable delayed src2 read for all cat3 instructions
|
||
- ir3: fix emitting descriptor prefetches at end of preamble
|
||
|
||
John Anthony (2):
|
||
|
||
- panvk: Enable storageBuffer16BitAccess
|
||
- panvk: Enable VK_KHR_vertex_attribute_divisor
|
||
|
||
Jordan Justen (6):
|
||
|
||
- intel/dev: Add PTL 0xb0b0 PCI ID
|
||
- intel/dev: Split hwconfig warning check into hwconfig_item_warning()
|
||
- intel/dev: Split apply and check paths for hwconfig
|
||
- intel/dev: Don't process hwconfig table to apply items when not required
|
||
- intel/dev: Add intel_check_hwconfig_items()
|
||
- iris: Check that mem_fence_bo was created
|
||
|
||
Jose Maria Casanova Crespo (9):
|
||
|
||
- v3d: Enable Early-Z with discards when depth updates are disabled
|
||
- rpi4/ci: mark another flaky timeline_semaphore test
|
||
- rpi4/ci: another detected flaky timeline_semaphore test
|
||
- vc4/ci: fails udpate after last piglit uprev
|
||
- rpi4/ci: Increase timeout for rusticl jobs.
|
||
- v3d: Don't load/store if rasterizer discard is enabled
|
||
- v3d/ci: update rpi expectations by last piglit uprev
|
||
- v3d: Apply FBO resources invalidations on job creation
|
||
- Revert "ci: take igalia farm offline"
|
||
|
||
Joshua Duong (1):
|
||
|
||
- gfxstream: update auto-generated comments.
|
||
|
||
José Roberto de Souza (16):
|
||
|
||
- intel/dev/xe: Fix access to eu_per_dss_mask
|
||
- intel/dev/xe: Fix size of eu_per_dss_mask
|
||
- intel/genxml/xe2: Add STATE_SYSTEM_MEM_FENCE_ADDRESS instruction
|
||
- anv: Always create anv_async_submit in init_copy_video_queue_state()
|
||
- anv: Emit STATE_SYSTEM_MEM_FENCE_ADDRESS
|
||
- iris: Emit STATE_SYSTEM_MEM_FENCE_ADDRESS
|
||
- iris: Add support for damage region
|
||
- anv: Allow larger SLM sizes for task and mesh shader
|
||
- anv: Check VkResult of perf query batch buffer
|
||
- anv: Check VkResult main batch buffer before start companion batch buffer
|
||
- iris: Drop BO_ALLOC_COHERENT from iris_utrace_create_ts_buffer()
|
||
- iris: Rename BO_ALLOC_COHERENT to BO_ALLOC_CACHED_COHERENT
|
||
- anv: Return scanout PAT entry for scanout and external buffers in discrete GPUs
|
||
- anv: Allow WSI blit_src Image to be kept compressed when transitioning to VK_IMAGE_LAYOUT_PRESENT_SRC_KHR
|
||
- iris: Make sure a uncached heap is choosen for scanout and shared buffers when LLC is not available
|
||
- iris: Pick scanout PAT entry for scanout buffers
|
||
|
||
Juan A. Suarez Romero (26):
|
||
|
||
- util/format: nr_channels is always <= 4
|
||
- v3dv: remove unused assignments
|
||
- v3dv: fix BO allocation
|
||
- v3dv: free pointers on multisync error
|
||
- v3dv: ensure there is always a perfmon and counter
|
||
- broadcom/compiler: ensure offset source exists
|
||
- broadcom/compiler: fix fp16 conversion operations
|
||
- v3d: make v3d_flush_resource reallocate non-shareable resources
|
||
- vc4: ensure sharing tiled resources are of proper format
|
||
- v3d: fix BO allocation
|
||
- v3d: remove intermediate variable
|
||
- v3d: find linear modifier when required
|
||
- vc4: find linear modifier when required
|
||
- v3d/ci: clean some asan failures
|
||
- v3d: avoid 0-size variable length array
|
||
- v3dv: fix assigned value is garbage or undefined
|
||
- vc4: initialize variable
|
||
- v3dv: check requirements for USAGE_INPUT_ATTACHMENT
|
||
- freedreno: a2xx: fix maybe uninitialized variable
|
||
- radeonsi/vcn: fix maybe uninitialized
|
||
- v3d: fix format overflow error
|
||
- virgl: fix member access to a NULL pointer struct
|
||
- etnaviv: cast assertion
|
||
- ci/build: add ubsan build jobs
|
||
- broadcom/ci: add ubsan jobs for broadcom drivers
|
||
- ci: take igalia farm offline
|
||
|
||
Jung-uk Kim (1):
|
||
|
||
- FreeBSD: Disable support for "-mtls-dialect" for FreeBSD
|
||
|
||
Juston Li (1):
|
||
|
||
- util/cache_test: Fix racey Cache.List test
|
||
|
||
Kai Wasserbäch (1):
|
||
|
||
- fix(FTBFS): clc/clover: pass a VFS instance explicitly
|
||
|
||
Karmjit Mahil (21):
|
||
|
||
- tu: Fix push_set host memory leak on command buffer reset
|
||
- tu: Fix potential alloc of 0 size
|
||
- nir: Fix \`no_lower_set` leak on early return
|
||
- tu: Fix memory leaks on VK_PIPELINE_COMPILE_REQUIRED
|
||
- nir/algebraic: turn \`u{ge,lt} a, 1` to \`i{ne,eq} a, 0`
|
||
- nir,ir3: Add icsel_eqz
|
||
- nir: Fix the spelling of compare
|
||
- freedreno/rddecompiler: clang-format fix
|
||
- freedreno/rddecompiler: Fix some unsused function warnings
|
||
- ir3: Fix some Wsign-compare when compiling a generate-rd.cc
|
||
- util/idalloc: Fix util_idalloc_foreach() build issue
|
||
- util/idalloc: Minor refactor of util_idalloc_foreach()
|
||
- tu: Fix \`clear_values` leak
|
||
- tu: Fix FDM patchpoint memory leak
|
||
- tu: Fix leaking of some descriptor sets
|
||
- tu: Initialize tu_tiling_config even when tiling isn't possible
|
||
- tu: Free pre_chain patchpoint data
|
||
- util/simple_mtx: Add ASSERTED to parameter used only in an assert
|
||
- vulkan: Add inital vram-report-limit layer
|
||
- freedreno/replay: Define __user for msm_kgsl
|
||
- loader/wayland: Fix missing timespec.h include
|
||
|
||
Karol Herbst (77):
|
||
|
||
- nv/codegen: Do not use a zero immediate for tex instructions
|
||
- nvc0: return NULL instead of asserting in nvc0_resource_from_user_memory
|
||
- clover: drop support for nir drivers
|
||
- gallium: drop PIPE_SHADER_IR_NIR_SERIALIZED
|
||
- rusticl/kernel: fix kernel variant selection
|
||
- vtn: handle struct kernel arguments passed by value
|
||
- nir/lower_cl_images: lower scalar image_loads to vec4
|
||
- rusticl/mem: add restrictions for CL_DEPTH, CL_DEPTH_STENCIL and msaa images
|
||
- rusticl/image: fix clEnqueueFillImage for CL_DEPTH
|
||
- rusticl/device: advertize cl_khr_depth_images if supported
|
||
- rusticl: enable cl_khr_depth_images
|
||
- rusticl: check for overrun status when deserializing
|
||
- rusticl/kernel: convert name and type_name to Option<CString>
|
||
- rusticl/mesa: make driver_name() return a &CStr
|
||
- rusticl/program: check if provided binary pointers are null
|
||
- rusticl: rework query APIs
|
||
- rusticl/api: add a write_len_only variant for writing API properties
|
||
- rusticl/api: add a write_iter variant for writing API properties
|
||
- rusticl/program: use write_len_only for CL_PROGRAM_BINARIES
|
||
- rusticl/program: use write_iter for CL_PROGRAM_DEVICES
|
||
- rusticl/program: pass the slice directly for CL_PROGRAM_IL
|
||
- rusticl/program: use write_len_only for CL_PROGRAM_IL
|
||
- rusticl/platform: pass the slice directly for CL_PLATFORM_EXTENSIONS_WITH_VERSION
|
||
- rusticl/api: use constant arrays instead of Vecs for queries
|
||
- rusticl/context: use write_iter for CL_DEVICES_FOR_GL_CONTEXT_KHR
|
||
- rusticl/proc: make generated entry points unsafe
|
||
- rusticl/api: mark get_info and get_info_obj as unsafe
|
||
- rusticl/util: add Properties::is_empty() and len()
|
||
- rusticl/util: add Properties::iter()
|
||
- rusticl/util: make Properties::props private
|
||
- rusticl/util: reimplement Properties over Vec of scalars
|
||
- rusticl/api: simplify CLProp implementation of Properties
|
||
- rusticl/api: use Properties for 0 terminated arrays consistently
|
||
- rusticl/util: make Properties::from_ptr unsafe
|
||
- rusticl/api: remove Option around Properties
|
||
- rusticl/util: rename Properties::from_ptr to new
|
||
- rusticl/util: fix duplicate key detection in Properties::new
|
||
- rusticl/platform: silence static_mut_refs warning
|
||
- rusticl/util: fix ptr_to_integer_transmute_in_consts warning
|
||
- rusticl: fix clippy::needless-lifetimes
|
||
- rusticl: fix clippy::doc-lazy-continuation
|
||
- rusticl/queue: add a life check to prevent applications dead locking
|
||
- rusticl: stop using system headers for CL and GL
|
||
- include: Update the OpenCL headers to latest
|
||
- rusticl/mesa: remove PipeTransfer::res
|
||
- rusticl/mem: remove mem_type argument from new_image
|
||
- rusticl/device: remove unused functions
|
||
- rusticl/mesa/context: use Default for pipe_grid_info initialization
|
||
- rusticl/mesa: add missing files to meson.build
|
||
- rusticl/queue: make QueueContext::dev public
|
||
- rusticl/mem: pass around QueueContext instead of PipeContext
|
||
- rusticl/mesa/resource: port to NonNull
|
||
- rusticl/device: fix CL_DEVICE_HALF_FP_CONFIG query
|
||
- rusticl/device: fix default device enumeration
|
||
- rusticl/kernel: take set kernel arguments into account for CL_KERNEL_LOCAL_MEM_SIZE
|
||
- rusticl/kernel: fix image_size of 1D buffer images
|
||
- rusticl/mesa: set take_ownership to true for set_sampler_views
|
||
- rusticl/mesa: add PipeSamplerView wrapper
|
||
- rusticl/mesa: use PipeSamplerView over the raw type
|
||
- rusticl/kernel: create the sampler views earlier
|
||
- rusticl/mem: add functions to create sampler and image views to Image
|
||
- rusticl/mesa: rework image and sampler view creation APIs
|
||
- rusticl/kernel: store memory arguments as Weak references
|
||
- rusticl/device: add unsynchronized mapping functions to helper context
|
||
- rusticl/mem: simplify is_svm implementation
|
||
- rusticl/mem: add Allocation type
|
||
- rusticl/mem: reimplement has_same_parent and rename it to backing_memory_eq
|
||
- rusticl/mem: rework last user of get_parent() and remove it
|
||
- rusticl/mem: add Allocation::is_user_alloc_for_dev
|
||
- rusticl/mem: use get_res_for_access instead of get_res_of_dev
|
||
- trace: copy pipe_caps
|
||
- trace: add get_compute_state_info
|
||
- rusticl/mem: set bind flags for gl imports
|
||
- rusticl/mesa: add PipeContext::device_reset_status
|
||
- rusticl/queue: check device error status
|
||
- rusticl/kernel: call nir_lower_variable_initializers earlier
|
||
- rusticl/mem: do not apply offset with in copy_image_to_buffer
|
||
|
||
Kenneth Graunke (35):
|
||
|
||
- brw: Fix emit_a64_oword_block_header UNIFORM -> VGRF copies
|
||
- brw: Fix try_rebuild_source's ult32/ushr handling to use unsigned types
|
||
- nir: Use load_global_constant for reorderable nir_var_mem_global access
|
||
- nir/algebraic: Reassociate fadd into fmul in DP4-like pattern
|
||
- brw: Drop image deref handling from brw_analyze_ubo_ranges
|
||
- brw: Drop "regular uniform" concept from UBO push analysis
|
||
- brw: Drop a few crocus references in comments
|
||
- brw: Use nir_combined_align in brw_nir_should_vectorize_mem
|
||
- brw: Only consider components read for UBO loads
|
||
- brw: Only consider components read for UBO push analysis
|
||
- brw: Simplify choose_oword_block_size_dwords()
|
||
- nir: Allow large overfetching holes in the load store vectorizer
|
||
- anv: Don't consider nir_var_mem_global for vectorizer robustness checks
|
||
- brw: Tune vectorizer conditions to allow overfetching with holes
|
||
- brw: Fix register unit calculation in SIMD32 LOAD_PAYLOAD lowering
|
||
- brw: Allow SIMD32 math instructions on Xe2
|
||
- brw: Combine convergent texture buffer fetches into fewer loads
|
||
- iris: Tune the BO cache's bucket sizes
|
||
- brw: Don't rely on SIMD splitting in opt_combine_convergent_txfs
|
||
- brw: Limit maximum push UBO ranges to 64 registers in the NIR pass.
|
||
- brw: Don't shrink UBO push ranges in the backend
|
||
- brw: Delete pull constant lowering
|
||
- brw: Delete assign_constant_locations and push_constant_loc[]
|
||
- brw: Fix vectorizer hole_size condition after signedness change
|
||
- nir: Add a nir_def_first_component_read() helper
|
||
- brw: Add more safeguards against misaligned OWord Block messages
|
||
- brw: Skip fetching unread leading components of UBO loads
|
||
- brw: Make get_nir_src_imm() usable for non-32-bit-sizes.
|
||
- brw: Skip unnecessary work for trivial emit_uniformize of IMMs
|
||
- brw: Skip unread leading/trailing components in convergent block loads
|
||
- brw: Add a new MEMORY_MODE_CONSTANT option
|
||
- brw: Allow CSE of MEMORY_MODE_CONSTANT loads
|
||
- brw: Align and combine constant-offset UBO loads in NIR
|
||
- brw: Always use MEMORY_LOAD for load_ubo_uniform_block_intel intrinsics
|
||
- brw: Fix Xe2 spilling code to limit to SIMD32 rather than SIMD16
|
||
|
||
Kevin Chuang (3):
|
||
|
||
- anv: Implement encode shader to fit in ANV BVH
|
||
- anv: Add INTEL_DEBUG for bvh dump and visualization tools
|
||
- anv/bvh: Dump BVH synchronously upon command buffer completion
|
||
|
||
Kevron Rees (1):
|
||
|
||
- anv, drirc: Add workaround to speed up Spiderman reg allocation
|
||
|
||
Konstantin (5):
|
||
|
||
- nir/lower_non_uniform_access: Group accesses using the same resource
|
||
- radv/printf: Guard against helper invocations
|
||
- radv: Do not overwrite VRS rates when doing fast clears
|
||
- vulkan/meta: Add a pipeline cache
|
||
- vulkan: Fix the argument order of update_as
|
||
|
||
Konstantin Seurer (39):
|
||
|
||
- util: Fix some brackets in util_dynarray\_.*_ptr
|
||
- nir: Add missing access flags to print_access
|
||
- radv: Lower non-uniform access after vectorization
|
||
- amd: Add ac_shader_debug_info
|
||
- aco: Handle nir_debug_info_instr
|
||
- aco: Pass debug information to the driver
|
||
- radv: Add a helper for accessing the shader binary
|
||
- radv: Store debug info inside radv_shader
|
||
- radv: Dump nir shaders before compiling
|
||
- nir: Add a first_line parameter to gather_debug_info
|
||
- nir: Do not gather source locations for phis
|
||
- radv: Add RADV_DEBUG=nirdebuginfo
|
||
- gallivm: Add float operation behavior flags to lp_type
|
||
- gallivm: Preserve -0 and nan
|
||
- lavapipe: Implement VK_KHR_shader_float_controls2
|
||
- gallivm: Use an accurate log2 implementation for lodq
|
||
- lavapipe: Implement VK_KHR_compute_shader_derivatives
|
||
- radv: Fix encoding empty acceleration structures
|
||
- llvmpipe: Disable anisotropic filtering for explicit lod
|
||
- llvmpipe: Use a simpler and faster AF implementation
|
||
- llvmpipe: Remove unused AF code
|
||
- llvmpipe: Move max_anisotropy to static sampler state
|
||
- lavapipe: Advertise vulkan 1.4
|
||
- meson: Require glslangValidator when building lavapipe
|
||
- lavapipe: Check the pool type in handle_reset_query_pool
|
||
- meson: Include the loader subdir when building lavapipe
|
||
- gallivm: Take helper invocations into account when skipping branches
|
||
- nir/print: Print less unused shader info
|
||
- nir/tests: Improve shader creation
|
||
- nir/tests: Add a helper for comparing a shader against a string
|
||
- nir/tests: Add reference shaders
|
||
- nir: Add a test runner
|
||
- nir/print: Do not print trailing spaces after preds/succs
|
||
- docs: Add documentation for NIR unit testing
|
||
- llvmpipe: Fix half-pixel sample offset with AF
|
||
- llvmpipe: Avoid a crash when using 5 coords with AF
|
||
- radv/rmv: Use radv_rmv_log_resource_destroy more
|
||
- radv/meta: Stop using strings for meta keys
|
||
- gallivm: Remove loop limiting
|
||
|
||
Koo, Anthony (1):
|
||
|
||
- amd/vpelib: Add system event logging
|
||
|
||
Lars-Ivar Hesselberg Simonsen (26):
|
||
|
||
- panvk: Set fs.multisampled sysval for v10+
|
||
- panvk: Add frag->frag barrier before resolve
|
||
- panvk: update expectations for G610
|
||
- pan/genxml: Fix decode of exception_handler 0x0
|
||
- pan/cs: Add mask support for reg_perm
|
||
- panvk: Build cmd_fb_preload on explicit fb_info
|
||
- panvk: Add incremental rendering support on v10+
|
||
- panfrost: Disable AFRC texture/sampler reswizzle
|
||
- panvk: Disable AFBC for mutable formats on v7
|
||
- panfrost: Only allow AFBC(RGB) and AFBC(BGR) on v7
|
||
- panfrost: Limit reswizzle to AFBC formats
|
||
- panfrost: Decouple reswizzling from texture build
|
||
- panfrost: Standardize naming of sampler reswizzle
|
||
- panvk: Remove ZS texture_swizzle_replicate_x
|
||
- panvk: Fix descriptor decode
|
||
- panvk: Fix valgrind issue in nir_lower_descriptors
|
||
- panvk: Fix valgrind issue in panvk_compile_shaders
|
||
- pan/genxml: Fix vertex_packet Attribute on v9+
|
||
- panvk: Use LD_VAR[_IMM] + ADs for varyings
|
||
- panvk: Limit AD allocation to max var loads in v9+
|
||
- panvk: Use LD_VAR_BUF[_IMM] when possible
|
||
- panvk: Fix barriers in secondary cmdbufs w/o rp's
|
||
- panfrost: Do not evaluate_per_sample for non-MSAA
|
||
- Revert "panfrost: remove is_blit flag"
|
||
- Revert "panfrost: fix hang by using MALI_PIXEL_KILL_WEAK_EARLY in color preload"
|
||
- panvk: Set missing shader_modifies_coverage flag
|
||
|
||
Leder, Brendan Steve (2):
|
||
|
||
- amd/vpelib: Refactor OCSC and update missing check
|
||
- amd/vpelib: Move bg color
|
||
|
||
Leonard Göhrs (1):
|
||
|
||
- ci/lava: update lavacli from version 1.5.2 to 2.2.0
|
||
|
||
Lina Versace (3):
|
||
|
||
- anv: Sort extensions in enablement table
|
||
- anv: Update features.txt
|
||
- anv: Fix feature pipelineProtectedAccess
|
||
|
||
LingMan (10):
|
||
|
||
- mesa: Bump required Rust version to 1.78
|
||
- nak/hw_test: Use std::mem::offset_of!()
|
||
- compiler/rust: Use std::mem::offset_of!()
|
||
- mesa: Add rustfmt.toml
|
||
- rusticl: Use C-string literals
|
||
- rusticl: Use C-string literals for spirv extension names
|
||
- rusticl/cl_prop: Use C-string literals
|
||
- rusticl/core: Use C-string literals for XPlatManager::get_proc_address_func
|
||
- rusticl: Use C-string literals for NirShader::add_var
|
||
- rusticl: Use C-string literals for DiskCache::new
|
||
|
||
Lionel Landwerlin (96):
|
||
|
||
- anv: fix extent computation in image->image host copies
|
||
- anv: update shader descriptor resource limits
|
||
- anv: split generated draw flags from mocs/dword-count
|
||
- intel: make sure intel_wa.h can be included by opencl code
|
||
- anv: implement Wa_16011107343/22018402687 for generated draws
|
||
- brw: allocate physical register sizes for spilling
|
||
- anv: fix descriptor asserts
|
||
- anv: fix incorrect aspect flag for depth/stencil formats
|
||
- anv: fix missing push constant reallocation
|
||
- anv: prevent access to destroyed vk_sync objects post submission
|
||
- anv: track allocated descriptor pool sizes
|
||
- anv: indent driconf code
|
||
- anv: add a workaround for X4 Foundations
|
||
- anv: document the X4 Foundations workaround a bit more
|
||
- anv: move helpers out of genX_pipeline.c/anv_private.h
|
||
- anv: remove 3DSTATE_RASTER from pipeline
|
||
- anv: remove 3DSTATE_MULTISAMPLE from the pipeline
|
||
- anv: remove 3DSTATE_VF_STATISTICS from pipeline
|
||
- anv: pass anv_device to batch_set_preemption
|
||
- anv: rework vertex input helper
|
||
- anv: split vertex buffer emission in a different function
|
||
- anv: move gfx tracking values to anv_cmd_graphics_state
|
||
- anv: move tracking of tcs_input_vertices/fs_msaa_flags to hw state
|
||
- anv: split runtime flushing code for reuse
|
||
- brw: change fs_msaa flags checks to test compiled flag first
|
||
- brw: rename brw_sometimes to intel_sometimes
|
||
- brw: move barycentric_mode enum to intel_shader_enums.h
|
||
- brw: move fs_msaa_flags logic to intel_shader_enums.h
|
||
- fix
|
||
- Revert in correct commit "fix"
|
||
- anv: move primitive_topology to anv_gfx_dynamic_state
|
||
- anv: try to avoid using cmd_buffer in gfx runtime flushing
|
||
- anv: reuse device local variable in hw state emission
|
||
- anv: rework Wa_18038825448 to track state on anv_gfx_dynamic_state
|
||
- anv: avoid using cmd_buffer for TBIMR state computation
|
||
- anv: avoid using cmd_buffer for flushing runtime
|
||
- anv/iris: leave 4k alignments for clear colors with modifiers
|
||
- brw: use transpose unspill messages when possible
|
||
- anv: report formats supported by the common bvh framework
|
||
- anv: fix missing bindings valid dynamic state change check
|
||
- anv: set pipeline flags correct for imported libs
|
||
- vulkan: make acceleration structure debug markers virtual
|
||
- vulkan: add an enum for the build step
|
||
- vulkan: track encode step of the BVH building
|
||
- anv: add BVH building tracking through u_trace
|
||
- intel/decoder: fix COMPUTE_WALKER handling
|
||
- anv: document UBO descriptor range alignments
|
||
- blorp: use 2D dimension for 1D tiled images
|
||
- hk: fix timeline value type
|
||
- anv: fix index buffer size changes
|
||
- anv: limit the memcpy data for push constants
|
||
- vulkan/runtime: avoid emitting empty build_leaves
|
||
- anv: add tracepoints timestamp mode for empty dispatches
|
||
- anv: rework tbimr push constant workaround
|
||
- anv: ensure null-rt bit in compiler isn't used when there is ds attachment
|
||
- anv: use the correct MOCS for depth destinations
|
||
- intel: fix generation shader on Gfx9
|
||
- brw: introduce a new register type for the address register
|
||
- brw: use phys_nr() more in generation
|
||
- brw: split validation iteration into blocks
|
||
- brw: add infra to make use of the address register in the IR
|
||
- brw: add scheduler support for address registers
|
||
- brw: avoid having the scratch surface handle partially written
|
||
- brw: move final send lowering up into the IR
|
||
- brw: fix coarse_z computation on Xe2+
|
||
- brw: handle load_printf_buffer_size intrinsic
|
||
- anv: handle printf buffer size relocations
|
||
- nir: make lower-level printf helper respect buffer size
|
||
- anv: update debug printf example code
|
||
- anv: remove print lowering
|
||
- blorp: disable PS shaders with depth/stencil HiZ ops
|
||
- brw: fix CSE with negation
|
||
- anv: don't look at pipelines to figure out CPS values
|
||
- compiler: add VARYING_BIT_PRIMITIVE_COUNT
|
||
- anv/Wa_18019110168: copy the primitive count writes
|
||
- anv/brw: rework primitive count writing
|
||
- libcl: add MIN2/MAX2 macros
|
||
- libcl_vk: add some vulkan enums/structures for DGC
|
||
- spirv: build vtn_bindgen for Anv/Iris
|
||
- brw/elk: move internal kernel parsing out of intel_clc
|
||
- meson: build mesa_clc for Anv/Iris
|
||
- intel/cl: switch to SPIRV as shader storage
|
||
- meson: rework mesa-clc=system handling
|
||
- intel: rework CL pre-compile
|
||
- meson: required SPIRV-Tools LLVM workaround on LLVM17+
|
||
- intel: fix dependency for internal CL shaders
|
||
- anv: use flags for format capabilities
|
||
- anv: pass physical device to format helpers
|
||
- anv: add a drirc to disable border colors without format
|
||
- anv: expose A4B4G4R4_UNORM_PACK16 support with CBCWF is disabled
|
||
- anv: dirty pipeline & push constants after internal CS shaders
|
||
- anv: reduce alignment for small heaps
|
||
- brw: fixup scoreboarding for find_live_channels
|
||
- anv,driconf: Add sampler coordinate precision workaround for Dynasty Warriors
|
||
- anv: disable VF statistics for memcpy
|
||
- anv: ensure Wa_16012775297 interacts correctly with Wa_18020335297
|
||
|
||
Lorenzo Rossi (1):
|
||
|
||
- nvk: fix preprocess buffer alignment
|
||
|
||
Louis-Francis Ratté-Boulianne (3):
|
||
|
||
- panfrost: Split up allocation and packing of tiler descriptor
|
||
- panfrost: Select the effective tile size as part of pan_fb_info
|
||
- panfrost: Re-emit texture descriptor if the data size has changed
|
||
|
||
Lu Yao (1):
|
||
|
||
- zink: fix decomposed_attrs val error when zink_vs_key->size is 4
|
||
|
||
Lucas De Marchi (1):
|
||
|
||
- intel/tools: Fix Xe KMD error dump parser
|
||
|
||
Lucas Stach (26):
|
||
|
||
- etnaviv: drm: properly handle BO list member
|
||
- etnaviv: drm: assert mutual exclusivity between cache and zombie list
|
||
- etnaviv: drm: use list_first_entry
|
||
- etnaviv: stall after RS/BLT operation when draw_stall debug option is enabled
|
||
- etnaviv: Update headers from rnndb
|
||
- etnaviv: add debug switch to disable texture descriptor usage
|
||
- etnaviv: fix polygon offset for 24bpp depth buffers
|
||
- ci/etnaviv: drop gl-1.4-polygon-offset fail
|
||
- etnaviv: isa: fix typo in SRC2_USE map
|
||
- etnaviv: Update headers from rnndb
|
||
- etnaviv: clean up component use setting in linker
|
||
- etnaviv: fix flatshading
|
||
- etnaviv: emit full varying component use
|
||
- ci/etnaviv: drop GC2000 flat shading fails
|
||
- etnaviv: split dummy RT backing store from reloc
|
||
- etnaviv: fix rendering without vertex buffers/attributes
|
||
- ci/etnaviv: drop failures caused by missing vertex attributes
|
||
- etnaviv: fix polygon offset disable
|
||
- etnaviv: memcpy varying setup from stack
|
||
- etnaviv: emit varying interpolation state on halti5
|
||
- etnaviv: fix flatshading on halti5 GPUs
|
||
- etnaviv: only emit used PA_SHADER_ATTRIBUTES states
|
||
- etnaviv: track TS flushed status as bool
|
||
- etnaviv: dynamically partition the constant memory in unfied uniform mode
|
||
- etnaviv: allow more constants in unified uniform mode
|
||
- etnaviv: hwdb: fix lookup of GC3000 in i.MX6QP
|
||
|
||
Lukas Lipp (1):
|
||
|
||
- wsi: Fix wrong function name for lvp wsi metal surface
|
||
|
||
M Henning (6):
|
||
|
||
- nvk/cmd_buffer: Pass count to set_root_array
|
||
- nvk: Fix invalidation of NVK_CBUF_TYPE_DYNAMIC_UBO
|
||
- nvk: Remove params for dirty_cbufs_for_descriptors
|
||
- nvk: Fix two typos in comments
|
||
- nvk: Fix uninitialized var warnings in host_copy
|
||
- nak/hw_runner: Skip copy call for empty buffer
|
||
|
||
Manuel (1):
|
||
|
||
- gfxstream: Avoid repeated functionality
|
||
|
||
Manuel Dun (4):
|
||
|
||
- gfxstream: Using DETECT_OS_ANDROID from util instead of __ANDROID__
|
||
- gfxstream: Using DETECT_OS_FUCHSIA from util instead of __Fushsia__
|
||
- gfxstream: Using DETECT_OS_LINUX from util instead of __linux__
|
||
- Gfxstream: Initial mingw "compilable" Windows version of mesa/gfxstream
|
||
|
||
Marc Herbert (5):
|
||
|
||
- docs: add "apt-get build-dep" and "dnf buildep"
|
||
- docs: cross-compile: add useful "apt" and "dnf" builddep commands
|
||
- docs: show how to use ccache when cross-compiling
|
||
- docs: show which pkg-config Fedora uses for cross-compilation
|
||
- docs: move cross c*_args from [properties] to [built-in options]
|
||
|
||
Marek Olšák (353):
|
||
|
||
- gallium/radeon: import libdrm_radeon source code, drop the dependency
|
||
- aco: remove unused TCS fields from aco_shader_info
|
||
- ac/nir: get pass_tessfactors_by_reg from nir_gather_tcs_info
|
||
- radeonsi: fix passing TCS wave ID from LS to HS for monolithic LS+HS
|
||
- radeonsi: don't overwrite info.tess._primitive mode when it can be correct
|
||
- radeonsi: get the value for load_tcs_primitive_mode_amd from shader info
|
||
- radeonsi: replace are_tessfactors_def_in_all_invocs with nir_gather_tcs_info
|
||
- radeonsi: reduce si_shader_key_ge::tes_prim_mode size to 2 bits
|
||
- radeonsi: remove unused function si_get_tcs_out_patch_stride
|
||
- radeonsi: don't set tess level outputs in patch_outputs_written unconditionally
|
||
- radeonsi: remove unused si_shader_info::output_readmask
|
||
- radeonsi: set \*outputs_written in scan_io_usage instead of later
|
||
- radeonsi: split outputs_written_before_tes_gs into ls_es_* and tcs_* masks
|
||
- radeonsi/ci: update navi31 failures
|
||
- glsl: add a helper for duplicated code calling nir_opt_varyings
|
||
- gallium: use struct nir_shader * type in finalize_nir instead of void *
|
||
- st/mesa: call pipe_screen::finalize_nir outside of st_finalize_nir
|
||
- gallium: add PIPE_CAP_CALL_FINALIZE_NIR_IN_LINKER
|
||
- st/mesa: add ST_DEBUG=xfb printing xfb info
|
||
- mesa: capture shaders to disk before invoking the linker
|
||
- nir/opt_varyings: add nir_io_always_interpolate_convergent_fs_inputs
|
||
- nir/opt_varyings: add nir_io_compaction_rotates_color_channels
|
||
- nir/opt_varyings: fix packing color varyings
|
||
- nir/opt_varyings: implement compaction without flexible interpolation
|
||
- nir/opt_varyings: don't count the cost of the same instruction multiple times
|
||
- radeonsi: fix buffer_size for emulated GS statistics
|
||
- radeonsi: fix an assertion failure in si_shader_ps with AMD_DEBUG=mono
|
||
- radeonsi: handle nir_intrinsic_component in kill_ps_outputs
|
||
- radeonsi: fix gl_FrontFace elimination when one side is culled
|
||
- radeonsi/ci: add options to test llvmpipe, softpipe, virgl, zink
|
||
- nir/print: print fb_fetch_output for variables
|
||
- nir/lower_pntc_ytransform: handle lowered IO
|
||
- nir/lower_clip: fixes for lowered IO without compact arrays
|
||
- nir/lower_clip: rewrite find_output to handle vec2/3 and make it readable
|
||
- nir/lower_fragcoord_wtrans: handle trimmed fragcoord loads
|
||
- nir/lower_two_sided_color: fix for lowered IO
|
||
- nir: add nir_io_semantics::fb_fetch_output_coherent
|
||
- nir: rename nir_io_glsl_opt_varyings to nir_io_dont_optimize and deprecate it
|
||
- nir: add nir_io_separate_clip_cull_distance_arrays to replace PIPE_CAP
|
||
- vc4/lower_blend: don't read non-existent channels
|
||
- nir: make use_interpolated_input_intrinsics a nir_lower_io parameter
|
||
- ac/surface: adjust HiZ enablement
|
||
- radeonsi: prepare for making SI_NGG_CULL_TRIANGLES/LINES VS only, rename them
|
||
- radeonsi: optionally return MESA_PRIM_UNKNOWN from si_get_input_prim
|
||
- radeonsi: rewrite/replace gfx10_ngg_get_vertices_per_prim
|
||
- radeonsi: return a better value for load_initial_edgeflags_amd
|
||
- radeonsi: clean up and rename gfx10_edgeflags_have_effect
|
||
- radeonsi: add helper si_shader_culling_enabled
|
||
- radeonsi: only compute and use min_direct_count on gfx7-8
|
||
- radeonsi: enable NGG culling for non-monolithic TES and GS
|
||
- radeonsi: don't use nir_io_dont_optimize because it's deprecated
|
||
- r300: don't lower sin/cos in finalize_nir
|
||
- nir/opt_varyings: use a hash table to make cloning SSA faster
|
||
- amd: import libdrm_amdgpu ioctl wrappers
|
||
- util,amd: add inlinable versions of drmIoctl/drmCommandWrite*
|
||
- nir: allow cloning indirect array derefs in nir_clone_deref_instr
|
||
- nir/lower_io_to_temporaries: fix interp_deref_at_* lowering
|
||
- radeonsi: don't call set_framebuffer_state in si_destroy_context
|
||
- radeonsi: handle a failure to create gfx_cs
|
||
- winsys/amdgpu: fix FD mismatch
|
||
- Revert "gbm: mark surface buffers as explicit flushed"
|
||
- nir/lower_clip: don't set cursor to fix crashes due to removed instructions
|
||
- nir/lower_clip: separate code for IO variables and intrinsics
|
||
- nir/lower_clip: set clip_distance_array_size outside of create_clipdist_vars
|
||
- nir/lower_clip: convert nir_lower_clip_gs to nir_shader_intrinsics_pass
|
||
- nir/lower_clip: implement ClipVertex lowering for GS + lowered IO correctly
|
||
- vc4: lower clip planes in st/mesa
|
||
- nir/opt_varyings: always call remove_dead_varyings in init_linkage
|
||
- nir/opt_varyings: add a default callback for varying_estimate_instr_cost
|
||
- nir/opt_varyings: replace options::lower_varying_from_uniform with a cost number
|
||
- nir/algebraic: use is_used_once in a few iand/ior patterns
|
||
- nir/algebraic: optimize (a & b) & (a & c) ==> (a & b) & c
|
||
- nir/algebraic: optimize (a | b) | (a | c) ==> (a | b) | c
|
||
- nir/algebraic: optimize (a & b) | (a | c) => a | c, (a & b) & (a | c) => a & b
|
||
- gallium: replace PIPE_SHADER_CAP_INDIRECT_INPUT/OUTPUT_ADDR with NIR options
|
||
- st/mesa: replace EmitNoIndirectInput / EmitNoIndirectOutput with NIR options
|
||
- util/bitset_test: test the return value of BITSET_TEST_RANGE_INSIDE_WORD better
|
||
- util/bitset: add BITSET_GET_RANGE_INSIDE_WORD
|
||
- nir/linking_helpers: don't promote interpolated varyings to flat
|
||
- nir/opt_varyings: remove redundant conditions from a while loop
|
||
- nir/opt_varyings: fix compaction with sparse indirect FS inputs
|
||
- nir/opt_varyings: count the number of unused components for compaction correctly
|
||
- nir/opt_varyings: fix max_slot for color varying compaction
|
||
- nir/opt_varyings: make top-level compaction code for TES, TCS, GS separate
|
||
- nir/opt_varyings: change try_move_postdominator param to nir_instr type
|
||
- amd,zink: remove options.varying_estimate_instr_cost callbacks
|
||
- nir/opt_varyings: propagate indirect uniform/UBO loads into the next shader
|
||
- nir/opt_varyings: add inter-shader code motion for uniform/UBO indexing
|
||
- nir/opt_varyings: fix getting deref variables for sysvals
|
||
- nir/opt_varyings: remove rare dead output stores after inter-shader code motion
|
||
- nir/opt_varyings: fix compile failures in the disabled PRINT code
|
||
- amd/ci: add piglit failures due to a overzealous test
|
||
- nir/lower_io_passes: lower indirect IO for TCS
|
||
- radeonsi: pass cull face state via user SGPRs for shader culling
|
||
- radeonsi: revert to always returning true for load_cull_any_enabled_amd
|
||
- radeonsi: try to fix Navi14 regression in debug builds
|
||
- radeonsi: don't compute total_direct_count in si_draw if it's unused
|
||
- radeonsi/ci: handle glinfo errors better
|
||
- radeonsi/ci: stop using a global flakes list, only use a per-chip flakes list
|
||
- radeonsi/ci: remove most flakes and some skips, update navi31 failures
|
||
- radeonsi/ci: remove --slow
|
||
- radeonsi/ci: update navi31 failures
|
||
- r600: fix a constant buffer memory leak for u_blitter
|
||
- ac/lower_ngg: improve streamout code generation for gfx12/ACO to match LLVM
|
||
- ac: update SPI_GRP_LAUNCH_GUARANTEE_* register values for gfx12
|
||
- ac/surface/gfx12: enable DCC 256B compressed blocks and reorder modifiers
|
||
- radeonsi/gfx12: set DB_RENDER_OVERRIDE based on stencil state
|
||
- radeonsi/gfx12: adjust HiZ/HiS logic
|
||
- ac/nir: reserve the first LDS vec4 for the HS tf0/1 group vote in TCS
|
||
- ac/nir: use s_sendmsg(HS_TESSFACTOR) to optimize writing tess factors for gfx11
|
||
- ac/nir: allow a TCS input to be available from both VGPRs and LDS
|
||
- ac,radv,radeonsi: enable TCS input reads from VGPRs for all compatible loads
|
||
- ac/nir: add new helpers for computing the TCS LDS/offchip size accurately
|
||
- radeonsi: remove unused parameter tcs_vgpr_only_inputs from si_get_nir_shader
|
||
- radeonsi: switch to the new TCS LDS/offchip size computation
|
||
- radv: switch to the new TCS LDS/offchip size computation
|
||
- ac/nir: call nir_gather_tcs_info only once for RADV
|
||
- nir/opt_varyings: set all IO types to float to facilitate full vectorization
|
||
- nir/opt_varyings: clear info->clip/cull_distance_array_size if relocated
|
||
- st/mesa: don't use nir_opt_fragdepth because it's incorrect with MSAA
|
||
- mesa: set correct XFB prim mode for draw validation after resuming XFB
|
||
- mesa: fix printing _NEW_* flags
|
||
- gallium: pass XFB primitive mode to set_stream_output_targets
|
||
- st/mesa: add a pass that unlowers IO intrinsics to variables
|
||
- glsl,st/mesa: always lower IO for GLSL, unlower IO for drivers
|
||
- v3d: enable uniform expression propagation from outputs to the next shader
|
||
- ci: update fail lists and trace checksums
|
||
- virgl/ci: disable virgl-traces because it doesn't upload results
|
||
- radeonsi/ci: don't copy skips.csv to the results directory
|
||
- radeonsi/ci: update failures and flakes
|
||
- radeonsi: fix a gfx10.3 regression due to a gfx12 change
|
||
- radeonsi: kill Z and stencil PS outputs if depth or stencil is disabled
|
||
- radeonsi/gfx11: fix alpha-to-coverage + alpha-to-one used together
|
||
- radeonsi: fix alpha-to-coverage + alpha-to-one used together for gfx6-10.3
|
||
- radeonsi: implement nir_opt_frag_depth using kill_z instead of the NIR pass
|
||
- radeonsi: eliminate shader code computing killed Z/S/samplemask PS outputs
|
||
- radeonsi: make NGG streamout output primitive type known at compile time
|
||
- radeonsi/gfx12: fix DrawTransformFeedback(stream != 0)
|
||
- radeonsi/gfx12: tune streamout performance
|
||
- radeonsi: make nir->info and si_shader_info::base identical
|
||
- radeonsi: remove some uses of enum pipe_shader_type
|
||
- radeonsi: make si_init_shader_args static
|
||
- radeonsi: call si_init_shader_args in si_get_nir_shader
|
||
- radeonsi: use nir->info instead of sel->info.base
|
||
- radeonsi: disable luminance alpha formats on gfx6
|
||
- radeonsi,radv: fix incorrect min_esverts for NGG subgroup calculation
|
||
- ac: remove unused code
|
||
- ac/llvm: remove unused code
|
||
- radeonsi/ci: update failures
|
||
- radeonsi: fix a TCS regression
|
||
- radeonsi: switch si_get_blitter_vs to IO intrinsics
|
||
- radeonsi: remove unused code
|
||
- amd: update addrlib
|
||
- radeonsi: fix a front face regression (crash)
|
||
- nir/opt_load_store_vectorize: make hole_size signed to indicate overlapping loads
|
||
- radv: reduce maxGeometryShaderInvocations to 32
|
||
- ac/nir: handle disabled PS VGPRs in ac_nir_load_arg_at_offset
|
||
- amd: lower load_pixel_coord in NIR
|
||
- amd: lower load_frag_coord in NIR
|
||
- amd: lower load_local_invocation_id in NIR
|
||
- amd: lower load_first_vertex/base_instance/draw_id/view_index in NIR
|
||
- amd: lower load_invocation_id in NIR
|
||
- amd: lower load_sample_id in NIR
|
||
- amd: lower load_sample_pos in NIR
|
||
- amd: lower load_frag_shading_rate in NIR
|
||
- amd: lower load_front_face in NIR
|
||
- ac,radeonsi: move load_vector_arg flags to common code
|
||
- amd: lower load_barycentric_pixel/centroid/sample in NIR
|
||
- amd: lower load_barycentric_at_offset in NIR
|
||
- amd: lower load_gs_wave_id_amd in NIR
|
||
- amd: lower load_vertex_id/instance_id and overwrite_vs_arguments in NIR
|
||
- radeonsi: don't return 0 from si_get_max_workgroup_size
|
||
- ac/nir: extract a load_subgroup_id lowered helper
|
||
- amd: lower load_local_invocation_index in NIR
|
||
- amd: lower load_subgroup_invocation in NIR
|
||
- amd: lower load_tess_rel_patch_id/primitive_id/tess_coord and overwrite.. in NIR
|
||
- ac/llvm: remove already lowered cases
|
||
- ac/nir: lower more loads in ac_nir_lower_intrinsics_to_args instead of drivers
|
||
- ac/nir: clean up ac_nir_lower_indirect_derefs
|
||
- ac/nir: add helper ac_nir_load_arg_upper_bound
|
||
- ac/nir: set arg_upper_bound_u32 for vs_rel_patch_id
|
||
- ac/nir: split local_invocation_ids to 3 separate VGPR inputs
|
||
- ac/nir: set upper ranges for range analysis while lowering system values
|
||
- radeonsi: lower sysval intrinsics as late as possible
|
||
- amd: optimize atomics before lowering intrinsics
|
||
- radeonsi: use nir_opt_sink
|
||
- radeonsi: use nir_opt_move
|
||
- vulkan: silence an unused variable warning
|
||
- llvmpipe: silence an unused result warning
|
||
- util/disk_cache: silence unused result warnings
|
||
- nir: set nir_io_semantics::num_slots to at least 1 in build helpers
|
||
- nir: set src_type and dest_type to float implicitly for IO build helpers
|
||
- nir: don't set num_slots/src/dest_type/write_mask when they're set automatically
|
||
- nir: flip the early exit condition in nir_lower_io_temporaries
|
||
- nir: remove redundant option linker_ignore_precision
|
||
- nir: use IO intrinsics in nir_lower_bitmap
|
||
- nir: use IO intrinsics in nir_lower_drawpixels
|
||
- mesa: remove unused PROGRAM_SYSTEM_VALUE
|
||
- mesa: remove unused PROGRAM_WRITE_ONLY
|
||
- st/mesa: fold st_translate_prog_to_nir into prog_to_nir
|
||
- st/mesa: run DCE before st_unlower_io_to_vars
|
||
- st/mesa: use IO intrinsics in st_nir_lower_fog
|
||
- st/mesa: use IO intrinsics in st_nir_lower_position_invariant
|
||
- st/mesa: switch ATI_fs to IO intrinsics
|
||
- st/mesa: unlower IO for internal shaders if needed
|
||
- st/mesa: switch Z/S DrawPixels shaders to IO intrinsics
|
||
- st/mesa: switch GL_SELECT shader to IO intrinsics
|
||
- st/mesa: switch st_nir_make_passthrough_shader to IO intrinsics
|
||
- st/mesa: switch st_pbo_create_vs and st_pbo_create_gs to IO intrinsics
|
||
- st/mesa: switch PBO create_fs to IO intrinsics
|
||
- st/mesa: switch st_nir_make_clearcolor_shader to IO intrinsics
|
||
- st/mesa: don't use nir_copy_var
|
||
- st/mesa: recompute IO bases for ARB_vp/fp
|
||
- glsl: remove unused code
|
||
- glsl: fix corruption due to blake3 hash not being set for nir_opt_undef
|
||
- radeonsi: ignore PIPE_RESOURCE_FLAG_TEXTURING_MORE_LIKELY for TC-compatible HTILE
|
||
- radeonsi: simplify and fix enable_tc_compatible_htile_next_clear logic
|
||
- radeonsi: re-enable non-TC-compatible HTILE for write-only Z/S
|
||
- mesa: switch ARB_vp/fp to IO intrinsics
|
||
- mesa: switch fixed-func fragment program to IO intrinsics
|
||
- nir/algebraic: use is_used_once for comparison patterns
|
||
- nir/algebraic: add and improve pack/unpack patterns
|
||
- nir/algebraic: optimize pack_split(unpack(a).x, unpack(a).y) -> a
|
||
- radeonsi: fix a perf regression due to slow reply from GEM_WAIT_IDLE for timeout=0
|
||
- radeonsi: always use RADEON_USAGE_DISALLOW_SLOW_REPLY
|
||
- ac: update ATOMIC_MEM definitions
|
||
- ac/nir: sort xfb info to facilitate vectorization of xfb stores
|
||
- ac/nir: vectorize streamout stores for legacy pipeline optimally
|
||
- ac/nir/ngg: vectorize streamout stores for NGG optimally
|
||
- ac/nir/ngg: fold so_vertex_index * so_stride into immediate offset
|
||
- ac/nir/ngg: export positions after streamout to improve performance
|
||
- ac,radeonsi: scalarize overfetching loads
|
||
- radeonsi: lower descriptors sooner to allow vectorizing descriptor loads
|
||
- amd: vectorize SMEM loads aggressively, allow overfetching for ACO
|
||
- radeonsi: don't set BREAK_PRIMGRP/WAVE_AT_EOI when tessellation is disabled
|
||
- radeonsi: only set BREAK_PRIMGRP/WAVE_AT_EOI when TES/GS need PrimID sysval after TES
|
||
- radeonsi/gfx12: enable alt_hiz_logic
|
||
- radeonsi/gfx12: set DIS_PG_SIZE_ADJUST_FOR_STRIP after shader compilation
|
||
- radeonsi/gfx12: use ACO if LLVM is 19 or older
|
||
- radeonsi/gfx12: use ACO for streamout because it's faster
|
||
- mesa: rework enablement of force_gl_names_reuse
|
||
- mesa: enable GL name reuse by default for all drivers except virgl
|
||
- ac/nir: remove broadcast_last_cbuf because it can be deduced from NIR
|
||
- ac/nir: split ac_nir_lower_ps into 2 passes
|
||
- nir: add barycentric coordinates src to load_point_coord_maybe_flipped
|
||
- ac: use Z_EXPORT_FORMAT=32_AR for Z + Alpha mrtz exports
|
||
- ac/llvm: lower vector load_const in NIR
|
||
- ac/llvm: remove the low-optimizing compiler option
|
||
- radeonsi: add si_screen::use_aco to shader cache key to fix shader cache failures
|
||
- radeonsi: remove unused variables from si_shader_context (LLVM)
|
||
- radeonsi: make many shader functions static or move them to .c files
|
||
- radeonsi: remove unused functions
|
||
- nir: add next_stage param to nir_slot_is_varying & nir_remove_sysval_output
|
||
- Revert "ac/llvm: enable wqm for ac_build_quad_swizzle from ac_build_fs_interp_mov"
|
||
- nir: add a pass that moves output stores to the end of the shader
|
||
- st/mesa: move VS & TES output stores to the end before unlowering IO
|
||
- mesa: switch fixed-func vertex program to IO intrinsics
|
||
- st/mesa: assert that all incoming shaders use lowered IO
|
||
- st/mesa: remove dead/no-op code due to IO being always lowered
|
||
- glsl: remove dead code due to IO being always lowered
|
||
- glsl: simplify nir_lower_io_to_temporaries logic
|
||
- nir: remove dead code due to IO being always lowered in st/mesa
|
||
- st/mesa: inline st_finalize_nir_before_variants
|
||
- nir: remove handling IO variables from passes used by st/mesa
|
||
- gallium/u_threaded: move tc_batch_execute after all call functions
|
||
- gallium/u_threaded: make the execute function table private
|
||
- gallium/u_threaded: use TC_END_BATCH to terminate the loop
|
||
- gallium/u_threaded: replace the function table with a switch and direct calls
|
||
- gallium/u_threaded: inline all tc_call functions
|
||
- gallium/u_threaded: sort cases in batch_execute by their occurrence
|
||
- zink/ci: skip KHR-Single-GL46...SizedDeclarationsPrimitive due to random timeout
|
||
- dri: put shared-glapi into libgallium.*.so
|
||
- glapi: stop using the remap table
|
||
- glapi: remove the remap table
|
||
- loader: improve the existing loader-libgallium non-matching version error
|
||
- glapi: rename exported symbols so as not to conflict with old libglapi
|
||
- freedreno/ci: skip a dmat3 div test timing out
|
||
- radv: don't call ac_nir_lower_ps_early
|
||
- ac/nir: optimize front_face in ac_nir_lower_ps_early
|
||
- ac/nir: lower sample_pos in ac_nir_lower_ps_early
|
||
- ac/nir: lower barycentric_at_offset/sample in ac_nir_lower_ps_early
|
||
- ac/nir: lower fbfetch_output in ac_nir_lower_ps_early
|
||
- ac/nir: return progress from ac_nir_lower_ps_early
|
||
- ac/nir: return progress from ac_nir_lower_ps_late
|
||
- ac/nir: handle FRAG_RESULT_COLOR with dual src blending in ac_nir_lower_ps_early
|
||
- ac/nir: switch passes to use nir_shader_intrinsics_pass
|
||
- ac/nir: drop 16x EQAA support from ac_get_ps_iter_mask
|
||
- ac/nir: clamp vertex color outputs in the right place
|
||
- radeonsi: sample shading state fixes
|
||
- ac,aco,radeonsi: replace SampleMaskIn with 1 << SampleID if full sample shading
|
||
- ac/nir: simplify force_*_sample_interp options in ac_nir_lower_ps_early
|
||
- ac/nir: simplify force_*_center_interp options in ac_nir_lower_ps_early
|
||
- ac/nir: optimize barycentric_at_sample(sample_id) in ac_lower_ps_early
|
||
- ac/nir: optimize frag_coord <-> pixel_coord in ac_nir_lower_ps_early
|
||
- ac/nir: eliminate sample_mask_in without MSAA in ac_nir_lower_ps_early
|
||
- ac/nir: cosmetic stuff for ac_nir_lower_ps
|
||
- aco: implement replacing frag_coord with pixel_coord in PS prolog
|
||
- aco: simplify how broadcast_last_cbuf is implemented in PS epilog
|
||
- aco: implement replacement of sample_mask_in with helper_invocation in PS prolog
|
||
- ac/nir: compute ddx/ddy for barycentric_at_offset at the beginning of shaders
|
||
- ac/nir: lower sample_pos to load_sample_positions_amd when frag_coord is center
|
||
- nir/opt_varyings: handle user barycentrics
|
||
- mesa: enable GL name reuse for virgl
|
||
- radeonsi: disallow compute queues on Raven/Raven2 due to hangs
|
||
- ac/nir: clamp vertex color outputs in the right place
|
||
- radeonsi: get sample positions from user SGPRs instead of memory
|
||
- radeonsi: fix PS prolog not counting used fragcoord VGPRs correctly
|
||
- radeonsi: implement replacing frag_coord with pixel_coord at draw time
|
||
- radeonsi: don't set the alpha ref user SGPR if alpha test doesn't use it
|
||
- radeonsi: simplify how broadcast_last_cbuf is implemented for PS epilogs
|
||
- radeonsi: use load_pixel_coord for polygon stipple lowering
|
||
- radeonsi: remove si_nir_kill_ps_outputs and use ac_nir_lower_ps_early instead
|
||
- radeonsi: add load_polygon_stipple_buffer_amd instead of using si_shader_args
|
||
- radeonsi: call si_init_gs_output_info in si_get_nir_shader
|
||
- radeonsi: add si_nir_shader_ctx holding parameters from si_get_nir_shader
|
||
- radeonsi: call si_nir_late_opts unconditionally
|
||
- radeonsi: set the "first" parameter of si_nir_opts correctly
|
||
- radeonsi: simplify how the NIR name of shader variants is modified
|
||
- radeonsi: cosmetic changes in get_nir_shader
|
||
- radeonsi: reorder NIR passes in get_nir_shader (part 1)
|
||
- radeonsi: reorder NIR passes in get_nir_shader (part 2)
|
||
- radeonsi: reorder NIR passes in get_nir_shader (part 3)
|
||
- radeonsi: split and restructure get_nir_shader
|
||
- radeonsi: get LS+HS and ES+GS together in get_nir_shader instead of separately
|
||
- radeonsi: set uses_vmem_load/sampler in get_nir_shaders
|
||
- radeonsi: move/rewrite PS color input gathering for shader variants
|
||
- radeonsi: use barycentrics from load_point_coord_maybe_flipped
|
||
- radeonsi: lower indirect indexing sooner
|
||
- radeonsi: move spi_ps_input_config functions up
|
||
- radeonsi: split si_fixup_spi_ps_input_config
|
||
- radeonsi: get SPI_PS_INPUT_ENA from shader variant NIR for ACO
|
||
- radeonsi: minor restructuring of si_llvm_compile_shader
|
||
- radeonsi: verify that SPI_PS_INPUT_ENA from LLVM is equal to ACO
|
||
- radeonsi: remove ac_shader_config from si_shader_part
|
||
- radeonsi: precompute COMPUTE_PGM_RSRC3
|
||
- radeonsi: set SHARED_VGPR_CNT for compute for ACO
|
||
- radeonsi: set SHARED_VGPR_CNT for gfx shaders for ACO
|
||
- radeonsi: gather PS inputs from shader variant NIR
|
||
- radeonsi: don't set BASE in si_nir_lower_ps_color_input
|
||
- radeonsi: remove si_shader_info code that is no longer needed
|
||
- radeonsi: implement replacement of sample_mask_in with helper_invocation
|
||
- radeonsi: ignore pipe_rasterizer_state::force_persample_interp
|
||
- radeonsi: fix interpolateAt* with non-GL4 ARB_sample_shading
|
||
- radeonsi/ci: add more gfx11 flakes
|
||
- radeonsi: set gl_FragCoord to pixel center to fix GLCTS failures
|
||
- radeonsi: validate BITSET_TEST_RANGE_INSIDE_WORD assertion at compile time
|
||
- radeonsi: remove SI_TRACKED__UNUSED_GAP
|
||
- radeonsi: dead code removal and move some code out of headers
|
||
- radeonsi: remove redundant divergence analysis and smem flagging
|
||
- radeonsi: remove an incorrectly defined modifier
|
||
- winsys/amdgpu: disable DCC for gfx12 when using AMD_FORCE_FAMILY
|
||
- ac/fake_hw_db: deobfuscate GPU name strings
|
||
- gallium,st/mesa: allow reporting compile failures from create_vs/fs/.._state
|
||
|
||
Mark Collins (5):
|
||
|
||
- util: Add file modification notifier utility
|
||
- tu/util: Support toggling TU_DEBUG options at runtime
|
||
- tu/lrz: Check for TU_DEBUG(nolrz) late
|
||
- freedreno/docs: Document TU_DEBUG_FILE
|
||
- util/u_debug: Ignore newlines in \`parse_*_string`
|
||
|
||
Martin Krastev (7):
|
||
|
||
- svga/ci: enable vmware farm
|
||
- svga/ci: set vmware piglit job parallelism to 2
|
||
- svga/ci: triage piglit failures
|
||
- svga/ci: update svga/ci KERNEL_TAG
|
||
- svga/ci: drop FDO_CI_CONCURRENT to 1
|
||
- svga/ci: disable vmware farm
|
||
- svga/ci: enable vmware farm
|
||
|
||
Martin Roukala (né Peres) (39):
|
||
|
||
- zink/ci: document new-ish vangogh flakes
|
||
- ci: disable mupuf's farm
|
||
- Revert "ci: disable mupuf's farm"
|
||
- ci: disable mupuf's farm
|
||
- Revert "ci: disable mupuf's farm"
|
||
- freedreno-ci: document more a618-gl flakes
|
||
- freedreno-ci: document a a750-gl flake
|
||
- turnip/ci: document the a750-vkcts expectations
|
||
- turnip/ci: bump the vkcts a750 timeout by 15 minutes
|
||
- turnip/ci: skip a vkd3d test that causes a GPU hang on a750
|
||
- nvk/ci: update the ga106 expectations
|
||
- zink/ci: update the nvk-ga106 expectations
|
||
- zink/ci: update the radv expectations
|
||
- radv/ci: update the vkcts expectations
|
||
- ci/test: make the .b2c-${arch}-test-* jobs provide a default b2c
|
||
- ci/tests: de-duplicate the b2c version between architectures
|
||
- ci/test: uprev to b2c v0.9.14
|
||
- freedreno/ci: use the default b2c
|
||
- r300/ci: use the default b2c
|
||
- i915g/ci: use the default b2c version
|
||
- ci/b2c: modernize the job description to use run_*
|
||
- ci/b2c: run the machine registration check before the test container
|
||
- radeonsi/ci: update the vangogh expectations
|
||
- radeonsi/ci: run on ACO changes
|
||
- radeonsi/ci: run a fraction of glcts-vangogh in pre-merge
|
||
- ci/init-stage2: use the common scripts from the build artifact
|
||
- ci/b2c: use the runner description rather than ID
|
||
- ci/b2c: allow defining a boot watchdog
|
||
- freedreno/ci: use the boot watchdog to ensure the a750 boots
|
||
- zink/ci: update nvk expectations
|
||
- zink/ci: update RADV expectations
|
||
- radeonsi/ci: update the vangogh expectations
|
||
- ci/b2c: allow jobs to select a file in the dtb url
|
||
- ci/b2c: allow using another initrd that contains firmware
|
||
- freedreno/ci: uprev the a750 kernel to msm-next
|
||
- ci: fix the artifact name
|
||
- zink/ci: use the debian-built-testing for nvk
|
||
- ci/b2c: fix the S3 artifact for amd64 manual vk/gl
|
||
- turnip/ci: re-introduce the \`multiviewport` flakes
|
||
|
||
Mary Guillemard (56):
|
||
|
||
- agx: Add support for EGL_NV_context_priority_realtime
|
||
- panfrost: Report default value for GROUP_PRIORITIES_INFO in drm-shim
|
||
- pan/kmod: Expose medium priority on panfrost
|
||
- panvk: Implement global priority extensions
|
||
- panvk: Advertise VK_EXT_tooling_info
|
||
- panvk: Advertise VK_KHR_shader_non_semantic_info
|
||
- panvk: Advertise VK_KHR_shader_relaxed_extended_instruction
|
||
- panvk: Implement VK_KHR_zero_initialize_workgroup_memory
|
||
- bi: Execute nir_opt_algebraic after nir_lower_pack
|
||
- panvk: Implement VK_EXT_sampler_filter_minmax for v10
|
||
- panvk: Only flag rw_nc pool as uncached on v10+
|
||
- panvk: Take rasterization samples into account in draw
|
||
- panfrost: Remove faulty assert in cs_loop_conditional_*
|
||
- panvk: Wire occlusion queries to internals
|
||
- panvk: Implement occlusion queries for JM
|
||
- panvk: Implement occlusion queries for CSF
|
||
- panvk: Expose precise occlusion queries
|
||
- panvk: Advertise VK_EXT_host_query_reset
|
||
- panvk: Enable depthClamp and depthBiasClamp
|
||
- panvk: Enable shaderInt16
|
||
- panvk: Advertise VK_KHR_index_type_uint8
|
||
- panvk: Advertise VK_KHR_map_memory2
|
||
- panvk: Disable integer array indices clamping
|
||
- panvk: Advertise VK_EXT_image_robustness
|
||
- panvk: Advertise VK_EXT_pipeline_robustness
|
||
- panvk: Call vk_free on queue array instead of vk_object_free
|
||
- panvk: Use vk_zalloc for queue array allocation
|
||
- panvk: Update Mali-G52 CI baseline
|
||
- panvk: Add a nightly job for Mali-G52
|
||
- nak: Fix 8-bit selection for vectors
|
||
- nak: Simplify 16-bit vector selection to not use try_from
|
||
- meson: Add mesa-clc and install-mesa-clc options
|
||
- meson: Add precomp-compiler and install-precomp-compiler options
|
||
- asahi: Remove unneeded dependencies for asahi_clc
|
||
- util/bitpack_helpers: Use UINT64_MAX instead of ~0ULL
|
||
- util/bitpack_helpers: Make fixed packs CL safe
|
||
- nir,agx: Allow nir_precomp_print_blob to print a static array
|
||
- libcl: Respect NDEBUG for assert
|
||
- panforst: Update ForEachMacros
|
||
- pan/genxml: Move pack_header to an external file
|
||
- libcl: Add VkQueryType and VkQueryResultFlagBits definitions
|
||
- pan/genxml: Switch unpack to use uint32_t
|
||
- pan/genxml: Emit struct details before pack function
|
||
- pan/genxml: Move [un]pack internals to use packed structs
|
||
- pan/genxml: Enforce explicit packed types on pan_[un]pack
|
||
- pan/genxml: Switch pan_section_ptr to cast to packed type
|
||
- pan/genxml: Switch [un]pack codegen to macros
|
||
- pan/genxml: Switch __gen_unpack to macros
|
||
- panfrost: Fix group priorities in drm-shim
|
||
- panfrost: Fix PROGRESS_LOAD destination register
|
||
- pan/bi: Properly encode LEA_BUF_IMM
|
||
- pan/bi: Remove shift lanes invalid encodings
|
||
- pan/bi: Fix invalid CLPER encoding
|
||
- pan/bi: Use 2D dimension with TEX_FETCH with CUBE on Valhall
|
||
- pan/decode: Fix indirect branch calculation for 64-bit
|
||
- panvk: Disallow unknown GPU models early in physical device init
|
||
|
||
Matt Turner (16):
|
||
|
||
- anv: Align anv_descriptor_pool::host_mem
|
||
- vulkan: Skip memcpy() call if passed null pointers
|
||
- anv: Protect memcpy/memset/qsort calls against NULL arguments
|
||
- anv: Avoid null ptr dereference
|
||
- intel: Avoid unaligned pointer access
|
||
- vulkan: Avoid pointer aliasing
|
||
- nir: Get correct number of components
|
||
- intel/decoder: Avoid duplicate symbols when expat is not available
|
||
- brw: Avoid reading past the end of \`p->store`
|
||
- brw: Pass brw_codegen to next_offset
|
||
- brw: Bounds check access to \`p->store`
|
||
- brw: Pass number and sizeof separately to calloc
|
||
- elk: Avoid reading past the end of \`p->store`
|
||
- elk: Pass brw_codegen to next_offset
|
||
- elk: Bounds check access to \`p->store`
|
||
- elk: Pass number and sizeof separately to calloc
|
||
|
||
Matthew Brost (1):
|
||
|
||
- anv/xe: Bind queue per anv_queue
|
||
|
||
Mauro Rossi (4):
|
||
|
||
- nvk/android: Avoid building error in nak bindings
|
||
- nvk/android: Advertise Vulkan 1.1 for Android 12L and lower
|
||
- nvk/android: Add support for ANDROID_native_buffer
|
||
- android: remove shared-glapi building rules
|
||
|
||
Maíra Canal (3):
|
||
|
||
- v3dv: Check multiple DRM primary nodes before picking the display fd
|
||
- v3dv: delete \`v3dv_debug.h`
|
||
- v3dv: use Mesa log infrastructure instead of using stderr
|
||
|
||
Mel Henning (27):
|
||
|
||
- nak: Fix two warnings of elided_named_lifetimes
|
||
- gallium/winsys/nouveau: Don't mark the api PUBLIC
|
||
- nak: Add nak_nir_mark_lcssa_invariants
|
||
- compiler/rust/bitset: Fix the bitset iterator
|
||
- compiler/rust: Fix running tests
|
||
- compiler/rust/bitset: Add a basic test
|
||
- compiler/rust/bitset: Removed unused start param
|
||
- compiler/rust/bitset: Make BitSetIter private
|
||
- compiler/rust/bitset: impl FromIterator
|
||
- compiler/rust/bitset: Remove impl Not
|
||
- compiler/rust/bitset: Add a lazy expression API
|
||
- compiler/rust/bitset: Take a stream in union_with
|
||
- nak: Migrate liveness to new bitset expression api
|
||
- compiler/rust/bitset: Don't expose words
|
||
- compiler/rust/bitset: Test next_unset()
|
||
- nak: Add ShaderModel::hw_reserved_gprs()
|
||
- nak: Add gpr_limit_from_local_size
|
||
- nir_validate: Handle unstructured control flow
|
||
- nak: lower_load_ssbo_descriptor modifies cf
|
||
- nir: Update num_blocks in sort_unstructured_blocks
|
||
- nvk: Fix an assertion in nvk_slm_area_ensure
|
||
- nak: Return VK_ERROR_UNKNOWN on assertion failure
|
||
- nak: Fix a spelling error
|
||
- nak/opt_copy_prop: Fix IAdd3 overflow check
|
||
- nak/opt_copy_prop: Add force_alu_src_type
|
||
- nak/opt_copy_prop: Force alu src for IAdd2X/IAdd3X
|
||
- driconf: force_vk_vendor on Deep Rock Galactic+NVK
|
||
|
||
Mi, Yanfeng (2):
|
||
|
||
- anv:Fix memory grow calculation overflow issue
|
||
- anv:increase instruction heap to 3Gb
|
||
|
||
Michael Cheng (2):
|
||
|
||
- anv : Add tracepoint for as_build
|
||
- intel : Expose Shader hashes for utrace and Perfetto
|
||
|
||
Michel Dänzer (4):
|
||
|
||
- Revert "util/mesa-db: Further simplify mesa_db_compact"
|
||
- Revert "util: Use persistent array of index entries"
|
||
- Revert "winsys/amdgpu: fix FD mismatch"
|
||
- winsys/amdgpu: Always use amdgpu_device_get_fd for aws->fd
|
||
|
||
Michel Zou (1):
|
||
|
||
- ac/gpu_info: Fix missing prototype mingw error
|
||
|
||
Mike Blumenkrantz (38):
|
||
|
||
- zink: restrict implicit feedback loop detection using miplevels/layers
|
||
- mesa: use default params for clearbuffer functions
|
||
- zink: rework query result checking
|
||
- zink: use internal map flag for qbos
|
||
- glsl: make gl_ViewID_OVR visible to all shader stages
|
||
- glsl: enable OVR_multiview if OVR_multiview2 is enabled
|
||
- lavapipe: stop storing texture handle for samplers
|
||
- vk/sampler: split out sampler init from create
|
||
- lavapipe: split out sampler init from create
|
||
- lavapipe: split out bda descriptor function params from struct
|
||
- lavapipe: fix bitmask type for sampler updating
|
||
- lavapipe: move workgraph lowering up and delete pipeline param
|
||
- lavapipe: unsupport NV_device_generated_commands
|
||
- lavapipe: stop using pipeline layouts in some places
|
||
- lavapipe: handle VK_REMAINING_ARRAY_LAYERS with HIC
|
||
- lavapipe: fix 3D->2D blitting
|
||
- lavapipe: abort on unsupported depth copy ops
|
||
- lavapipe: support zs<->color copies
|
||
- lavapipe: maintenance8
|
||
- zink: enable maintenance8
|
||
- glsl: plumb num_views down to shader_info::view_mask
|
||
- zink: fix viewport detection when switching last stage shaders
|
||
- zink: add radv ci fail
|
||
- zink: disable shader objects when viewmask is set
|
||
- zink: fix replacing incompatible pipelines
|
||
- egl: never select swrast for vmwgfx
|
||
- zink: deduplicate VkDevice and VkInstance
|
||
- aco: exclude novalidateir from codegen flags
|
||
- zink: check for bound gfx stages before dereferencing
|
||
- zink: add zink_resource_reference() util function
|
||
- zink: refcount needs_present resource
|
||
- ci: mark radv-raven-traces-restricted with allow_failure
|
||
- zink: emit SpvCapabilityDemoteToHelperInvocation for IsHelperInvocation
|
||
- zink: also refcount needs_present from frontbuffer flush
|
||
- zink: guard rebar check against fallback heap detection
|
||
- radv: fix error reporting for VkExternalMemoryTypeFlagBitsKHR
|
||
- zink: only enable unsynchronized_texture_subdata with HIC
|
||
- zink: never try to oom flush during unsync texture upload
|
||
|
||
Mike Lothian (1):
|
||
|
||
- gallium/radeon: Fix r600_pci_ids.h include
|
||
|
||
Mykhailo Skorokhodov (1):
|
||
|
||
- drirc/anv: force_vk_vendor=-1 for Bellwright
|
||
|
||
Nanley Chery (22):
|
||
|
||
- anv: Support non-0/1 sRGB fast-clear colors on gfx9
|
||
- anv: Store fast-clear colors with the view swizzle
|
||
- anv: Drop fast-clear value conversion check
|
||
- intel/blorp: Assert 3D Ys fast-clear restriction
|
||
- intel/isl: Allow CCS on 3D 64bpp+ Tile64
|
||
- intel: Allow CCS on 3D surfaces for gfx120
|
||
- intel/isl: Fix DecompressInL3 assignment on gfx12.5
|
||
- anv: Enable storage accesses with modifiers on gfx12+
|
||
- anv: Enable more storage compression on gfx12+
|
||
- anv: Only consider R32 image formats as supporting atomics
|
||
- anv: Allow compressed memtypes with default buffer types
|
||
- anv: Slow clear if fast-clear cost is not mitigated
|
||
- iris: Reduce fast-clear post-amble flushes
|
||
- iris: Use L3 Fabric flush in fast-clear post-amble on TGL
|
||
- anv: Reduce fast-clear post-amble synchronization
|
||
- anv: Use L3 Fabric flush in fast-clear post-amble on TGL
|
||
- anv: Drop bpc check for non-zero fast clears
|
||
- Revert "anv: turn off non zero fast clears for CCS_E"
|
||
- anv: Inline can_fast_clear_with_non_zero_color
|
||
- anv: Allow more single subresource fast-clears with FCV
|
||
- anv: Drop can_fast_clear_with_non_zero_color()
|
||
- anv: Limit slow clear heuristic to ACM and prior
|
||
|
||
Patrick Lerda (8):
|
||
|
||
- r600: fix the evergreen sampler when the minification and the magnification are not identical
|
||
- r600: restructure r600_create_vertex_fetch_shader() to remove memcpy()
|
||
- r600: ensure that the last vertex is always processed on evergreen
|
||
- r600: evergreen stencil/depth mipmap blit workaround
|
||
- r600: reverse fix spec ext_packed_depth_stencil getteximage
|
||
- winsys/radeon: fix radeon_winsys_bo_from_handle() related race condition
|
||
- r600: fix r600_init_screen_caps() has_streamout issue
|
||
- r600: fix r600_init_shader_caps() has_atomics issue
|
||
|
||
Paulo Zanoni (3):
|
||
|
||
- brw: don't forget the base when emitting SHADER_OPCODE_MOV_RELOC_IMM
|
||
- brw: don't read past the end of old_src buffer in resize_sources()
|
||
- brw: increase brw_reg::subnr size to 6 bits
|
||
|
||
Pavel Ondračka (27):
|
||
|
||
- r300: group KIL for R300/R400
|
||
- r300: run nir_opt_algebraic in the backend
|
||
- r300: always transform sin/cos input for fs
|
||
- r300/ci: update RV410 CI expectations
|
||
- ci: bring back some i915g testing
|
||
- i915/ci: update CI expectations
|
||
- r300: disable ATI2N textures on R400
|
||
- r300: disable microtiling for scanout buffers
|
||
- r300/ci: update CI expectations
|
||
- r300: fix uninitialized use in transform_vertex_ROUND
|
||
- nir: add support for clamping in nir_lower_tex_shadow
|
||
- etnaviv: always clamp shadow sampler comparison reference value
|
||
- r300: fix presubtract assert
|
||
- r300: move shadow lowering to NIR
|
||
- r300: reswizzle some shadow texture calculations to use w channel
|
||
- r300: delete backend shadow lowering code
|
||
- r300: use ssa-like form for gl_FragCoord transformation
|
||
- r300: add some more nir cleanup compiler passes
|
||
- r300: use ssa-like form for backend texture lowering
|
||
- r300: don't allocate fs registers when translating from NIR
|
||
- r300: get rid of the register rename pass
|
||
- r300: get rid of some texture fixups
|
||
- r300: remove support for register arrays from nir_to_rc
|
||
- r300: fix memory leak in contant remaping
|
||
- ci: fix debian-build-testing BUILDTYPE
|
||
- i915/ci: use debian-build-testing instead of debian-testing
|
||
- i915: rework shader compile failures reporting
|
||
|
||
Peyton Lee (5):
|
||
|
||
- frontends/va: add support for VAProcColorStandardExplicit
|
||
- frontends/va: add support for VAProcColorStandardExplicit
|
||
- frontends/va: function process_frame has return value
|
||
- radeonsi/vpe: optimize software functions
|
||
- radeonsi/vpe: add destroy_fence function
|
||
|
||
Philipp Zabel (11):
|
||
|
||
- teflon: Use correct convolution params struct
|
||
- teflon: Mark dilated convolutions and fused activation as not supported
|
||
- teflon: Support fused ReLU activation
|
||
- etnaviv/nn: Enable fused ReLU activation
|
||
- teflon: Add is_signed parameter to ml_subgraph_invoke and ml_subgraph_read_output
|
||
- etnaviv/nn: Add support for signed 8-bit tensors
|
||
- teflon/tests: prep test executor for signed convolutions
|
||
- teflon/tests: Enable int8 tests
|
||
- etnaviv/ml: Create combined input tensors for addition first
|
||
- teflon: Reject per-axis quantization
|
||
- teflon: Support fused ReLU6 activation via output saturation
|
||
|
||
Pierre-Eric Pelloux-Prayer (40):
|
||
|
||
- radv: set info->family_overridden when RADV_FORCE_FAMILY is used
|
||
- ac/surface: add flags to surface metadata
|
||
- radeonsi: refuse to import texture with family_overriden being set
|
||
- ac: rename ac_surface_test_common -> ac_fake_hw_db
|
||
- ac: add 'polaris12' gpu to ac_fake_hw_db
|
||
- ac: switch AMD_FORCE_FAMILY handling to using ac_fake_hw_db
|
||
- radeonsi/tests: update expected results
|
||
- ac/perfcounter: fix buffer overflow
|
||
- dri: Remove unused function
|
||
- radeonsi/gfx12: disable display dcc for front buffer rendering
|
||
- radeonsi: disable DCC for PIPE_BIND_USE_FRONT_RENDERING
|
||
- glx: return BadMatch for invalid reset notification strategy
|
||
- ac/nir: remove prim_stride_ret arg from ngg_build_streamout_buffer_info
|
||
- radeonsi: use bytes units in streamout
|
||
- DEPENDENCY: ac/llvm: fix sparse code handling
|
||
- radeonsi: fallback to util_blitter_draw_rectangle
|
||
- radeonsi/tests: update results
|
||
- gl/spirv: update subgroup_size if GroupNonUniform is used
|
||
- amd: move all uses of libdrm_amdgpu to ac_linux_drm
|
||
- amd: amdgpu-virtio implementation
|
||
- ac/virtio: disable userptr and local buffers
|
||
- ac/virtio: disable timeline syncobj support
|
||
- radeonsi: enable virtio native context support
|
||
- radv: enable virtio native context support
|
||
- radv/virtio: disable syncobj timeline support
|
||
- ac/virtio: add virtio-only AMDGPU_GEM_CREATE flag
|
||
- radeonsi, radv, virtio: use AMDGPU_GEM_CREATE_VIRTIO_SHARED
|
||
- radeonsi: clear the debug callback on ctx destroy
|
||
- ttn: init source_blake3 and name from tgsi_shader_info
|
||
- ac/llvm: add wqm param to ac_build_quad_swizzle
|
||
- ac/llvm: enable wqm for ac_build_quad_swizzle from ac_build_fs_interp_mov
|
||
- radeonsi: do not use std::max
|
||
- glx: fix glx-create-context-invalid-es-version
|
||
- dri: use _checked variants of xcb requests
|
||
- dri: deal with ARGB1555
|
||
- egl/wayland: validate dri_screen_display_gpu before use
|
||
- amd: add ac_drm_device_get_cookie
|
||
- radeonsi: use ac_drm_device_get_cookie
|
||
- radeonsi: update si_need_gfx_cs_space upper bound
|
||
- radeonsi: disable dcc when external shader stores are used
|
||
|
||
Qiang Yu (81):
|
||
|
||
- ac/surface/tests: support all block sizes
|
||
- ac/surf: add more modifiers to gfx12 supported list
|
||
- radeonsi: disable use_gfx12_xfb_intrinsic when use ACO
|
||
- util/blake3: add _mesa_blake3_from_printed_string
|
||
- radeonsi: add AMD_FORCE_SHADER_USE_ACO for debug
|
||
- nir: do not generate b2i64 when driver want to lower it
|
||
- aco: enable gfx12 support for radeonsi
|
||
- radeonsi: fix unigine heaven crash when use aco on gfx8/9
|
||
- aco: fix voffset missing when buffer store base >=4096
|
||
- radeonsi: fix OpenCL shader compile fail
|
||
- ac/nir: lower access for shared and scratch memory
|
||
- ac,radv: move ac_nir_lower_bit_size_callback to common place
|
||
- radeonsi: fix OpenCL piglit tests fails when using ACO
|
||
- radeonsi: replace ac_nir_lower_subdword_loads
|
||
- ac: remove ac_nir_lower_subdword_loads
|
||
- radeonsi: fix global access ACO compile fail when OpenCL
|
||
- radeonsi: enable ACO by default for pre-GFX10 GPUs
|
||
- radeonsi: unify disk cache id no matter use_aco or not
|
||
- gallium: add pipe_caps struct definition
|
||
- gallium: add u_init_pipe_screen_caps
|
||
- asahi: add agx_init_screen_caps
|
||
- crocus: add crocus_init_screen_caps
|
||
- d3d12: add d3d12_init_screen_caps
|
||
- etnaviv: add etna_init_screen_caps
|
||
- freedreno: add fd_init_screen_caps
|
||
- i915: add i915_init_screen_caps
|
||
- iris: add iris_init_screen_caps
|
||
- lima: add lima_init_screen_caps
|
||
- llvmpipe: add llvmpipe_init_screen_caps
|
||
- nouveau/nv30: add nv30_init_screen_caps
|
||
- nouveau/nv50: add add nv50_init_screen_caps
|
||
- nouveau/nvc0: add nvc0_init_screen_caps
|
||
- panfrost: add panfrost_init_screen_caps
|
||
- r300: add r300_init_screen_caps
|
||
- r600: add r600_init_screen_caps
|
||
- radeonsi: add si_init_screen_caps
|
||
- softpipe: add softpipe_init_screen_caps
|
||
- svga: add svga_init_screen_caps
|
||
- tegra: init screen caps
|
||
- v3d: add v3d_init_screen_caps
|
||
- vc4: add vc4_init_screen_caps
|
||
- virgl: add virgl_init_screen_caps
|
||
- zink: add zink_init_screen_caps
|
||
- nine: change cap macros to use pipe_caps access
|
||
- egl,gallium,glx: replace dri_get_screen_param with pipe_caps access
|
||
- mesa/st: enable extension use pipe_caps access
|
||
- egl,gallium,gbm,mesa: replace get_param with pipe_caps access
|
||
- gallium,mesa: replace get_paramf with pipe_caps access
|
||
- rusticl: use pipe_caps access
|
||
- asahi: remove agx_get_param and agx_get_paramf
|
||
- crocus: remove crocus_get_param and crocus_get_shader_paramf
|
||
- d3d12: remove d3d12_get_param and d3d12_get_paramf
|
||
- etnaviv: remove etna_screen_get_param and etna_screen_get_paramf
|
||
- freedreno: remove fd_screen_get_param and fd_screen_get_paramf
|
||
- i915: remove i915_get_param and i915_get_paramf
|
||
- iris: remove iris_get_param and iris_get_paramf
|
||
- lima: remove lima_screen_get_param and lima_screen_get_paramf
|
||
- llvmpipe: remove llvmpipe_get_param and llvmpipe_get_paramf
|
||
- nouveau/nv30: remove nv30_screen_get_param and nv30_screen_get_paramf
|
||
- nouveau/nv50: remove nv50_screen_get_param and nv50_screen_get_paramf
|
||
- nouveau/nvc0: remove nvc0_screen_get_param and nvc0_screen_get_paramf
|
||
- panfrost: remove panfrost_get_param and panfrost_get_paramf
|
||
- r300: remove r300_get_param and r300_get_paramf
|
||
- r600: remove r600_get_param and r600_get_paramf
|
||
- radeonsi: remove si_get_param and si_get_paramf
|
||
- softpipe: remove softpipe_get_param and softpipe_get_paramf
|
||
- svga: remove svga_get_param and svga_get_paramf
|
||
- tegra: remove tegra_screen_get_param and tegra_screen_get_paramf
|
||
- v3d: remove v3d_screen_get_param and v3d_screen_get_paramf
|
||
- vc4: remove vc4_screen_get_param and vc4_screen_get_paramf
|
||
- virgl: remove virgl_get_param and virgl_get_paramf
|
||
- zink: remove zink_get_param and zink_get_paramf
|
||
- gallium: remove get_param and get_paramf
|
||
- docs,src: replace doc and comments for PIPE_CAP with pipe_caps
|
||
- gallium,mesa: remove uint surffix from pipe_caps
|
||
- radeonsi: remove si_screen.max_texel_buffer_elements
|
||
- etnaviv: remove min/max_texture_gather_offset init
|
||
- lavapipe: fix min_vertex_pipeline_param
|
||
- gallium: fix ddebug and noop screen caps init
|
||
- radeonsi: fix has_non_uniform_tex_access info
|
||
- radeonsi: fix GravityMark corruption when use aco
|
||
|
||
Rebecca Mckeever (14):
|
||
|
||
- panvk: Use vk_image::drm_format_mod instead of pan_image::layout.modifier
|
||
- panvk: Replace tab with spaces
|
||
- panvk: Enable multiplane images and image views
|
||
- pan/texture: s/pan_image_view_get_zs_image/pan_image_view_get_zs_plane/
|
||
- pan/texture: s/pan_image_view_get_rt_image/pan_image_view_get_color_plane/
|
||
- pan/texture: Accept holes in the pan_image_view::planes array
|
||
- pan/desc: Pass an image to pan_force_clean_write_rt()
|
||
- pan/desc: Add a pan_image_view_get_s_plane() helper and use it
|
||
- panvk: Support D32_S8 as a multiplanar format
|
||
- pan/format: Use HW version to determine siting for YUV 422 formats
|
||
- pan/texture: Only use plane_chroma_2p for chroma planes
|
||
- util/hash_table: Add _mesa_hash_table_u64_replace()
|
||
- panvk: Allow a 32-bit binding value in desc id key and use 64-bit keys
|
||
- panvk: Fix assertion in is_disjoint()
|
||
|
||
Rhys Perry (72):
|
||
|
||
- nir: add more intrinsics to nir_intrinsic_can_reorder
|
||
- nir/algebraic: optimize bcsel(ieq(b, 0), a, shift(a, b))
|
||
- nir/algebraic: optimize ushr(a, ishl(iand(b, 3), 3))
|
||
- ac/nir: add ACCESS_CAN_REORDER to lowered load_global_constant
|
||
- aco: optimize nir_op_shfr with <32 src1
|
||
- nir,aco,ac/llvm: add nir_op_alignbyte_amd
|
||
- nir_lower_mem_access_bit_sizes: support 64-bit offsets
|
||
- nir_lower_mem_access_bit_sizes: add nir_mem_access_shift_method
|
||
- nir_lower_mem_access_bit_sizes: pass access to callback
|
||
- nir_lower_mem_access_bit_sizes: support load_constant
|
||
- aco,ac/nir: flag loads to use smem in NIR
|
||
- radv,ac/nir: lower sub-dword loads using nir_lower_mem_access_bit_sizes
|
||
- aco: remove load byte_align
|
||
- radv,ac/nir: split global access using nir_lower_mem_access_bit_sizes
|
||
- nir/algebraic: fix iabs(ishr(iabs(a), b)) optimization
|
||
- nir/algebraic: check bit sizes in lowered unpack(pack()) optimization
|
||
- nir/lcssa: fix premature exit of loop after rematerializing derefs
|
||
- glsl/list: add comments above foreach macros
|
||
- glsl/list: add and use helpers in foreach_list_typed macros
|
||
- glsl/list: remove parenthesis in foreach_list_typed macros
|
||
- glsl/list: remove underscores in foreach_list_typed macros
|
||
- nir/opt_move_discards_to_top: use nir_tex_instr_has_implicit_derivative
|
||
- nir: fix return value of nir_instr_move for some cases
|
||
- nir/opt_move_discards_to_top: remove recursion
|
||
- nir/opt_move_discards_to_top: update variable name
|
||
- nir/opt_move_discards_to_top: use nir_intrinsic_can_reorder
|
||
- nir/opt_move_discards_to_top: add more intrinsics to add_src_to_worklist
|
||
- nir/opt_move_discards_to_top: allow multiple discards to be moved
|
||
- nir/lcssa: use nir_intrinsic_can_reorder
|
||
- nir/algebraic: add ddxy to is_only_used_as_float
|
||
- nir/algebraic: add is_used_once to bcsel(, bcsel()) opts
|
||
- nir/algebraic: optimize more bcsel(, bcsel())
|
||
- aco: add SSA repair pass
|
||
- aco: use repair pass for LCSSA workaround
|
||
- aco: require WQM after demote in control flow
|
||
- aco: skip code if exec is empty
|
||
- aco/tests: add tests for empty exec masks
|
||
- aco: don't use uniform continues if exec might be empty
|
||
- aco: make small_vec copyable
|
||
- aco: use small_vec in RegCounterMap
|
||
- nir/tests: fix SSA dominance in opt_if_merge tests
|
||
- aco/gfx12: insert wait between VMEM WaW
|
||
- aco: force linear for event_vmem_sample and event_vmem_bvh
|
||
- aco: don't CSE p_shader_cycles_hi_lo_hi
|
||
- radv: constant fold after lowering memory accesses
|
||
- radv: fix expanded push constant loads when all are inlined
|
||
- radv: skip loading unused push constants
|
||
- ac/nir: have ac_nir_lower_mem_access_bit_sizes preserve >128 bit SMEM
|
||
- nir: make load_helper_invocation non-reorderable
|
||
- nir/move_discards_to_top: don't move across more intrinsics
|
||
- nir: make ballot ALU and mbcnt_amd operations reorderable
|
||
- aco: fix max_workgroup_count[0]
|
||
- aco: decrease max_workgroup_size
|
||
- radv: increase maxComputeWorkGroupCount[0]
|
||
- aco/tests: fix skip_lines=True with remaining characters in matches
|
||
- aco/util: fix bit_reference::operator&=
|
||
- aco: use VOP3 v_mov_b16 if necessary
|
||
- v3dv: fix SSA dominance error
|
||
- microsoft/compiler: invalidate loop analysis in dxil_nir_lower_double_math
|
||
- microsoft/compiler: repair SSA in dxil_nir_split_tess_ctrl
|
||
- d3d12: fix phi handling in d3d12_lower_primitive_id
|
||
- d3d12: store only once in d3d12_emit_points
|
||
- nir: rerun loop analysis if the parameters change
|
||
- nir/loop_analyze: use a sparse array and stop indexing SSA defs
|
||
- nir/gcm: stop preserving nir_metadata_loop_analysis
|
||
- nir/liveness: stop requiring instr indices
|
||
- nir/validate: validate metadata
|
||
- nir/validate: preserve dominance during SSA validation
|
||
- nir/validate: validate ssa dominance by default
|
||
- radv: set has_image_bvh_intersect_ray for null winsys
|
||
- aco: don't use divergence information for most ALU defs
|
||
- nir/divergence: assume all instructions are loop invariant if no continues
|
||
|
||
Rob Clark (11):
|
||
|
||
- vdrm+tu+fd: Make cross-device optional
|
||
- freedreno/registers: Add GMU_CORE_FW_VERSION
|
||
- freedreno/a6xx: Align lrz setup with tu
|
||
- freedreno/a6xx: Add nolrzfc debug option
|
||
- freedreno/a6xx: Align lrz height to 32
|
||
- tu: Align lrz height to 32
|
||
- freedreno/a6xx: Use LATE_Z with OC + discard
|
||
- freedreno/a6xx: Fix timestamp emit
|
||
- ir3: Add preamble instr count metric
|
||
- freedreno/pps: Fix multiple counter collection runs
|
||
- tu: Fix raytracing query with vdrm
|
||
|
||
Robert Mader (2):
|
||
|
||
- v3d: Support SAND128 base modifier
|
||
- freedreno: Support offset query for multi-planar planes
|
||
|
||
Rohan Garg (5):
|
||
|
||
- intel/compiler: disable mesh autostrip for WA 16020916187
|
||
- iris: use CALLOC_STRUCT instead of calloc for readability
|
||
- isl: disable aux when creating uncompressed TileY/Tile64 surfaces from compressed ones
|
||
- anv: refactor choose_isl_tiling_flags to pass fewer arguments
|
||
- iris: assert that we're not exporting a TILE64 surface
|
||
|
||
Roland Scheidegger (1):
|
||
|
||
- llvmpipe: Fix overflow issues calculating loop iterations for aniso
|
||
|
||
Roman Stratiienko (1):
|
||
|
||
- v3dv/android: Suppress AHB-related log spam
|
||
|
||
Ruijing Dong (2):
|
||
|
||
- radeosi/vcn: enable EFC for VCN5.0+ when gfx >= 12
|
||
- radeonsi/vcn: center mv map buffer changed in vcn5.x
|
||
|
||
Russell Greene (1):
|
||
|
||
- perfetto: fix macos compile
|
||
|
||
Sagar Ghuge (30):
|
||
|
||
- anv: Enable MCS_CCS compression on Gfx12+
|
||
- blorp: Use the calculated execution mask
|
||
- anv: Update include dir for anv_tests
|
||
- anv: Split GRL code path in separate file
|
||
- anv: Add header to track BVH data structures
|
||
- anv: Add shader to build BVH header
|
||
- anv: Add shader to copy acceleration structures
|
||
- anv: Implement cmd_fill_buffer_addr callback
|
||
- anv: Move update buffer code in helper
|
||
- anv: Implement write_buffer_cp callback
|
||
- anv: Implement flush_buffer_write_cp callbck
|
||
- anv: Implement cmd_dispatch_unaligned callback
|
||
- anv: Implement acceleration structure API
|
||
- anv: Add helper to copy data from src to dest anv_address
|
||
- intel: Use the common RT BVH framework
|
||
- intel/compiler: Extend nir_intrinsic_load_topology_id_intel for xe3
|
||
- intel/genxml: Drop morton walk field from Xe2
|
||
- intel/genxml: Update COMPUTE_WALKER_BODY
|
||
- intel: Use Morton compute walk order
|
||
- intel/genxml: Update SAMPLER_STATE structure
|
||
- anv: Switch to ANISOTROPIC_FAST filter mode
|
||
- iris: Switch to ANISOTROPIC_FAST filter mode
|
||
- intel: Set correct maxComputeSharedMemorySize for Xe3+
|
||
- intel/genxml: Add coarse pixel related changes
|
||
- anv: Add pipelined coarse pixel state
|
||
- intel/genxml: Update URB related instructions and structures
|
||
- iris: Use 3DSTATE_URB_ALLOC_* instructions
|
||
- blorp: Use 3DSTATE_URB_ALLOC_* instructions
|
||
- anv: Use 3DSTATE_URB_ALLOC_* instructions
|
||
- intel/brw/xe3+: Don't compile SIMD32 if there is ray queries
|
||
|
||
Sam Lantinga (1):
|
||
|
||
- util: Fixed crash in HEVC encoding on 32-bit systems
|
||
|
||
Samuel Pitoiset (241):
|
||
|
||
- aco: cleanup using fixed registers in the trap handler shader
|
||
- aco: save/restore SCC in the trap handler shader
|
||
- aco: use scalar buffer stores for dumping SGPRS from the trap on GFX8
|
||
- aco: add a helper to dump SGPR to memory for the trap handler
|
||
- aco: fix storing SQ_WAVE_STATUS in the trap handler shader
|
||
- aco: declare phys regs for tba_hi/tma_hi
|
||
- radv,aco: dump m0 and exec from the trap handler
|
||
- vulkan/runtime: return same cmdbuf level from the command pool freelist
|
||
- docs: add missing documentation for RADV_DEBUG=psocachestats
|
||
- radv: remove unused parameter to radv_fill_nir_compiler_options()
|
||
- radv: dump the trap handler shader with RADV_DEBUG=dump_trap_handler
|
||
- aco: do not reorder s_trap instructions
|
||
- radv: cleanup printing SGPRS dumped from the trap handler
|
||
- radv,aco: dump more SQ_WAVE regs from the trap handler
|
||
- radv,aco: add a separate function to compile the trap handler shader
|
||
- aco: simplify postprocessing the trap handler shader
|
||
- radv,aco: use the trap handler layout struct while compiling the shader
|
||
- radv: fix the TMA descriptor size
|
||
- radv: compute the TMA BO size instead of using a constant
|
||
- radv,aco: save/restore overwritten VGPRs in the trap handler shader
|
||
- nir: add nir_intrinsic_debug_break instruction
|
||
- spirv: handle NonSemantic.DebugBreak to emit nir_debug_break()
|
||
- aco: emit nir_intrinsic_debug_break
|
||
- radv: emit nir_debug_break instructions when the trap handler is enabled
|
||
- radv: do not always invalidate L2 for GPUs with non-coherent RBs on GFX10+
|
||
- radv: move the GFX11 special case for mips to radv_image_is_pipe_misaligned()
|
||
- radv: determine the first mip that is pipe misaligned on GFX10+
|
||
- radv: use vk_image_view_subresource_range() when possible
|
||
- radv: pass the image subresource range to radv_{src,dst}_access_flush()
|
||
- radv: optimize the pipe misaligned L2 cache invalidation on GFX11
|
||
- aco: fix saving/restoring VGPRS in the trap handler on GFX9
|
||
- aco: use a 64-bit mov to save exec in the trap handler shader
|
||
- aco: add a new variant for vop1() with two operands
|
||
- aco: fix validation for v_movrels_b32 and friends
|
||
- aco: restore m0/exec before exiting the trap handler
|
||
- aco: use all invocations from the current wave in the trap handler
|
||
- aco: save/restore VGPRS on GFX8 in the trap handler shader
|
||
- aco: drop the second M0 operand for s_set_gpr_idx_on
|
||
- radv,aco: dump VGPRS from the trap handler shader
|
||
- radv: mark live invocations when dumping VGPRS with the trap handler
|
||
- radv: dump SPIR-V and NIR for the faulty shader detected with the trap
|
||
- radv: fix ignoring src stage mask when dst stage mask is BOTTOM_OF_PIPE
|
||
- radv: consider VK_PIPELINE_STAGE_2_NONE like BOTTOM_OF_PIPE
|
||
- radv: destroy meta resources properly when creating the device failed
|
||
- radv: add a helper to destroy a logical device
|
||
- radv: add a new drirc option to disable DCC for mips and enable it for RDR2
|
||
- radv,aco: dump LDS from the trap handler
|
||
- radv: remove VK_VALVE_descriptor_set_host_mapping
|
||
- radv: fix skipping on-disk shaders cache when not useful
|
||
- radv: mark VERDE (GFX6) as Vulkan 1.3 conformant
|
||
- radv: fix dumping debug/perftest options when there are holes
|
||
- radv: add a pipeline helper to skip shaders cache
|
||
- radv: fix dumping the trap handler shader disassembly
|
||
- radv: fix printing with RADV_DEBUG=psocachestats
|
||
- radv: only pass relevant stages when emitting DGC push constants
|
||
- radv: capture shader executable info at shader creation time
|
||
- radv: allow shaders caching with RADV_DEBUG=hang and the trap handler
|
||
- vulkan: add MESA_VK_TRACE_PER_SUBMIT
|
||
- radv: finish tools after cleaning meta resources
|
||
- radv: add new start/stop sqtt helpers for capturing with SQTT
|
||
- radv: add support for capturing RGP per-submit
|
||
- radv: add address binding report support for BOs imported with a fd
|
||
- radv: add address binding report support for BOs imported with a ptr
|
||
- radv: add a small helper to dump VM fault with the GPU hang report
|
||
- radv: dump address binding report with RADV_DEBUG=hang
|
||
- radv: try to detect use-after-free with address binding report
|
||
- zink/ci: skip one more modifier test on POLARIS10
|
||
- radv: promote VK_KHR_dynamic_rendering_local_read to core 1.4 API
|
||
- radv: promote VK_KHR_global_priority to core 1.4 API
|
||
- radv: promote VK_KHR_index_type_uint8 to core 1.4 API
|
||
- radv: promote VK_KHR_line_rasterization to core 1.4 API
|
||
- radv: promote VK_KHR_maintenance5 to core 1.4 API
|
||
- radv: promote VK_KHR_maintenance6 to core 1.4 API
|
||
- radv: promote VK_KHR_map_memory2 to core 1.4 API
|
||
- radv: promote VK_KHR_push_descriptor to core 1.4 API
|
||
- radv: promote VK_KHR_shader_subgroup_rotate to core 1.4 API
|
||
- radv: promote VK_EXT_pipeline_robustness to core 1.4 API
|
||
- radv: add new Vulkan 1.4 features/properties
|
||
- radv: advertise Vulkan 1.4 on GFX8+
|
||
- radv: bump VKCTS conformance version to 1.4.0.0 for some GFX8+ GPUs
|
||
- radv/ci: mark few tests as expected failures
|
||
- ac/parse_ib: fix parsing SDMA CONSTANT_FILL packet
|
||
- ac/parse_ib: print VA for the SDMA CONSTANT_FILL/WRITE packets
|
||
- radv: fix stencil only copies of depth/stencil images with SDMA
|
||
- radv: enable DGC IES for compute with ESO
|
||
- radv: fix initializing HTILE when the image has VRS rates
|
||
- ci: update VKCTS main to a9f7069b9a5ba94715a175cb1818ed504add0107
|
||
- radv: remove redundant drirc for incorrect dual-source blending
|
||
- radv: add radv_disable_dcc_stores and enable for Indiana Jones: The Great Circle
|
||
- radv: only dump device name info on Linux with RADV_DEBUG=hang
|
||
- radv: dump the Mesa version with RADV_DEBUG=hang
|
||
- radv/meta: add missing vk_meta_device_finish()
|
||
- radv/meta: move vk_meta_device_init() to radv_device_init_meta()
|
||
- radv: disable alphaToOne except for Zink
|
||
- ac/nir: export alpha to MRTZ.a and one to MRT0.a for alpha-to-one on GFX11
|
||
- aco: export alpha to MRTZ.a and one to MRT0.a for alpha-to-one on GFX11
|
||
- radv: fix alpha-to-coverage with alpha-to-one when MRTZ is also exported
|
||
- radv: remove remaining discard to demote options
|
||
- radv: fix disabling DCC for stores with drirc
|
||
- radv: simplify determining some fragment shader info with epilogs
|
||
- radv: fix alpha-to-coverage with alpha-to-one without MRTZ
|
||
- Revert "radv: disable alphaToOne except for Zink"
|
||
- spirv: add an options to lower SpvOpTerminateInvocation to OpKill
|
||
- radv: add radv_lower_terminate_to_discard and enable for Indiana Jones
|
||
- radv: mark HAWAII (GFX7) as Vulkan 1.3 conformant
|
||
- radv: report same buffer aligment for DGC preprocessed buffer
|
||
- Revert "radv: fix creating unlinked shaders with ESO when nextStage is 0"
|
||
- radv/ci: fix expected list of failures for TAHITI
|
||
- radv: fix missing variants for the last VGT stage with shader object
|
||
- ci: uprev vkd3d-proton to c965c1351fd6915a65bb7f647319536252a24a93
|
||
- radv: fix capturing RT pipelines that return VK_OPERATION_DEFERRED_KHR for RGP
|
||
- radv: reorganize query code by adding separate begin/end helpers
|
||
- radv: remove dead code in radv_CmdCopyQueryPoolResults()
|
||
- radv: add few more query helpers for copying results
|
||
- radv: only enable emulated mesh/task shader queries on GFX10.3
|
||
- radv/nir: fix checking if task shader invocations query is enabled
|
||
- radv: fix getting the number of vertices per prim for the last VGT stage
|
||
- radv: rename GDS queries to emulated queries
|
||
- radv/nir: simplify lowering of query intrinsics
|
||
- radv: cleanup enabling the global BO list when BDA is used
|
||
- radv: check descriptor indexing features for enabling the global BO list
|
||
- radv: rework emitting SPI_SHADER_Z_FORMAT
|
||
- radv: rename color output state to fragment output state
|
||
- radv: add support for VK_PRIMITIVE_TOPOLOGY_META_RECT_LIST_MESA
|
||
- radv: use VK_PRIMITIVE_TOPOLOGY_META_RECT_LIST_MESA for meta pipelines
|
||
- radv: pass extra graphics pipeline create info using pNext
|
||
- radv/meta: rework creating meta pipelines for query resolves
|
||
- radv/meta: convert the copy/fill pipelines to vk_meta
|
||
- radv/meta: convert the copy VRS to HTILE pipelines to vk_meta
|
||
- radv/meta: convert the FMASK expand pipelines to vk_meta
|
||
- radv/meta: convert the FMASK copy pipelines to vk_meta
|
||
- radv/meta: convert the DCC retile pipelines to vk_meta
|
||
- radv/meta: convert the HTILE expand CS pipelines to vk_meta
|
||
- radv/meta: convert the DCC decompress CS pipelines to vk_meta
|
||
- radv/meta: convert the clear HTILE mask pipelines to vk_meta
|
||
- radv/meta: convert the DCC comp-to-single pipelines to vk_meta
|
||
- radv/meta: convert DGC pipeline layout to vk_meta
|
||
- radv/meta: convert the query resolve pipelines to vk_meta
|
||
- radv/meta: convert the image-to-buffer pipelines to vk_meta
|
||
- radv/meta: convert the buffer-to-image pipelines to vk_meta
|
||
- radv/meta: convert the image-to-image pipelines to vk_meta
|
||
- radv/meta: convert the clear image pipelines to vk_meta
|
||
- radv/meta: convert the compute resolve pipelines to vk_meta
|
||
- radv/meta: remove radv_meta_create_compute_pipeline()
|
||
- vulkan: add a new vk_meta option to use the rect list pipeline path
|
||
- vulkan: use the meta pipeline cache for graphics pipelines
|
||
- radv/meta: convert the HTILE expand GFX pipelines to vk_meta
|
||
- radv/meta: convert the HW resolve GFX pipelines to vk_meta
|
||
- radv/meta: convert the fast-clear GFX pipelines to vk_meta
|
||
- radv/meta: convert the blit GFX pipelines to vk_meta
|
||
- radv/meta: convert the clear GFX pipelines to vk_meta
|
||
- radv/meta: convert the resolve GFX pipelines to vk_meta
|
||
- radv/meta: use only one push constant range for blit2d pipelines
|
||
- radv/meta: convert the blit2d GFX pipelines to vk_meta
|
||
- radv/meta: remove unused radv_meta_create_xxx() helpers
|
||
- radv: fix destroying DGC pipelines
|
||
- radv: disable RT with LLVM completely
|
||
- radv/meta: remove a workaround for building accel structs with LLVM
|
||
- radv/meta: always initialize emulated etc2 on-demand
|
||
- radv/meta: move initializing emulated astc to radv_device_init_meta()
|
||
- radv/meta: stop initializing RT accel structs
|
||
- radv: fix adding the BO to cmdbuf list when emitting buffer markers
|
||
- radv/meta: fix loading the meta pipeline cache
|
||
- radv/meta: reduce length of some cache keys
|
||
- radv/meta: add radv_meta_get_noop_pipeline_layout()
|
||
- radv/meta: do not create redundant pipeline layout objects
|
||
- radv: disable logic op for float/srgb formats
|
||
- ac/descriptors: fix configuring NBC views on GFX12
|
||
- aco: fix VS prologs on GFX12
|
||
- radv: disable VRS coarse shading with 8x MSAA on GFX12
|
||
- radv: configure the VRS surface swizzle mode on GFX12
|
||
- radv: fix programming WALK_ALIGN8_PRIM_FITS_ST on GFX12
|
||
- radv: program DB_RENDER_OVERRIDE correctly on GFX12
|
||
- ac/nir: fix lowering subgroup ID for compute shaders on GFX12
|
||
- ac/nir: fix a comment typo in load_subgroup_id_lowered()
|
||
- ac/gpu_info: add cp_dma_use_L2
|
||
- radv: fix CP DMA clears/copies on GFX12
|
||
- aco: always use ds_bpermute for shuffle/rotate on GFX12
|
||
- radv: fix configuring the attribute ring size on GFX12
|
||
- radv: rename attr_ring to ge_rings
|
||
- radv: change the BASE_HI field for VGT_TF_MEMORY_BASE_HI on GFX12
|
||
- ac/surface: honor RADEON_SURF_PREFER_xxx_ALIGNMENT on GFX12
|
||
- radv: advertise VK_MESA_image_alignment_control on GFX12
|
||
- radv: fix emitting SPI_SHADER_GS_OUT_CONFIG_PS with NULL FS on GFX12
|
||
- radv: fail to initialize when the AMD GPU generation is unsupported
|
||
- radv: mark AMD CDNA as unsupported
|
||
- radv: add GFX12 support to the null winsys
|
||
- ac/nir: fix skipping streamout when no buffers are bound on GFX12
|
||
- vulkan: Update XML and headers to 1.4.305
|
||
- radv: promote VK_EXT_depth_clamp_zero_one to KHR
|
||
- radv: bump maxViewportDimensions to 32K on GFX12
|
||
- radv: add a helper to report if cooperative matrix is enabled
|
||
- zink/ci: add lists for RADV/GFX1200
|
||
- radv: remove duplicate definition of SQTT_BUFFER_ALIGN_SHIFT
|
||
- ac/sqtt: update programming SQTT on GFX12
|
||
- radv: add support for VkMemoryBarrierAccessFlags3KHR
|
||
- radv: adjust the source aspect for color to depth/stencil image copies
|
||
- radv: advertise VK_KHR_maintenance8
|
||
- radv: do not overallocate the number of exports for streamout on GFX12
|
||
- radv: fix transform feedback on GFX12
|
||
- radv: declare a new user SGPR for emulating queries on GFX12
|
||
- radv: lower emulated queries with global atomics on GFX12
|
||
- radv: allocate memory for the shader query buffer on GFX12
|
||
- radv: emit the shader buffer query VA on GFX12
|
||
- radv: use global atomics for generated/written primitives query on GFX12
|
||
- radv: re-emit streamout state for GFX12 when the user SGPR changes
|
||
- radv: exclude layer when recomputing FS input bases
|
||
- ac/cmdbuf: program SPI_SHADER_GS_MESHLET_CTRL to 0 in the GFX12 preamble
|
||
- radv: program COMPUTE_DISPATCH_INTERLEAVE on GFX12
|
||
- radv: add support for BO metadata on GFX12
|
||
- radv: add a new helper to set image BO metadata
|
||
- ac/gpu_info: add gfx12_supports_display_dcc
|
||
- radv: fix an assertion about DCC and modifier on GFX12
|
||
- radv: fix the number of drm modifier planes for DCC on GFX12
|
||
- ci: update VKCTS main to a9988483c0864d7190e5e6264ccead95423dfd00
|
||
- radv/ci: update descriptor buffer skipped tests
|
||
- radv: fix disabling logic op for srgb/float formats when blending is enabled
|
||
- radv: disable video support on GFX12
|
||
- radv: disable VK_KHR_cooperative_matrix on GFX12
|
||
- radv: fix programming pitches for LINEAR_SUB_WINDOW on GFX12
|
||
- radv: fix programming mip level for TILED_SUB_WINDOWS on GFX12
|
||
- radv/ci: add expected list of failures for GFX1200
|
||
- radeonsi: fix programming DCC for SDMA on GFX12
|
||
- radv: use stage instead of entrypoint to determine valid gfx stages
|
||
- docs: add a note about GFX12 (RDNA4) on RADV
|
||
- ac,radeonsi: add SDMA DCC tiling for GFX12+
|
||
- ac/descriptors: allow to configure DCC for buffer descriptors
|
||
- radv/amdgpu: add support for AMDGPU_GEM_CREATE_GFX12_DCC
|
||
- radv/meta: add missing pipeline lookups
|
||
- radv/meta: stop using string keys also for DGC and query objects
|
||
- util/disk_cache: add a new helper to create a disk cache
|
||
- vulkan/runtime: allow to use a different disk cache
|
||
- radv: fix caching on-demand meta shaders
|
||
- radv: fix adding the BO to cmdbuf list when starting conditional rendering
|
||
- radv: fix fetching draw vertex data from counter buffers with transform feedback
|
||
- radv/meta: disable conditional rendering for fill/update buffer operations
|
||
- radv: fix adding the VRS image BO to the cmdbuf list on GFX11
|
||
- ac,radv,radeonsi: add new GFX12_DCC_WRITE_COMPRESS_DISABLE tiling flag
|
||
- ac/gpu_info: add gfx12_supports_dcc_write_compress_disable
|
||
- radv: add initial DCC support on GFX12
|
||
- radv: fix adding the BO for unaligned SDMA copies to the cmdbuf list
|
||
|
||
Saroj Kumar (1):
|
||
|
||
- ac/surface: fix missing NULL check in gfx12_select_swizle_mode()
|
||
|
||
Sathishkumar S (1):
|
||
|
||
- radeonsi/vcn: enable roi decode and rgb targets on JPEG_5_0_1
|
||
|
||
Scott Moreau (1):
|
||
|
||
- dri: Fix hardware cursor for cards without modifier support
|
||
|
||
Serdar Kocdemir (4):
|
||
|
||
- Change C style cast on extension structs
|
||
- Wrap queue related functions on codegen
|
||
- The BumpPool of VkStream is not freeAll'ed
|
||
- gfxstream: add VK_DRIVER_FILES to devenv
|
||
|
||
Sergi Blanch Torne (6):
|
||
|
||
- 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: disable Collabora's farm due to unexpected power cut
|
||
- Revert "ci: disable Collabora's farm due to unexpected power cut"
|
||
|
||
Shashank Sharma (1):
|
||
|
||
- amd: add new AMDGPU_INFO subquery for userqueue metadata
|
||
|
||
Sil Vilerino (26):
|
||
|
||
- vl/vl_winsys: Add missing include for function declaration
|
||
- u_dynarray.h: Fix warning C4267 conversion from 'size_t' to 'type', possible loss of data
|
||
- u_math.h: Change power of two assert to fix warning C4146: unary minus operator applied to unsigned type, result still unsigned
|
||
- src/gallium/auxiliary/util/u_draw.h: Fix C4244 'argument' : conversion from 'type1' to 'type2', possible loss of data
|
||
- util: Fix warning C4244 'argument' : conversion from 'type1' to 'type2', possible loss of data
|
||
- src/compiler: Fix warning C4244 'argument' : conversion from 'type1' to 'type2', possible loss of data
|
||
- src/compiler: Fix warning C4389: An == or != operation involved signed and unsigned variables. This could result in a loss of data.
|
||
- d3d12: Fix warning C4267 conversion from 'size_t' to 'type', possible loss of data
|
||
- d3d12: Fix warning C4244 'argument' : conversion from 'type1' to 'type2', possible loss of data
|
||
- d3d12: Fix warning C4389: An == or != operation involved signed and unsigned variables. This could result in a loss of data.
|
||
- d3d12: Fix warning C4018 signed/unsigned mismatch
|
||
- d3d12: Add offset limit check to d3d12_resource_from_memobj
|
||
- d3d12_bufmgr.cpp: Fix warning C4244 for x86 builds assign uint64_t to size_t
|
||
- util: cpu_detect.c Fix warning C5274: behavior change: _Alignas no longer applies to the type '<unnamed-tag>' (only applies to declared data objects)
|
||
- d3d12_video_encoder_bitstream_builder_h264: Fix warning C4244 for x86 builds assign uint64_t to size_t
|
||
- d3d12_resource: Fix warning C4244 for x86 builds assign uint64_t to uintptr_t
|
||
- d3d12_video_dec_h264: Fix warning C4244 uint64_t to size_t cast
|
||
- d3d12_video_dec_vp9.cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'const unsigned int', possible loss of data
|
||
- d3d12_video_dec_hevc.cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'const unsigned int', possible loss of data
|
||
- d3d12_video_proc.h/cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'const unsigned int', possible loss of data
|
||
- d3d12_video_enc_av1.cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'unsigned int', possible loss of data
|
||
- d3d12_video_enc_h264.cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'unsigned int', possible loss of data
|
||
- d3d12_video_enc_hevc.cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'unsigned int', possible loss of data
|
||
- d3d12_video_dec.h/cpp: Fix warning C4244: 'argument': conversion from 'uint64_t' to 'unsigned int', possible loss of data
|
||
- d3d12_video_enc.h/cpp Fix warning C4244: 'argument': conversion from 'uint64_t' to 'unsigned int', possible loss of data
|
||
- d3d12: Enable Warnings C4267, C4996, C4146, C4244, C4389, C4838, C4302, C4018 in src/gallium/drivers/d3d12 subtree
|
||
|
||
Simon Perretta (70):
|
||
|
||
- pvr: add initial pco stub/boilerplate
|
||
- pvr, pco: Add new compiler framework and shader gen stubs
|
||
- pco: add env debug option parsing
|
||
- pco: stubs for SPIR-V/NIR compilation options
|
||
- pvr: connect basic pco functions to the driver
|
||
- pvr: remove pipeline shader hard-coding support
|
||
- pvr: add device info and functions for calculating available temps
|
||
- pvr: add shader compilation stubs
|
||
- pvr: track pipeline flags
|
||
- pvr: add device info for additional iterator features
|
||
- pvr: fix GetInstanceProcAddr ubsan warning when _instance == NULL
|
||
- pvr: drop PVRX macro
|
||
- pco: suppress warning for functions passing structs
|
||
- pco: pygen stubs
|
||
- pco, pygen: enum emit support, define some enums and op/ref mods/types
|
||
- pco, pygen: define basic isa field types
|
||
- pco, pygen: define and emit isa instruction group header variant fields
|
||
- pco, pygen: isa instruction group header validation and encoding support
|
||
- pco, pygen: isa lower source definitions
|
||
- pco, pygen: isa upper sources definitions
|
||
- pco, pygen: isa internal source selector definitions
|
||
- pco, pygen: isa destination definitions
|
||
- pco, pygen: isa main alu ops
|
||
- pco, pygen: isa backend alu ops
|
||
- pco, pygen: isa bitwise alu ops
|
||
- pco, pygen: isa control alu ops
|
||
- pco, pygen: query bytes required for each variant
|
||
- pco, pygen: generate op and mod info
|
||
- pco: define data structures and basic builder implementation with ops
|
||
- pco: NIR translation and PCO IR pass boilerplate
|
||
- pco: printing and validation boilerplate
|
||
- pco, pygen: generate string representations of enum elements
|
||
- pco: basic instruction printing
|
||
- pco, pygen: move unnamed tuple structs into classes
|
||
- pco, pygen: add bitset support for op mods
|
||
- pco, pygen: common underscore replacement for op names
|
||
- pco: add verbose printing debug option
|
||
- pco, pygen: distinguish hw ops that are built directly into instruction groups
|
||
- pco, pygen: instruction to instruction group mapping, printing
|
||
- pco: additional ref functions
|
||
- pco: boilerplate nir lowering passes
|
||
- pco, pygen: add initial uvsw op boilerplate
|
||
- pco, pygen: add better exception messages
|
||
- pco: adjust align padding to be per-function instead of per-shader
|
||
- pco, pygen: support querying ref mods, if op/ref mods have been set
|
||
- pco: set up and tear down glsl type singleton with context
|
||
- pco, pygen: add support for instructions with variable srcs/dests
|
||
- pco, pygen: re-order some mods to match their evaluation order
|
||
- pco: print ranges of non-ssa refs with >1 channel, datatypes for immediates
|
||
- pco, pygen: drop unspecified bit sizes for references
|
||
- pco, pygen: add defs and mappings for common ops
|
||
- pco, pygen: restructure igrp alu components into arrays
|
||
- pco, pygen: amend bitfield assertion messages
|
||
- pco, pygen: isa ditr op
|
||
- pco, pygen: isa itrsmp op
|
||
- pco: initial implementation of translation and passes
|
||
- pco: add public print wrappers
|
||
- pco: vector component tracking, vector collation when ingesting NIR
|
||
- pco: re-indexing debug option and additional vector and component tracking
|
||
- pco: add mappings and translation for ditr
|
||
- pco: temporarily add hardcoded vs/fs I/O for testing, BXS-4-64 iteration support
|
||
- pco: add helpers for overriding ref chans and offsetting vals
|
||
- pco: vec coalescing improvement to register allocation
|
||
- pco: add opt subpass for propagating comps referencing hw regs
|
||
- pco: track the number of bytes encoded for each function
|
||
- pvr, pco: rewrite compiler/driver interface for vs & fs I/O
|
||
- pco: modifier propagation optimization, shared opt context boilerplate
|
||
- pco: initial validation boilerplate and SSA checks
|
||
- CODEOWNERS: update for new pco compiler tree
|
||
- pco: fix x86 build
|
||
|
||
Simon Ser (6):
|
||
|
||
- dri: revert INVALID modifier special-casing
|
||
- llvmpipe: handle llvmpipe_resource_map() errors
|
||
- dri: don't fetch X11 modifiers if we don't support them
|
||
- egl/wayland: only supply LINEAR modifier when supported
|
||
- egl/wayland: fallback to implicit modifiers if advertised by compositor
|
||
- gbm: fix get_back_bo() failure with gbm_surface and implicit modifiers
|
||
|
||
Sonny Jiang (1):
|
||
|
||
- radeonsi/vcn: Add vcn_5_0_1 support
|
||
|
||
Tapani Pälli (21):
|
||
|
||
- intel/dev: update mesa_defs.json from workaround database
|
||
- anv: utilize ray query bo per queue for Wa_14022863161
|
||
- anv: extend Wa_14017794102 with lineage Wa_14023061436
|
||
- isl: modify existing assert by allowing CCS_E aux usage
|
||
- intel/dev: update mesa_defs.json from workaround database
|
||
- intel/dev: lower amount of max gs threads for Wa_18040209780
|
||
- anv/android: always create 2 graphics and compute capable queues
|
||
- iris: allow bo cache for compressed bos on verx10 == 200
|
||
- drirc/anv: force_vk_vendor=-1 for Marvel Rivals
|
||
- intel/dev: update mesa_defs.json from internal database
|
||
- dri: remove GLsync typedef
|
||
- anv: handle mesh in sbe_primitive_id_override
|
||
- iris: initialize whole pipe_box struct for memcmp
|
||
- intel/compiler: take reg_unit size into account with ubo ranges
|
||
- anv: set dependency between SF_CLIP and CC_PTR states
|
||
- mesa/st: take pixelmaps in to account in drawpixels cache
|
||
- intel/dev: update mesa_defs.json from internal database
|
||
- isl: use workaround framework for Wa_1207137018
|
||
- mesa: enable GL_EXT_conservative_depth extension
|
||
- anv: tighten condition for changing barrier layouts
|
||
- anv: apply cache flushes on pipeline select with gfx20
|
||
|
||
Thomas H.P. Andersen (2):
|
||
|
||
- drirc/nvk: force_vk_vendor=-1 for Artifact Classic
|
||
- nvk: follow naming convention for devices
|
||
|
||
Tim Huang (1):
|
||
|
||
- amd: add GFX v11.5.3 support
|
||
|
||
Tim Keller (1):
|
||
|
||
- dril: Check for null config in dril_target.c
|
||
|
||
Timothy Arceri (24):
|
||
|
||
- glsl/nir: fix function cloning at link time
|
||
- glsl: fix compiler global temp collisions
|
||
- glsl: tidy up glsl_to_nir() params
|
||
- glsl: remove unused member
|
||
- Revert "glsl: Move ForceGLSLAbsSqrt handling to glsl-to-nir."
|
||
- glsl: remove more now unused params from glsl_to_nir()
|
||
- glsl: don't copy symbol table to shaders
|
||
- glsl: drop _mesa_glsl_copy_symbols_from_table()
|
||
- glsl: use symbol table directly for builtin functions
|
||
- glsl: drop unused symbol table from gl_shader
|
||
- glsl: disable function return lowering in glsl ir
|
||
- glsl: remove return lowering from glsl ir
|
||
- glsl: drop last remaining lower jump test
|
||
- glsl: remove now unused ir reader
|
||
- glsl: move _mesa_glsl_compile_shader() declaration
|
||
- glsl: remove glsl/program.h
|
||
- nir: allow loops with unknown induction var initialiser to unroll
|
||
- glsl: drop unused ir_equals.cpp
|
||
- glsl: drop unused array refcount code and tests
|
||
- glsl: drop opt_dead_code_local
|
||
- glsl: enable layout qualifier if OVR_multiview enabled
|
||
- glsl: fix num_views validation message
|
||
- glsl: fix num_views linker error
|
||
- glsl: fix return value for subgroupBallot()
|
||
|
||
Timur Kristóf (109):
|
||
|
||
- radv: Mark GS copy shaders as internal.
|
||
- radv: Add ability to dump shaders based on stage.
|
||
- aco: Separate options for printing IR and recording disassembly.
|
||
- radv: Separate option to dump NIR.
|
||
- radv: Separate option to print shader disassembly.
|
||
- radv: Separate option to dump backend IR.
|
||
- radv: Refactor RADV_DEBUG=shaders to be a combination of other options.
|
||
- radv: Slightly reword preoptir debug flag.
|
||
- radv: Also allow filtering SPIR-V dump per stage.
|
||
- radv: Set dump flags in a smarter way by default.
|
||
- amd: Rename GFX1103_R1/R2 to PHOENIX/2
|
||
- radv: Add a flush postamble on GFX6.
|
||
- radv: Don't flush at the end of each command buffer on GFX6.
|
||
- ac/nir/ngg: Don't emit dead code with dot_op.
|
||
- ac/nir/ngg: Trade 1 VALU shift for 2 SALU add.
|
||
- ac/nir/cull: Slightly refactor control flow for small primitive culling.
|
||
- ac/nir/ngg: Slightly refactor workgroup scan.
|
||
- ac/nir/ngg: Pass wg_repack_result as pointer instead of returning it.
|
||
- ac/nir/ngg: Workgroup scan over two bools.
|
||
- ac/nir/ngg: Implement optional primitive compaction.
|
||
- ac/nir/ngg: Remove erroneous NUW addition from workgroup scan.
|
||
- radv: Reorder potentially per-primitive FS builtins.
|
||
- radv: Slightly simplify potentially per-primitive FS inputs.
|
||
- radv, aco: Consolidate num_interp + num_prim_interp into num_inputs.
|
||
- radv: Emit SPI_PS_IN_CONTROL when emitting PS inputs on GFX10.3.
|
||
- radv: Remove now unused num_prim_interp from shader_info.
|
||
- radv: Use default 0 for undefined builtin PS inputs.
|
||
- radv: Only set NGG_DISABLE_PROVOK_REUSE for VS.
|
||
- ac/nir/ngg: Add ability to store primitive ID as per-primitive.
|
||
- radv: Reorder FS primitive ID input after layer and viewport.
|
||
- radv: Configure implicit VS primitive ID to be per-primitive.
|
||
- ac/nir/ngg: Use ac_nir_prerast_out in mesh shader lowering.
|
||
- ac/nir/ngg: Simplify updating mesh shader output info.
|
||
- ac/nir: Pass ac_nir_prerast_out to ac_nir_export_parameters.
|
||
- ac/nir: Pass ac_nir_prerast_out to ac_nir_export_position.
|
||
- ac/nir: Introduce ac_nir_store_parameters_to_attr_ring.
|
||
- ac/nir/ngg: Refactor VS/TES attribute ring stores.
|
||
- ac/nir/ngg: Refactor GS attribute ring stores.
|
||
- ac/nir/ngg: Refactor export_pos0_wait_attr_ring.
|
||
- ac/nir/ngg: Remove dead code for attribute ring stores.
|
||
- ac/nir/ngg: Move wait attr ring workaround for GS to better place.
|
||
- ac/nir/ngg: Move emitting GS vertex param exports to if.
|
||
- ac/nir/ngg: Refactor storing per-primitive primitive ID to attribute ring.
|
||
- ac/nir: Mark when pre-rast output is used as varying or sysval.
|
||
- ac/nir: Split GS output usage masks to varying and sysval masks.
|
||
- ac/nir: Only export positions when they are really system values.
|
||
- ac/nir: Only export parameters when they are actually varying.
|
||
- ac/nir: Only store params to attribute ring that are varying.
|
||
- aco: Update documentation
|
||
- radv: Add some documentation.
|
||
- radv: Implement FS layer ID input as a system value.
|
||
- Revert "nir/opt_varyings: Add workaround for RADV mesh shader multiview."
|
||
- ac/nir/ngg: Don't mark multiview layer output as varying.
|
||
- amd: Set lower_layer_fs_input_to_sysval in common code, not in drivers.
|
||
- radv: Rename layer_input to reads_layer in PS info.
|
||
- radv: Only print "testing use only" message on GFX12+.
|
||
- ac/nir: Move ac_nir_lower_bit_size_callback to ac_nir.c
|
||
- ac/nir: Move ac_nir_get_mem_access_flags to ac_nir.c
|
||
- ac/nir: Move ac_nir callback functions to ac_nir.c
|
||
- ac/nir: Move ac_set_nir_options to ac_nir.c
|
||
- ac: Stop including nir.h in ac_shader_util.h
|
||
- ac/nir: Rename emit_streamout to ac_nir_emit_legacy_streamout
|
||
- ac: Move ac_nir_config struct to ac_nir.h
|
||
- ac/nir: Move ac_nir_create_gs_copy_shader to separate file.
|
||
- ac/nir: Expose ac_nir_unpack_value in ac_nir_helpers.h
|
||
- ac/nir: Move ac_nir_lower_intrinsics_to_args to separate file.
|
||
- ac/nir: Move ac_nir_lower_legacy_vs to separate file.
|
||
- ac/nir: Move ac_nir_lower_legacy_gs to separate file.
|
||
- ac/nir: Move ac_nir_gs_shader_query declaration to ac_nir_helpers.h
|
||
- ac/nir: Move ac_nir_opt_pack_half to separate file.
|
||
- ac/nir: Move ac_nir_lower_mem_access_bit_sizes to separate file.
|
||
- ac/nir: Move ac_nir_lower_sin_cos to separate file.
|
||
- ac/nir: Move pre-rasterization related utilities in separate file.
|
||
- ac/nir: Rename ac_nir_lower_ngg_ms to ac_nir_lower_ngg_mesh.
|
||
- ac/nir: Move ac_nir_lower_ngg_mesh to separate file.
|
||
- ac: Move AC_HS_MSG_VOTE_LDS_BYTES to ac_shader_util.h
|
||
- ac: Stop including ac_nir.h from ac_shader_util.c
|
||
- ac/nir: Move all ac_nir_* files to a new folder.
|
||
- radv: Lower array derefs of vectors outside of shader linking.
|
||
- ac/nir/ngg: Mitigate NGG fully culled bug when GS output is compile-time zero.
|
||
- ac/nir/ngg: Mitigate attribute ring wait bug when primitive ID is per-primitive.
|
||
- aco: Move NGG pos export scheduling determination to drivers.
|
||
- ac/nir/ngg: Remove some superfluous variables from culling code.
|
||
- ac/nir/ngg: Add a few comments explaining some variables.
|
||
- ac/nir/ngg: Remove unused vs_output struct.
|
||
- ac/nir/ngg: Carve out ac_nir_ngg_alloc_vertices_and_primitives.
|
||
- ac/nir/ngg: Use ac_nir_ngg_alloc_vertices_and_primitives in mesh shader lowering.
|
||
- ac/nir/ngg: Carve out ac_nir_create_output_phis.
|
||
- ac/nir/ngg: Carve out NGG streamout code.
|
||
- ac/nir/ngg: Carve out ac_nir_repack_invocations_in_workgroup.
|
||
- ac/nir/ngg: Slightly refactor emitting vertex parameters.
|
||
- ac/nir/ngg: Add radeon_info to NGG lowering options.
|
||
- ac/nir/ngg: Add and use a has_attr_ring_wait_bug field to ac_gpu_info.
|
||
- ac/nir/ngg: Add and use a has_attr_ring field to ac_gpu_info.
|
||
- ac/nir/ngg: Add and use a has_ngg_fully_culled_bug field to ac_gpu_info.
|
||
- ac/nir/ngg: Add and use a has_ngg_passthru_no_msg field to ac_gpu_info.
|
||
- ac/nir/ngg: Use gfx_level from radeon_info.
|
||
- ac/nir/ngg: Remove gfx_level and family from NGG lowering options.
|
||
- ac/nir/ngg: Pass radeon_info to mesh shader lowering.
|
||
- ac/nir/ngg: Use has_attr_ring and has_attr_ring_wait_bug in mesh shader lowering too.
|
||
- ac/nir/ngg: Rework attribute ring wait workaround in VS/TES.
|
||
- ac/nir/ngg: Carve out ngg_gs_process_out_primitive.
|
||
- ac/nir/ngg: Carve out ngg_gs_process_out_vertex.
|
||
- ac/nir/ngg: Rework GS output code for better attribute ring handling.
|
||
- ac/nir/ngg: Remove now unused export_pos0_wait_attr_ring.
|
||
- ac/nir/ngg: Don't call has_input_primitive in GS lowering.
|
||
- ac/nir/ngg: Move GS lowering to separate file.
|
||
- radv, radeonsi: Disable early prim export on GFX11+.
|
||
- ac/nir/ngg: Use SALU to calculate which threads store to attribute ring in GS.
|
||
|
||
Tomeu Vizoso (42):
|
||
|
||
- etnaviv/ml: Fix includes
|
||
- etnaviv/nn: Fix use of etna_core_info
|
||
- etnaviv/ci: Add expectation files for the VIPNano-SI+ NPU
|
||
- etnaviv/ml: Rework the dumping of tensors
|
||
- etnaviv: Add script to decode weights in Huffman format
|
||
- etnaviv/ml: Split V7 coefficient encoding to a new file
|
||
- etnaviv/ml: Add encoding of coefficients for V8
|
||
- etnaviv/ml: Fix padding for convolutions in V8
|
||
- etnaviv/ml: Implement tiling for V8
|
||
- etnaviv/ml: Set two bits in the NN instruction for V8
|
||
- etnaviv/ml: Disable caching on V8
|
||
- etnaviv/ml: Fix reshuffle TP jobs on V8
|
||
- etnaviv/ml: Only reshuffle when needed on V8
|
||
- etnaviv/ml: Make use of the new depthwise support in V8
|
||
- etnaviv/ci: Update expectations for the NPU in the A311D
|
||
- etnaviv/ml: Zero out the NN config
|
||
- etnaviv/ml: Zero all BOs
|
||
- teflon: Support multiple graph inputs and outputs
|
||
- etnaviv/ml: Adapt to changes in teflon regarding multiple inputs
|
||
- etnaviv/ml: Support addition operations on V8
|
||
- teflon: Add files mentioned in the docs for image classification
|
||
- teflon/docs: Update performance measurements on LibreComputer Alta
|
||
- teflon/docs: Add i.MX8MP to list of supported NPUs
|
||
- teflon/docs: Clarify smoke test instructions
|
||
- teflon: Add tests for the YOLOX model
|
||
- teflon: Support tests with inputs with less than 4 dims
|
||
- teflon: Rename model tests so they aren't skipped by gtest-runner
|
||
- teflon: Don't crash when a tensor isn't quantized
|
||
- teflon/tests: Add support for models with float inputs and outputs
|
||
- teflon/tests: Also use the cache for models in the test suite
|
||
- etnaviv/ml: Specify which of the input tensors need transposing.
|
||
- etnaviv/ml: Fix in_image_slice in transposes when width != height
|
||
- etnaviv/ml: Take offsets into account in TP operations
|
||
- teflon: Add support for tensor split and concatenation operations
|
||
- etnaviv/ml: Add support for tensor split and concatenation operations
|
||
- teflon: Limit support for Add to two unpopulated tensors
|
||
- etna/ml: Write out the size of the requested tensor
|
||
- teflon: Add support for tensor padding operations
|
||
- etnaviv/ml: Add support for tensor padding operations
|
||
- teflon: Add support for FullyConnected
|
||
- teflon: Add tests for FullyConnected
|
||
- etnaviv/ml: Implement FullyConnected
|
||
|
||
Valentine Burley (99):
|
||
|
||
- amd/ci: Drop x86_64 suffix from job names
|
||
- amd/ci: Merge and convert Raven piglit testing
|
||
- amd/ci: Convert LAVA RADV jobs to deqp-runner suites
|
||
- amd/ci: Increase fraction for radeonsi-raven-piglit
|
||
- panfrost/ci: Turn redundant GLESCTS-full run into disabled Piglit job
|
||
- svga/ci: Convert to deqp-runner suite
|
||
- panfrost/ci: Convert to deqp-runner suite
|
||
- ci: Drop lava-piglit:(x86_64|arm64) definitions
|
||
- radv/ci: Convert Valve RADV jobs to deqp-runner suites
|
||
- turnip/ci: Bump the number of tests per group for a618
|
||
- turnip/ci: Bump the number of tests per group for a630
|
||
- turnip/ci: Bump the number of tests per group for a660
|
||
- turnip/ci: Decrease fraction for a630-vk-asan
|
||
- turnip/ci: Adjust some timeouts
|
||
- turnip/ci: Remove a630-vk-asan skip
|
||
- turnip/ci: Update expectations
|
||
- freedreno/ci: Drop redundant DEQP_VER
|
||
- turnip/ci: Ony increase hangcheck timer for spilling tests on a630
|
||
- lavapipe/ci: Convert lavapipe-vk-asan to a deqp-runner suite
|
||
- etnaviv/ci: Convert to deqp-runner suites
|
||
- softpipe/ci: Convert softpipe-asan-gles31 to a deqp-runner suite
|
||
- radv/ci: Use deqp-vk-main in Raven and Stoney RADV jobs
|
||
- turnip/ci: Enable ASan leak detection in a630-vk-asan
|
||
- ci/deqp: Remove non-suite support
|
||
- llvmpipe/ci: Move Piglit timeout inside the suite
|
||
- ci/deqp: Simplify conditional arguments
|
||
- ci/deqp: Add a DEQP_FORCE_ASAN option
|
||
- llvmpipe/ci: Actually enable ASan testing for llvmpipe-deqp-asan
|
||
- anv/ci: Fix GPU_VERSION configuration for anv-jsl and anv-jsl-full
|
||
- anv/ci: Bump the number of tests per group for ADL
|
||
- anv/ci: Bump the number of tests per group for JSL
|
||
- anv/ci: Bump the number of tests per group for TGL
|
||
- anv/ci: Re-enable TGL and JSL manual jobs
|
||
- anv/ci: Remove fails that are in .gitlab-ci/all-skips.txt
|
||
- anv/ci: Update expectations
|
||
- ci/lava: Use CI_JOB_TIMEOUT instead of separate variable
|
||
- ci/windows: Bump the number of tests per group
|
||
- ci/windows: Add a manual full job
|
||
- ci/windows: Update expectations
|
||
- turnip/ci: Update expectations
|
||
- ci/windows: Always include windows-msvc in scheduled pipelines
|
||
- panvk/ci: Move the fractions out of suites
|
||
- panvk/ci: Bump the number of tests per group for G52
|
||
- lavapipe/ci: Bump the number of tests per group
|
||
- lavapipe/ci: Update expectations
|
||
- venus/ci: Bump the number of tests per group
|
||
- venus/ci: Update expectations
|
||
- angle/ci: Update expectations
|
||
- zink/ci: Update expectations for ANV
|
||
- turnip/ci: Document flake
|
||
- lavapipe/ci: Update expectations
|
||
- lavapipe/ci: Re-enable lavapipe-vk-asan
|
||
- ci: Uprev vkd3d-proton to b121e6d746341e0aaba7663e3d85f3194e8e20e1
|
||
- virgl/ci: Disable virgl-iris-traces-performance
|
||
- virgl/ci: Migrate the two iris jobs to 1130g7-volteer
|
||
- anv/ci: Increase anv-tgl-angle parallelism to 2
|
||
- zink/ci: Migrate the two TGL traces jobs to 1130g7-volteer
|
||
- zink/ci: Increase zink-anv-tgl parallelism to 4
|
||
- ci: Add Valentine to the restricted traces access list
|
||
- freedreno/ci: Update a630-traces-restricted checksums
|
||
- zink/ci: Skip crashing trace in zink-anv-tgl-traces-restricted
|
||
- turnip/ci: Decrease the fraction on a660-vk-full
|
||
- ci: Fix trace update script reading GitLab token from default location
|
||
- pan/ci: Document some flakes
|
||
- android/ci: Allow specifying Vulkan driver in cuttlefish-runner.sh
|
||
- android/ci: Build ANV for Android
|
||
- freedreno/ci: Update expectations
|
||
- panfrost/ci: Revert to 6.6 kernel on G57
|
||
- amd/ci: Add lava-hp-x360-14a-cb0001xx-zork and use it for VA-API testing
|
||
- amd/ci: Run full radeonsi-raven-va job pre-merge
|
||
- freedreno/ci: Update expectations again
|
||
- turnip/ci: Bump the number of tests per group for a630-vk-asan
|
||
- anv/ci: Move a test to common anv-skips
|
||
- ci: Uprev VKCTS to 1.4.1.0
|
||
- pan/ci: Properly wire up DRIVER_NAME
|
||
- panvk/ci: Skip waived tests
|
||
- ci: Uprev VKCTS to 1.4.1.1
|
||
- ci: Skip broken PenumbraOverture trace for zink and freedreno
|
||
- zink/ci: Update checksum for Osmos trace on TGL
|
||
- anv/ci: Revert to 6.6 kernel on anv-jsl
|
||
- iris/ci: Decrease iris-glk-deqp paralellism
|
||
- panfrost/ci: Move panfrost-g52-piglit to nightly
|
||
- zink/ci: Increase zink-anv-adl parallelism
|
||
- turnip/ci: Increase a660-vk fraction
|
||
- freedreno/ci: Decrease a660-gl paralellism
|
||
- freedreno/ci: Disable a618-gl, a618-egl, and a618-piglit
|
||
- turnip/ci: Disable a630-vk
|
||
- freedreno/ci: Decrease a630-gl parallelism
|
||
- freedreno/ci: Re-enable some traces on a618 and disable a630-traces
|
||
- zink/ci: Increase parallelism of zink-tu-a618
|
||
- freedreno/ci: Don't automatically retry manual jobs
|
||
- freedreno/ci: Migrate a618-piglit-full to kingoftown
|
||
- amd/ci: Migrate amd-raven-skqp from lenovo-zork to hp-zork
|
||
- anv/ci: Decrease anv-jsl-angle parallelism
|
||
- virgl/ci: Skip flaky trace
|
||
- amd/ci: Increase amd-raven-skqp parallelism
|
||
- freedreno/ci: Document flakes
|
||
- venus/ci: Skip flaky test due to intermittent timeouts
|
||
- amd/ci: Revert to 6.6 kernel on Raven
|
||
|
||
Vignesh Raman (6):
|
||
|
||
- ci: Uprev crosvm
|
||
- ci: Force db410c to host mode
|
||
- ci: Uprev kernel to 6.13
|
||
- ci: update expectation files
|
||
- ci: export RESULTS_DIR in crosvm-script.sh
|
||
- ci: use CI_PROJECT_NAME for artifacts name
|
||
|
||
Vinson Lee (4):
|
||
|
||
- hk: Fix hk_ia_update arguments order
|
||
- vulkan: Add missing va_end
|
||
- intel/elk: Fix assert with side effect
|
||
- hk: Fix build error with static_assert
|
||
|
||
Visan, Tiberiu (3):
|
||
|
||
- amd/vpelib: patch to match shader (#456)
|
||
- amd/vpelib: remove luma offset (#459)
|
||
- amd/vpelib: fixed file headers for Palamida scan
|
||
|
||
Vldly (1):
|
||
|
||
- freedreno: Fix resource tracking on repeated map with discard
|
||
|
||
Xaver Hugl (1):
|
||
|
||
- vulkan/wsi: unset GAMMA_LUT, CTM and DEGAMMA_LUT when doing a modeset
|
||
|
||
Yinjie Yao (3):
|
||
|
||
- radeonsi/vcn: Indentation fix
|
||
- radeonsi/vcn: Fix compile warnings with previously uninitialized variables.
|
||
- radeonsi/vcn: Disable 2pass encode for VCN 5.0.
|
||
|
||
Yiwei Zhang (4):
|
||
|
||
- venus: enable VK_EXT_external_memory_acquire_unmodified if needed
|
||
- venus: use dedicated allocation for ANB image memory import
|
||
- venus: fix to handle pipeline flags2 from maint5
|
||
- venus: fix maintenance5 props init and create flags2
|
||
|
||
Yogesh Mohan Marimuthu (25):
|
||
|
||
- amd: update amdgpu_drm.h for new userq ioctl
|
||
- amd: include amdgpu_drm.h from mesa instead of system for ac_fake_hw_db.h
|
||
- winsys/amdgpu: add DOORBELL domain to bo
|
||
- winsys/amdgpu: add CLEAR_VRAM flag to zero vram when creating bo
|
||
- winsys/amdgpu: add userq helper functions
|
||
- ac/gpuinfo: add use_userq and AMD_USERQ variable
|
||
- winsys/amdgpu: call userq init and destroy functions
|
||
- ac: add new userq signal and wait packet id
|
||
- ac: add inherit vmid field to indirect buffer packet
|
||
- winsys/amdgpu: use bo_va_op_raw() function instead of bo_va_op()
|
||
- winsys/amdgpu: use timeline syncobj for userq vm operations
|
||
- winsys/amdgpu: destroy bo_fence_lock late in do_winsys_deinit()
|
||
- winsys/amdgpu: pass job fences to VM ioctl
|
||
- winsys/amdgpu: wait for vm syncobj before creating userq
|
||
- winsys/amdgpu: move noop and ib_bytes adjustment to cs_flush
|
||
- winsys/amdgpu: move legacy chunk init and submission to new function
|
||
- winsys/amdgpu: add userq cmd submission support in amdgpu_cs_submit_ib()
|
||
- winsys/amdgpu: don't add fence dependency of other queues for userq
|
||
- winsys/amdgpu: send hdp flush packet for userq
|
||
- winsys/amdgpu: keep has_local_buffers true for userq
|
||
- winsys/amdgpu: use VM_ALWAYS_VALID for all VRAM and GTT allocations
|
||
- ac/gpu_info: populate fw info using new fw info ioctl for userq
|
||
- winsys/amdgpu: ring doorbell before calling userq_signal ioctl
|
||
- winsys/amdgpu: use next_wptr as cache for userq
|
||
- winsys/amdgpu: ensure strict order in updating mqd wptr and doorbell
|
||
|
||
You, Min-Hsuan (1):
|
||
|
||
- amd/vpelib: fix coverity defects
|
||
|
||
Zan Dobersek (8):
|
||
|
||
- fd/pps: specify counter group for each countable
|
||
- fd/pps: provide derived counters on a7xx
|
||
- freedreno/registers: update RB_BLIT_INFO, RB_CCU_CNTL
|
||
- tu/a7xx: use concurrent resolve groups
|
||
- tu: ensure completion of generic-clear resolves for color, depth/stencil clears
|
||
- tu/a7xx: support 8x MSAA
|
||
- freedreno/registers: fix RBBM_PRIMCTR understanding and usage
|
||
- freedreno/a7xx: fix fd_lrzfc_layout
|
||
|
||
Zhao, Jiali (1):
|
||
|
||
- amd/vpelib: 420 and 422 Output Single Segment cositing support
|
||
|
||
Zoltán Böszörményi (3):
|
||
|
||
- features.txt: Add Vulkan 1.4 section
|
||
- docs/features: Mark VK_EXT_host_image_copy as implemented on Turnip
|
||
- docs/features: Mark more Vulkan 1.4 features as done for drivers
|
||
|
||
duncan.hopkins (9):
|
||
|
||
- glx: change \`#if` guard around \`dri_common.h` to stop missing 'driDestroyConfigs' symbol on MacOS builds.
|
||
- glx: ignore zink check for has_explicit_modifiers and DRI3 on MacOS.
|
||
- kopper: Add '#if' guard around \`loader_dri3_get_pixmap_buffer` to stop missing symbol on MacOS.
|
||
- glx: Guard some of the bind_extensions() code with the same conditions as \`glx_screen`s `frontend_screen` member.
|
||
- glx: Add back in \`applegl_create_display()` so the OpenGL.framework, on MacOS, pointer get setup.
|
||
- zink: MoltenVk has conditional VK_DYNAMIC_STATE_VERTEX_INPUT_BINDING_STRIDE support.
|
||
- zink: Avoid optimalDeviceAccess on MoltenVK when creating depth taregts.
|
||
- zink, kopper: Conitionally add VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT to swap chain imageUsage.
|
||
- zink: stop zink_set_primitive_emulation_keys producing geometry shaders on platforms that do not support them.
|
||
|
||
liuqiang (2):
|
||
|
||
- lavapipe: Resolved write to pointer after free
|
||
- d3d10umd: Modify comment
|
||
|
||
nyanmisaka (1):
|
||
|
||
- frontends/vdpau: Get AV1 decode subsampling_x/y
|
||
|
||
sergiuferentz (1):
|
||
|
||
- Use try_unbox in VkDescriptorBufferInfo
|