Repository navigation

AGX correctness series — fast-track for upstream submission - #2

Closed

joshuaswarren wants to merge 10000 commits into

omacom:mainfrom

joshuaswarren:upstream/correctness

Closed

AGX correctness series — fast-track for upstream submission#2
joshuaswarren wants to merge 10000 commits into
omacom:mainfrom
joshuaswarren:upstream/correctness

This pull request is big! We're only showing the most recent 250 commits

Commits on Aug 31, 2026
- nvk/meta: Allow gfx for CmdCopyMemoryToImage
  
  drawoc
  authored and
  Marge Bot
  committed
  c1d3b2d
- vk/meta: Match b2i and i2i copy offset calculation
  
  drawoc
  authored and
  Marge Bot
  committed
  580d2dd
- vk/meta: Expand comment in copy_image_prepare_gfx_push_const
  
  drawoc
  authored and
  Marge Bot
  committed
  e3145f1
- panvk/jm: Fix unbound attribute reads
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  44a7a09
Commits on Sep 1, 2026
- nv30: add noticed flake
  
  thegawin
  authored and
  Marge Bot
  committed
  fb95d60
- nv30/ci: add nv40 KHR results
  
  thegawin
  authored and
  Marge Bot
  committed
  3b57800
- nv30/ci: remove tests fixed by 56058ab73f564e4f7baf99f09fdb4d133660f205
  
  thegawin
  authored and
  Marge Bot
  committed
  aeebf22
- nv30/ci: add recent regression
  
  thegawin
  authored and
  Marge Bot
  committed
  22d5601
- kk: Workaround for intermittent memory_model failures in macOS 26 with M3+
  
  aitor-lunarg
  authored and
  Marge Bot
  committed
  b9c45fd
- kk: add base to shared data access
  
  aitor-lunarg
  authored and
  Marge Bot
  committed
  4197710
- kk: move constant data out of scratch
  
  aitor-lunarg
  authored and
  Marge Bot
  committed
  24a6ecd
- radv/ci: update list of expected failures for TAHITI/HAWAII
  
  hakzsam
  authored and
  Marge Bot
  committed
  66a662b
- radv/ci: document more flakes in stoney-angle
  
  mupuf
  authored and
  Marge Bot
  committed
  becbbc4
- radeonsi/ci: update the vangogh expectations
  
  mupuf
  authored and
  Marge Bot
  committed
  55a5195
- amd/ci: Use gfx-ci/firmware repo for CI-tron jobs
  
  valentineburley
  authored and
  Marge Bot
  committed
  bd292a2
- radv/ci: switch to the gfx-ci/linux kernels
  
  mupuf
  authored and
  Marge Bot
  committed
  55a45a5
- panvk/csf: fix layered FB barrier in secondary cmdbufs
  
  olivia-fl
  authored and
  Marge Bot
  committed
  abe2c03
- panvk/csf: support layered indirect draw in primary cmdbufs
  
  olivia-fl
  authored and
  Marge Bot
  committed
  57811fc
- panvk/csf: minor register assignment refactor in launch_draw
  
  olivia-fl
  authored and
  Marge Bot
  committed
  7b4c444
- panvk/csf: support layered draws in inherited secondary cmdbufs
  
  olivia-fl
  authored and
  Marge Bot
  committed
  3583503
- panvk/csf: advertise shaderOutputLayer
  
  olivia-fl
  authored and
  Marge Bot
  committed
  d6e734f
- radv: ignore VGT_STREAMOUT_SYNC on compute queue
  
  hakzsam
  authored and
  Marge Bot
  committed
  c735afa
- radv: ignore non compute cache flushes also for gang CS
  
  hakzsam
  authored and
  Marge Bot
  committed
  267c85c
- radv: cleanup determining when DCC retile is needed
  
  hakzsam
  authored and
  Marge Bot
  committed
  3188359
- radv: slightly refactor the DCC retile path
  
  hakzsam
  authored and
  Marge Bot
  committed
  83f04c7
- radv: fallback to compute for DCC retiles on transfer queue
  
  hakzsam
  authored and
  Marge Bot
  committed
  d2bba11
- radv: stop excluding all QFOTs on the transfer queue
  
  hakzsam
  authored and
  Marge Bot
  committed
  29e6f44
- docs/panfrost: document PAN_I_WANT_A_BROKEN_VULKAN_DRIVER
  
  TheFakeMontyOnTheRun
  authored and
  Marge Bot
  committed
  fc284af
- compiler/rust: Add U8EnumSet::remove()
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  c9aec4e
- kraid/v9: Disallow LD_PKA.b1/b3
  
  gfxstrand
  authored and
  Marge Bot
  committed
  2272d1a
- kraid/hw_tests: Add tests for byte LOAD
  
  gfxstrand
  authored and
  Marge Bot
  committed
  102b9a1
- intel: fixup mcs sampler handle lowering
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  79900cb
- iris: lower gl_FragColor away
  
  alyssarosenzweig
  authored and
  Marge Bot
  committed
  e420c68
- brw: remove unused helper
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  01b3325
- brw: pass shader key to pass tracker in compile_cs
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  340312b
- blorp: stop using FRAG_RESULT_COLOR
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  e5e1eed
- intel: rename a NIR pass that is not tied to BRW
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  0c81842
- intel: extract lower fragment outputs out of jay
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  c4da8b3
- intel/nir: fixup RT write logic with missing alpha
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  5cd7368
- intel/nir: disable sample mask write if we're skipping it
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  35cb8e2
- intel/nir: replicate alpha with alpha_to_coverage & no sample mask
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  934888c
- intel/nir: deal with lower_io with dual source
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  f776640
- brw: help copy propagation for src0 alpha
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  a1f598a
- brw: reuse jay's lowering fs output pass
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  0e6117f
- anv: move non push constant related value out of the push constant section
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  406434f
- anv: someone doesn't know its powers of 2
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  3075b93
- intel/brw: Replace tab by spaces to fix identation on brw_shader::assign_curb_setup()
  
  zehortigoza
  authored and
  Marge Bot
  committed
  bedcb46
- brw: use constant offset in extended descriptor for spill/fill
  
  llandwerlin-intel
  authored and
  Marge Bot
  committed
  58a18fb
- nir/opt_algebraic: remove some redundant iands after explicit io lowering
  
  DadSchoorse
  authored and
  Marge Bot
  committed
  0ca69f7
- nir: add a pass to optimize scalar arrays to vec arrays
  
  DadSchoorse
  authored and
  Marge Bot
  committed
  1d1f212
- radv: use nir_opt_scalar_array_vars_to_vec for shared
  
  DadSchoorse
  authored and
  Marge Bot
  committed
  b2fbb4a
- radv: use nir_opt_scalar_array_vars_to_vec for temps
  
  DadSchoorse
  authored and
  Marge Bot
  committed
  ce0ae0e
- radeonsi/mm: Set PIPE_BIND_SHARED for video buffers
  
  nowrep
  authored and
  Marge Bot
  committed
  6718beb
- radeonsi: Don't copy next pointer from template when creating texture
  
  nowrep
  authored and
  Marge Bot
  committed
  ebca92b
- radeonsi: Support reallocate inplace for multi plane textures
  
  nowrep
  authored and
  Marge Bot
  committed
  2dfdafb
- radeonsi: Don't force PIPE_BIND_SHARED for multi plane textures
  
  nowrep
  authored and
  Marge Bot
  committed
  112ec48
- tu: Fix single texel alignment for non-POT texel buffers
  
  HansKristian-Work
  authored and
  Marge Bot
  committed
  1a83501
- freedreno/computerator: Fix perfcounter calculations
  
  werman
  authored and
  Marge Bot
  committed
  6b6f50e
- nv30/ci: update nv34 results
  thegawin
  committed
  2d220a1
- nv30/ci: add missing ",Fail" to nv43's fails file
  thegawin
  committed
  3327cd8
- panfrost/model: Add Bifrost pixel/texel/FMA rates
  
  Chris Healy
  authored and
  Marge Bot
  committed
  67c350e
- pan/bi: Make bi_gather_stats arith and texel rates model-aware
  
  Chris Healy
  authored and
  Marge Bot
  committed
  9654df5
- radv: fix sync events for STAGE_2_COPY_INDIRECT_BIT_KHR
  
  hakzsam
  authored and
  Marge Bot
  committed
  e60c43e
- radv: only consider PS_PARTIAL_FLUSH for STAGE_2_BLIT_BIT
  
  hakzsam
  authored and
  Marge Bot
  committed
  ad08de4
- radv: use vk_expand_src_stage_flags2() in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  8eba113
- radv: remove redundant STAGE_2_VERTEX_INPUT in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  75daa13
- radv: remove redundant STAGE_2_BOTTOM_OF_PIPE in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  a1d5646
- radv: remove redundant STAGE_2_PRE_RASTERIZATION_SHADERS in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  157e119
- radv: cleanup STAGE_2_ALL_TRANSFER stages in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  6630663
- radv: remove redundant STAGE_2_ALL_{GRAPHICS,COMMANDS} for VS sync in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  8d2ed0f
- radv: remove redundant STAGE_2_ALL_COMMANDS for CS sync in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  55ee4bd
- radv: remove redundant STAGE_2_ALL_COMMANDS for PS sync in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  a4a0341
- radv: re-organize radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  3297a1e
- radv: ignore stage flags that are expanded for radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  e1832af
- radv: validate pipeline stage flags in radv_stage_flush()
  
  hakzsam
  authored and
  Marge Bot
  committed
  81479d1
- intel/tools: Enable surface and sampler decoding in aubinator_error_decode
  
  cmsanta
  authored and
  Marge Bot
  committed
  34b9d9a
- pan/perf: Don't count the blend prologue
  
  daivikbhatia
  authored and
  Marge Bot
  committed
  fd6d94b
- fdl: generalize UBWC block dimensions for all 2-plane YUV formats
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  f4eccfc
- freedreno/a6xx: allow YUV plane-format aliases in fd6_check_valid_format
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  d7528cb
- fdl/fd6_view: fix NV12 texture descriptor format and chroma offset for UBWC
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  797fdd5
- fdl: add fdl6_layout_multiplanar_image() for NV12-family plane stitching
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  2da769e
- freedreno/a6xx: add layout_multiplanar_resource screen hook
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  5cf96c1
- freedreno: allocate NV12 Y and UV planes into a single GEM BO
  
  Tupili Krishna Gowtham Reddy
  authored and
  Marge Bot
  committed
  9e1f8d4
- lavapipe: reset compute shader info in emit_compute_state()
  
  zmike
  authored and
  Marge Bot
  committed
  a02947f
- lavapipe: zero compute state for every mesh draw
  
  zmike
  authored and
  Marge Bot
  committed
  83bd6bc
- gfxstream: Fixes for extensions with type erased objects
  
  jmacnak
  authored and
  gurchetansingh
  committed
  fc6e859
- dri: bail out when missing swrast loader in drisw_init_screen()
  
  emersion
  authored and
  Marge Bot
  committed
  84bf3b5
- docs: add a link to gpu-ratemeter
  
  marekolsak
  authored and
  Marge Bot
  committed
  b65f0d1
- kraid/ops: Fix Display for OpF32ToI32
  Lorenzo Rossi
  committed
  4755cce
- kraid/dst_mod_prop: Fix wrong F32ToF16 dst data-type
  Lorenzo Rossi
  committed
  7d53db9
- kraid/nir: Fix miscompile for shift_and(x, shift, 0)
  Lorenzo Rossi
  committed
  9e70938
- kraid/opt_copy_prop: Fix hardcoded 4 bytes in swizzle read source
  Lorenzo Rossi
  committed
  db87f26
- kraid/opt_promote_consts: Respect inputs.promote_immediates
  Lorenzo Rossi
  committed
  6a036f1
- kraid/isa: Fix is_message not set for hardcoded messages
  Lorenzo Rossi
  committed
  ab56224
- nir/intrinsics: Rename preloaded register for v15 compatibility
  Lorenzo Rossi
  committed
  f90336b
- kraid: Clean up preloads
  Lorenzo Rossi
  committed
  25b5927
- kraid/ra: Clean up clippy complaints
  Lorenzo Rossi
  committed
  bfa915e
Commits on Sep 2, 2026
- tu: Fix gen8 TEX_LINE_OFFSET
  
  Rob Clark
  authored and
  Marge Bot
  committed
  25b5f40
- anv/rt: Skip scratch AABB init
  
  sagarghuge20
  authored and
  Marge Bot
  committed
  2684b3a
- intel/isl: Bump HALIGN from 16B to 32B on Xe2+
  
  nchery-intel
  authored and
  Marge Bot
  committed
  8f5b13d
- anv: make it possible to toggle 3DSTATE_RASTER::ForceMultisampling
  
  tpalli
  authored and
  Marge Bot
  committed
  d756b36
- amd: only enable the addrlib SIMD path when supported
  
  hakzsam
  authored and
  Marge Bot
  committed
  c6f13bc
- etnaviv: hwdb: Add PE_A8B8G8R8 feature
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  ff6f518
- etnaviv: Rename RB_SWAP PE format defines
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  17e5ead
- etnaviv: Update headers from rnndb
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  4de432d
- etnaviv: Handle native RGBA PE format
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  c9730af
- radv/sqtt: fix RGP capture cancel when the SQTT buffer hits its limit
  
  mythilam
  authored and
  Marge Bot
  committed
  1562915
- loader: consider radeon in amd_predicate
  
  peppsac
  authored and
  Marge Bot
  committed
  25e1859
- radeonsi: remove reachable assert
  
  peppsac
  authored and
  Marge Bot
  committed
  b4e3a67
- radeonsi: pass si_shader to si_get_ir_cache_key
  
  peppsac
  authored and
  Marge Bot
  committed
  ae3a8cd
- ac/info: disable userq support for SDMA
  
  peppsac
  authored and
  Marge Bot
  committed
  69688d2
- r600: Convert TGSI selectors to NIR earlier
  
  gerddie
  authored and
  Marge Bot
  committed
  4f63e65
- r600: Cache NIR selector info
  
  gerddie
  authored and
  Marge Bot
  committed
  6eddc80
- r600: Cache fragment color output count
  
  gerddie
  authored and
  Marge Bot
  committed
  0ca89d2
- r600: Cache viewport state on shader selectors
  
  gerddie
  authored and
  Marge Bot
  committed
  31fa503
- r600/sfn: Stop updating TGSI shader info
  
  gerddie
  authored and
  Marge Bot
  committed
  c20eaef
- r600: Cache LDS output masks from NIR
  
  gerddie
  authored and
  Marge Bot
  committed
  fa884a6
- r600: Drop NIR to TGSI shader info scan
  
  gerddie
  authored and
  Marge Bot
  committed
  3f1669c
- r600: Switch interpolation to glsl_interp_mode
  
  gerddie
  authored and
  Marge Bot
  committed
  f239600
- r600/sfn: get rid of useless indirect_file tracking
  
  gerddie
  authored and
  Marge Bot
  committed
  23ad4f9
- panfrost/csf: initialize flush_id to zero in the tiler OOM handler
  
  xianzhongli
  authored and
  Marge Bot
  committed
  5b3c6d8
- etnaviv: Compute resource render compatibility once
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  808c581
- radv: fix RADV_FORCE_VRS (Persona 3 Reload)
  
  marekolsak
  authored and
  Marge Bot
  committed
  80aca79
- kraid/ir: Fix indentation for phis
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  1790e3e
- kraid: Fix misc printing issues
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  d174115
- kraid: Add max_reg_count function to model
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  4f6eded
- kraid: Add MemoryEffect property to ops
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  6b0fc2c
- kraid: Helper for register range instead of bytes
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  86e4172
- kraid: Add scheduling barriers
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  c1d093d
- kraid: Implement improved message slot assignment
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  bdfa2b4
- kraid: Add serial debug flag
  
  arm-sinclair
  authored and
  Marge Bot
  committed
  21d401b
- pipe: Add reference to pipe_video_codec
  
  nowrep
  authored and
  Marge Bot
  committed
  16f638a
- va: Use refcounting for pipe_video_codec
  
  nowrep
  authored and
  Marge Bot
  committed
  bc7566d
- va: Keep codec ref in surface/buffer for fence wait/destroy
  
  nowrep
  authored and
  Marge Bot
  committed
  6984534
- va: Don't allow NULL codec in vaEndPicture
  
  nowrep
  authored and
  Marge Bot
  committed
  f72bdfa
- anv: fix 64bit values promoted to inline register
  llandwerlin-intel
  committed
  cc7d76d
- docs: update calendar for 26.2.2
  1ace
  committed
  bf2bc4c
- docs: add release notes for 26.2.2
  1ace
  committed
  3122d85
- docs: add sha sum for 26.2.2
  1ace
  committed
  6ad22f9
- freedreno/a5xx: implement dual-source blending
  
  lumag
  authored and
  Marge Bot
  committed
  4e2f81b
- nir: derive color_is_dual_source from declared vars
  
  shadeslayer
  authored and
  Marge Bot
  committed
  50d4bae
- ci/deqp: backport GLES2 CTS fixes
  
  ondracka
  authored and
  Marge Bot
  committed
  aade1a7
- docs: discourage AI-generated code comments, commit messages, gitlab comments
  
  marekolsak
  authored and
  Marge Bot
  committed
  96d992a
- etnaviv: hwdb: Add PE_32BPC_COLORMASK_FIX feature
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  36e19e6
- etnaviv: Widen 32bpc RT colormasks without PE_32BPC_COLORMASK_FIX
  
  Christian Gmeiner
  authored and
  Marge Bot
  committed
  d7f4354
Commits on Sep 3, 2026
- ci: store the output of vulkaninfo for all stages that use Vulkan
  
  hakzsam
  authored and
  Marge Bot
  committed
  b79af6f
- radv: refactor common pipeline stage masks
  
  hakzsam
  authored and
  Marge Bot
  committed
  d4f7667
- radv: remove unneccesary handling of transfer stages in write_event()
  
  hakzsam
  authored and
  Marge Bot
  committed
  4710a73
- radv: use radv_get_src_stage_flags2() in write_event()
  
  hakzsam
  authored and
  Marge Bot
  committed
  72a15c1
- radv: use radv_get_src_stage_flags2() for radv_cp_dma_wait_for_stages()
  
  hakzsam
  authored and
  Marge Bot
  committed
  050e270
- radv: remove redundant STAGE_2_xx bits when waiting for CP DMA
  
  hakzsam
  authored and
  Marge Bot
  committed
  74aba19
- radv: move CP DMA wait stages close to the other pipeline stage masks
  
  hakzsam
  authored and
  Marge Bot
  committed
  1b67785
- amd: fix GFX12 SPM GCEA_SE counter aliasing with GCEA_CPWD
  
  JoeWangAMD
  authored and
  Marge Bot
  committed
  b7e179e
- zink/ci: mark spec@arb_timer_query@timestamp-get as flakes for RADV
  
  hakzsam
  authored and
  Marge Bot
  committed
  427056f
- r300: don't propagate negation through absolute values
  
  ondracka
  authored and
  Marge Bot
  committed
  a16f6ce
- r300: emulate base-level NPOT cube sampling
  
  ondracka
  authored and
  Marge Bot
  committed
  cb4f7e1
- r300: split tiled NPOT 3D uploads by layer
  
  ondracka
  authored and
  Marge Bot
  committed
  5436402
- r300: lower NPOT wrap modes per coordinate
  
  ondracka
  authored and
  Marge Bot
  committed
  ec96518
- r300: clamp NPOT 3D coordinates to the logical edge
  
  ondracka
  authored and
  Marge Bot
  committed
  c36ff7b
- r300: clamp NPOT cube sampling to the logical face edge
  
  ondracka
  authored and
  Marge Bot
  committed
  e4f65a5
- r300: correct NPOT cube filter selection at the logical edge
  
  ondracka
  authored and
  Marge Bot
  committed
  c461b99
- pvr: add missing GPU architecture to device info for some BVNCs
  
  frankbinns
  authored and
  Marge Bot
  committed
  0adc6cc
- pvr: fixup ordering of (has_)usc_itr_parallel_instances in dev info
  
  frankbinns
  authored and
  Marge Bot
  committed
  41a2125
- pan/ci: document recent regression in xfails
  
  Erik Faye-Lund
  authored and
  Marge Bot
  committed
  c9e74db
- util/video: Account for MaxFS in h264 level derivation
  
  lolzballs
  authored and
  Marge Bot
  committed
  08de4d6
- kk: Add missing Vulkan stages to vkCmdWriteTimestamp() handling
  
  jeremyg-lunarg
  authored and
  Marge Bot
  committed
  caf453a
- kk: Add bridge functions for resolve texture
  
  K0bin
  authored and
  Marge Bot
  committed
  8b3743a
- kk: Do resolves using Metal render passes when possible
  
  K0bin
  authored and
  Marge Bot
  committed
  02767b6
- pvr: add BXM-4-64 (36.52.104.182) to list of conformant GPUs
  
  frankbinns
  authored and
  Marge Bot
  committed
  e82c68d
- nir: Add access flags to shared and global atomics
  
  cmarcelo
  authored and
  Marge Bot
  committed
  5149b96
- subprojects/rustix: enable the features an Apple backend needs
  
  utzcoz
  authored and
  Marge Bot
  committed
  db5c334
- subprojects: add a wrap for mach2
  
  utzcoz
  authored and
  Marge Bot
  committed
  e915c7f
- util: implement the sys layer for Darwin
  
  utzcoz
  authored and
  Marge Bot
  committed
  16ed392
- gfxstream/guest: stub the virtio-gpu backend on Apple
  
  utzcoz
  authored and
  Marge Bot
  committed
  aa96564
- tu: Stop writing two copies of the FDM viewport/scissor state.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  2f3ecdb
- tu: Fix handling of pImmutableSamplers in tu_GetDescriptorSetLayoutSupport().
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  97e2f7c
- tu: Free through the right allocator on vkCreateDescriptorSetLayout fail.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  4b5f174
- tu: Free the new set on descriptor set allocation fragmentation failure.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  0ff13bb
- tu: Properly distinguish pool fragmentation from pool exhaustion.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  8a542d9
- tu: Use memcpy instead of assuming opaque capture data pointers are aligned.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  226684a
- tu: Use STACK_ARRAY() for unbounded attachment_count-based allocations.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  198307a
- tu: Don't advertise sample count > 1 for FDM/FSR usage.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  891c9f8
- tu: Don't advertise buffer support for compressed formats.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  08a72d2
- tu: Don't advertise sparse residency when the pdev can't do it.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  163073f
- tu: Drop TRANSFER_SRC/DST from bufferFeatures.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  5da874c
- tu: Fix 3D subresourceLayout.size and memcpy size.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  58562e0
- tu: Fix the sType of the returned VkFragmentShadingRateKHRs.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  3799ded
- tu: Drop !filter_width assertion for 8xx.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  4417723
- tu: Fix leaks of a few device mutexes.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  53e4630
- tu: Fix destroy order of autotune vs perfcounters in failure path.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  b7844f8
- tu: Fix failure unwind ordering error for bin preamble.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  1c12bdb
- tu/kgsl: Fix double-close on load failure.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  0838bc1
- tu/kgsl: Don't leak the device/dma_fd on UBWC init failure.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  2d68ec2
- tu/msm: Fix leak of fd on tu_try_enable_vm_bind() failure.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  aece59e
- tu: Fix data race on memory_budget_props->heapUsage.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  8340c75
- tu: Don't try to use lazy allocation for MSRTSS if the device can't.
  
  Emma Anholt
  authored and
  Marge Bot
  committed
  0b0634a
- intel/compiler: Add atomic branch optimizations
  
  cmarcelo
  authored and
  Marge Bot
  committed
  084fece
- anv/intel: plumb atomic-branch control through drirc prog key
  
  jthiagar
  authored and
  Marge Bot
  committed
  8a0fa27
- anv: enable atomic-branch optimization for Mount And Blade
  
  jthiagar
  authored and
  Marge Bot
  committed
  6a31317
- kraid/ra: Don't unnecessarily over-write source byes
  
  gfxstrand
  authored and
  Marge Bot
  committed
  b770bb5
- kraid/ra: Add a helper for getting a source alignment
  
  gfxstrand
  authored and
  Marge Bot
  committed
  04e0387
- kraid: Add a SSAValue indexed vector struct
  
  gfxstrand
  authored and
  Marge Bot
  committed
  6d1cd5e
- kraid/ra: Use max_live.reg as the arena limit
  
  gfxstrand
  authored and
  Marge Bot
  committed
  2a1c945
- kraid/ra: Add an alloc_regs helper function
  
  gfxstrand
  authored and
  Marge Bot
  committed
  8dcfe91
- compiler/rust/enum_as_u8: Implement some ops for U8EnumSet
  
  gfxstrand
  authored and
  Marge Bot
  committed
  29b7391
- kraid/ra: Add an is_aligned_pinned_range() helper
  
  gfxstrand
  authored and
  Marge Bot
  committed
  c7d70b0
- kraid/ra: Rework widening and over-allocation
  
  gfxstrand
  authored and
  Marge Bot
  committed
  47138d1
- kraid/ra: Rework choose_*_bytes()
  
  gfxstrand
  authored and
  Marge Bot
  committed
  fbf5f46
- kraid/ra: Use round robin allocation for registers
  
  gfxstrand
  authored and
  Marge Bot
  committed
  d98d03e
- compiler/rust: Break the guts of UnionFind into a helper struct
  
  gfxstrand
  authored and
  Marge Bot
  committed
  47b7b26
- compiler/rust: Implement IntoIterator for UnionFind
  
  gfxstrand
  authored and
  Marge Bot
  committed
  c85e2d8
- kraid/ra: Try to place SSAValues in phi webs in the same register
  
  gfxstrand
  authored and
  Marge Bot
  committed
  84b091f
- kraid/ra: Try to pre-collect vectors
  
  gfxstrand
  authored and
  Marge Bot
  committed
  ac1aeef
- kraid/ra: Use a loop continue for special RA cases
  
  gfxstrand
  authored and
  Marge Bot
  committed
  05696cf
- kraid/ra: Try to coalesce MKVEC when we can
  
  gfxstrand
  authored and
  Marge Bot
  committed
  958e3d6
- freedreno/drm-shim: add Adreno 735 (FD735) for shader-db
  
  compilersutra
  authored and
  Marge Bot
  committed
  25e648a
- ac/nir: add ac_nir_lower_fs_input_loads for indirect FS input indexing
  
  marekolsak
  authored and
  Marge Bot
  committed
  cf5c6ec
- radv: scalarize IO even if optimizations are disabled
  
  marekolsak
  authored and
  Marge Bot
  committed
  d089ea5
- aco,radv,radeonsi: use ac_nir_lower_fs_input_prim_mask
  
  marekolsak
  authored and
  Marge Bot
  committed
  e0817b7
- radv: enable subgroup-convergent indirect FS input indexing
  
  marekolsak
  authored and
  Marge Bot
  committed
  395c00c
- radv: always allow RADV_FORCE_VRS on gfx11+
  
  marekolsak
  authored and
  Marge Bot
  committed
  02632c0
- radv: fix sRGB color clears via compute
  
  marekolsak
  authored and
  Marge Bot
  committed
  f23858d
- radv: add a drirc option to select the image clear/copy/blit/resolve path
  
  marekolsak
  authored and
  Marge Bot
  committed
  f952c7d
- kraid/hw_tests: Apply Clippy suggestions
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  02ea260
- kraid/hw_tests: Switch to builder pattern for args
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  6023a04
- kraid/hw_tests: Move read-write data buffer layout to Rust
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  bfc60f1
- kraid/hw_tests: Implement simple buffer descriptors
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  74d0245
- kraid/hw_tests: Fix 64-bit test data address support
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  67c6e19
- kraid/ir: Add RegRef::new utility
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  261d7a9
- kraid/hw_tests: Add infrastructure for low-level hardware tests
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  5a781ac
- kraid/hw_tests: Add LD_PKA hardware bug test
  
  Lorenzo Rossi
  authored and
  Marge Bot
  committed
  4eacb58
- brw: Refactor a big chunk of brw_opt_fill_and_spill to its own function
  
  ianromanick
  authored and
  Marge Bot
  committed
  58ce042
- brw: Try harder at fill-to-fill propagation
  
  ianromanick
  authored and
  Marge Bot
  committed
  87cbec6
- freedreno/registers: Name WAIT_CACHE_FLUSH
  
  Rob Clark
  authored and
  Marge Bot
  committed
  9735d4f
- freedreno/registers: Name ONE_REG_WR
  
  Rob Clark
  authored and
  Marge Bot
  committed
  172d435
- freedreno/registers: Drop bogus variants attr
  
  Rob Clark
  authored and
  Marge Bot
  committed
  2e9d766
- tu: Add tu_pkt7 builder
  
  Rob Clark
  authored and
  Marge Bot
  committed
  3691b85
- tu: Move reg_pair definition
  
  Rob Clark
  authored and
  Marge Bot
  committed
  9ece9d1
- tu: Add scratch-ram helper
  
  Rob Clark
  authored and
  Marge Bot
  committed
  8d3d4e2
- tu: Use scratch-ram helper for tu_dispatch()
  
  Rob Clark
  authored and
  Marge Bot
  committed
  546c67d
- tu: Add CP_REG_RMW helper
  
  Rob Clark
  authored and
  Marge Bot
  committed
  d677199
- tu: Add CP_REG_TO_MEM / CP_MEM_TO_REG helpers
  
  Rob Clark
  authored and
  Marge Bot
  committed
  aef4384
- tu: Add CP_COND_WRITE helper
  
  Rob Clark
  authored and
  Marge Bot
  committed
  194d334
- magma-gpu-rs: bump buffer size
  
  gurchetansingh
  authored and
  Marge Bot
  committed
  37bc0d5
Commits on Sep 19, 2026

Commits

Commits on Aug 31, 2026

Commits on Sep 1, 2026

Commits on Sep 2, 2026