4365 lines
193 KiB
ReStructuredText
4365 lines
193 KiB
ReStructuredText
Mesa 20.1.0 Release Notes / 2020-05-27
|
||
======================================
|
||
|
||
Mesa 20.1.0 is a new development release. People who are concerned with
|
||
stability and reliability should stick with a previous release or wait
|
||
for Mesa 20.1.1.
|
||
|
||
Mesa 20.1.0 implements the OpenGL 4.6 API, but the version reported by
|
||
glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) /
|
||
glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being
|
||
used. Some drivers don't support all the features required in OpenGL
|
||
4.6. OpenGL 4.6 is **only** available if requested at context creation.
|
||
Compatibility contexts may report a lower version depending on each
|
||
driver.
|
||
|
||
Mesa 20.1.0 implements the Vulkan 1.2 API, but the version reported by
|
||
the apiVersion property of the VkPhysicalDeviceProperties struct depends
|
||
on the particular driver being used.
|
||
|
||
SHA256 checksum
|
||
---------------
|
||
|
||
::
|
||
|
||
2109055d7660514fc4c1bcd861bcba9db00c026119ae222720111732dba27c83 mesa-20.1.0.tar.xz
|
||
|
||
New features
|
||
------------
|
||
|
||
- GL_ARB_compute_variable_group_size on i965.
|
||
- GL_EXT_depth_bounds_test on Iris.
|
||
- GL_EXT_texture_shadow_lod on radeonsi, nvc0.
|
||
- GL_NV_alpha_to_coverage_dither_control on radeonsi
|
||
- GL_NV_copy_image on all gallium drivers.
|
||
- GL_NV_pixel_buffer_object on all gallium drivers, i915, i965, swrast.
|
||
- GL_NV_viewport_array2 on nvc0 (GM200+).
|
||
- GL_NV_viewport_swizzle on nvc0 (GM200+).
|
||
- VK_AMD_memory_overallocation_behavior on RADV.
|
||
- VK_KHR_shader_non_semantic_info on Intel, RADV.
|
||
- GL_EXT_draw_instanced on gles2
|
||
- VK_KHR_8bit_storage for ACO on GFX8+
|
||
- VK_KHR_16bit_storage for ACO on GFX8+ (storageInputOutput16 is still
|
||
unsupported)
|
||
- shaderInt16 for ACO on GFX9+
|
||
- VK_KHR_shader_float16_int8 for ACO on GFX8+ (shaderFloat16 is still
|
||
unsupported)
|
||
- VK_EXT_robustness2 on Intel, RADV.
|
||
- Add Rocket Lake (RKL) support on anvil and iris.
|
||
|
||
Bug fixes
|
||
---------
|
||
|
||
- Reproduceable i915 gpu hang Intel Iris Plus Graphics (Ice Lake 8x8
|
||
GT2)
|
||
- glsl: regression affecting shader compilation time
|
||
- freedreno: glamor issue with x11 desktops
|
||
- [gles3] supertuxkart: some textures are incorrect
|
||
- Double lock in fbobject.c
|
||
- [bisected] Steam crashes when newest Iris built with LTO
|
||
- i965/vec4: opt_cse_local cause the out of bound array access
|
||
- NIR: Regression on shader using 8/16-bit integers
|
||
- lp_bld_intr.c:70:16: error: use of undeclared identifier
|
||
'LLVMFixedVectorTypeKind'; did you mean 'LLVMVectorTypeKind'?
|
||
- Deadlock in anv_timelines_wait()
|
||
- post_version.py does not work with release candidates
|
||
- post_version.py does not work with release candidates
|
||
- radv regression on android
|
||
- src\util\meson.build:294:4: ERROR: Program or command 'winepath' not
|
||
found or not executable
|
||
- debug builds are massively broken on Windows
|
||
- heavy glitches on amd ryzen 5 since version 20.x
|
||
- zink asserts with 32-bit boolean
|
||
- Dirt: Showdown bad performance and broken rendering with enabled
|
||
advanced lightning
|
||
- gravit & Firefox WebGL broken since
|
||
3dc2ccc14c0e035368fea6ae3cce8c481f3c4ad2 "ac/surface: replace
|
||
RADEON_SURF_OPTIMIZE_FOR_SPACE with !FORCE_SWIZZLE_MODE"
|
||
- mesa 20.0.5 causing kitty to crash
|
||
- radeonsi: "Torchlight II" trace showing regression on mesa-20.0.6
|
||
[bisected]
|
||
- [RADV/LLVM/ACO/Regression] After mesa commit
|
||
a3dc7fffbb7be0f1b2ac478b16d3acc5662dff66 all games stucks at start
|
||
- Android building error after commit 2ab45f41
|
||
- iris: Crash when trying to capture window in OBS Studio
|
||
- Properly annotate control flow convergence points
|
||
- intel/compiler: Register coalesce doesn't move conditional modifiers
|
||
- [bisected] [iris] mpv under wayland: failed to import supplied
|
||
dmabufs: Unsupported buffer format 808669784
|
||
- [Bisected][Iris] piglit.spec.!opengl 1_1.max-texture-size crashes on
|
||
x32 platform
|
||
- anv : android deqp assert
|
||
dEQP-VK.api.external.memory.android_hardware_buffer.dedicated.image#export_import_bind_bind
|
||
- GL cts gtf30.GL3Tests.sgis_texture_lod.sgis_texture_lod_basic_getter
|
||
failure
|
||
- freedreno/a6xx: texture cache vs realloc_bo()
|
||
- [Bisected]
|
||
dEQP-VK.subgroups.ballot_mask.ext_shader_subgroup_ballot.\* failures
|
||
- dEQP-VK.subgroups.size_control.compute.\* crashes on HSW and TGL
|
||
- zink: framebuffer and pipeline caches accumulate due to
|
||
zink_create_surface()
|
||
- FTBFS due to LLVM commit 2dea3f129878 (LLVMVectorTypeKind is gone)
|
||
- [r600/Turks] 20.0.2: modesetting/radeon driver SIGABRT at loading X
|
||
(kernel 5.5.10, ppc64)
|
||
- piglit spec.!opengl 1.0.gl-1.0-fpexceptions crash on Iris
|
||
- ci: Update the Wine version
|
||
- SPIR-V: Failure in dEQP-VK.graphicsfuzz.control-flow-switch
|
||
- SPIR-V: OpConvertUToPtr from spec constant fails to compile
|
||
- ACO: Regression: Texture corruption
|
||
- radv: Reading ViewportIndex in fragment shader returns garbage
|
||
- piglit
|
||
spec.arb_gpu_shader_fp64.execution.arb_gpu_shader_fp64-vs-non-uniform-control-flow-ssbo
|
||
crash on Iris
|
||
- piglit
|
||
spec/arb_gpu_shader_fp64/execution/built-in-functions/vs-sign-neg-abs.shader_test
|
||
failure on IVB
|
||
- [ANV] gfxbench Aztec Ruins misrenders on gen11+
|
||
- glxinfo cmd crashed
|
||
- radeonsi: GL_LINES rendering is affected by GL_POINT_SPRITE
|
||
- nir: nir_lower_returns can't handle nested loops
|
||
- Graphic artifacts with Mesa 20.0.4 on intel HD 510 GPU
|
||
- [Iris] [Bisected] Some KHR-GL46.arrays_of_arrays_gl. tests are
|
||
failing
|
||
- Mesa 20 regression makes Lightsprint demos crash
|
||
- metro redux games crash upon loading certain levels on amdgpu
|
||
- dri_common.h:58:8: error: unknown type name '__GLXDRIdrawable'
|
||
- Graphical glitches on Intel Graphics when Xorg started on Iris driver
|
||
- GL/GLES test crashes on G33/i915 platforms
|
||
- GL/GLES test crashes on G33/i915 platforms
|
||
- GL/GLES test crashes on G33/i915 platforms
|
||
- SIGSEGV src/compiler/glsl/ast_function.cpp:53
|
||
- manywin aborts with "i965: Failed to submit batchbuffer: Invalid
|
||
argument"
|
||
- manywin aborts with "i965: Failed to submit batchbuffer: Invalid
|
||
argument"
|
||
- manywin aborts with "i965: Failed to submit batchbuffer: Invalid
|
||
argument"
|
||
- manywin aborts with "i965: Failed to submit batchbuffer: Invalid
|
||
argument"
|
||
- v3d: transform feedback issue
|
||
- radv: Enable TC-compat HTILE in VK_IMAGE_LAYOUT_GENERAL.
|
||
- radv:
|
||
dEQP-VK.binding_model.descriptorset_random.sets4.noarray.ubolimitlow.sbolimitlow.imglimitlow.noiub.comp.noia.0
|
||
segfault
|
||
- radv: RAVEN fails
|
||
dEQP-VK.pipeline.timestamp.misc_tests.reset_query_before_copy
|
||
- buffer overflow in nouveau driver on mesa 20.0.2
|
||
- xmlconfig sha1 code has overflow and possible bug
|
||
- enable storageBuffer16BitAccess feature in radv for SI and CIK
|
||
- Build Fails with Clang Shared Library
|
||
- Thousands of 32 bit regressions in VulkanCTS and GL test suites due
|
||
to handling of cross-invocation
|
||
- anv: isl assert when running dEQP-VK.geometry.layered.3d.*.readback
|
||
- Weston drm-backend.so seems to fail with Mesa master and
|
||
LIBGL_ALWAYS_SOFTWARE=1
|
||
- freedreno/turnip: Don't request pixlodenable when we don't use it
|
||
- VulkanCTS uniform_buffer_block_geom spins forever
|
||
- freedreno: dEQP-GLES3.functional.fbo.msaa.4_samples.r16f flakiness in
|
||
CI
|
||
- src\util\meson.build:291:4: ERROR: Program or command 'winepath' not
|
||
found or not executable
|
||
- RADV: flickering textures in Q.U.B.E. 2 through Proton
|
||
- Missing ENDBR in entry_x86-64_tls.h, entry_x86_tls.h and
|
||
entry_x86_tsd.h
|
||
- [regression][bisected] Android build test fails:
|
||
marshal_generated.c', missing and no known rule to make it
|
||
- Missing ENDBR in rtasm_x86sse.c
|
||
- src/intel/tools/aubinator_viewer.cpp:383:52: error: format ‘%lx’
|
||
expects argument of type ‘long unsigned int’, but argument 5 has type
|
||
‘uint64_t {aka long long unsigned int}’ [-Werror=format=]
|
||
- src/compiler/glsl/ast_to_hir.cpp:2134: ir_rvalue\*
|
||
ast_expression::do_hir(exec_list*, \_mesa_glsl_parse_state*, bool):
|
||
Assertion \`result != NULL \|\| !needs_rvalue' failed.
|
||
- process_test fails on macOS
|
||
- Vulkan Overlay is blinking
|
||
- Regression: 9d64ad2fe79 broke Rocket League
|
||
- GameMaker games (Memoranda and Undertale) + amdgpu — Segmentation
|
||
fault on launch
|
||
- Civilization VI - Animated leader characters small black squares
|
||
artifacts
|
||
- [ACO] Reliable crash with RPCS3 that is not present with LLVM
|
||
- [RADV] vkCmdBindTransformFeedbackBuffersEXT pSizes optional parameter
|
||
not handled
|
||
- [RadeonSI] - Curse of the Dead Gods (1123770) - Lighting is not
|
||
rendering correctly.
|
||
- soft-fp64: \__fsat64 incorrectly returns NaN for a NaN input. It
|
||
should return zero.
|
||
- Hang when using glWaitSync with multithreaded shared GL contexts
|
||
- RPCS3 / Persona 5 - Performance regression [RADV / Navi]
|
||
- [ANV] Rendering corruption in Shadow of the Tomb Raider
|
||
- src/compiler/glsl/glcpp/glcpp-parse.y:1297: \_token_print: Assertion
|
||
\`!"Error: Don't know how to print token."' failed.
|
||
- [CTS] dEQP-VK.descriptor_indexing.\* fails on RADV/LLVM
|
||
- Unigine Valley failure / assert
|
||
- [Gen9/icl] [Bisected] [Regression]
|
||
dEQP-GLES3.functional.shaders.loops.short_circuit.do_while_fragment
|
||
fail
|
||
- [RadeonSI][gfx10/navi] Kerbal Space Program crash: si_draw_vbo:
|
||
Assertion \`0' failed
|
||
- Budget Cuts hits VK_AMD_shader_fragment_mask assert
|
||
- Follow-up from "i965/blorp: Don't resolve HiZ unless we're
|
||
reinterpreting"
|
||
- crash in vc4_write_uniforms with shaders involving YUV textures
|
||
- Corrupted output with vaapi 10 bit -> 8 bit transcoding on AMD RAVEN
|
||
- tessellator.cpp:78:7: error: 'fmin' is missing exception
|
||
specification 'noexcept'
|
||
- Please add Raspberry Pi 4 to features.txt
|
||
- Build failure with bison 2.3.
|
||
- Mesa build fails on 32 bit architecture
|
||
- Mesa build fails on 32 bit architecture
|
||
- Incorrect rendering with vaapi + uyvy422
|
||
- V3D/Broadcom (Raspberry Pi 4) - GLES 3.1 - GL_EXT_texture_norm16
|
||
advertised, but not usable
|
||
- mesa-20.0.0/src/amd/compiler/aco_instruction_selection.cpp:7221:55:
|
||
style: Same expression on both sides of '&&
|
||
- i965 assertion failure in fallback_rgbx_to_rgba
|
||
- vaapi bob deinterlacer produces wrong output height on AMD
|
||
- Compute copies do not handle SUBSAMPLED formats
|
||
- Please document RADV_TEX_ANISO variable in envvars.html
|
||
- unexpected CI failure
|
||
- Multiple glapi_mapi_tmp.h
|
||
- drisw crashes on calling NULL putImage on EGL surfaceless platform
|
||
(pbuffer EGLSurface)
|
||
- VRAM leak with vuilkan external memory + opengl memory objects
|
||
- [radeonsi][vaapi][bisected] invalid VASurfaceID when playing
|
||
interlaced DVB stream in Kodi
|
||
- [RADV] GPU hangs while the cutscene plays in the game Assassin's
|
||
Creed Origins
|
||
- ACO: The Elder Scrolls Online crashes on startup (Navi)
|
||
- Broken rendering of glxgears on S/390 architecture (64bit, BigEndian)
|
||
- aco: sun flickering with Assassins Creeds Origins
|
||
- !1896 broke ext_image_dma_buf_import piglit tests with radeonsi
|
||
- aco: wrong geometry with Assassins Creed Origins on GFX6
|
||
- valgrind errors since commit a8ec4082a41
|
||
- src/broadcom/qpu/qpu_pack.c:962:25: error: implicit declaration of
|
||
function 'ffs' is invalid in C99
|
||
[-Werror,-Wimplicit-function-declaration] mux_b =
|
||
ffs(desc->mux_b_mask) - 1;
|
||
- X fails to start with amdgpu and Mesa 20.1 on Fedora
|
||
- GPU hangs in Factorio on Radeon RX 5700 XT (MSI GAMING X)
|
||
- OSMesa osmesa_choose_format returns a format not supported by
|
||
st_new_renderbuffer_fb
|
||
- Build error with VS on WIN
|
||
- Using EGL_KHR_surfaceless_context causes spurious "libEGL warning:
|
||
FIXME: egl/x11 doesn't support front buffer rendering."
|
||
- !3460 broke texsubimage test with piglit on zink+anv
|
||
- VERSION needs to be bumped for trunk master
|
||
- The screen is black when using ACO
|
||
|
||
Changes
|
||
-------
|
||
|
||
Abhishek Kumar (1):
|
||
|
||
- anv/android: fix assert in anv_import_ahw_memory
|
||
|
||
Adam Jackson (1):
|
||
|
||
- gallium: enable EGL_EXT_image_dma_buf_import_modifiers
|
||
unconditionally
|
||
|
||
Albert Astals Cid (5):
|
||
|
||
- cube_face_coord: Use fabsf instead of fabs since we know it's floats
|
||
- cube_face_index: Use fabsf instead of fabs since we know it's floats
|
||
- aco: Minor optimization in spill_ctx constructor
|
||
- aco: pass vars by const &
|
||
- Fix promotion of floats to doubles
|
||
|
||
Alejandro Piñeiro (7):
|
||
|
||
- docs/features: add v3d driver
|
||
- nir/linker: remove reference to just SPIR-V linking
|
||
- v3d/tex: don't configure tmu config 1 if not needed
|
||
- v3d/tex: Configuration Parameter 1 can be only skipped if P2 can be
|
||
skipped too
|
||
- v3d/packet: fixing TMU_Config_Parameter_2 definition
|
||
- nir: add nir_tex_instr_need_sampler helper
|
||
- v3d: support for textureQueryLOD
|
||
|
||
Alexandros Frantzis (3):
|
||
|
||
- gitlab-ci: Automated testing with OpenGL traces
|
||
- gitlab-ci: Fix traces caching in tracie
|
||
- gitlab-ci: Check the Mesa version used for tracie tests
|
||
|
||
Alyssa Rosenzweig (505):
|
||
|
||
- pan/midgard: Break out one-src read_components
|
||
- pan/midgard: Implement mixed-type constant packing
|
||
- panfrost: Avoid overlapping copy
|
||
- pan/midgard: Check for null consts
|
||
- pan/midgard: Remove unused variable
|
||
- panfrost: Use size0 when calculating the offset to a depth level
|
||
- pan/midgard: Fix scheduling issue with csel + render target reference
|
||
- panfrost: Simplify swizzle translation
|
||
- panfrost: Update comment about magic number relating to barriers
|
||
- panfrost: Ensure compute shader_meta is zeroed
|
||
- panfrost: Identify mali_shared_memory structure
|
||
- panfrost: Unify bifrost_scratchpad with mali_shared_memory
|
||
- panfrost: Rename bifrost_framebuffer->mali_framebuffer
|
||
- panfrost: Rename unknown2_8 to padding
|
||
- panfrost: Allocate RAM backing of shared memory
|
||
- pan/midgard: Track pressure when scheduling ld/st
|
||
- pan/midgard: Fix missing prefixes
|
||
- pan/midgard: Fix swizzles harder
|
||
- pan/midgard: Implement barriers
|
||
- pan/midgard: Allow jumping out of a shader
|
||
- pan/midgard: Fix 32/64 mixed swizzle packing
|
||
- pan/midgard: Use dummy tag for empty shaders
|
||
- pan/midgard: Improve barrier disassembly
|
||
- pan/midgard: Overhaul tag handling
|
||
- pan/midgard: Imply next tags
|
||
- pan/midgard: Infer tags entirely
|
||
- pan/midgard: Set xyzx swizzle for load_compute_arg
|
||
- pan/midgard: Identify stack barrier flag
|
||
- pan/midgard: Don't crash with constants on unknown ops
|
||
- pan/midgard: Use fprintf instead of printf for constants
|
||
- pan/decode: Remove extraneous newline
|
||
- pan/decode: Add \`minimal\` mode
|
||
- pan/decode: Cleanup pandecode_jc
|
||
- panfrost: Implement PAN_DBG_SYNC with pandecode/minimal
|
||
- panfrost: Print synced traces to stderr
|
||
- panfrost: Rewrite scoreboarding routines
|
||
- panfrost: Update scoreboarding notes
|
||
- panfrost: Cleanup transfer_map
|
||
- panfrost: Avoid reading GPU memory when packing vertices
|
||
- panfrost: Debitfieldize mali_uniform_buffer_meta
|
||
- panfrost: Remove enum panfrost_memory_layout
|
||
- panfrost: Remove dirty tracking
|
||
- panfrost: Remove old comment
|
||
- panfrost: Remove old hack
|
||
- panfrost: Remove flush_frontbuffer
|
||
- pan/midgard: Identify clamp(x, -1.0, 1.0) flag
|
||
- panfrost: Move checksum routines to root panfrost
|
||
- panfrost: Move pan_afbc.c to root
|
||
- panfrost: Move format translation to root
|
||
- panfrost: Rewrite texture descriptor creation logic
|
||
- nir: Add SSBO->global lowering pass
|
||
- pan/midgard: Lower SSBOs in NIR
|
||
- pan/midgard: Implement nir_intrinsic_get_buffer_size
|
||
- pan/midgard: Implement load/store_shared
|
||
- panfrost: Combine get_index_buffer with bound computation
|
||
- panfrost: Implement index buffer cache
|
||
- pan/decode: Dump scratchpad size if present
|
||
- pan/midgard: Don't spill near a branch
|
||
- panfrost: Fix gl_VertexID/InstanceID
|
||
- panfrost: Fix padded_vertex_count generation
|
||
- panfrost: Update spilling comment framebuffer->shared
|
||
- panfrost: Don't set shared->unk0
|
||
- panfrost: Fix param getting
|
||
- panfrost: Default to 256 threads for TLS
|
||
- panfrost: Reserve an extra page for spilling
|
||
- panfrost: Simplify stack shift calculation
|
||
- panfrost: Expose PIPE_CAP_PRIMITIVE_RESTART
|
||
- panfrost: Add PAN_MESA_DEBUG=gles3 option
|
||
- panfrost: Increase SSBO/image limit from 4->8
|
||
- pan/midgard: Allow inverted inverted ops
|
||
- pan/midgard: Allow fusing inverted sources for inverted ops
|
||
- pan/midgard: Partially fix 64-bit swizzle alignment
|
||
- pan/midgard: Extract nir_ssa_index helper
|
||
- pan/midgard: Add LDST_ADDRESS property
|
||
- pan/midgard: Fix load/store argument sizing
|
||
- pan/midgard: Round up bytemasks when promoting uniforms
|
||
- pan/midgard: Force address alignment
|
||
- pan/midgard: Add address analysis framework
|
||
- pan/midgard: Use address analysis for globals, etc
|
||
- pan/decode: Calm an assert to a pandecode error
|
||
- pan/decode: Restore bifrost sample_locations
|
||
- pan/decode: Fix tiler weights printing
|
||
- pan/decode: Skip analysis for Bifrost tiler structures
|
||
- pan/bi: Add discard ops
|
||
- pan/bi: Add ICMP.GL.NEQ op
|
||
- pan/bi: Move notes on FMA opcodes from disassembler
|
||
- pan/bi: Introduce CSEL4 class
|
||
- pan/bi: Move notes on ADD ops to notes file
|
||
- pan/bi: Decode FMA_SHIFT properly
|
||
- pan/bi: Add v4i8 mode to FMA_SHIFT
|
||
- pan/bi: Identify extended FMA opcodes
|
||
- pan/bi: Decode ADD_SHIFT properly
|
||
- pan/bi: Combine LOAD_VARYING_ADDRESS instructions by type
|
||
- pan/bi: Squash LD_ATTR ops together
|
||
- pan/bi: Structify FMA_FADD
|
||
- pan/bi: Move some definitions from disasm to bifrost.h
|
||
- panfrost: Add note about preloaded varyings
|
||
- pan/bi: Gut old compiler
|
||
- pan/bi: Stub out new compiler
|
||
- pan/bi: Add the control flow graph
|
||
- pan/bi: Add src/dest fields to bifrost_instruction
|
||
- pan/bi: Add class properties
|
||
- pan/bi: Add modifiers to bi_instruction
|
||
- pan/bi: Add BI_GENERIC property
|
||
- pan/bi: Factor out enum bifrost_minmax_mode
|
||
- pan/bi: Add a bifrost_roundmode field
|
||
- pan/bi: Add bifrost_minmax_mode field
|
||
- pan/bi: Add bi_load structure
|
||
- pan/bi: Pull out bifrost_load_var
|
||
- pan/bi: Add bi_load_vary structure
|
||
- pan/bi: Add PAN_SCHED\_\* flags
|
||
- pan/bi: Add bi_clause, bi_bundle abstractions
|
||
- pan/bi: Add dest_type field to bifrost_instruction
|
||
- pan/bi: Add special indices
|
||
- pan/bi: Add constant field to bi_instruction
|
||
- pan/bi: Add class-specific ops
|
||
- pan/bi: Add clause header fields to bi_clause
|
||
- pan/bi: Clarify special op scheduling
|
||
- pan/bi: Add swizzles
|
||
- pan/bi: Add source type for conversions
|
||
- pan/bi: Add EXTRACT, MAKE_VEC synthetic ops
|
||
- pan/bi: Add constants to bi_clause
|
||
- pan/bi: Add pred/successors to build CFG
|
||
- pan/bi: Extract bifrost_branch structure
|
||
- pan/bi: Add bi_branch data
|
||
- pan/bi: Add CSEL condition
|
||
- pan/bi: Add high-latency property for classes
|
||
- pan/bi: Add quirks system
|
||
- pan/bi: Add IR iteration macros
|
||
- pan/bi: Move some print routines out of the disasm
|
||
- pan/bi: Add BIR manipulation routines to bir.c
|
||
- pan/bi: Move bi_interp_mode_name to bi_print
|
||
- pan/bi: Add bi_instruction printing
|
||
- pan/bi: Add bi_print_bundle for printing bi_bundle
|
||
- pan/bi: Add bi_print_clause
|
||
- pan/bi: Add bi_print_block
|
||
- pan/bi: Add bi_print_shader
|
||
- pan/bi: Lower and optimize NIR
|
||
- pan/bi: Walk through the NIR control flow graph
|
||
- pan/bi: Improve block printing
|
||
- pan/bi: Don't print types for unconditional branches
|
||
- pan/bi: Print branch target
|
||
- pan/bi: Add instruction emit/remove helpers
|
||
- pan/bi: Call nir_lower_io_to_temporaries in cmdline
|
||
- pan/bi: Add support for if-else blocks
|
||
- pan/bi: Handle loops when ingesting CFG
|
||
- pan/bi: Handle jumps (breaks, continues)
|
||
- pan/bi: Fix destination printing
|
||
- pan/bi: Implement nir_intrsinic_load_interpolated_input
|
||
- pan/bi: Add blend_location to IR for BI_BLEND
|
||
- pan/bi: Add bi_schedule_barrier helper
|
||
- pan/bi: Implement store_output for fragment shaders
|
||
- pan/bi: Implement load_input for vertex shaders
|
||
- pan/bi: Add helpers for creating temporaries
|
||
- pan/bi: Implement store_vary for vertex shaders
|
||
- pan/bi: Add preliminary LOAD_UNIFORM implementation
|
||
- pan/bi: Implement load_const
|
||
- pan/bi: Add dummy scheduler
|
||
- pan/bi: Rename next-wait to simply 'wait'
|
||
- pan/bi: Fix Android.mk
|
||
- panfrost: Move mir_to_bytemask to common code
|
||
- pan/bi: Generalize swizzles to avoid extracts
|
||
- pan/bi: Introduce writemasks
|
||
- pan/bi: Remove bi_load
|
||
- pan/bi: Lower vec\* to writemasks in NIR
|
||
- pan/bi: Add initial handling of ALU ops
|
||
- pan/bi: Allow inlining constants
|
||
- pan/bi: Implement fsat as mov.sat
|
||
- pan/bi: Add a bunch of ALU ops
|
||
- pan/bi: Add BI_SPECIAL\_\* enum
|
||
- pan/bi: Handle special ops in NIR->BIR
|
||
- pan/bi: Implement fabs, fneg as fmov with mods
|
||
- pan/bi: Disable lower_sub
|
||
- pan/bi: Add isub op
|
||
- pan/bi: Import algebraic pass from midgard
|
||
- pan/bi: Implement nir_op_bcsel
|
||
- pan/bi: Lower b2f to bcsel
|
||
- pan/bi: Specify comparison op for BI_CMP
|
||
- pan/bi: Print source types unconditionally
|
||
- pan/bi: Implement comparison opcodes via BI_CMP
|
||
- panfrost: Promote midgard_program to panfrost/util
|
||
- pan/midgard: Remove unused iterators
|
||
- pan/midgard: Adjust sysval-related prototypes
|
||
- pan/midgard: Remove indexing dependency of sysvals
|
||
- pan/midgard: Decontextualize midgard_nir_assign_sysval_body
|
||
- pan/midgard: Remove dest_override sysval argument
|
||
- panfrost: Move Midgard sysval code to common Panfrost
|
||
- pan/bi: Switch to panfrost_program
|
||
- pan/bi: Implement sysvals
|
||
- pan/midgard: Localize \`visited\` tracking
|
||
- pan/midgard: Decontextualize liveness analysis core
|
||
- pan/midgard: Sync midgard_block field names with Bifrost
|
||
- pan/midgard: Subclass midgard_block from pan_block
|
||
- panfrost: Move liveness analysis to root panfrost/
|
||
- panfrost: Sync Midgard/Bifrost control flow
|
||
- pan/bi: Paste over bi_has_arg
|
||
- pan/bi: Add bi_bytemask_of_read_components helpers
|
||
- pan/bi: Add bi_next/prev_op helpers
|
||
- pan/bi: Add bi_max_temp helper
|
||
- pan/bi: Add liveness analysis pass
|
||
- pan/bi: Add dead code elimination pass
|
||
- pan/bi: Implement nir_op_ffma
|
||
- pan/bi: Fix swizzle for second argument to ST_VARY
|
||
- panfrost: Move lcra to panfrost/util
|
||
- pan/midgard: Remove incorrect comment in RA
|
||
- pan/bi: Minor fixes in iteration macros
|
||
- pan/bi: Fix vector handling of readmasks
|
||
- pan/bi: Fix missing src_types
|
||
- pan/bi: Add register allocator
|
||
- pan/bi: Interpret register allocation results
|
||
- pan/bi: Setup initial clause packing
|
||
- pan/bi: Sketch out instruction word packing
|
||
- pan/bi: Add packing for register control field
|
||
- pan/bi: Pack register fields
|
||
- pan/bi: Add missing \__attribute__((packed))
|
||
- pan/bi: Assign registers to ports
|
||
- pan/bi: Route through first_instruction field
|
||
- pan/bi: Model 3-bit Bifrost srcs in IR
|
||
- pan/bi: Add struct bifrost_fma_fma
|
||
- pan/bi: Pack BI_FMA ops
|
||
- pan/bi: Pack fadd32
|
||
- pan/bi: List ADD classes in bi_pack_add
|
||
- pan/bi: Generalize bi_get_src a bit
|
||
- pan/bi: Pass second src for load_vary ops
|
||
- pan/bi: Emit load_vary ops
|
||
- pan/bi: Skip over data registers in port assignment
|
||
- pan/bi: Route through clause header
|
||
- pan/bi: Pretty-print clause types in disassembler
|
||
- pan/bi: Don't hide SCHED_ADD inside HI_LATENCY
|
||
- pan/bi: Track clause types during scheduling
|
||
- pan/bi: Flesh out ATEST in IR
|
||
- pan/bi: Add ATEST packing
|
||
- pan/bi: Flesh out BI_BLEND
|
||
- pan/bi: Pack BI_BLEND
|
||
- pan/bi: Implement FMA/MOV without modifiers
|
||
- pan/bi: Add bi_emit_before helper
|
||
- pan/bi: Add move lowering pass
|
||
- pan/bi: Pack a constant quadword
|
||
- pan/bi: Document constant related errata(?)
|
||
- pan/bi: Index out constants in instructions
|
||
- pan/bi: Include UBO index for sysval reads
|
||
- pan/bi: Add bi_load32_components helper
|
||
- pan/bi: Pack ld_ubo ops
|
||
- pan/bi: Pack ld_var_addr
|
||
- pan/bi: Flesh out st_vary IR
|
||
- pan/bi: Generalize data register setting
|
||
- pan/bi: Add store_channels property
|
||
- pan/bi: Pack st_vary
|
||
- pan/bi: Pack LD_ATTR
|
||
- pan/bi: Lower bool to ints
|
||
- pan/bi: Remove hacks for 1-bit booleans in IR
|
||
- pan/bi: Add \`soft\` NIR->BIR condition translation
|
||
- pan/bi: Implement csel fusing
|
||
- pan/bi: Respect shift when printing immediates
|
||
- pan/bi: Use bi_lookup_immediate when packing
|
||
- pan/bi: Default csel to "!= 0" mode
|
||
- pan/bi: Pack csel4 opcodes
|
||
- pan/bi: Ingest vecN directly (again)
|
||
- pan/bi: Lower combines to rewrites for scalars
|
||
- pan/bi: Rewrite aligned vectors as well
|
||
- panfrost: Split panfrost_device from panfrost_screen
|
||
- panfrost: Isolate panfrost_bo_access_for_stage to pan_cmdstream.c
|
||
- panfrost: Inline reference counting routines
|
||
- panfrost: Move pan_bo to root panfrost
|
||
- pan/bit: Link standalone compiler with en/decoder
|
||
- panfrost: Move device open/close to root panfrost
|
||
- pan/bit: Open up the device
|
||
- panfrost: Stub out G31/G52 quirks
|
||
- pan/bit: Submit a WRITE_VALUE job as a sanity check
|
||
- pan/bit: Begin generating a vertex job
|
||
- pan/bi: Fix overzealous write barriers
|
||
- pan/bi: Fix off-by-one in scoreboarding packing
|
||
- pan/bi: Enable precision lowering in standalone compiler
|
||
- panfrost: Enable PIPE_SHADER_CAP_FP16 on Bifrost
|
||
- pan/bi: Handle f2f\* opcodes
|
||
- pan/bi: Ignore swizzle in unwritten component
|
||
- pan/bi: Finish FMA structures
|
||
- pan/bi: Fix missing type for fmul
|
||
- pan/bi: Add FMA16 packing
|
||
- pan/bi: Pack outmod and roundmode with FMA
|
||
- pan/bi: Expand out FMA conversion opcodes
|
||
- pan/bi: Enumerate conversions
|
||
- pan/bi: Handle standard FMA conversions
|
||
- pan/bi: Add bifrost_fma_2src generic
|
||
- pan/bi: Add one-source f32->f16 op
|
||
- pan/bi: Assert out i16 related converts for now
|
||
- pan/bi: Handle round opcodes in frontend
|
||
- pan/bi: Add v2f16 versions of rounding ops
|
||
- pan/bi: Structify fadd/min/max16
|
||
- pan/bi: Handle core faddminmax16 packing
|
||
- pan/bi: Handle abs packing for fp16/FMA add/min
|
||
- pan/bi: Handle fp16/abs scheduling restriction
|
||
- pan/bi: Fix handling of constants with COMBINE
|
||
- pan/bit: Add \`run\` mode to the cmdline
|
||
- pan/bit: Wire through I/O
|
||
- pan/bi: Fix writes_component for VECTOR
|
||
- pan/bi: Use STAGE srcs for scheduler nops
|
||
- pan/bi: Don't set the back-to-back bit yet
|
||
- pan/bi: Add cmdline option for verbose disassembly
|
||
- pan/bi: Fix unused port swapping
|
||
- pan/bi: Handle fmov class ops
|
||
- pan/bi: Fix outmod/roundmode flip
|
||
- pan/bi: Export bi_class_name
|
||
- pan/bi: Fix duplicated source in ADD.v2f16
|
||
- pan/bi: Fix negation in ADD.v2f16
|
||
- pan/bi: Don't gobble zero ports
|
||
- pan/bi: Allow BI_FMA to take mods
|
||
- pan/bi: Handle BIFROST_FIRST_WRITE_FMA_P2_READ_P3
|
||
- pan/bi: Add helper to debug port assignment
|
||
- pan/bi: Match CSEL argument order with hw
|
||
- pan/bit: Stub out BIR interpreter
|
||
- pan/bit: Handle read/write
|
||
- pan/bit: Add preliminary FMA/ADD/MOV implementations
|
||
- pan/bit: Implement outmods
|
||
- pan/bit: Implement floating source mods
|
||
- pan/bit: Add packing test framework
|
||
- pan/bit: Add helper for generating floating mod tests
|
||
- pan/bit: Add verbose printing for tests
|
||
- pan/bit: Add 16-bit fmod tests
|
||
- pan/bit: Add FMA tests
|
||
- pan/bit: Add CSEL to interpreter
|
||
- pan/bit: Add csel tests
|
||
- pan/bit: Make run more useful
|
||
- pan/bit: Add mode to run unit tests
|
||
- pan/bi: Remove nontrivial SPECIAL ops
|
||
- pan/bi: Add 32-bit \_FAST packing
|
||
- pan/bi: Add fp16 support for frcp/frsq
|
||
- pan/bit: Add special op interpreting
|
||
- pan/bit: Add special unit test
|
||
- pan/bi: Implement min/max on FMA
|
||
- pan/bi: Structify ADD unit add/min/max
|
||
- pan/bi: Add ADD add/min/max fp32 packing
|
||
- pan/bi: Set BI_MODS for MINMAX
|
||
- pan/bi: Fix incorrect abs flip in fma/fadd16
|
||
- pan/bi: Force ADD scheduling for MINMAX
|
||
- pan/bit: Unify test frontends
|
||
- pan/bit: Add min/max support to interpreter
|
||
- pan/bit: Enable more debug for \`run\`
|
||
- pan/bit: Add fmin/max16 tests
|
||
- pan/bit: Wire up add/add op+test
|
||
- panfrost: Add IS_BIFROST quirk
|
||
- panfrost: Populate bifrost-specific structs within mali_shader_meta
|
||
- panfrost: Staticize a few cmdstream functions
|
||
- panfrost: Unify vertex/tiler structures
|
||
- panfrost: Set mfbd.msaa.sample_locations on Bifrost
|
||
- panfrost: Call the Bifrost compiler on bi devices
|
||
- pan/bi: Fix nondeterministic register packing
|
||
- pan/midgard: Remove unused max_varying variable
|
||
- panfrost: Move varying linking to cmdstream
|
||
- panfrost: Move uniform_count to pan_assemble
|
||
- panfrost: Pass compiler-appropriate options
|
||
- pan/bi: Fix backwards registers ports
|
||
- panfrost: Fix BI_BLEND packing
|
||
- pan/bi: Let !b2b imply branch_cond
|
||
- pan/decode: Print Bifrost blend descriptor
|
||
- panfrost: Drop dependency on nonexistant write_value
|
||
- pan/bi: Lower fsqrt
|
||
- pan/midgard: Fix f2u naming confusion
|
||
- pan/bi: Set BI_ROUNDMODE for BI_CONVERT
|
||
- pan/bi: Fix incorrect swizzle packing assert
|
||
- pan/bi: Rewrite conversion packing
|
||
- pan/bi: ADD packing for CONVERT
|
||
- pan/bit: Add BI_CONVERT interpretation
|
||
- pan/bit: Add BI_CONVERT tests
|
||
- pan/bi: Add disasm for ADD.i8
|
||
- pan/bi: Disable FMA scheduling for CONVERT
|
||
- pan/bi: Add BI_TABLE for fast table accesses
|
||
- pan/bi: Add special op for exp2
|
||
- pan/bi: Add op for ADD_FREXPM
|
||
- pan/bi: Add FLOG2_U op to disassembler
|
||
- pan/bi: Add log_frexpe op to IR
|
||
- pan/bi: Add frexp_log packing
|
||
- pan/bi: Add bi_pack_fma_2src helper
|
||
- pan/bi: Pack ADD_FREXPM
|
||
- pan/bi: Add log2_help packing
|
||
- pan/bi: Add \_MSCALE flag for FMA/ADD
|
||
- pan/bi: Structify FMA_MSCALE
|
||
- pan/bi: Pack FMA_MSCALE
|
||
- pan/bi: Add fexp2_fast packing
|
||
- pan/bi: Split src/dest index printing
|
||
- pan/bi: Ensure CONSTANT srcs have types
|
||
- pan/bi: Fix bi_get_immediate with multiple imms
|
||
- pan/bi: Fix packing with multiple constants
|
||
- pan/bi: Fix packing with low-nibble-set on hi constant
|
||
- pan/bi: Fix lower_combine swizzle rewrite
|
||
- pan/bi: Add fexp2 implementation
|
||
- pan/bi: Implement flog2
|
||
- pan/bi: Fix vec2/3 handling
|
||
- pan/bi: Handle st_vary with <4 components
|
||
- pan/bi: Try to reuse constants in ALU
|
||
- pan/bi: Workaround constant packing errata
|
||
- pan/bi: Structify add and min/max fp16 ADD
|
||
- pan/bi: Pack ADD.v2f16
|
||
- pan/bi: Pack MAX.v2f16
|
||
- pan/bi: Dump extra bits for disasm
|
||
- pan/bi: Round constants to 32-bit
|
||
- pan/bi: Lower special ops to 32-bit
|
||
- pan/bit: Add FREXP interp support
|
||
- pan/bit: Add frexp_log test
|
||
- pan/bit: Add BI_REDUCE_FMA interp
|
||
- pan/bit: Add FMA_REDUCE test
|
||
- pan/bit: Add log2 helper interp
|
||
- pan/bit: Add BI_TABLE test
|
||
- pan/bit: \_MSCALE interp
|
||
- pan/bit: Add FMA_MSCALE test
|
||
- pan/bit: Add fexp2_fast interp
|
||
- pan/bit: Add fexp2_fast test
|
||
- pan/bit: Add constants test
|
||
- pan/bit: Add fp16 min/max tests
|
||
- pan/bi: Print tex_compact coordinates
|
||
- pan/bi: Document when dual-tex is triggered
|
||
- pan/bi: Disassemble f16 dual tex
|
||
- pan/bi: Structify TEX compact
|
||
- pan/bi: Include TEX_COMPACT f16 opcode
|
||
- pan/bi: Feed data register to BI_TEX
|
||
- pan/bi: Add normal/compact/dual switch to IR
|
||
- pan/bi: Stub out tex_compact logic
|
||
- pan/bi: Generate TEX_COMPACT instruction
|
||
- pan/bi: Pack TEX compact instructions
|
||
- pan/bi: Assert out multiple textures
|
||
- panfrost: Fix crashes with small BOs
|
||
- panfrost: Assert on unimplemented fragcoord etc
|
||
- panfrost: Set clear_color_[12] in the extra fb desc
|
||
- panfrost: Add tentative bifrost_texture_descriptor
|
||
- panfrost: decode textures and samplers on bifrost
|
||
- pan/decode: Remove is_zs weirdness
|
||
- panfrost: Identify texture layout field
|
||
- panfrost: The texture descriptor has a pointer to a trampoline
|
||
- pan/bi: Pack fp16 ATEST
|
||
- pan/bi: Passthrough type for ATEST
|
||
- pan/bi: Passthrough blend types
|
||
- pan/bi: Assign blend descriptor for BLEND op
|
||
- pan/bi: Add missing BI_VECTOR
|
||
- pan/bi: Fix ADD.v4i8 opcode
|
||
- pan/bi: Eliminate writemasks in the IR
|
||
- pan/bi: Rename BI_SWIZZLE to BI_SELECT
|
||
- pan/bi: Pack FMA SEL16
|
||
- pan/bi: Pack FMA SEL8
|
||
- pan/bi: Pack ADD SEL16
|
||
- pan/bi: Force BI_SELECT arguments scalar
|
||
- pan/bit: Interpret BI_SELECT
|
||
- pan/bit: Add SELECT tests
|
||
- pan/bi: Fix RA wrt 16-bit swizzles
|
||
- pan/bi: Implement 16-bit COMBINE lowering
|
||
- nir: Move nir_lower_mediump_outputs from ir3
|
||
- ir3: Use shared mediump output lowering
|
||
- pan/bi: Add bool->float opcodes
|
||
- pan/bi: Add CSEL.64 opcode
|
||
- pan/bi: Add some 8-bit compares
|
||
- pan/bi: Add 64-bit int compares
|
||
- pan/bi: Add FCMP.GL.v2f16 on ADD opcode
|
||
- pan/bi: Add CSEL.8 opcode
|
||
- pan/bi(t): Fix SELECT tests
|
||
- pan/bi: Deduplicate csel/cmp cond
|
||
- pan/bi: Remove bi_round_op
|
||
- pan/bi: Structify FMA FCMP
|
||
- pan/bi Strucitfy ADD FCMP 32
|
||
- pan/bi: Structify FMA FCMP16
|
||
- pan/bi: Structify ADD FCMP16
|
||
- pan/bi: Structify FMA ICMP 32
|
||
- pan/bi: Structify FMA ICMP 16
|
||
- pan/bi: Structify ADD ICMP 32
|
||
- pan/bi: Fix source mod testing for CMP
|
||
- pan/bi: Pack FMA 32 FCMP
|
||
- pan/bi: Factor out fp16 abs logic
|
||
- pan/bi: Pack fma.fcmp16
|
||
- pan/bi: Relax double-abs condition
|
||
- pan/bit: Prepare condition evaluation for vectors
|
||
- pan/bit: Interpret CMP
|
||
- pan/bi: Add initial fcmp test
|
||
- pan/bi: Add bitwise modifiers
|
||
- pan/bi: Pack BI_BITWISE
|
||
- pan/bi: Handle iand/ior/ixor in NIR->BIR
|
||
- pan/bit: Interpret BI_BITWISE
|
||
- pan/bit: Add BITWISE test
|
||
- panfrost: Fix BO reference counting
|
||
- panfrost: Move Bifrost IR indexing to common
|
||
- pan/bi: Use common IR indices
|
||
- pan/mdg: Remove nir_alu_src_index
|
||
- pan/mdg: Use PAN_IS_REG
|
||
- pan/mdg: SSA_FIXED_MINIMUM already covered by PAN_IS_REG
|
||
- pan/mdg: Don't break SSA
|
||
- pan/mdg: Remove goofy 16-bit comment
|
||
- pan/mdg: Remove old hack
|
||
- pan/mdg: Set lower_flrp16
|
||
- pan/bi: Share ALU type printing
|
||
- pan/mdg: Add type fields to IR
|
||
- pan/mdg: Track ALU src types
|
||
- pan/mdg: Track ALU dest type
|
||
- pan/mdg: Another goofy comment gone
|
||
- pan/mdg: Track a primary type for I/O
|
||
- pan/mdg: Denoise prints
|
||
- pan/mdg: Track v_mov type (force uint32 for now?)
|
||
- pan/mdg: Track texture types
|
||
- pan/mdg: Set texture full fields at pack time
|
||
- pan/mdg: Move sampler_type emission to pack time
|
||
- pan/mdg: Lower specials to 32-bit
|
||
- pan/mdg: Specialize swizzle to type
|
||
- pan/mdg: Always print the mask
|
||
- pan/mdg: Make some branch targets more explicit
|
||
- pan/mdg: Don't crash on unknown branch target
|
||
- pan/mdg: Pass through some types from scheduling
|
||
- pan/mdg: Move condense_writemask to disasm
|
||
- pan/mdg: Ensure fdot is scalar out in disasm
|
||
- pan/mdg: Replicate 16-bit swizzles
|
||
|
||
Andreas Baierl (8):
|
||
|
||
- lima/parser: Fix RSW depth test parsing
|
||
- lima/parser: Extend AUX0 findings
|
||
- lima/parser: Change value name in RSW parser
|
||
- lima/parser: Extend rsw parsing showing strings instead of numbers
|
||
- gitlab-ci: lima: Add flaky tests to the skips list
|
||
- gitlab-ci: Enable the lima job again
|
||
- gitlab-ci: Add add a set of lima flakes
|
||
- lima: Add etc1 support
|
||
|
||
Andres Gomez (27):
|
||
|
||
- tracie: correct typo
|
||
- gitlab-ci: add missing popd to the build-deqp-vk.sh script
|
||
- gitlab-ci: build gfxreconstruct into the Vulkan testing container
|
||
- gitlab-ci: build VulkanTools into the Vulkan testing container
|
||
- gitlab-ci: Change devices format to <api-vendor-deviceId>
|
||
- gitlab-ci: Add gfxreconstruct traces support
|
||
- gitlab-ci: Add jobs to be able to test Vulkan
|
||
- gitlab-ci: Fix indentation and dangerous "\" in the last multiline
|
||
line
|
||
- gitlab-ci: Remove unneeded python3-pilkit dependency
|
||
- gitlab-ci: Sort packages to install alphabetically
|
||
- gitlab-ci: add python3-requests to the test-vk container
|
||
- gitlab-ci/traces: Add Vulkan sample entries for POLARIS10
|
||
- gitlab-ci: Don't use buster-backports packages by default for
|
||
x86_test-vk
|
||
- gitlab-ci: add Wine, win64's apitrace and DXVK to the Vulkan testing
|
||
container
|
||
- gitlab-ci: add apitrace's DXGI traces support
|
||
- gitlab-ci: replay apitrace traces in headless mode
|
||
- gitlab-ci: add Wine and DXVK env variables to Vulkan's tracie runner
|
||
- gitlab-ci/traces: Add D3D11 sample entry for POLARIS10
|
||
- gitlab-ci: Vulkan tracie runner to return last command exit code
|
||
- gitlab-ci: protect usage of shell variables with double quotes
|
||
- gitlab-ci: make explicit tracie is gitlab specific
|
||
- gitlab-ci: adapt query_traces_yaml to gitlab specific changes
|
||
- gitlab-ci: install winehq-stable to get 5.0 instead of 4.0
|
||
- Revert "meson,ci: Disable sparse_array tests on windows"
|
||
- gitlab-ci: update tracie README after changes in main script
|
||
- gitlab-ci: create always the "results" directory with tracie
|
||
- gitlab-ci: correct tracie behavior with replay errors
|
||
|
||
Andrii Simiklit (2):
|
||
|
||
- Revert "glx: convert glx_config_create_list to one big calloc"
|
||
- i965/vec4: Ignore swizzle of VGRF for use by var_range_end()
|
||
|
||
Anuj Phogat (2):
|
||
|
||
- intel/gen12+: Reserve 4KB of URB space per bank for Compute Engine
|
||
- intel/gen12+: Set way_size_per_bank to 4
|
||
|
||
Arcady Goldmints-Orlov (7):
|
||
|
||
- compiler/nir: Add support for variable initialization from a pointer
|
||
- compiler/spirv: Add support for non-constant initializers
|
||
- Rename nir_lower_constant_initializers to
|
||
nir_lower_variable_initalizers
|
||
- spirv: Remove outdated SPIR-V decoration warnings
|
||
- nir: Lower returns correctly inside nested loops
|
||
- anv: increase minUniformBufferOffsetAlignment to 64
|
||
- intel/compiler: fix alignment assert in nir_emit_intrinsic
|
||
|
||
Axel Davy (1):
|
||
|
||
- gallium/util: Fix leak in the live shader cache
|
||
|
||
Bas Nieuwenhuizen (29):
|
||
|
||
- radv: Allow non-dedicated linear images and buffer.
|
||
- radv: Do not set SX DISABLE bits for RB+ with unused surfaces.
|
||
- radv: Optimize emitting index buffer changes.
|
||
- radv: Do not redundantly set the RB+ regs on pipeline switch.
|
||
- radeonsi: Fix compute copies for subsampled formats.
|
||
- amd/llvm: Fix divergent descriptor indexing. (v3)
|
||
- amd/llvm: Fix divergent descriptor regressions with radeonsi.
|
||
- radv: Store 64-bit availability bools if requested.
|
||
- radv: Consider maximum sample distances for entire grid.
|
||
- radv: Whitespace fixup.
|
||
- radv: Use correct buffer count with variable descriptor set sizes.
|
||
- winsys/amdgpu: Retrieve WC flags from imported buffers.
|
||
- drm-uapi,radv,radeonsi: Add amdgpu_drm.h header.
|
||
- vulkan/wsi: Add callback to set ownership of buffer.
|
||
- radv: Add WSI buffers to BO list only if they can be used.
|
||
- st/dri: Set next in template instead of after creation. (v2)
|
||
- radeonsi: Count planes for imported textures.
|
||
- radv: Use actual memory type count for setting app-visible bitset.
|
||
- radv: Stop using memory type indices.
|
||
- radv/winsys: Add function to get domains/flags from fd.
|
||
- radv: Determine memory type for import based on fd.
|
||
- radv: Expose 4G element texel buffers.
|
||
- radv: Fix implicit sync with recent allocation changes.
|
||
- radv: Extend tiling flags to 64-bit.
|
||
- radv: Provide a better error for permission issues with priorities.
|
||
- radv/winsys: Remove extra sizeof multiply.
|
||
- radv: Handle failing to create .cache dir.
|
||
- radv: Do not close fd -1 when NULL-winsys creation fails.
|
||
- radv: Implement vkGetSwapchainGrallocUsage2ANDROID.
|
||
|
||
Bernd Kuhls (1):
|
||
|
||
- util/os_socket: Include unistd.h to fix build error
|
||
|
||
Blaž Tomažič (1):
|
||
|
||
- radeonsi: Fix omitted flush when moving suballocated texture
|
||
|
||
Boris Brezillon (45):
|
||
|
||
- pan/midgard: Add an enum to describe the render targets
|
||
- pan/midgard: Make sure we pass the right RT id to
|
||
emit_fragment_store()
|
||
- pan/midgard: Lower bitfield extract to shifts
|
||
- pan/midgard: Don't check 'branch && branch->writeout' twice in
|
||
mir_schedule_alu()
|
||
- pan/midgard: Stop leaking instruction objects in mir_schedule_alu()
|
||
- panfrost: Fix the damage box clamping logic
|
||
- pan/midgard: Turn Z/S stores into zs_output_pan intrinsics
|
||
- pan/midgard: Add nir_intrinsic_store_zs_output_pan support
|
||
- panfrost: Z24 variants should be sampled as R32UI
|
||
- panfrost: Add the MALI_WRITES_{Z,S} flags
|
||
- panfrost: Set the MALI_WRITES_{Z,S} flags when needed
|
||
- Revert "panfrost: Z24 variants should be sampled as R32UI"
|
||
- panfrost: Pass the sampler view format when creating a tex descriptor
|
||
- panfrost: Assign primitive_size.pointer only if writes_point_size()
|
||
returns true
|
||
- panfrost: Add an helper to retrieve the currently active shader state
|
||
- panfrost: Move the batch stack size adjustment out of
|
||
panfrost_queue_draw()
|
||
- panfrost: Move viewport desc emission out of panfrost_emit_for_draw()
|
||
- panfrost: Move the const buf emission logic out of
|
||
panfrost_emit_for_draw()
|
||
- panfrost: Move shared mem desc emission out of panfrost_launch_grid()
|
||
- panfrost: Dissociate shader meta patching from the desc emission
|
||
- panfrost: Move panfrost_attach_vt_framebuffer() to pan_cmdstream.c
|
||
- panfrost: Stop using panfrost_emit_for_draw() for compute jobs
|
||
- panfrost: Simplify panfrost_emit_for_draw() and make it private
|
||
- panfrost: Add an helper to update the occclusion query part of a
|
||
tiler job desc
|
||
- panfrost: Add an helper to update the rasterizer part of a tiler job
|
||
desc
|
||
- panfrost: Prepare things to get rid of panfrost_shader_state.tripipe
|
||
- panfrost: Prepare shader_meta descriptors at emission time
|
||
- panfrost: Add a panfrost_sampler_desc_init() helper
|
||
- panfrost: Move sampler/tex descs emission helpers to pan_cmdstream.c
|
||
- panfrost: Add an helper to emit a pair of vertex/tiler jobs
|
||
- panfrost: Drop initial mali_attr_meta.src_offset assignment
|
||
- panfrost: Ignore BO start addr when adjusting src_offset
|
||
- panfrost: Prepare attribute for builtins at state creation time
|
||
- panfrost: Emit attribute descriptors after patching the templates
|
||
- panfrost: Move the mali_attr.src_offset adjustment to a sub-function
|
||
- panfrost: Rename panfrost_stage_attributes()
|
||
- panfrost: Move streamout offset update out of panfrost_draw_vbo()
|
||
- panfrost: Move vertex/tiler payload initialization out of
|
||
panfrost_draw_vbo()
|
||
- panfrost: Inline panfrost_queue_draw() and panfrost_emit_for_draw()
|
||
- panfrost: Move panfrost_emit_vertex_data() to pan_cmdstream.c
|
||
- panfrost: Move panfrost_emit_varying_descriptor() to pan_cmdstream.c
|
||
- panfrost: Re-init the VT payloads at draw/launch_grid() time
|
||
- panfrost: Use ctx->active_prim in panfrost_writes_point_size()
|
||
- panfrost: Get rid of ctx->payloads[]
|
||
- vtn/opencl: add rint-support
|
||
|
||
Brian Ho (17):
|
||
|
||
- turnip: Promote tu_cs_get_size/is_empty to header
|
||
- turnip: Execute main cs for secondary command buffers
|
||
- turnip: Advertise 8 bit subpixel precision
|
||
- ir3: Disable copy prop for immediate ldlw offsets
|
||
- turnip: Set has_gs in ir3_shader_key
|
||
- turnip: Emit geometry shader obj and related consts
|
||
- turnip: Configure VPC for geometry shaders
|
||
- turnip: Configure VFD_CONTROL with gsheader and primitiveid
|
||
- turnip: Set up REG_A6XX_SP_GS_CONFIG
|
||
- turnip: Selectively configure GRAS_LAYER_CNTL
|
||
- turnip: Update maxGeometryShaderInvocations to match blob
|
||
- turnip: Populate tu_pipeline.active_stages
|
||
- turnip: Enable geometry shaders for CP_DRAWs
|
||
- turnip: Enable geometryShader device feature
|
||
- turnip: Correctly set layer stride for 3D images
|
||
- turnip: Emit geometry shader descriptor consts
|
||
- freedreno/turnip: Update GRAS_LAYER_CNTL to GRAS_MAX_LAYER_INDEX
|
||
|
||
Caio Marcelo de Oliveira Filho (46):
|
||
|
||
- anv: Advertise VK_KHR_shader_non_semantic_info
|
||
- radv: Advertise VK_KHR_shader_non_semantic_info
|
||
- intel/gen12: Take into account opcode when decoding SWSB
|
||
- spirv: Be consistent when checking for Shader/Kernel
|
||
- anv: Use intel_debug_flag_for_shader_stage()
|
||
- anv: Add pipe_state_for_stage() helper
|
||
- nir/builder: Add nir_scoped_memory_barrier()
|
||
- nir: Add the alias NIR_MEMORY_ACQ_REL
|
||
- nir/tests: Use nir_scoped_memory_barrier() helper
|
||
- nir, intel: Move use_scoped_memory_barrier to nir_options
|
||
- anv: Remove unused field xfb_used from anv_pipeline
|
||
- anv: Remove unused field \`urb.total_size\`
|
||
- nir: Don't skip a bit in nir_memory_semantics
|
||
- nir: Reorder nir_scopes so wider scope has larger numeric value
|
||
- nir: Add pass to combine adjacent scoped memory barriers
|
||
- intel/fs: Combine adjacent memory barriers
|
||
- anv: Add a new enum to identify the pipeline type
|
||
- anv: Use pipeline type to decide whether or not lower multiview
|
||
- anv: Use a dynamic array for storing executables in pipeline
|
||
- anv: Keep the shader stage in anv_shader_bin
|
||
- anv: Pass the right pipe_state to flush_descriptor_sets()
|
||
- anv: Remove redundant check in flush_descriptor_sets() helpers
|
||
- anv: Decouple flush_descriptor_sets() helpers from pipeline struct
|
||
- anv: Decouple flush_descriptor_sets() from pipeline struct
|
||
- anv: Use a separate field in the pipeline for compute shader
|
||
- anv: Split graphics and compute bits from anv_pipeline
|
||
- anv: Reduce compute pipeline batch_data size
|
||
- anv: Remove duplicate code in anv_cmd_buffer_bind_descriptor_set
|
||
- intel/blorp: Plumb the stage through blorp upload_shader
|
||
- mesa/main: Fix overflow in validation of DispatchComputeGroupSizeARB
|
||
- nir: Add per_view attribute to nir_variable
|
||
- intel/gen12: Add XML description for 3DSTATE_PRIMITIVE_REPLICATION
|
||
- intel/fs: Allow multiple slots for position
|
||
- anv/gen12: Lower VK_KHR_multiview using Primitive Replication
|
||
- intel/compiler: Replace cs_prog_data->push.total with a helper
|
||
- anv: Stop using cs_prog_data->threads
|
||
- iris: Stop using cs_prog_data->threads
|
||
- intel/compiler: Remove cs_prog_data->threads
|
||
- intel/fs,vec4: Properly account SENDs in IVB memory fence
|
||
- spirv: Fix propagation of OpVariable access flags
|
||
- spirv: Handle instruction aliases in vtn_gather_types
|
||
- spirv: Update the headers from latest Khronos master
|
||
- intel/fs: Allow FS_OPCODE_SCHEDULING_FENCE stall on registers
|
||
- intel/fs,vec4: Pull stall logic for memory fences up into the IR
|
||
- intel/fs: Only stall after sending all memory fence messages
|
||
- i965: Use correct constant for max_variable_local_size
|
||
|
||
Chad Versace (12):
|
||
|
||
- anv: Drop unused anv_image_get_surface_for_aspect_mask()
|
||
- anv: Rename param make_surface::dev to device
|
||
- anv: Delete anv_image::ccs_e_compatible
|
||
- anv: Clarify behavior of anv_image_aspect_to_plane()
|
||
- anv: Respect ISL_SURF_USAGE_DISABLE_AUX_BIT in make_surface()
|
||
- turnip: Add magic register values to tu_physical_device
|
||
- turnip: Add a618 support
|
||
- anv: Drop anv_image.c:get_surface()
|
||
- anv: Add anv_image_plane_needs_shadow_surface() (v2)
|
||
- anv: Refactor creation of aux surfaces (v2)
|
||
- anv: Flatten the logic add_aux_surface_if_supported (v3)
|
||
- anv: Use isl_drm_modifier_get_default_aux_state()
|
||
|
||
Chia-I Wu (2):
|
||
|
||
- egl/android: require ANDROID_native_fence_sync for buffer age
|
||
- egl/android: enable/disable KHR_partial_update correctly
|
||
|
||
Chris Lord (2):
|
||
|
||
- vc4: fix vc4_yuv_blit overwriting fragment constant buffer slot 0
|
||
- vc4: Fix query_dmabuf_modifiers mis-reporting external_only property
|
||
|
||
Chris Wilson (1):
|
||
|
||
- iris: Fix import sync-file into syncobj
|
||
|
||
Christian Gmeiner (44):
|
||
|
||
- etnaviv: enable texture upload memory throttling
|
||
- etnaviv: update headers from rnndb
|
||
- etnaviv: fix alpha test on GC3000
|
||
- etnaviv: add etna_constbuf_state object
|
||
- etnaviv: ask kernel for max number of supported varyings
|
||
- etnaviv: update headers from rnndb
|
||
- etnaviv: increase number of supported varyings to 16
|
||
- etnaviv: implement emit_string_marker
|
||
- etnaviv: get rid of etna_spec in etna_context
|
||
- etnaviv: enable shareable shaders
|
||
- freedreno: calculate modified bit mask only once
|
||
- freedreno: simplify fd_set_shader_buffers(..)
|
||
- freedreno: ssbo: keep track if a buffer gets written
|
||
- freedreno: ssbo: mark resource read or written depending on usage
|
||
- etnaviv: get rid of SE_CLIP\_\*
|
||
- etnaviv: rework clippling calculation to be a derived state
|
||
- etnaviv: do the left shift by 16 at emit time
|
||
- etnaviv: get rid of struct compiled_scissor_state
|
||
- etnaviv: s/scissor_s/scissor
|
||
- etnaviv: compiled_framebuffer_state: get rid of SE_SCISSOR\_\*
|
||
- etnaviv: rename hw queries to acc queries
|
||
- etnaviv: rework etna_acc_sample_provider
|
||
- etnaviv: explicitly call resource_written(..)
|
||
- etnaviv: reset no_wait_cnt after triggered flush
|
||
- etnaviv: rework wait/flush logic
|
||
- etnaviv: extend acc query provider with supports(..) function
|
||
- etnaviv: make use of a fixed size array to track of all acc query
|
||
provider
|
||
- etnaviv: extend result(..) to return if data is ready
|
||
- etnaviv: extend acc sample provide with an allocate(..)
|
||
- etnaviv: move generic perfmon functionality into own file
|
||
- etnaviv: convert perfmon queries to acc queries
|
||
- etnaviv: drop redundant calls to etna_acc_query_suspend(..)
|
||
- etnaviv: change begin_query(..) to a void function
|
||
- etnaviv: remove the "active" member of queries
|
||
- etnaviv: anisotropic filtering is supported starting with HALTI0
|
||
- etnaviv: update headers from rnndb
|
||
- etnaviv: add anisotropic filter support
|
||
- docs/features: mark GL_ARB_texture_filter_anisotropic as done for
|
||
etnaviv
|
||
- etnaviv: drop default state for FE_HALTI5_ID_CONFIG
|
||
- etnaviv: call util_blitter_save_fragment_constant_buffer_slot(..)
|
||
- etnaviv: support for using generic blit path
|
||
- ci: bare-metal: power down device after tests
|
||
- etnaviv: fix SAMP_ANISOTROPY register value
|
||
- etnaviv: do not use int filter when anisotropic filtering is used
|
||
|
||
Christopher Egert (1):
|
||
|
||
- radv: use util_float_to_half_rtz
|
||
|
||
Christopher James Halse Rogers (1):
|
||
|
||
- egl/wayland: Fix zwp_linux_dmabuf usage
|
||
|
||
Connor Abbott (55):
|
||
|
||
- freedreno: Fix CP_COND_REG_EXEC bit positions
|
||
- freedreno: Add CP_REG_WRITE documentation
|
||
- freedreno: Fix CP_COND_EXEC
|
||
- tu: Move vsc_data and vsc_data2 allocation into the device
|
||
- tu: Don't emit initial render target state in tile_load_ib
|
||
- tu: Properly set UBWC flags in RB_RENDER_CNTL
|
||
- tu/blit: Support blits in secondary cmdstreams
|
||
- tu: Support multisample image clears
|
||
- tu: Disable linear depth attachments
|
||
- tu: Sysmem rendering
|
||
- tu: Add helper for CP_COND_REG_EXEC
|
||
- tu: Handle vkCmdClearAttachments() with sysmem
|
||
- tu: Support resolve ops with sysmem rendering
|
||
- tu: Support input attachments with sysmem
|
||
- tu: Force sysmem with mipmapped non-aligned linear stores
|
||
- tu: Rewrite border color handling
|
||
- lima/gpir: Make lima_gpir_node_insert_child() useful
|
||
- lima/gpir: Optimize conditional break/continue
|
||
- lima/gpir: Optimize nots created from branch lowering
|
||
- tu: Fix border color with compute shaders
|
||
- freedreno/fdl: Add base_align
|
||
- tu: Return the correct alignment for images
|
||
- freedreno: Cleanup event names
|
||
- freedreno: Rename RB_DONE_TS
|
||
- tu: Dump out shader assembly when requested
|
||
- tu: ir3: Emit push constants directly
|
||
- freedreno/a6xx: Add UBO size field
|
||
- freedreno/a6xx: Add registers for the bindless model
|
||
- ir3: Add bindless instruction encoding
|
||
- ir3: Plumb through support for a1.x
|
||
- ir3: Also don't propagate immediate offset with LDC
|
||
- ir3: LDC also has a destination
|
||
- ir3: Plumb through bindless support
|
||
- ir3: Rewrite UBO push analysis to support bindless
|
||
- tu: Switch to the bindless descriptor model
|
||
- tu: Emit CP_LOAD_STATE6 for descriptors
|
||
- tu: Add missing code for immutable samplers
|
||
- tu: Implement descriptor set update templates
|
||
- ir3: Fix txs with bindless
|
||
- ir3: Fix LDC offset units
|
||
- ir3: Handle load_ubo_ir3 when promoting to constants
|
||
- tu: Align GMEM resolve blit scissor
|
||
- tu: Use tu_cs_add_entries() with non-render-pass secondaries
|
||
- ir3/ra: Fix off-by-one issues with live-range extension
|
||
- freedreno/a6xx: Expand various varying-count bitfields
|
||
- tu: Fix the advertised maxFragmentInputComponents
|
||
- ir3: Don't double-insert the first block
|
||
- ir3: Fix bug with shaders that only exit via discard
|
||
- freedreno/a6xx: Document PrimID passthrough registers
|
||
- ir3: Skip missing VS outputs in VS out map when linking
|
||
- tu: Implement PrimID passthrough
|
||
- freedreno/a6xx: Implement PrimID passthrough
|
||
- st/nir: Fix assigning PointCoord location with !PIPE_CAP_TEXCOORD
|
||
- ir3: Remove VARYING_SLOT_PNTC remapping hack
|
||
- tu: Don't invert point coords
|
||
|
||
D Scott Phillips (6):
|
||
|
||
- intel/tools/aubinator_error_decode: read HW Context before other
|
||
batches
|
||
- intel/tools/aubinator_error_decode: Decode ring buffers from HEAD to
|
||
TAIL
|
||
- util/sparse_array: don't stomp head's counter on pop operations
|
||
- intel/fs: Update location of Render Target Array Index for gen12
|
||
- anv,iris: Fix input vertex max for tcs on gen12
|
||
- anv/gen11+: Disable object level preemption
|
||
|
||
Daniel Schürmann (73):
|
||
|
||
- aco: fix image_atomic_cmp_swap
|
||
- nir: gather info whether a shader uses demote_to_helper
|
||
- nir: add pass to lower discard() to demote()
|
||
- amd/llvm: implement nir_intrinsic_demote(_if) and
|
||
nir_intrinsic_is_helper_invocation
|
||
- radeonsi: lower discard to demote when FS_CORRECT_DERIVS_AFTER_KILL
|
||
is enabled
|
||
- radv: use nir_lower_discard_to_demote to work around game bugs
|
||
- amd: join emit_kill() from radv and radeonsi in ac_nir_to_llvm
|
||
- nir: fix unpack_64_4x16 in lower_alu_to_scalar()
|
||
- aco: add comparison operators for PhysReg
|
||
- aco: add sub-dword regclasses
|
||
- aco: refactor regClass setup for subdword VGPRs
|
||
- aco: validate p_create_vector with subdword elements properly
|
||
- aco: validate register alignment of subdword operands and definitions
|
||
- aco: validate uninitialized operands
|
||
- aco: validate RA of subdword assignments
|
||
- aco: print subdword registers
|
||
- aco: fix Temp and assignment of renamed operands during RA
|
||
- aco: remove unnecessary reg_file.fill() operation in
|
||
get_reg_create_vector()
|
||
- aco: add notion of subdword registers to register allocator
|
||
- aco: create helper function to collect variables from register area
|
||
- aco: adapt register allocation for subdword registers
|
||
- aco: align subdword registers during RA when necessary
|
||
- aco: small refactoring of shuffle code lowering
|
||
- aco: add builder function for subdword copy()
|
||
- aco: lower subdword shuffles correctly.
|
||
- aco: don't propagate SGPRs into subdword PSEUDO instructions
|
||
- aco: don't assume split_vector(create_vector) has the same number of
|
||
elements when optimizing
|
||
- aco: don't vectorize 8/16bit load/store_ssbo
|
||
- aco: add missing conversion operations for small bitsizes
|
||
- aco: add byte_align_scalar() & trim_subdword_vector() helper
|
||
functions
|
||
- aco: prepare helper functions for subdword handling
|
||
- aco: implement vec2/3/4 with subdword operands
|
||
- aco: implement storagePushConstant8 & storagePushConstant16
|
||
- aco: implement 8bit/16bit load_buffer
|
||
- aco: implement 8bit/16bit store_ssbo
|
||
- aco: use MUBUF to load subdword SSBO
|
||
- aco: guarantee that Temp fits in 4 bytes
|
||
- aco: add explicit padding for all Instruction sub-structs
|
||
- aco: improve hashing for value numbering
|
||
- aco: improve register assignment when live-range splits are necessary
|
||
- aco: replace assignment hashmap by std::vector in register allocation
|
||
- aco: during RA only insert into renames table if a variable got
|
||
renamed
|
||
- aco: improve speed of live_var_analysis
|
||
- aco: refactor try_remove_trivial_phi() in RA
|
||
- aco: change some std::map to std::unordered_map in
|
||
register_allocation
|
||
- aco: change live_out variables to std::unordered_set
|
||
- aco: move all needed helper containers to ra_ctx
|
||
- aco: RA - move all std::function objects into proper functions
|
||
- aco: setup subdword regclasses for ssa_undef & load_const
|
||
- aco: ensure correct bit representation of subdword constants
|
||
- aco: don't constant-propagate into subdword PSEUDO instructions
|
||
- aco: lower subdword phis with SGPR operands
|
||
- aco: rename aco_lower_bool_phis() -> aco_lower_phis()
|
||
- aco: make some reg_file helpers private and fix their uses
|
||
- aco: fix p_extract_vector optimization in presence of unequally sized
|
||
vector operands
|
||
- aco: use v_subrev_f32 for fsub with an sgpr operand in src1
|
||
- aco: fix 64bit fsub
|
||
- aco: move src1 to vgpr instead of using VOP3 for VOP2 instructions
|
||
during isel
|
||
- aco: simplify operand handling in RA
|
||
- aco: refactor get_reg() to take Temp instead of RegClass
|
||
- aco: refactor get_reg() to also handle affinities
|
||
- aco: create pseudo dummy instruction in RA to be used for live-range
|
||
splits
|
||
- aco: create and use DefInfo struct in RA
|
||
- aco: use DefInfo in more places to simplify RA
|
||
- aco: move attempt to find strided register into get_reg_simple()
|
||
- aco: allocate full register for subdword definitions if HW doesn't
|
||
support it
|
||
- aco: don't create vector affinities for operands which are not killed
|
||
or are duplicates
|
||
- aco: refactor get_reg_simple() to return early on exact matches
|
||
- aco: stop get_reg_simple after reaching max_used_gpr
|
||
- aco: try to always find a register with stride for even sizes
|
||
- aco: use upper part of gap in register file if it is beneficial for
|
||
striding
|
||
- aco: coalesce v_mad's accumulator with definition's affinities
|
||
- aco: either copy-propagate or inline create_vector operands
|
||
|
||
Daniel Stone (15):
|
||
|
||
- Revert "gitlab-ci: disable panfrost runners"
|
||
- egl/wayland: Don't invalidate buffers on no-op resize
|
||
- util/test: Use MAX_PATH on Windows
|
||
- CI: Add native Windows VS2019 build
|
||
- CI: Windows: Fix Docker tag argument inversion
|
||
- CI: Disable Panfrost Mali-T820 jobs
|
||
- CI: Avoid htz4 runner for VS2019
|
||
- meson: Add VS 4624 warning exclusion to remove piles of LLVM warnings
|
||
- CI: Re-enable Windows VS2019 builds
|
||
- EGL: Add eglSetDamageRegionKHR to GLVND dispatch list
|
||
- meson: Make shared-llvm into a tri-state boolean
|
||
- CI: Disable Windows/VS2019 builds
|
||
- Revert "CI: Disable Windows/VS2019 builds"
|
||
- ci/windows: Make Chocolatey installs more reliable
|
||
- CI: Disable Lima jobs due to lab unhealthiness
|
||
|
||
Danylo Piliaiev (29):
|
||
|
||
- i965: Do not set front_buffer_dirty if there is no front buffer
|
||
- st/mesa: Handle the rest renderbuffer formats from OSMesa
|
||
- osmesa/tests: Cover OSMESA_RGB GL_UNSIGNED_BYTE case
|
||
- st/nir: Unify inputs_read/outputs_written before serializing NIR
|
||
- brw_nir: Cast bitshift to unsigned
|
||
- brw_fs: Avoid zero size vla
|
||
- intel/compiler: Do not qsort zero sized array
|
||
- intel/bufmgr: Cast bitshift to unsigned
|
||
- glsl/blob: Do not call memcpy if there is nothing to copy
|
||
- iris: Do not dereference nullptr with pipe_reference
|
||
- i965: Do not generate D16 B5G6R5_UNORM configs on gen < 8
|
||
- intel/tools: Fix compilation with UBSan
|
||
- glsl: do not crash if string literal is used outside of
|
||
#include/#line
|
||
- st/mesa: Fix signed integer overflow when using
|
||
util_throttle_memory_usage
|
||
- intel/aub_viewer: Fix format specifier for uint64_t
|
||
- nir: Fix breakage of foreach_list_typed_safe assumptions in loop
|
||
unrolling
|
||
- anv: Do not sample from 3d depth image with HiZ
|
||
- glsl/list: Fix undefined behaviour of foreach\_\* macros
|
||
- st/mesa: Update shader info of ffvp/ARB_vp after translation to NIR
|
||
- st/mesa: Re-assign vs in locations after updating nir info for
|
||
ffvp/ARB_vp
|
||
- spirv: Expand workaround for OpControlBarrier on old GLSLang
|
||
- st/mesa: Treat vertex inputs absent in inputMapping as zero in
|
||
mesa_to_tgsi
|
||
- iris/bufmgr: Check if iris_bo_gem_mmap failed
|
||
- i965: Fix out-of-bounds access to brw_stage_state::surf_offset
|
||
- anv: Translate relative timeout to absolute when calling
|
||
anv_timelines_wait
|
||
- anv: Fix deadlock in anv_timelines_wait
|
||
- meson: Disable GCC's dead store elimination for memory zeroing custom
|
||
new
|
||
- mesa: Fix double-lock of Shared->FrameBuffers and usage of wrong
|
||
mutex
|
||
- intel/fs: Work around dual-source blending hangs in combination with
|
||
SIMD16
|
||
|
||
Dave Airlie (69):
|
||
|
||
- llvmpipe/query: add support for indexed queries
|
||
- gallivm/swr: add stream_id to geom epilogue emit
|
||
- gallivm/nir: add support for multiple vertex streams
|
||
- draw: change geom shader output to an array of outputs.
|
||
- draw/gs: track emitted prims + verts per stream.
|
||
- draw: emit multiple streams to streamout.
|
||
- draw: don't emit vertex to streams with no outputs
|
||
- llvmpipe: advertise 4 vertex streams
|
||
- gallivm/s390: fix pass init order on s390 with llvm 8 (v2)
|
||
- ci: bump debian image and change llvm deps to 8
|
||
- dri: add another get shm variant.
|
||
- glx/drisw: add getImageShm2 path
|
||
- glx/drisw: return false if shmid == -1
|
||
- glx/drisw: fix shm put image fallback
|
||
- gallivm/tgsi: fix stream id regression
|
||
- gallivm/nir: fix integer divide SIGFPE
|
||
- gallivm/nir: handle mod 0 better.
|
||
- gallium/auxiliary: add the microsoft tessellator and a pipe wrapper.
|
||
- gallivm/nir: split out 64-bit splitting code
|
||
- gallivm/nir: add support for tess system values
|
||
- gallivm/nir: align store_var param order with load_var
|
||
- gallivm/tgsi/swr: add mask vec to the tcs store
|
||
- gallivm/nir: add tessellation i/o support.
|
||
- draw: add JIT context/functions for tess stages.
|
||
- draw: add main tessellation code
|
||
- draw: hook up final bits of tessellation
|
||
- gallium/nir/tgsi: only scan fragment shader inputs for usage_mask
|
||
- llvmpipe: add support for tessellation shaders
|
||
- gallivm/tessellator: use private functions for min/max to avoid
|
||
namespace issues
|
||
- gallium: fix build with latest meson and gcc10
|
||
- gallivm/s3tc: split out dxt5 alpha code
|
||
- gallivm: add support for rgtc/latc fetches.
|
||
- gallium/llvmpipe: add an optimised 32-bit memset
|
||
- gallivm/rgtc: fix the truncation to 8-bit
|
||
- gallivm/rgtc: enable fast path for snorm types.
|
||
- Revert "gallivm: disable rgtc/latc SNORM accellerated fetches"
|
||
- llvmpipe: fixup context leaks.
|
||
- draw: collect tessellation invocations statistics
|
||
- llvmpipe: report tessellation shader statistics.
|
||
- llvmpipe/query: fix transform feedback overflow any queries.
|
||
- gallivm: fix left over shader vote debug
|
||
- gallivm/nir: lower implicit lod to tex.
|
||
- gallivm/draw: calloc prim id toavoid undef
|
||
- llvmpipe: fix no tokens detections.
|
||
- draw: fix tessellation stats query
|
||
- llvmpipe/setup: move line stats collection earlier.
|
||
- draw/cull: run pipeline for culled points.
|
||
- draw: fix user culling pipeline order. (v2)
|
||
- u_blitter: fix stencil blitting
|
||
- draw: free the NIR IR.
|
||
- draw/tess: free the NIR
|
||
- llvmpipe/nir: free the nir shader
|
||
- nir/linking: fix issue with two compact variables in a row. (v2)
|
||
- gallivm/nir: fix image store conversions
|
||
- gallivm/nir: add helper invocation support
|
||
- util/indirect: handle stride less than number of parameters.
|
||
- llvmpipe: bump max images to 16
|
||
- llvmpipe: fix ssbo alignment
|
||
- draw/tess: fix TES patch vertices in.
|
||
- llvmpipe: fix d32 unorm depth conversions.
|
||
- llvmpipe/setup: add point size clamping
|
||
- llvmpipe: enable stencil only formats. (v2)
|
||
- llvmpipe: clamp color storage for integer types.
|
||
- gallivm: fix stencil border
|
||
- vulkan: add initial device selection layer. (v6.1)
|
||
- ci: add llvmpipe paths to virgl rules
|
||
- draw/tess: free tessellation control shader i/o memory.
|
||
- llvmpipo/nir: free compute shader NIR
|
||
- llvmpipe: compute shaders work better with all the threads.
|
||
|
||
David Stevens (1):
|
||
|
||
- egl/android: set window usage flags
|
||
|
||
Denys (1):
|
||
|
||
- gitlab: add bug report template
|
||
|
||
Dominik Behr (1):
|
||
|
||
- meson: fix debug build on Android
|
||
|
||
Drew Davenport (1):
|
||
|
||
- radv: Filter extensions not whitelisted for Android
|
||
|
||
Duncan Hopkins (2):
|
||
|
||
- zink. Added storage CISto descriptor pool. Added storage in
|
||
descriptor pool for combined image samplers as well as uniform
|
||
buffers. Stops some shaders from running through a pools storage
|
||
faster than zinks internal tracking.
|
||
- zink: zero out zink_render_pass_state
|
||
|
||
Dylan Baker (48):
|
||
|
||
- docs/release-calendar: 20.0.0-rc1 has been released
|
||
- docs: Mark 20.0-rc2 as done
|
||
- docs: Add release notes for 19.3.4
|
||
- docs: Add SHA256 sum for 19.3.4
|
||
- docs: Mark 19.3.4 as done
|
||
- docs: Mark 20.0.0-rc3 as done
|
||
- Docs: Add 20.0.0 release notes
|
||
- docs: Update index, relnotes, and release-calendar for 20.0
|
||
- docs: Update stable process around using fixes: and gitlab
|
||
- docs/submittingpatches: Fix confusing typo + missing pronoun
|
||
- docs: Update release notes with current process
|
||
- bin/post_version.py: Update the release calendar as well
|
||
- bin/post_version.py: Pretty print the html
|
||
- bin/post_version.py: Make the git commit as well.
|
||
- docs: update releasing to cover updated post_version.py
|
||
- docs: add relnotes for 20.0.1
|
||
- docs: Add sha256sums for 20.0.1
|
||
- docs: update news, calendar, and link release notes for 20.0.1
|
||
- Docs: Add release notes for 20.0.2
|
||
- docs/relnotes: Add sha256 sums for 20.0.2
|
||
- docs: update calendar, add news item, and link releases notes for
|
||
20.0.2
|
||
- docs/release-calendar: Add calendar for 20.1 Release candidates
|
||
- bin/gen_release_notes.py: Fix version detection for .0 release
|
||
- bin/pick-ui: Add a new maintainer script for picking patches
|
||
- replace \_mesa_is_pow_two with util_is_power_of_two\_\*
|
||
- replace \_mesa_next_pow_two\_\* with util_next_power_of_two\_\*
|
||
- replace \_mesa_logbase2 with util_logbase2
|
||
- replace LOG2 with util_fast_log2
|
||
- u_math: add x86 optimized version of ifloor
|
||
- replace IFLOOR with util_ifloor
|
||
- Replace IROUND_POS with \_mesa_roundevenf
|
||
- mesa/main: remove unused IROUNDD
|
||
- replace IROUND with util functions
|
||
- move windows strtok_r define to u_string
|
||
- Replace IS_INF_OR_NAN with util_is_inf_or_nan
|
||
- replace malloc macros in imports.h with u_memory.h versions
|
||
- util: Add an aligned realloc function
|
||
- replace imports memory functions with utils memory functions
|
||
- mesa|mapi: replace \_mesa_[v]snprintf with [v]snprintf
|
||
- mesa: move ADD_POINTERS to macros.h
|
||
- dri/nouveau: replace assert with unreachable
|
||
- remove final imports.h and imports.c bits
|
||
- meson: update llvm dependency logic for meson 0.54.0
|
||
- docs: Add relnotes for 20.0.5
|
||
- docs: Add sha256 sums for 20.0.5
|
||
- docs: update calendar, add news item, and link releases notes for
|
||
20.0.5
|
||
- mesa: Follow OpenGL conversion rules for values that exceed storage
|
||
size
|
||
- tests: Make tests aware of meson test wrapper
|
||
|
||
Edmondo Tommasina (1):
|
||
|
||
- radv/sqtt: fix RADV_THREAD_TRACE_BUFFER_SIZE spelling
|
||
|
||
Eduardo Lima Mitev (3):
|
||
|
||
- turnip/pipeline: Don't assume tu_shader is a valid object
|
||
- turnip: Instance can be NULL resolving 'GetInstanceProcAddr' entry
|
||
point
|
||
- anv/radv: Resolving 'GetInstanceProcAddr' should not require a valid
|
||
instance
|
||
|
||
Eli Schwartz (1):
|
||
|
||
- docs: fix typo in v20 release notes
|
||
|
||
Elie Tournier (3):
|
||
|
||
- spirv2nir: print nir shader if translation succed
|
||
- spirv2nir: Add kernel spirv support
|
||
- docs/features: Update virgl OpenGL 4.5 features GL_ARB_clip_control
|
||
and GL_KHR_robustness are now expose in the guest.
|
||
|
||
Emil Velikov (11):
|
||
|
||
- meson: glx: drop with_glx == dri check
|
||
- glx: set the loader_logger early and for everyone
|
||
- egl/drm: reinstate (kms\_)swrast support
|
||
- Revert "egl/dri2: Don't dlclose() the driver on
|
||
dri2_load_driver_common failure"
|
||
- loader: use a maximum of 64 drmDevices
|
||
- loader: simplify loader_get_user_preferred_fd()
|
||
- loader: simplify codeflow in drm_get_pci_id_for_fd
|
||
- loader: move "using driver..." message to
|
||
loader_get_kernel_driver_name
|
||
- loader: fallback to kernel name, if PCI fails
|
||
- glx: omit loader_loader() for macOS
|
||
- egl: simplify client/platform extension handling
|
||
|
||
Emmanuel Gil Peyrot (1):
|
||
|
||
- Expose EGL_KHR_platform\_\* when EXT is supported
|
||
|
||
Eric Anholt (144):
|
||
|
||
- gallium/osmesa: Fix a typo in the unit test's test names.
|
||
- gallium/osmesa: Fix MakeCurrent of non-8888 contexts.
|
||
- gallium/osmesa: Fill out other format tests.
|
||
- gallium/osmesa: Try to fix the test for big-endian.
|
||
- util: Make helper functions for pack/unpacking pixel rows.
|
||
- mesa/st: Use direct util_format_pack/unpack instead of u_tile.
|
||
- gallium/util: Remove pipe_get_tile_z/put_tile_z.
|
||
- softpipe: Drop the raw_to\* part of the tile cache interface.
|
||
- softpipe: Refactor pipe_get/put_tile_rgba\_\* paths.
|
||
- gallium: Add and use a helper for packing uc from a color_union.
|
||
- gallium: Refactor some single-pixel util_format_read/writes.
|
||
- util: Drop unpacking from int signed to unsigned and vice versa.
|
||
- freedreno: Move the layout debug under FD_MESA_DEBUG=layout.
|
||
- freedreno: Include the layer size in layout debug.
|
||
- freedreno: Rename the UBWC layer size field and store it as bytes.
|
||
- freedreno/a6xx: Disable the core layer-size setup.
|
||
- freedreno: Swap the whole resource layout in shadowing.
|
||
- freedreno: Blit all array levels when uncompressing UBWC.
|
||
- freedreno: Disable UBWC on Z24S8 if not TEXTURE_2D.
|
||
- freedreno: Allow UBWC on textures with multiple mipmap levels.
|
||
- mesa: Clean up some endianness adapters for shader image formats.
|
||
- intel/isl: Move iris's pipe-to-isl format function to isl.
|
||
- glsl,nir: Switch the enum representing shader image formats to
|
||
PIPE_FORMAT.
|
||
- mesa/st: Move the SYSTEM_VALUE -> TGSI_SEMANTIC map to
|
||
tgsi_from_mesa.
|
||
- nouveau: Reuse tgsi_get_sysval_semantic().
|
||
- nouveau: reuse tgsi_get_gl_frag_result_semantic().
|
||
- nouveau: Reuse tgsi_get_gl_varying_semantic().
|
||
- u_tile: Skip the packed temporary and just store tiles directly.
|
||
- ci: Disable a bunch of tests on freedreno a630.
|
||
- ci: Bump the GLES CTS version to 3.2.6.1.
|
||
- Revert "gallium: Fix big-endian addressing of non-bitmask array
|
||
formats."
|
||
- ci: Extend the a630 flake list to reduce spurious failures.
|
||
- radv: Squelch possibly-undefined warning
|
||
- llvmpipe: Fix real uninitialized use of "atype" for SEMANTIC_FACE
|
||
- llvmpipe: Silence "possibly uninitialized value" warning for
|
||
ssbo_limit.
|
||
- llvmpipe: Silence uninitialized variable warning about "chan"
|
||
- llvmpipe: Fix warning about uninitialized "op" in the NIR path.
|
||
- llvmpipe: Silence uninitialized variable warning about "vals"
|
||
- llvmpipe: Silence uninitialized variable warning about "scissor"
|
||
- llvmpipe: Fix another uninitialized value warning, on init_val.
|
||
- gallium: Only define PIPE_ALIGNSTACK on x86.
|
||
- ci: prepare-artifacts: Make the indent here match previously in the
|
||
file
|
||
- ci: Make sure that we have a proper shell prompt for LAVA.
|
||
- ci: Make LAVA job fails emit the full list of unexpected test
|
||
results.
|
||
- ci: Document how LAVA runners work.
|
||
- ci: Don't bother generating deqp junit results since we don't present
|
||
it.
|
||
- ci: Remove a useless filtering of the lava logs.
|
||
- nir: Rename gl_nir_lower_bindless_images.c in preparation for
|
||
extending it.
|
||
- nir: Make image lowering optionally handle the !bindless case as
|
||
well.
|
||
- gallium: Add a cap for enabling lowering of image load/store
|
||
intrinsics.
|
||
- v3d: Ask the state tracker to lower image accesses off of derefs.
|
||
- glsl: Factor out the sampler dim coordinate components switch
|
||
statement.
|
||
- spirv_to_nir: Reuse glsl_sampler_dim_coordinate_components().
|
||
- freedreno/ir3: Reuse glsl_get_sampler_dim_coordinate_components() in
|
||
tex_info.
|
||
- tgsi_to_nir: Reuse glsl_get_sampler_dim_coordinate_components().
|
||
- prog_to_nir: Reuse glsl_get_sampler_dim_coordinate_components().
|
||
- freedreno/ir3: Fix the arg to
|
||
ir3_get_num_components_for_image_format()
|
||
- nir: Move intel's intrinsic_image_coordinate_components() to core
|
||
nir.
|
||
- freedreno: Switch to using lowered image intrinsics.
|
||
- ci: Blacklist another freedreno flaky test.
|
||
- meson: Disable bison's -Wdeprecated since we still support old bison.
|
||
- turnip: Fix compiler warning about casting a nondispatchable handle.
|
||
- freedreno/computerator: Fix defined-but-not-used warnings from
|
||
lex/yacc.
|
||
- ci: Remove LLVM from ARM test drivers.
|
||
- ci: Stop disabling ACPI in the LAVA arm64 kernel build.
|
||
- ci: Shrink the arm64 kernel build a bit.
|
||
- ci: Include db410c support in the ARM container.
|
||
- aco: Fix signed-vs-unsigned warning.
|
||
- ci: Enable -Werror on meson-vulkan and meson-testing.
|
||
- ci: Switch testing on db410c over to LAVA.
|
||
- ci: Add a disabled-by-default job for GLES3 testing on db410c.
|
||
- ci: Flip db410c back to docker mode.
|
||
- ci: Print the renderer/version that our dEQP invocation is using.
|
||
- ci: Fix installation of firmware for db410c's nic.
|
||
- ci: Make a simple little bare-metal fastboot mode for db410c.
|
||
- glsl/tests: Catch mkdir errors to help explain when they happen.
|
||
- glsl/tests: Fix waiting for disk_cache_put() to finish.
|
||
- ci: Update the ci-templates commit.
|
||
- ci: Enable ccache in the container builds.
|
||
- ci: Enable ccaching of CMake builds as well.
|
||
- ci: Enable testing GLES2-3 on a530 (Dragonboard 820c).
|
||
- freedreno/a5xx: Fix min-vs-mag filtering decisions on non-mipmap tex.
|
||
- gallium/util: Switch util_float_to_half to \_mesa_float_to_half()'s
|
||
impl.
|
||
- ci: Ban the recent popular freedreno a630 flakes.
|
||
- ci: Disable tests that showed intermittent fails on a530 in day 1.
|
||
- ci: Only run the freedreno baremetal tests when freedreno/core
|
||
changes.
|
||
- freedreno: Switch to exposing only half-integer pixel centers.
|
||
- ci: Move db820c and db410c's gles3 tests to manual, like radv did.
|
||
- glsl: Restore the IsES flag on the shader when reading from cache.
|
||
- ci: Ban the recent popular freedreno a630 intermittent failure.
|
||
- freedreno: Remove always-true return from per-gen begin_query.
|
||
- freedreno: Remove the "active" member of queries.
|
||
- freedreno: Fix acc query handling in the presence of batch
|
||
reordering.
|
||
- freedreno: Associate the acc query bo with the batch.
|
||
- freedreno: Count blits in GL_TIME_ELAPSED and perf counter queries.
|
||
- freedreno/a6xx: Fix timestamp queries.
|
||
- freedreno: Rename "is_blit" to "is_discard_blit"
|
||
- freedreno: Fix detection of being in a blit for acc queries.
|
||
- freedreno: Work around UBWC flakiness.
|
||
- freedreno: Drop an unnecessary include marked "this should go away"
|
||
- freedreno/turnip: Use the NIR info to decide if we need helper
|
||
invocations.
|
||
- loader: Warn when we fail to open a device node due to permissions.
|
||
- ci: Consistently use -j4 across x86 build jobs and -j8 on ARM.
|
||
- freedreno/a6xx: Sink the per-level size temps inside the loop.
|
||
- freedreno/a6xx: Remove the "aligned_height" temporary.
|
||
- freedreno/a6xx: Drop the "alignment" layout temporary.
|
||
- freedreno: Add the outline of a test for a6xx texture layout.
|
||
- freedreno/a6xx: Set a level's pitch based on minified level0 pitch,
|
||
not width0.
|
||
- freedreno: Fix leak of binning shader variants.
|
||
- freedreno/ir3: Stop doing b2n on the SEL condition.
|
||
- freedreno/ir3: CSE the up/downconversion of SEL's cond's size.
|
||
- freedreno/a5xx+: Skip compiling the old gmem blit programs.
|
||
- freedreno/drm-shim: Add support for faking other adreno chips.
|
||
- freedreno/ir3: Drop handling FRAG_RESULT_DEPTH writing to .z
|
||
- freedreno: Introduce a "cpp_shift" value for cpp divs/muls.
|
||
- freedreno: Make the slice pitch be bytes, not pixels.
|
||
- drm-shim: Let the driver choose to overwrite the first render node.
|
||
- nir/lower_two_sided_color: Fix picking of new driver location.
|
||
- nir/lower_clip: Fix picking of unused driver locations.
|
||
- gallium: Fix setup of pstipple frag coord var.
|
||
- freedreno/ir3: Fix driver_location of the added vertex_flags varying.
|
||
- freedreno/ir3: Fix sizing of the inputs/outputs array.
|
||
- vc4: Use NIR shader's num_outputs for generating our new output.
|
||
- ci: Drop redundant freedreno stage specification.
|
||
- ci: Enable GLES3 testing on db410c/db820c (freedreno a306 and a530).
|
||
- freedreno: Fix derivatives without texturing on a3xx-a5xx.
|
||
- ci: Enable GLES 3.1 testing on db820c (a530).
|
||
- freedreno/ir3: Fix the disasm of half-float STG dests.
|
||
- freedreno/ir3: Print a space after nop counts, like qcom's disasm.
|
||
- freedreno/ir3: Add a unit test for our disassembler.
|
||
- freedreno/ir3: Convert remaining disasm src prints to reginfo.
|
||
- freedreno/ir3: Refactor out print_reg_src().
|
||
- freedreno/ir3: Add support for disasm of cat2 float32 immediates.
|
||
- ci: Enable --compact-display false on all dEQP runs.
|
||
- ci: Add sanity checking that dEQP gets the expected GL_RENDERER.
|
||
- freedreno: Fix calculation of the const buffer cmdstream size.
|
||
- ci: Allow namespacing of dEQP run results files.
|
||
- ci: Clean up some excessive use of pipes in dEQP results processing.
|
||
- ci/freedreno: Add a test run of a few driver options.
|
||
- util/ra: Sanity check that the driver selected a valid reg.
|
||
- util/ra: Sanity check that we're adding a valid reg to a class.
|
||
- util/ra: Use util_dynarray for the adjacency list.
|
||
- util/ra: Use util_dynarray for handling the conflict lists.
|
||
- util/ra: Improve ra_set_finalize() performance.
|
||
|
||
Eric Engestrom (58):
|
||
|
||
- VERSION: bump after 20.0 branch point
|
||
- egl: put full path to libEGL_mesa.so in GLVND json
|
||
- gitlab-ci: disable a630 tests as mesa-cheza is down
|
||
- util/os_socket: fix header unavailable on windows
|
||
- freedreno/perfcntrs: fix fd leak
|
||
- dri: delete gen-symbol-redefs.py
|
||
- util/disk_cache: check for write() failure in the zstd path
|
||
- meson: don't bother trying \`python2\`
|
||
- Revert "egl: put full path to libEGL_mesa.so in GLVND json"
|
||
- egl: directly access static members instead of using
|
||
\_egl{Get,Set}ConfigKey()
|
||
- meson: explicitly disallow unsupported build directory layout
|
||
- docs: fix typos in the release docs
|
||
- bin/gen_release_notes.py: fix commit list command
|
||
- gen_release_notes: fix vulkan version reported
|
||
- docs/relnotes/19.3: fix vulkan version reported
|
||
- docs/relnotes/20.0: fix vulkan version reported
|
||
- Revert "docs/relnotes/19.3: fix vulkan version reported"
|
||
- docs: trivial fix for html structure
|
||
- docs/releasing: add missing </li> tags
|
||
- docs: add release notes for 19.3.5
|
||
- docs: update calendar, add news item, and link releases notes for
|
||
19.3.5
|
||
- vulkan/wsi: fix cleanup when dup() fails
|
||
- gen_release_notes: fix version in "you should wait" message
|
||
- gen_release_notes: resolve ambiguity by renaming \`version\` to
|
||
\`previous_version\` and \`next_version\` to \`this_version\`
|
||
- meson: use existing variables in inc_common
|
||
- meson: inline \`inc_common\`
|
||
- vulkan: drop unused include directories
|
||
- intel: drop unused include directories
|
||
- scons: prune unused Makefile.sources
|
||
- docs: add release notes for 20.0.3
|
||
- docs/relnotes: add sha256sum for 20.0.3
|
||
- docs: update calendar, add news item, and link releases notes for
|
||
20.0.3
|
||
- docs: add release notes for 20.0.4
|
||
- docs/relnotes: add sha256sum for 20.0.4
|
||
- docs: update calendar, add news item, and link releases notes for
|
||
20.0.4
|
||
- glx: fix 630 times -Wlto-type-mismatch when building with LTO enabled
|
||
- glx: use anonymous namespace to avoid -Wodr issues when building with
|
||
LTO enabled
|
||
- pick-ui: auto-scroll the feedback window
|
||
- pick-ui: compute .pick_status.json path only once
|
||
- pick-ui: make .pick_status.json path relative to the git root instead
|
||
of the script
|
||
- pick-ui: show commit sha in the pick list
|
||
- VERSION: bump to 20.1.0-rc1
|
||
- .pick_status.json: Update to af55bdd05d94eda59ee1c9331a50045000da5db5
|
||
- .pick_status.json: Update to 57796946985de60204189426ca8eb7bbfa97c396
|
||
- .pick_status.json: Mark 3fac55ce0d066d767d6c6c8308f79d0c3e566ec0 as
|
||
denominated
|
||
- .pick_status.json: Update to 29da52128090a1ef8ef782188c0f67c7f5ec8d19
|
||
- VERSION: bump to 20.1.0-rc2
|
||
- .pick_status.json: Update to 772b15ad3227e08bb4e18932ac9ecf4c29271160
|
||
- .pick_status.json: Update to 56f955e4850035d915a2a87e2ebea7fa66ab5e19
|
||
- .pick_status.json: Update to c1c0cf7a66905e8d7ad506842a41b0ad0c5b10da
|
||
- VERSION: bump to 20.1.0-rc3
|
||
- .pick_status.json: Update to 5a6beb6a24aa084adfd6c57edd0a64f0a044611a
|
||
- post_version.py: fix branch name construction for release candidates
|
||
- post_version.py: invert \`is_point\` into \`is_first_release\` to
|
||
make its purpose clearer
|
||
- post_version.py: stop adding release candidates to the index and
|
||
relnotes
|
||
- VERSION: bump to 20.1.0-rc4
|
||
- .pick_status.json: Update to a91306677c613ba7511b764b3decc9db42b24de1
|
||
- tree-wide: fix deprecated GitLab URLs
|
||
|
||
Erik Faye-Lund (154):
|
||
|
||
- zink: enable texture-buffer objects
|
||
- zink: implement load_instance_id
|
||
- zink: implement support for derivative-control
|
||
- zink: be more careful about the mask-check
|
||
- zink: disallow depth-stencil blits with format-change
|
||
- st/mesa: use uint-result for sampling stencil buffers
|
||
- zink: lower away fdph
|
||
- zink: fixup sampler-usage
|
||
- zink: replace unset buffer with a dummy-buffer
|
||
- zink: emit blend-target index
|
||
- zink: only inspect dual-src limit if feature enabled
|
||
- Revert "nir: Add a couple trivial abs optimizations"
|
||
- zink: do not use SpvDimRect
|
||
- zink: fix binding-usage
|
||
- zink: do not report texture-samplers for unsupported stages
|
||
- zink/spirv: do not reinvent store_dest
|
||
- zink/spirv: prefer store_dest over store_dest_uint
|
||
- zink/spirv: rename functions a bit
|
||
- zink/spirv: unit_value -> raw_value
|
||
- zink/spirv: uint -> raw
|
||
- zink: do not convert bools to/from uint
|
||
- util: promote u_debug_memory.c to src/util
|
||
- util: move debug_memory_{begin,end} to os_memory_debug.h
|
||
- gallium/util: do not use debug_print_format
|
||
- gallium/util: remove unused debug_print_foo helpers
|
||
- zink/spirv: do not use bitwise operations on booleans
|
||
- pipebuffer: clean up cast-warnings
|
||
- rbug: clean up cast-warnings
|
||
- rbug: do not return void-value
|
||
- vtn/opencl: fully enable OpenCLstd_Clz
|
||
- compiler/nir: move build_exp helper into builtin-builder
|
||
- compiler/nir: move build_log helper into builtin-builder
|
||
- vtn/opencl: add native exp/log-support
|
||
- vtn/opencl: add native exp10/log10-support
|
||
- vtn/opencl: add native exp2/log2-support
|
||
- nv50: remove unused variable
|
||
- meson: disable some more warnings on msvc
|
||
- mesa/main: correct extension-checks for GL_BLACKHOLE_RENDER_INTEL
|
||
- mesa/main: clean-up extension-checks for point-sprites
|
||
- mesa/main: clean up extension-check for GL_VERTEX_PROGRAM
|
||
- mesa/main: clean up extension-check for GL_VERTEX_PROGRAM_TWO_SIDE
|
||
- mesa/main: clean up extension-check for GL_VERTEX_PROGRAM_POINT_SIZE
|
||
- mesa/main: clean up extension-check for GL_TEXTURE_RECTANGLE
|
||
- mesa/main: clean up extension-check for GL_STENCIL_TEST_TWO_SIDE
|
||
- mesa/main: clean up extension-check for GL_DEPTH_BOUNDS_TEST
|
||
- mesa/main: clean up extension-check for AMD_depth_clamp_separate
|
||
- mesa/main: clean up extension-check for GL_FRAGMENT_SHADER_ATI
|
||
- mesa/main: clean up extension-check for GL_TEXTURE_CUBE_MAP_SEAMLESS
|
||
- mesa/main: clean up extension-check for GL_RASTERIZER_DISCARD
|
||
- mesa/main: clean up extension-check for GL_TEXTURE_EXTERNAL
|
||
- mesa/main: remove unused macro
|
||
- wgl: drop pointless debug_printf
|
||
- wgl: drop unused member
|
||
- wgl: move screen-init to a helper
|
||
- wgl: do not create screen from DllMain
|
||
- st/dri: make sure software color-buffers are linear
|
||
- zink: be less picky about tiled resources
|
||
- .mailmap: add an alias for Alan Swanson
|
||
- .mailmap: add an alias for Alyssa Rosenzweig
|
||
- .mailmap: add an alias for Andrii Simiklit
|
||
- .mailmap: add an alias for Anuj Phogat
|
||
- .mailmap: add an alias for Axel Davy
|
||
- .mailmap: add an alias for Boris Brezillon
|
||
- .mailmap: add an alias for Bruce Cherniak
|
||
- .mailmap: update aliases for Carl-Philip Hänsch
|
||
- .mailmap: add an alias for Chad Versace
|
||
- .mailmap: add a couple of aliases for Chandu Babu Namburu
|
||
- .mailmap: add alias for Chenglei Ren
|
||
- .mailmap: add an alias for Christian Gmeiner
|
||
- .mailmap: add an alias for Christian Inci
|
||
- .mailmap: add a few aliases for Christoph Haag
|
||
- .mailmap: add an alias for Colin McDonald
|
||
- .mailmap: specify spelling for Constantine Kharlamov
|
||
- .mailmap: add an alias for Craig Stout
|
||
- .mailmap: add an alias for Daniel Schürmann
|
||
- .mailmap: add an alias for Danylo Piliaiev
|
||
- .mailmap: add an alias for Dave Airlie
|
||
- .mailmap: add an alias for Dylan Baker
|
||
- .mailmap: add a couple of aliases for Dylan Noblesmith
|
||
- .mailmap: add an alias for Emmanuel Gil Peyrot
|
||
- .mailmap: add an alias for Erik Faye-Lund
|
||
- .mailmap: specify spelling for Francesco Ansanelli
|
||
- .mailmap: specify spelling for Gurchetan Singh
|
||
- .mailmap: add an alias for Haihao Xiang
|
||
- .mailmap: add an alias for Harish Krupo
|
||
- .mailmap: specify spelling for Heinrich Fink
|
||
- .mailmap: specify spelling for Henri Verbeet
|
||
- .mailmap: add an alias for Igor Gnatenko
|
||
- .mailmap: add an alias for Illia Iorin
|
||
- .mailmap: specify spelling for James Zhu
|
||
- .mailmap: add an alias for Jan Beich
|
||
- .mailmap: clean up aliases for Jeremy Huddleston
|
||
- .mailmap: add an alias for Julien Isorce
|
||
- .mailmap: add a few aliases for Karol Herbst
|
||
- .mailmap: add a few aliases for Kevin Rogovin
|
||
- .mailmap: add a few aliases for Kristian Høgsberg
|
||
- .mailmap: add an alias for Lionel Landwerlin
|
||
- .mailmap: specify spelling for Liviu Prodea
|
||
- .mailmap: update aliases for Marc-André Lureau
|
||
- .mailmap: add alias for Matthias Groß
|
||
- .mailmap: add an alias for Neha Bhende
|
||
- .mailmap: add an alias for Neil Roberts
|
||
- .mailmap: specify spelling for Nian Wu
|
||
- .mailmap: add an alias for Nicholas Bishop
|
||
- .mailmap: update aliases for Nicolai Hähnle
|
||
- .mailmap: add an alias for Philipp Zabel
|
||
- .mailmap: update aliases for Pierre-Eric Pelloux-Prayer
|
||
- .mailmap: add an alias for Plamena Manolova
|
||
- .mailmap: add an alias for Qiang Yu
|
||
- .mailmap: specify spelling for Randy Xu
|
||
- .mailmap: add an alias for Renato Caldas
|
||
- .mailmap: add an alias for Rob Clark
|
||
- .mailmap: add an alias for Rodrigo Vivi
|
||
- .mailmap: add an alias for Samuel Li
|
||
- .mailmap: add an alias for Sergii Romantsov
|
||
- .mailmap: specify spelling for Sonny Jiang
|
||
- .mailmap: add a couple of aliases for Steinar H. Gunderson
|
||
- .mailmap: add a couple of aliases for Suresh Guttula
|
||
- .mailmap: add an alias for Thierry Reding
|
||
- .mailmap: add an alias for Timo Aaltonen
|
||
- .mailmap: add a couple of aliases for Timothy Arceri
|
||
- .mailmap: add an alias for Tim Wiederhake
|
||
- .mailmap: add an alias for Tom Stellard
|
||
- .mailmap: add an alias for Tomasz Figa
|
||
- .mailmap: add an alias for Topi Pohjolainen
|
||
- .mailmap: add an alias for Vadym Shovkoplias
|
||
- .mailmap: add an alias for Varad Gautam
|
||
- .mailmap: specify spelling for Vivek Kasireddy
|
||
- .mailmap: specify spelling for Wladimir J. van der Laan
|
||
- .mailmap: add an alias for Xavier Bouchoux
|
||
- .mailmap: add an alias for Yaakov Selkowitz
|
||
- .mailmap: add alias for Zhaowei Yuan
|
||
- .mailmap: add an alias for Zhongmin Wu
|
||
- meson: use override_options to change warning-level
|
||
- wgl: silence some cast-warnings
|
||
- util/tests: initialize variable
|
||
- mesa: fixup cast expression
|
||
- vbo: avoid including wingdi.h on win32
|
||
- meson: tell flex that we support c99
|
||
- gtest: Update to 1.10.0
|
||
- meson: do not disable incremental linking for debug-builds
|
||
- docs: remove outdated sentence
|
||
- mesa/gallium: do not use enum for bit-allocated member
|
||
- meson: correct windows-version define
|
||
- mesa/main: do not store unrecognized extensions in context
|
||
- mesa/main: do not pass context to one-time extension init
|
||
- mesa/main: do not init remap-table per api
|
||
- mesa/main: Do not pass context to one_time_init
|
||
- mesa/main: one_time_init() -> \_mesa_initialize()
|
||
- mesa/st: call \_mesa_initialize() early
|
||
- zink: lower b2b to b2i
|
||
- util/os_memory: never use os_memory_debug.h
|
||
- zink: implement i2b1
|
||
- zink: use general-layout when blitting to/from same resource
|
||
|
||
Francisco Jerez (57):
|
||
|
||
- intel/fs/cse: Make HALT instruction act as CSE barrier.
|
||
- intel/fs/gen7: Fix fs_inst::flags_written() for
|
||
SHADER_OPCODE_FIND_LIVE_CHANNEL.
|
||
- intel/fs: Add virtual instruction to load mask of live channels into
|
||
flag register.
|
||
- intel/fs/gen12: Workaround unwanted SEND execution due to broken
|
||
NoMask control flow.
|
||
- intel/fs/gen12: Fixup/simplify SWSB annotations of SIMD32 scratch
|
||
writes.
|
||
- intel/fs/gen12: Workaround data coherency issues due to broken NoMask
|
||
control flow.
|
||
- intel/fs: Set src0 alpha present bit in header when provided in
|
||
message payload.
|
||
- intel/fs/gen11: Work around dual-source blending hangs in combination
|
||
with SIMD32.
|
||
- intel/fs: Make sample_mask_reg() local to brw_fs.cpp and use it in
|
||
more places.
|
||
- intel/fs: Use helper for discard sample mask flag subregister number.
|
||
- intel/fs/gen7+: Swap sample mask flag register and FIND_LIVE_CHANNEL
|
||
temporary.
|
||
- intel/fs: Refactor predication on sample mask into helper function.
|
||
- intel/fs: Return consistent UW types from sample_mask_reg() in
|
||
fragment shaders.
|
||
- intel/fs/gen7+: Implement discard/demote for SIMD32 programs.
|
||
- intel/compiler: Move base IR definitions into a separate header file
|
||
- intel/compiler: Reverse inclusion dependency between brw_cfg.h and
|
||
brw_shader.h
|
||
- intel/compiler: Nest definition of live variables block_data
|
||
structures
|
||
- intel/compiler: Reverse inclusion dependency between
|
||
brw_fs_live_variables.h and brw_fs.h
|
||
- intel/compiler: Reverse inclusion dependency between
|
||
brw_vec4_live_variables.h and brw_vec4.h
|
||
- intel/compiler: Introduce simple IR analysis pass framework
|
||
- intel/compiler: Introduce backend_shader method to propagate IR
|
||
changes to analysis passes
|
||
- intel/compiler: Define more detailed analysis dependency classes
|
||
- intel/compiler: Pass detailed dependency classes to
|
||
invalidate_analysis()
|
||
- intel/compiler: Mark virtual_grf_interferes and vars_interfere as
|
||
const
|
||
- intel/compiler: Move all live interval analysis results into
|
||
fs_live_variables
|
||
- intel/compiler: Move all live interval analysis results into
|
||
vec4_live_variables
|
||
- intel/compiler: Restructure live intervals computation code
|
||
- intel/compiler: Pass single backend_shader argument to the
|
||
fs_live_variables constructor
|
||
- intel/compiler: Pass single backend_shader argument to the
|
||
vec4_live_variables constructor
|
||
- intel/compiler/fs: Add live interval validation pass
|
||
- intel/compiler/vec4: Add live interval validation pass
|
||
- intel/compiler/fs: Switch liveness analysis to IR analysis framework
|
||
- intel/compiler/vec4: Switch liveness analysis to IR analysis
|
||
framework
|
||
- intel/compiler: Drop invalidate_live_intervals()
|
||
- intel/compiler: Move idom tree calculation and related logic into
|
||
analysis object
|
||
- intel/compiler: Move dominance tree data structure into idom_tree
|
||
object
|
||
- entel/compiler: Simplify new_idom reduction in dominance tree
|
||
calculation
|
||
- intel/compiler: Move register pressure calculation into IR analysis
|
||
object
|
||
- intel/compiler: Calculate num_instructions in O(1) during register
|
||
pressure calculation
|
||
- intel/fs: Fix workaround for VxH indirect addressing bug under
|
||
control flow.
|
||
- intel/fs/gen12: Fix interaction of SWSB dependency combination with
|
||
EU fusion workaround.
|
||
- intel/fs/gen12: Fix hangs with per-sample SIMD32 fragment shader
|
||
dispatch.
|
||
- intel/fs/gen12: Work around dual-source blending hangs in combination
|
||
with SIMD32.
|
||
- intel/fs/gen12: Fix Render Target Read header setup for new thread
|
||
payload layout.
|
||
- intel/ir: Add missing initialization of backend_reg::offset during
|
||
construction.
|
||
- intel/fs: Rename half() helpers to quarter(), allow index up to 3.
|
||
- intel/fs: Fix constness of argument of
|
||
fs_instruction_scheduler::is_compressed().
|
||
- intel/fs: Replace fs_visitor::bank_conflict_cycles() with stand-alone
|
||
function.
|
||
- intel/vec4: Fix constness of vec4_instruction::reads_flag() and
|
||
::writes_flag().
|
||
- intel/ir: Import shader performance analysis pass.
|
||
- intel/fs: Heap-allocate fs_visitors in brw_compile_fs().
|
||
- intel/fs: Implement performance analysis-based SIMD32 heuristic for
|
||
fragment shaders.
|
||
- intel/fs: Add INTEL_DEBUG=no32 debugging flag.
|
||
- intel/ir: Use brw::performance object instead of CFG cycle counts for
|
||
codegen stats.
|
||
- intel/ir: Pass block cycle count information explicitly to
|
||
disassembler.
|
||
- intel/ir: Remove scheduling-based cycle count estimates.
|
||
- intel/ir: Update performance analysis parameters for memory fence
|
||
codegen changes.
|
||
|
||
Fritz Koenig (3):
|
||
|
||
- Revert "gitlab-ci: disable a630 tests as mesa-cheza is down"
|
||
- Revert "gitlab-ci: disable a630 tests as mesa-cheza is down (again)"
|
||
- freedreno: allow FMT6_8_UNORM as a UBWC format
|
||
|
||
Georg Lehmann (3):
|
||
|
||
- Correctly wait in the fragment stage until all semaphores are
|
||
signaled
|
||
- Vulkan Overlay: Don't try to change the image layout to present twice
|
||
- Vulkan overlay: use the corresponding image index for each swapchain
|
||
|
||
Gert Wollny (63):
|
||
|
||
- r600: force new CF with TEX only if any texture value is written
|
||
- r600: Increase space for IO values to agree with
|
||
PIPE_MAX_SHADER_IN/OUTPUTS
|
||
- r600: Add NIR compiler options
|
||
- r600: Update state code to accept NIR shaders
|
||
- r600/sfn: Add a basic nir shader backend
|
||
- r600: enable NIR backend DEBUG flag for supported architectures
|
||
- r600/sfn: Add the VS in and FS out vectorization
|
||
- r600/sfn: Add the WaitAck instruction
|
||
- r600/sfn: add live range evaluation for the GPR
|
||
- r600/sfn: add register remapping
|
||
- r600/sfn: Add lowering arrays to scratch and according instructions
|
||
- r600/sfn: Add a load GDS result instruction
|
||
- r600/sfn: Add MemRingOut instructions
|
||
- r600/sfn: add emitVertex instructions
|
||
- r600/sfn: Add support for geometry shader
|
||
- r600/sfn: Add VS for TCS shader skeleton
|
||
- r600/sfn: Add compute shader skeleton
|
||
- r600/sfn: Add GDS instructions
|
||
- r600/sfn: Add lowering UBO access to r600 specific codes
|
||
- r600: Make sure LLVM is not used for DRAW
|
||
- r600/sfn: Add support for atomic instructions
|
||
- r600/sfn: Add support for SSBO load and store
|
||
- r600/sfn: Add .editorconfig file
|
||
- r600/sfn: Add some documentation
|
||
- r600/sfn: Avoid using dynamic_cast to identify type
|
||
- r600/sfn: Use static_cast when type is already known
|
||
- r600/sfn: Don't try to catch exceptions, the driver doesn't throw any
|
||
- gallium/tgsi_to_nir: Set nir_intrinsic_align_mul to 16 and offset to
|
||
0
|
||
- r600: Dump a few more variables when requested
|
||
- r600/sfn: Reduce array limit for scratch usage
|
||
- r600/sfn: Fix setting alignments when lowering UBOs
|
||
- r600/sfn: Implementing instructions blocks
|
||
- r600/nir: Pin interpolation results to channel
|
||
- r600/sfn: Fix null pointer deref in live range evalation
|
||
- r600/sfn: Handle b2b1 like it was a mov
|
||
- r600/sfn: Fix handling of GS inputs
|
||
- r600/sfn: Fix using the result of a fetch instruction in next fetch
|
||
- r600/sfn: Count only literals that are not inline to split
|
||
instruction groups
|
||
- r600/sfn: use new temp register allocation when loading single value
|
||
temporaries
|
||
- nir: Add r600 specific intrinsics for tesselation shader IO
|
||
- nir: Add umad24 and umul24 opcodes
|
||
- r600: Handle texcoord semantics in LDS index evaluation
|
||
- r600/sfn: simplify UBO lowering pass
|
||
- r600/sfn: Don't emit inline constants in the r600 IR
|
||
- r600/sfn: Add LDS IO instructions to r600 IR
|
||
- r600/sfn: Add LDS instruction to assembly conversion
|
||
- r600/sfn: Add TF write instruction
|
||
- r600/sfn: Add IR instruction to fetch the TESS parameters
|
||
- r600/sfn: Handle umul24 and umad24
|
||
- r600/sfn: Emit some LDS instructions
|
||
- r600/sfn: Move emission of barrier from compute shader to shader base
|
||
- r600/sfn: Add methods to valuepool to get a vector of values
|
||
- r600/sfn: Move some shader base methods to the public interface
|
||
- r600/sfn: extract class to handle the VS export to different stages
|
||
- r600/sfn: derive the GS from the vertex stage for a common interface
|
||
- r600/sfn: Handle LDS output in VS
|
||
- r600/sfn: Move removing of unused variables
|
||
- r600/sfn: Add lowering passes for Tesselation IO
|
||
- r600/sfn: Add tesselation shaders
|
||
- r600: Enable tesselation for NIR
|
||
- r600: Fix nir compiler options, i.e. don't lower IO to temps for TESS
|
||
- r600/sfn: Fix printing vertex fetch instruction flags
|
||
- r600: Fix duplicated subexpression in r600_asm.c
|
||
|
||
Greg V (3):
|
||
|
||
- amd/addrlib: fix build on non-x86 platforms
|
||
- r600: add missing <array> include
|
||
- svga: fix build on FreeBSD
|
||
|
||
H.J. Lu (2):
|
||
|
||
- x86_init_func_common: Add ENDBR at function entry
|
||
- x86: Add ENDBR at function entries
|
||
|
||
Hanno Böck (1):
|
||
|
||
- Properly check mmap return value
|
||
|
||
Hyunjun Ko (27):
|
||
|
||
- freedreno/ir3: fix printing half constant registers.
|
||
- freedreno/ir3: Add cat4 mediump opcodes
|
||
- freedreno/ir3: put the conversion back for half const to the right
|
||
place.
|
||
- freedreno/ir3: Fold const only when the type is float
|
||
- freedreno/ir3: Add new ir3 pass to fold out fp16 conversions
|
||
- nir: Add optimization for doing removing f16/f32 conversions
|
||
- freedreno/ir3: handle half registers for arrays during register
|
||
allocation.
|
||
- turnip: support indirect draw
|
||
- glsl: Handle fp16 unary operations when lowering matrix operations
|
||
- glsl/lower_instructions: Handle fp16 for MOD_TO_FLOOR
|
||
- turnip: Gather information for transform feedback
|
||
- turnip: Define structs for transform feedback
|
||
- turnip: Setup stream-output when linking program
|
||
- turnip: Implement stream-out emit and vkApis for transform feedback
|
||
- turnip: Implement an empty function vkCmdDrawIndirectByteCountEXT
|
||
- turnip: Enable VK_EXT_transform_feedback
|
||
- turnip: Add tu6_control struct.
|
||
- turnip: Fix wrong assignment of xfb output's offset.
|
||
- turnip: Do gathering xfb info after nir_remove_dead_variables
|
||
- freedreno: Enable mediump lowering
|
||
- freedreno/ir3: enable nir_opt_loop_unroll on a6xx
|
||
- nir: fix wrong assignment to buffer in xfb_varyings_info
|
||
- turnip: make the struct slot_value of queries get 2 values
|
||
- turnip: Implement and enable
|
||
VK_QUERY_TYPE_TRANSFORM_FEEDBACK_STREAM_EXT
|
||
- turnip : Fix wrong offset calculation for xfb buffer.
|
||
- turnip: Skip unused regs when setting up streamout buffers
|
||
- turnip: Fix crashes when geometry shader constants aren't used
|
||
|
||
Iago Toral Quiroga (1):
|
||
|
||
- nir: add a bool bitsize lowering pass
|
||
|
||
Ian Romanick (62):
|
||
|
||
- intel/fs: Don't count integer instructions as being possibly coissue
|
||
- nir: Mark fmin and fmax as commutative and associative
|
||
- mesa/draw: Make sure all the unused fields are initialized to zero
|
||
- nir/search: Use larger type to hold linearized index
|
||
- intel/fs: Correctly handle multiply of fsign with a source modifier
|
||
- intel/fs: Do cmod prop again after scheduling
|
||
- intel/fs: Allow NOT instructions in conditional discard optimization
|
||
- intel/fs: Fix NULL destinations on 3-source instructions again after
|
||
late DCE
|
||
- nir/algebraic: Simplify logic to detect sign of an integer
|
||
- nir/algebraic: optimize ior(ine(a, 0), ine(b, 0)) to ine(ior(a, b),
|
||
0)
|
||
- nir/algebraic: Generalize some and-of-shift-right patterns [v2]
|
||
- nir/algebraic: Constant reassociation for bitwise operations too
|
||
- nir/algebraic: Simplify a contradiction that can occur in
|
||
\__flt64_nonnan
|
||
- soft-fp64/b2f: Reimplement using bitwise logic ops
|
||
- soft-fp64: Don't open-code umulExtended
|
||
- soft-fp64: Simplify \__countLeadingZeros32 function
|
||
- soft-fp64: Pick a single idiom for treating sign value as a Boolean
|
||
- soft-fp64: Store sign value as 0 or 0x80000000
|
||
- soft-fp64/fneg: Don't treat NaN specially
|
||
- soft-fp64/flt: Perform checks in a different order
|
||
- soft-fp64/fsat: Correctly handle NaN
|
||
- soft-fp64/fsat: Micro-optimize x < 0 test
|
||
- soft-fp64/fsat: Micro-optimize x >= 1 test
|
||
- soft-fp64: Relax the way NaN is propagated
|
||
- soft-fp64/ffloor: Simplify the >= 0 comparison
|
||
- soft-fp64: Optimize \__fmin64 and \__fmax64 by using different
|
||
evaluation order [v2]
|
||
- soft-fp64/fadd: Instead of tracking "b < a", track sign of the
|
||
difference
|
||
- soft-fp64/fadd: Massively split the live range of zFrac0 and zFrac1
|
||
- soft-fp64/fadd: Pick zero or non-zero result based on subtraction
|
||
result
|
||
- soft-fp64/fadd: Just let the subtraction happen when the result will
|
||
be zero
|
||
- soft-fp64/fadd: Delete a redundant condition check
|
||
- soft-fp64/fadd: Reformat after previous commit
|
||
- soft-fp64/fadd: Combine an if-statement into the preceeding
|
||
else-clause
|
||
- soft-fp64/fadd: Rename aFrac and bFrac variables
|
||
- soft-fp64/fadd: Use absolute value of expDiff
|
||
- soft-fp64/fadd: Move common code out of both branches of an
|
||
if-statement
|
||
- soft-fp64/fadd: Common code optimization for differing sign case
|
||
- soft-fp64: Split a block that was missing a cast on a comparison
|
||
- intel/vec4: Allow late copy propagation on vec4
|
||
- nir/algebraic: Change the default cursor location when replacing a
|
||
unary op
|
||
- nir/algebraic: Distribute source modifiers into instructions
|
||
- nir/algebraic: Use value range analysis to convert fmax to fsat
|
||
- nir/algebraic: Remove a redundant fabs pattern
|
||
- tnl: Don't dereference NULL obj pointer in bind_indices
|
||
- tnl: Don't dereference NULL obj pointer in replay_init
|
||
- tnl: Don't dereference NULL obj pointer in t_rebase_prims
|
||
- tnl: Silence unused parameter 'attrib' warning in
|
||
convert_half_to_float
|
||
- tnl: Silence unused parameter warnings in \_tnl_draw_prims
|
||
- tnl: Silence unused parameter warnings in dump_draw_info
|
||
- tnl: Silence unused parameter warnings in \_tnl_split_inplace
|
||
- tnl: Code formatting in t_draw.c
|
||
- tnl: Code formatting in t_rebase.c
|
||
- intel/compiler: Silence unused parameter warnings in vec4_tcs_visitor
|
||
- intel/compiler: Silence unused parameter warning in
|
||
fs_live_variables::setup_one_read
|
||
- intel/compiler: Silence unused parameter warning in
|
||
update_inst_scoreboard
|
||
- intel/compiler: Only GE and L modifiers are commutative for SEL
|
||
- intel/compiler: CSEL can do saturate
|
||
- intel/compiler: Fixup operands in fs_builder::emit() that takes array
|
||
- nir/algebraic: Detect some kinds of malformed variable names
|
||
- nir/algebraic: Require operands to iand be 32-bit
|
||
- nir/algebraic: Optimize ushr of pack_half, not ishr
|
||
- anv/tests: Don't rely on assert or changing NDEBUG in tests
|
||
|
||
Icecream95 (16):
|
||
|
||
- panfrost: Fix non-debug builds
|
||
- panfrost: Inline panfrost_get_default_swizzle
|
||
- panfrost: LogicOp support
|
||
- nir: Allow nir_format conversions to work on 32-bit values
|
||
- panfrost: LogicOp fixes and non 8-bit format support
|
||
- mesa/format_utils: Add a fast-path for RGBA to BGRA
|
||
- panfrost: Extend the tiled store fast-path to loads
|
||
- panfrost: Mark 64-bit formats as unsupported
|
||
- panfrost: Add support for B5G5R5X1
|
||
- st/mesa: Fall back on R3G3B2 for R3_G3_B2
|
||
- panfrost: Add support for R3G3B2
|
||
- panfrost: Correctly identify format 0x4c
|
||
- pan/midgard: Fix a divide by zero in emit_alu_bundle
|
||
- panfrost: Fix GL_EXT_vertex_array_bgra
|
||
- panfrost: Enable PIPE_CAP_VERTEX_COLOR_UNCLAMPED
|
||
- panfrost: Fix background showing when using discard
|
||
|
||
Icenowy Zheng (3):
|
||
|
||
- lima: remove its hash table entry when invalidating a resource
|
||
- lima: expose fragment shader derivatives capability
|
||
- lima: implement zsbuf reload
|
||
|
||
Ilia Mirkin (24):
|
||
|
||
- nv50: report max lod bias of 15.0
|
||
- gitlab-ci: disable panfrost runners
|
||
- mesa: fix \_mesa_draw_nonzero_divisor_bits to return nonzero divisors
|
||
- nv50,nvc0: add newly added PIPE_CAP's to list
|
||
- st/mesa: allow TXB2/TXL2 to work with cube array shadow textures
|
||
- nvc0: enable EXT_texture_shadow_lod
|
||
- st/vdpau: avoid asserting on new VDP_YCBCR\_\* formats
|
||
- st/vdpau: make query test for 2D support
|
||
- nv50: don't try to upload MSAA settings for BUFFER textures
|
||
- gallium: add viewport swizzling state and cap
|
||
- mesa: add GL_NV_viewport_swizzle support
|
||
- st/mesa: add NV_viewport_swizzle support
|
||
- nvc0: add NV_viewport_swizzle support for GM200+
|
||
- compiler: add VARYING_SLOT_VIEWPORT_MASK
|
||
- glsl: add NV_viewport_array2 support
|
||
- mesa: add NV_viewport_array2 enable, attach to glsl
|
||
- gallium: add TGSI_SEMANTIC_VIEWPORT_MASK
|
||
- gallium: add TGSI_PROPERTY_LAYER_VIEWPORT_RELATIVE
|
||
- gallium: add PIPE_CAP_VIEWPORT_MASK
|
||
- st/mesa: add support for GL_NV_viewport_array2
|
||
- nvc0: enable GL_NV_viewport_array2
|
||
- nv50,nvc0: update with latest caps
|
||
- docs: update for recently-added nvc0 features
|
||
- mesa: add interaction between compute derivatives and variable local
|
||
sizes
|
||
|
||
Indrajit Kumar Das (4):
|
||
|
||
- glapi/copyimage: Implement CopyImageSubDataNV
|
||
- gallium: prepare framework for supporting
|
||
AlphaToCoverageDitherControlNV
|
||
- mesa: add support for AlphaToCoverageDitherControlNV
|
||
- radeonsi: enable support for AlphaToCoverageDitherControlNV
|
||
|
||
Ivan Molodetskikh (1):
|
||
|
||
- egl: allow INVALID format for linux_dmabuf
|
||
|
||
James Xiong (2):
|
||
|
||
- iris: handle the failure of converting unsupported yuv formats to isl
|
||
- gallium: let the pipe drivers decide the supported modifiers
|
||
|
||
James Zhu (1):
|
||
|
||
- radeonsi: fix Segmentation fault during vaapi enc test
|
||
|
||
Jan Palus (1):
|
||
|
||
- targets/opencl: fix build against LLVM>=10 with Polly support
|
||
|
||
Jan Vesely (2):
|
||
|
||
- clover: Use explicit conversion from llvm::StringRef to std::string
|
||
- clover: Check if the detected clang libraries are usable
|
||
|
||
Jan Zielinski (8):
|
||
|
||
- gallium/swr: Fix various asserts and security issues
|
||
- gallium/swr: fix corruptions in Unigine Heaven
|
||
- gallium/swr: use ElementCount type arguments for getSplat()
|
||
- gallium/gallivm: Remove workaround disabling AVX code for newer CPUs
|
||
- gallium/gallivm: fix compilation issues with llvm 11
|
||
- gallium/gallivm: remove unused header include for newer LLVM
|
||
- gallium/swr: Fix LLVM 11 compilation issues
|
||
- gallium/swr: Fix crashes and failures in vertex fetch
|
||
|
||
Jason Ekstrand (202):
|
||
|
||
- genxml: Add a new 3DSTATE_SF field on gen12
|
||
- anv,iris: Set 3DSTATE_SF::DerefBlockSize to per-poly on Gen12+
|
||
- intel/genxml: Drop SLMEnable from L3CNTLREG on Gen11
|
||
- iris: Set SLMEnable based on the L3$ config
|
||
- iris: Store the L3$ configs in the screen
|
||
- iris: Use the URB size from the L3$ config
|
||
- i965: Re-emit l3 state before BLORP executes
|
||
- intel: Take a gen_l3_config in gen_get_urb_config
|
||
- intel/blorp: Always emit URB config on Gen7+
|
||
- iris: Consolodate URB emit
|
||
- anv: Emit URB setup earlier
|
||
- intel/common: Return the block size from get_urb_config
|
||
- intel/blorp: Plumb deref block size through to 3DSTATE_SF
|
||
- anv: Plumb deref block size through to 3DSTATE_SF
|
||
- iris: Plumb deref block size through to 3DSTATE_SF
|
||
- anv: Always fill out the AUX table even if CCS is disabled
|
||
- intel/eu/validate: Don't validate regions of sends
|
||
- intel/disasm: SEND has two sources on Gen12+
|
||
- intel/tools: Handle strides better when dumping buffers
|
||
- intel/fs: Write the address register with NoMask for MOV_INDIRECT
|
||
- anv/blorp: Use the correct size for vkCmdCopyBufferToImage
|
||
- anv: No-op submit and wait calls when no_hw is set
|
||
- anv: Reject modifiers on depth/stencil formats
|
||
- vulkan: Update the XML and headers to 1.2.133
|
||
- nir: Fix the nir_builder include path for nir_builtin_builder
|
||
- nir/builder: Return an integer from nir_get_texture_size
|
||
- intel/isl: Add isl_aux_info.c to Makefile.sources
|
||
- anv: Always enable the data cache
|
||
- nir: Drop nir_tex_instr::texture_array_size
|
||
- anv: Use the PIPE_CONTROL instead of bits for the CS stall W/A
|
||
- anv: Use a proper end-of-pipe sync instead of just CS stall
|
||
- anv: Do end-of-pipe sync around MCS/CCS ops instead of CS stall
|
||
- nir: Flush to zero with OOB low exponents in ldexp
|
||
- isl: Set 3DSTATE_DEPTH_BUFFER::Depth correctly for 3D surfaces
|
||
- iris: Allow HiZ on blit sources
|
||
- blorp: Write to depth/stencil images as depth/stencil when possible
|
||
- anv: Enable HiZ for VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL
|
||
- iris: Enable CCS for copies from HiZ+CCS depth buffers
|
||
- iris: Enable HiZ and stencil CCS for blorp blit destinations
|
||
- iris: Don't skip fast depth clears if the color changed
|
||
- anv: Parse VkPhysicalDeviceFeatures2 in CreateDevice
|
||
- anv: Mark max_push_range UNUSED and simplify the code
|
||
- anv: Pass buffer addresses into emit_push_constant\*
|
||
- anv: Delete some pointless break statements
|
||
- anv: Align UBO sizes to 32B
|
||
- anv: Add an align_down_u32 helper
|
||
- anv: Bounds-check pushed UBOs when robustBufferAccess = true
|
||
- vulkan/wsi: Don't leak the FD when
|
||
GetImageDrmFormatModifierProperties fails
|
||
- vulkan/wsi: Return an error if dup() fails
|
||
- intel/isl: Clean up some aux surface logic
|
||
- intel/isl: Add a separate ISL_AUX_USAGE_HIZ_CCS_WT
|
||
- intel/blorp: Allow HIZ_CCS_WT in copy sources
|
||
- iris: Use ISL_AUX_USAGE_HIZ_CCS_WT to indicate write-through HiZ
|
||
- intel/isl: Require ISL_AUX_USAGE_HIZ_CCS_WT for HZ+CCS WT mode
|
||
- intel/isl: Add a separate ISL_AUX_USAGE_STC_CCS
|
||
- intel/blorp: Allow STC_CCS in blit sources
|
||
- iris: Use ISL_AUX_USAGE_STC_CCS for stencil CCS
|
||
- intel: Require ISL_AUX_USAGE_STC_CCS for stencil CCS
|
||
- intel/isl: Set DepthStencilResource based on aux usage
|
||
- anv: Dump push ranges via VK_KHR_pipeline_executable_properties
|
||
- anv: Fix the comparison in an assert
|
||
- anv: Push UBO ranges relative to the start of the binding
|
||
- anv: Do an end-of-pipe sync before updating AUX table entries
|
||
- intel/isl: Don't align linear images to 64K on Gen12+
|
||
- intel/blorp: Add support for swizzling fast-clear colors
|
||
- anv: Swizzle fast-clear values
|
||
- intel/iris: Always initialize CCS to 0
|
||
- anv: Only add END_OF_PIPE_SYNC if we actually have AUX_INVAL
|
||
- util/sparse_array: Finish the sparse_array in the tests
|
||
- util/sparse_array: Add a node_size_log2 temporary
|
||
- meson,ci: Disable sparse_array tests on windows
|
||
- util/sparse_array: Stash the node level in the node pointer
|
||
- anv: Stop fetching the timestamp frequency ourselves
|
||
- intel/dump_gpu: Add an ensure_device_info helper
|
||
- intel/dump_gpu: Handle a bunch of getparam in the no-HW case
|
||
- intel/nir: Run copy-prop and DCE after lower_bool_to_int32
|
||
- nir: Add b2b opcodes
|
||
- aco: Implement b2b32 and b2b1
|
||
- nir: Use b2b opcodes for shared and constant memory
|
||
- nir: Insert b2b1s around booleans in nir_lower_to
|
||
- anv: Set alignments on descriptor and constant loads
|
||
- nir: Validate that memory load/store ops work on whole bytes
|
||
- nir: Set UBO alignments in lower_uniforms_to_ubo
|
||
- nir/opt_loop_unroll: Fix has_nested_loop handling
|
||
- nir/lower_int64: Lower 8 and 16-bit downcasts with nir_lower_mov64
|
||
- nir/algebraic: Add downcast-of-pack opts
|
||
- nir: Add a nir_op_is_vec helper
|
||
- nir: Copy propagate through vec8s and vec16s
|
||
- nir: Handle vec8/16 in bool_to_bitsize
|
||
- nir: Handle vec8/16 in gather_ssa_types
|
||
- nir: Handle vec8/16 in lower_phis_to_scalar
|
||
- nir: Handle vec8/16 in lower_regs_to_ssa
|
||
- nir: Handle vec8/16 in opt_split_alu_of_phi
|
||
- nir: Treat vec8/16 as select in opt_peephole_select
|
||
- nir: Handle vec8/16 in opt_undef_vecN
|
||
- nir: Handle vec8/16 in nir_shrink_array_vars
|
||
- anv: Account for the header in anv_state_stream_alloc
|
||
- anv/allocator: Use util_dynarray for blocks in anv_state_stream
|
||
- spirv: Implement OpCopyObject and OpCopyLogical as blind copies
|
||
- Revert "spirv: Implement OpCopyObject and OpCopyLogical as blind
|
||
copies"
|
||
- anv/image: Use align_u64 for image offsets
|
||
- nir/from_ssa: Only chain movs when a src is also a dest
|
||
- intel/fs: Choose memory message type based on bit size
|
||
- anv: Improve brw_nir_lower_mem_access_bit_sizes
|
||
- iris: Set alignments on cbuf0 and constant reads
|
||
- intel/nir: Lower memory access bit sizes later
|
||
- nir/load_store_vectorize: Fix shared atomic info
|
||
- nir/load_store_vectorize: Use nir_iadd_imm for offsets
|
||
- nir/load_store_vectorize: Add support for nir_var_mem_global
|
||
- intel/nir: Enable load/store vectorization
|
||
- spirv: Add a vtn_block() helper
|
||
- spirv: Add cast and loop helpers for vtn_cf_node
|
||
- spirv: Make vtn_case a vtn_cf_node
|
||
- spirv: Make vtn_function a vtn_cf_node
|
||
- spirv: Add a parent field to vtn_cf_node
|
||
- spirv: Rewrite CFG construction
|
||
- Revert "spirv: Rewrite CFG construction"
|
||
- nir: Assert memory loads are aligned
|
||
- anv: Advertise SEND count through
|
||
VK_EXT_pipeline_executable_properties
|
||
- anv: Fix UBO range detection in anv_nir_compute_push_layout
|
||
- nir: Add an alignment to nir_intrinsic_load_constant
|
||
- nir: Add some sanity assertions in opt_large_constants
|
||
- intel: Add \_const versions of prog_data cast helpers
|
||
- anv: Report correct SLM size
|
||
- intel/batch_decoder: Stop printing to stdout
|
||
- intel/cfg: Add first/last_block helpers
|
||
- anv: Emit pushed UBO bounds checking code in the back-end compiler
|
||
- intel/blorp: Delete an unused enum
|
||
- spirv: Handle OOB vector extract operations
|
||
- spirv,nir: Add a better vector_insert
|
||
- spirv: Error if OpCompositeInsert/Extract has OOB indices
|
||
- nir/builder: Handle any bit-size selector in nir_extract
|
||
- spirv: Call nir_builder directly for vector_extract
|
||
- spirv,nir: Move the SPIR-V vector insert code to NIR
|
||
- anv: Move vb_emit setup closer to where it's used in flush_state
|
||
- anv: Apply any needed PIPE_CONTROLs before emitting state
|
||
- nir/dominance: Better handle unreachable blocks
|
||
- nir/gcm: Loop over blocks in pin_instructions
|
||
- nir/gcm: Use an array for storing the early block
|
||
- nir/gcm: Move block choosing into a helper function
|
||
- nir/gcm: Add a real concept of "progress"
|
||
- nir/gcm: Delete dead instructions
|
||
- nir/gcm: Prefer the instruction's original block
|
||
- intel/fs: Rename block to scan_block in can_coalesce_vars
|
||
- intel/fs: Coalesce when the src live range is contained in the dst
|
||
- glsl: Hard-code noise to zero in builtin_functions.cpp
|
||
- nir: Delete the fnoise opcodes
|
||
- meta,i965: Rip GL_EXT_texture_multisample_blit_scaled support out of
|
||
meta
|
||
- spirv: Allow constants and NULLs in SpvOpConvertUToPtr
|
||
- anv: Properly handle all sizes of specialization constants
|
||
- radv: Properly handle all sizes of specialization constants
|
||
- turnip: Properly handle all sizes of specialization constants
|
||
- spirv: Use nir_const_value for spec constants
|
||
- nir/opt_deref: Remove certain sampler type casts
|
||
- spirv: Fix passing combined image/samplers through function calls
|
||
- anv: Drop an assert
|
||
- nir/lower_subgroups: Mask off unused bits in ballot ops
|
||
- anv: Add a vk_image_layout_to_usage_flags helper
|
||
- anv: Move vk_image_layout_is_read_only higher
|
||
- anv: Be more conservative about image view usage
|
||
- anv: Rework anv_layout_to_aux_state
|
||
- anv/blorp: Do less hard-coding of aux usages
|
||
- anv: Generalize some aux usage checks
|
||
- intel/blorp: Allow more HiZ usages in hiz_clear_depth_stencil
|
||
- anv: Simplify a case in layout_to_aux_usage
|
||
- anv/cmd_buffer: Move anv_image_init_aux_tt higher
|
||
- intel/isl: Delete a misleading comment
|
||
- intel/isl: Refactor isl_surf_get_ccs_surf
|
||
- anv: Add support for HiZ+CCS
|
||
- spirv: Rewrite CFG construction
|
||
- intel/devinfo: Compute the correct L3$ size for Gen12
|
||
- anv: Expose CS workgroup sizes based on a maximum of 64 threads
|
||
- anv: Return an error if allocating attachment memory fails
|
||
- anv: Add TRANSFER_SRC to pass usage not subpass usage
|
||
- anv: Stop filling out the clear color in compute_aux_usage
|
||
- anv: Assert surface states are valid
|
||
- anv: Use ANV_FROM_HANDLE for pInheritanceInfo fields
|
||
- anv: Mark images written in end_subpass
|
||
- anv: Split command buffer attachment setup in three
|
||
- anv: Allocate surface states per-subpass
|
||
- intel: Move swizzle_color_value from blorp to ISL
|
||
- anv: Disallow fast-clears which require format-reinterpretation
|
||
- anv: Stop allowing non-zero clear colors in input attachments
|
||
- anv: Refactor cmd_buffer_setup_attachments
|
||
- anv: Rework depth_stencil_attachment_compute_aux_usage
|
||
- anv: Split color_attachment_compute_aux_usage in two
|
||
- anv: Use anv_layout_to_aux_usage for color during render passes
|
||
- anv: Allow all clear colors for texturing on Gen11+
|
||
- vulkan: Update Vulkan XML and headers to 1.2.139
|
||
- nir/copy_prop_vars: Handle volatile better
|
||
- nir/copy_prop_vars: Report progress when deleting self-copies
|
||
- nir/dead_write_vars: Handle volatile
|
||
- nir/combine_stores: Handle volatile
|
||
- anv: Handle NULL descriptors
|
||
- anv: Handle null vertex buffer bindings
|
||
- anv: Claim VK_EXT_robustness2 support
|
||
- intel/fs: Don't delete coalesced MOVs if they have a cmod
|
||
- vulkan: Allow destroying NULL debug report callbacks
|
||
- anv:gpu_memcpy: Emit 3DSTATE_VF_INDEXING on Gen8+
|
||
- nir/lower_double_ops: Rework the if (progress) tree
|
||
- nir/opt_deref: Report progress if we remove a deref
|
||
- nir/copy_prop_vars: Record progress in more places
|
||
|
||
Jesse Natalie (3):
|
||
|
||
- wgl: add official gldrv.h header-file
|
||
- wgl: use gldrv.h instead of stw_icd.h
|
||
- util/ralloc: fix ralloc alignment on Win64
|
||
|
||
John Stultz (7):
|
||
|
||
- freedreno: Add ir3_cf.c and ir3_delay.c to Makefile.sources
|
||
- panfrost: Move pan_afbc.c file to the the right Makefile.source file
|
||
- gallium: hud_context: Fix scalar initializer warning.
|
||
- Android.mk: Tweak MESA_ENABLE_LLVM checks
|
||
- etnaviv: Avoid shift overflow
|
||
- vc4_bufmgr: Remove duplicative VC definition
|
||
- r600: Fix build error in sfn_nir_lower_fs_out_to_vector.cpp
|
||
|
||
Jon Turney (1):
|
||
|
||
- Fix util/process test on Cygwin
|
||
|
||
Jonathan Marek (79):
|
||
|
||
- freedreno/a6xx: use single format enum
|
||
- freedreno/a6xx: fix Z24_UNORM_S8_UINT_AS_R8G8B8A8
|
||
- freedreno: name sysmem color/depth flush events
|
||
- freedreno/a6xx: document some unknown bits
|
||
- turnip: add option to force use of hw binning
|
||
- turnip: fix COND_EXEC reserved size in tu_query
|
||
- turnip: add tu_device pointer to tu_cs
|
||
- turnip: automatically reserve cmdstream space in emit_pkt4/emit_pkt7
|
||
- turnip: remove marker seqno
|
||
- turnip: make cond_exec helper easier to use
|
||
- turnip: move tile_load_ib/sysmem_clear_ib into draw_cs
|
||
- hud: add GALLIUM_HUD_SCALE
|
||
- turnip: enable sampleRateShading feature
|
||
- turnip: enable
|
||
fullDrawIndexUint32/independentBlend/dualSrcBlend/logicOp
|
||
- etnaviv: disable INT_FILTER for ASTC
|
||
- util/format: add missing BC4/BC5 vulkan formats
|
||
- turnip: rework format table to support r5g5b5a1_unorm/b5g5r5a1_unorm
|
||
- turnip: add r5g5b5a1_unorm/b5g5r5a1_unorm formats
|
||
- turnip: check the right alignment requirement on shader iova
|
||
- turnip: move some constant state to tu6_init_hw
|
||
- turnip: remove unecessary MRT_CONTROL fill
|
||
- turnip: minify image_view extent
|
||
- turnip: fix hw binning + render_area offset interaction
|
||
- turnip: fix srgb MRT
|
||
- turnip: don't hardcode gmem base for input attachment
|
||
- turnip: remove unnecessary fb size check
|
||
- turnip: fall back to sysmem when attachments don't fit into gmem
|
||
- turnip: increase array sizes in tu_descriptor_map
|
||
- turnip: improve binning pipe layout config
|
||
- turnip: fix tile->slot calculation
|
||
- etnaviv: nir: add compile_check_limits
|
||
- freedreno/registers: more GRAS_CL_CNTL bits, Z_CLAMP
|
||
- turnip: fix znear clipping
|
||
- turnip: implement depth clamp
|
||
- turnip: implement timestamp query
|
||
- turnip: fix compute shaders crashing after geometry shader change
|
||
- turnip: improve vertex input handling
|
||
- turnip: use buffer size instead of bo size for VFD_FETCH_SIZE
|
||
- freedreno/registers: add RB_CCU_CNTL bitfields
|
||
- freedreno/a6xx: set bypass RB_CCU_CNTL value for blitter
|
||
- turnip: RB_CCU_CNTL fixes
|
||
- turnip: split up gmem/tile alignment
|
||
- turnip: fix nir validate failure from push constant lowering
|
||
- turnip: disable 8x msaa
|
||
- turnip: save attachment samples in renderpass state
|
||
- turnip: use dirty bits for dynamic viewport/scissor state
|
||
- turnip: rework format helpers
|
||
- turnip: add vk_format_is_snorm/is_float
|
||
- turnip: new clear/blit implementation with shader path fallback
|
||
- freedreno/computerator: support nop prefix
|
||
- freedreno/computerator: support bindless sampler instructions
|
||
- freedreno/ir3: fix emit_tex_info split_dest
|
||
- freedreno/ir3: don't overwrite wrmask in ir3_SAM
|
||
- turnip: compute render_components/srgb_cntl at renderpass creation
|
||
time
|
||
- turnip: don't limit framebuffer size to image size
|
||
- turnip: image_view rework
|
||
- nir: add common convert_ycbcr for vulkan csc
|
||
- nir: convert_ycbcr: preserve alpha channel
|
||
- anv: use common nir_convert_ycbcr
|
||
- radv: use common nir_convert_ycbcr
|
||
- turnip: fix GMEM resolve in CmdNextSubpass
|
||
- turnip: disable depth test for S8_UINT attachment
|
||
- turnip: improve GMEM load/store logic
|
||
- turnip: enable VK_FORMAT_S8_UINT as stencil format
|
||
- turnip: set shader key msaa field
|
||
- turnip: implement VK_EXT_sample_locations
|
||
- turnip: implement VK_EXT_filter_cubic
|
||
- turnip: enable cube arrays
|
||
- turnip: implement VK_EXT_sampler_filter_minmax
|
||
- turnip: divide cube map depth by 6
|
||
- freedreno/ir3: fix 16-bit ssbo access
|
||
- freedreno/ir3: set even bit for f2f16_rtne
|
||
- freedreno/ir3: fix incorrect conversion folding
|
||
- turnip: remove unused RB_UNKNOWN_8E04_blit
|
||
- turnip: use RESOLVE_TS event
|
||
- turnip: add adreno 650
|
||
- nir: add pack_32_2x16_split/unpack_32_2x16_split lowering
|
||
- freedreno/ir3: run nir_lower_pack
|
||
- turnip: fix wrong substream size in parse_multisample_and_color_blend
|
||
|
||
Jordan Justen (6):
|
||
|
||
- intel/compiler: Restrict cs_threads to 64
|
||
- intel: Update TGL PCI strings
|
||
- intel: Add TGL PCI ID
|
||
- intel/dev: Split .num_subslices out of GEN12_FEATURES macro
|
||
- intel/dev: Add device info for RKL
|
||
- docs/relnotes/new_features.txt: Add RKL to 20.1 release notes
|
||
|
||
Jose Maria Casanova Crespo (5):
|
||
|
||
- broadcom: Fix implicit declaration of ffs for Android build
|
||
- v3d: Sync on last CS when non-compute stage uses resource written by
|
||
CS
|
||
- v3d: Primitive Counts Feedback needs an extra 32-bit padding.
|
||
- v3d: Fix swizzle in DXT3 and DXT5 formats
|
||
- v3d: Include supported DXT formats to enable s3tc/dxt extensions
|
||
|
||
Joshua Ashton (3):
|
||
|
||
- radv: Use TRUNC_COORD on samplers
|
||
- radv: Pass logical device to si_emit_graphics
|
||
- radeonsi: Use TRUNC_COORD on samplers
|
||
|
||
José Fonseca (4):
|
||
|
||
- meson: Avoid duplicate symbols.
|
||
- scons: Prune out unnecessary targets.
|
||
- gitlab-ci: Prune all SCons jobs except scons-win64, and allows
|
||
failures.
|
||
- appveyor: Remove Meson job.
|
||
|
||
Juan A. Suarez Romero (6):
|
||
|
||
- nir/lower_double_ops: add note for lowering mod
|
||
- nir/lower_double_ops: relax lower mod()
|
||
- nir/algebraic: coalesce fmod lowering
|
||
- anv: use urb_setup_attribs in SBE
|
||
- intel/compiler: store the FS inputs in WM prog data
|
||
- anv/pipeline: allow more than 16 FS inputs
|
||
|
||
Karol Herbst (18):
|
||
|
||
- clover: add trivial clCreateCommandQueueWithProperties implementation
|
||
- nir/lower_ssbo: handle atomics
|
||
- gallium: make handles of set_global_binding 64 bit
|
||
- Revert "gallium: make handles of set_global_binding 64 bit"
|
||
- nv50, nvc0: fix must_check warning of util_dynarray_resize_bytes
|
||
- clover: fix build with single library clang build
|
||
- gallium: add PIPE_CAP_SYSTEM_SVM
|
||
- clover: add stubs for SVM
|
||
- clover: implement CL_DEVICE_SVM_CAPABILITIES
|
||
- clover: implement clSetKernelArgSVMPointer
|
||
- clover: implement SVM functions for devices with fine grained system
|
||
SVM support
|
||
- clover: implement cl_arm_shared_virtual_memory
|
||
- clover: expose cl_arm_shared_virtual_memory for devices with SVM
|
||
support
|
||
- nvc0: enable ASTC and ETC on GM20B
|
||
- mesa: fix enum value of VIEWPORT_SWIZZLE_POSITIVE_W_NV
|
||
- gallium: initialize viewport swizzle in cso_set_viewport_dims
|
||
- Revert "nvc0: fix line width on GM20x+"
|
||
- st/mesa: properly guard fallback_copy_texsubimage aginst failed maps
|
||
|
||
Kenneth Graunke (14):
|
||
|
||
- intel/genxml: Drop "reserved" enum
|
||
- isl: Fix the android build.
|
||
- iris: Dump frame markers with INTEL_DEBUG=submit
|
||
- iris: Trim "../../src/gallium/drivers/iris/" out of debug dump
|
||
filenames
|
||
- iris: Make mocs an inline helper in iris_resource.h
|
||
- iris: Fix BLORP vertex buffers to respect ISL MOCS settings
|
||
- iris: Set MOCS for constant packets on Gen12+
|
||
- intel/compiler: Drop nir_lower_to_source_mods() and related handling.
|
||
- intel/compiler: Put back saturate on [iu]add_sat opcodes
|
||
- intel/compiler: Don't copy prop source mods into PICK_HIGH_32BIT
|
||
- intel/compiler: Delete abs/neg handling in fsign code
|
||
- intel/compiler: Don't create 64-bit src1 immediates in
|
||
opt_peephole_sel
|
||
- nir: Actually do load/store vectorization beyond vec2
|
||
- iris: Fix downcast of bound_vertex_buffers from uint64_t to int
|
||
|
||
Konrad Dybcio (1):
|
||
|
||
- freedreno/a4xx: enable A405
|
||
|
||
Kristian Høgsberg (39):
|
||
|
||
- nir: Delete unused is_var_constant() helper
|
||
- nir: Make unroll pragma work on clang
|
||
- freedreno/fdperf: Cast away some ignored return values
|
||
- spirv/opencl: Cast opcode up front to avoid warnings
|
||
- glsl: Use 'using' to be explicit about visitor overloads
|
||
- nir: Remove always-true assert
|
||
- turnip: Be explicit about converting vk compare func to a6xx
|
||
- freedreno/a6xx: Add fd6_resource_screen_init()
|
||
- freedreno: Set up supported modifiers in fd*_resource_screen_init()
|
||
- freedreno: Add layout_resource_for_modifier screen vfunc
|
||
- freedreno/a6xx: Implement layout for DRM_FORMAT_MOD_QCOM_COMPRESSED
|
||
- turnip: Drop explicit configure opt-in for turnip
|
||
- ci: Drop turnip opt-in option
|
||
- freedreno/ir3: Set IR3_REG_HALF flag on src as well in immediate MOV
|
||
- Mark a few static inline helpers with ASSERTED
|
||
- main/get: Converted type conversion macros to inline functions
|
||
- nir/types: Add glsl_float16_type() helper
|
||
- freedreno/ir3: Lower output precision
|
||
- Revert "glsl: Use a simpler formula for tanh"
|
||
- Revert "spirv: Use a simpler and more correct implementaiton of
|
||
tanh()"
|
||
- freedreno/ir3: Don't fold conversions into sign
|
||
- glsl: Add ir_constant constructor for fp16
|
||
- glsl: Add fp16 case for ir_triop_lrp optimization
|
||
- glsl: Implement constant propagation for fp16
|
||
- glsl: Expand fp16 to float before constant expression evaluation
|
||
- glsl: Add type queries for fp16+float and fp16+float+double
|
||
- glsl/lower_instructions: Handle fp16 for FDIV_TO_MUL_RCP
|
||
- radeonsi: Stop exposing PIPE_SHADER_CAP_FP16
|
||
- turnip: Add missing VKAPI_ATTR annotations
|
||
- turnip: Stub out VK_KHR_external_{fence,semaphore}_fd
|
||
- turnip: Make Android platform build
|
||
- turnip: Drop dep_llvm from dependencies
|
||
- freedreno/ir3: Fix sz vs class confusion
|
||
- freedreno/computerator: Decouple ir3 assembler
|
||
- freedreno/ir3: Move ir3 assembler to backend compiler
|
||
- freedreno/ir3: Parse, but ignore @in, @out and @tex headers
|
||
- freedreno/ir3: Reset lex line number when we start parsing
|
||
- freedreno/ir3: Print @tex write mask using 0x%x
|
||
- freedreno: Use the right amount of &'s
|
||
|
||
Krzysztof Raszkowski (10):
|
||
|
||
- gallium/swr: fix gcc warnings
|
||
- gallium/swr: Fix gcc 4.8.5 compile error
|
||
- gallium/swr: Fix llvm11 compilation issues
|
||
- gallium/swr: simplify environmental variabled expansion code
|
||
- gallium/swr: fix rdtsc debug statistics mechanism
|
||
- gallium/swr: Fix min/max range index draw
|
||
- Revert "gallium/swr: Fix min/max range index draw"
|
||
- gallium/swr: Fix vcvtph2ps llvm intrinsic compile error
|
||
- gallium/swr: Fix array stride problem.
|
||
- gallium/swr: Re-enable scratch space for client-memory buffers
|
||
|
||
Leandro Ribeiro (1):
|
||
|
||
- i965: remove duplicated comment
|
||
|
||
Leo Liu (1):
|
||
|
||
- radeon/jpeg: fix the jpeg dt_pitch with YUYV format
|
||
|
||
Lepton Wu (1):
|
||
|
||
- virgl: Use ETC2 formats directly when possible.
|
||
|
||
Lionel Landwerlin (49):
|
||
|
||
- iris: implement gen12 post sync pipe control workaround
|
||
- anv: implement gen9 post sync pipe control workaround
|
||
- anv: implement gen12 post sync pipe control workaround
|
||
- anv: set MOCS on push constants
|
||
- mesa: add INTEL_blackhole_render
|
||
- i965: enable INTEL_blackhole_render
|
||
- st: add support for INTEL_blackhole_render
|
||
- iris: add support INTEL_blackhole_render
|
||
- intel/tools/aub_dump: move aub file initialization to maybe_init()
|
||
- intel/tools/aub_dump: fix crash when using the default legacy context
|
||
- intel/aub_dump: stub the waits when overriding the device
|
||
- intel/tools/dump_gpu: fix getparam values
|
||
- anv: stop storing prog param data into shader blobs
|
||
- intel/decoder: don't consider header fields past dword0
|
||
- isl: implement linear tiling row pitch requirement for display
|
||
- isl: properly filter supported display modifiers on Gen9+
|
||
- isl: only apply main surface ccs pitch constraint with CCS
|
||
- isl: drop min row pitch alignment when set by the driver
|
||
- intel: add new TGL pci ids
|
||
- i965/iris: fix crash when calling GetPerfQueryDataINTEL
|
||
- vulkan/overlay: Add a workaround semaphore for application presenting
|
||
without one
|
||
- intel/perf: move register definition to special file
|
||
- intel/perf: break GL query stuff away
|
||
- intel/perf: move mdapi query definitions to their own file
|
||
- intel/perf: document meaning of query field
|
||
- intel/perf: store the probed i915-perf version
|
||
- isl: set bpb for Y8_UNORM
|
||
- isl: don't warn in physical extent calculation for yuv formats
|
||
- intel/aub_viewer: fix access to freed memory
|
||
- drm-shim: return device platform as specified
|
||
- drm-shim: stub libdrm's use of realpath()
|
||
- iris: properly free resources on BO allocation failure
|
||
- iris: share buffer managers accross screens
|
||
- iris: make resources take a ref on the screen object
|
||
- i965: store DRM fd on intel_screen
|
||
- i965: share buffer managers across screens
|
||
- iris: drop cache coherent cpu mapping for external BO
|
||
- intel/perf: Enable MDAPI queries for Gen12
|
||
- anv: skip writing perfcntr in results on Gen12+
|
||
- util/sparse_free_list: manipulate node pointers using atomic
|
||
primitives
|
||
- iris: fail screen creation when kernel support is not there
|
||
- include/drm-uapi: bump headers
|
||
- intel/perf: store default sseu configuration
|
||
- intel/perf: specify sseu configuration when supported
|
||
- anv: force whole EU array to be powered for perf queries
|
||
- drm-shim: provide a valid fake syncobj handle at creation
|
||
- drm-shim: stub syncobj wait ioctl
|
||
- iris: don't assert on unfinished aux import in copy paths
|
||
- anv: don't expose VK_INTEL_performance_query without kernel support
|
||
|
||
Liviu Prodea (2):
|
||
|
||
- scons/windows: Support build with LLVM 10.
|
||
- util: Make process_test path compatible with mingw native toolchains
|
||
|
||
Louis-Francis Ratté-Boulianne (7):
|
||
|
||
- glsl/linker: add DisableTransformFeedbackPacking workaround
|
||
- glsl/linker: handle array/struct members for DisableXfbPacking
|
||
- glsl/linker: add xfb workaround for modified built-in variables
|
||
- gallium: add PIPE_CAP_PACKED_STREAM_OUTPUT
|
||
- gallium: add PIPE_CAP_VIEWPORT_TRANSFORM_LOWERED
|
||
- gallium: add PIPE_CAP_PSIZ_CLAMPED
|
||
- panfrost: fix transform feedback
|
||
|
||
Lucas Stach (1):
|
||
|
||
- etnaviv: retarget transfer to render resource when necessary
|
||
|
||
Marek Olšák (254):
|
||
|
||
- vbo: move GLvertexformat initialization into a template header file
|
||
for reuse
|
||
- vbo: use the template for noop GLvertexformat initialization
|
||
- vbo: use the template for save GLvertexformat initialization
|
||
- vbo: move reusable code from vbo_attrib_tmp.h into vbo_util.h
|
||
- mesa: implement missing display list functions while switching to the
|
||
template
|
||
- radeonsi: don't report that multi-plane formats are supported
|
||
- radeonsi: fix the DCC MSAA bug workaround
|
||
- radeonsi: don't update states for the DCC MSAA bug on GFX6-7
|
||
- glx: print FPS with 2 decimal places
|
||
- mesa: fix incorrect uses of FLUSH_CURRENT
|
||
- mesa: remove FLUSH_CURRENT calls that have no effect
|
||
- mesa: import PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET handling
|
||
- vbo: create the immediate mode buffer only in vbo_exec_vtx_map
|
||
- vbo: skip FlushMappedBufferRange for glBegin/End by using a
|
||
persistent mapping
|
||
- vbo: don't unmap persistent buffer mappings for glBegin/End
|
||
- vbo: remove immediate mode code that doesn't do anything and simplify
|
||
stuff
|
||
- vbo: interleave attrsz, attrtype, and active_sz in memory
|
||
- vbo: remove a funky recursive call in glBegin
|
||
- vbo: don't check ctx->NewState twice in glBegin
|
||
- vbo: keep the immediate mode buffer always mapped for simplicity
|
||
- vbo: don't set FLUSH_UPDATE_CURRENT for glVertex
|
||
- vbo: pass only either uint32_t or uint64_t into ATTR_UNION
|
||
- vbo: don't store glVertex values temporarily into exec
|
||
- vbo: optimize resizing vertex attributes during immediate mode
|
||
- vbo: fix resizing 64-bit vertex attributes
|
||
- vbo: use FlushVertices flags properly and clear NeedFlush correctly
|
||
- vbo: increase the size of the immediate mode buffer to decrease draw
|
||
count
|
||
- vbo: add/update unlikely statements in ATTR_UNION
|
||
- vbo: delay flagging FLUSH_STORED_VERTICES until glEnd
|
||
- vbo: also map the immediate mode buffer for read
|
||
- vbo: clean up resetting vertex attribs
|
||
- vbo: merge use_buffer_objects into vbo_CreateContext to skip the big
|
||
malloc
|
||
- í965: don't use \_mesa_prim::is_indirect
|
||
- mesa: remove unused \_mesa_prim::is_indirect
|
||
- mesa: don't use bitfields in \_mesa_prim
|
||
- st/mesa: optimize st_update_array with ALWAYSINLINE
|
||
- radeonsi: don't wait for shader compilation to finish when destroying
|
||
a context
|
||
- mesa: translate into gallium vertex formats in mesa/main
|
||
- mesa: remove unused \_mesa_draw_indirect
|
||
- st/mesa: always inline the code setting non-64bit vertex elements
|
||
- st/mesa: simplify determination whether a draw has user vertex
|
||
buffers
|
||
- st/mesa: simplify determination whether a draw needs min/max index
|
||
- st/mesa: change some loops from while to do..while in st_atom_array.c
|
||
- st/mesa: make st_setup_current static
|
||
- st/mesa: simplify releasing the current attrib buffer
|
||
- gallium/u_upload_mgr: reduce dereferences by adding buffer_size
|
||
- gallium/u_upload_mgr: don't do align twice in the u_upload_alloc fast
|
||
path
|
||
- gallium/u_vbuf: adjust the heuristic for unrolling indices
|
||
- gallium/cso_hash: inline a bunch of functions
|
||
- gallium/cso_hash: make cso_hash declared within structures instead of
|
||
alloc'd
|
||
- gallium/cso_hash: remove always constant variable nodeSize
|
||
- gallium/cso_hash: cosmetic changes, no behavior changes
|
||
- gallium/cso_hash: remove another layer of pointer indirection
|
||
- st/mesa: try to fix MSVC build failure due to ALWAYS_INLINE
|
||
- vbo: remove dead code in vbo_can_merge_prims
|
||
- vbo: remove redundant code in vbo_exec_fixup_vertex
|
||
- mesa: document \_mesa_prim::begin/end
|
||
- mesa: don't use memset in glDrawArrays
|
||
- mesa: fix immediate mode with tessellation and varying patch vertices
|
||
- gallium/util: remove unused u_surfaces.c/h
|
||
- util: remove the dependency on kcmp.h
|
||
- nir: fix gl_nir_lower_images for bindless images
|
||
- tgsi_to_nir: set num_images and num_samplers with holes correctly
|
||
- gallium/hash_table: consolidate hash tables with pointer keys
|
||
- gallium/hash_table: consolidate hash tables with FD keys
|
||
- gallium/hash_table: use the same callback signatures as
|
||
util/hash_table
|
||
- gallium/hash_table: turn it into a wrapper around util/hash_table
|
||
- gallium/hash_table: remove some function wrappers
|
||
- mesa: remove leftovers from ARB_shadow_ambient
|
||
- mesa: call FLUSH_VERTICES before updating CoordReplace
|
||
- i965: stop using "indirect" parameter from Driver.Draw (non-indirect)
|
||
- mesa: remove unused "indirect" parameter from Driver.Draw
|
||
- gallium/cso_hash: pack cso_node better
|
||
- gallium/cso_hash: inline struct cso_hash_data
|
||
- gallium: pass cso_velems_state into cso_context instead of
|
||
pipe_vertex_element
|
||
- gallium/u_threaded: fix uploading user indices with start != 0
|
||
- gallium/u_threaded: convert dividing by index_size to a bit shift
|
||
- mesa/i965: remove \_mesa_prim::indirect_offset
|
||
- mesa: remove redundant \_mesa_prim::is_indexed
|
||
- mesa: move num_instances and base_instance out of \_mesa_prim
|
||
- mesa: clean up glMultiDrawElements code, use alloca for small draw
|
||
count (v2)
|
||
- mesa: don't unroll glMultiDrawElements if one count is 0
|
||
- mesa: optimize glMultiDrawArrays, call Draw only once (v2)
|
||
- mesa: fix incorrect prim.begin/end for glMultiDrawElements
|
||
- nir: replace GCC unroll with an option that works on GCC < 8.0
|
||
- gallivm: fix 5 warnings
|
||
- nir: fix 5 warnings
|
||
- mesa: fix 11 warnings
|
||
- gallium/u_vbuf: silence a warning by using unreachable
|
||
- mesa: add index_size_shift = log2(index_size) into
|
||
\_mesa_index_buffer
|
||
- mesa: replace some index_size multiplications and divisions with
|
||
shifts
|
||
- vbo: don't look at the second draw's count when merging 2 glBegin/End
|
||
draws
|
||
- vbo: deduplicate copy_vertices functions
|
||
- vbo: clean up vbo_copy_vertices
|
||
- vbo: handle GS and tess primitive types when splitting Begin/End
|
||
- vbo: clean up conditional blocks in ATTR_UNION
|
||
- vbo: fold code from vbo_exec_fixup_vertex to
|
||
vbo_exec_wrap_upgrade_vertex
|
||
- Revert "mesa: check for z=0 in \_mesa_Vertex3dv()"
|
||
- mesa: remove \_mesa_index_buffer::index_size in favor of
|
||
index_size_shift
|
||
- mesa: optimize get_index_size
|
||
- mesa: deduplicate draw indirect functions
|
||
- vbo: merge more primitive types for glBegin/End (v2)
|
||
- vbo: merge draws even when begin==0 or end==0
|
||
- glthread: don't generate the sync fallback if the call size is not
|
||
variable
|
||
- glthread: don't prefix variable_data with const
|
||
- glthread: inline \_mesa_unmarshal_dispatch_cmd and convert the switch
|
||
to a table
|
||
- glthread: reduce pointer dereferences in glthread_unmarshal_batch
|
||
- glthread: use int instead of size_t where it's OK
|
||
- glthread: simplify repeated function sequences in marshal_generated.c
|
||
- glthread: don't insert \_mesa_post_marshal_hook into every function
|
||
- glthread: don't increment variable_data if it's the last
|
||
variable-size param
|
||
- glthread: add GL_DRAW_INDIRECT_BUFFER tracking and generator support
|
||
- glthread: add/update count and marshal fields for many GL functions
|
||
- glthread: handle complex pointer parameters and support GL functions
|
||
with strings
|
||
- glthread: check the size of all variable params and clean up the code
|
||
- glthread: replace custom ClearBuffer marshalling with generated one
|
||
- glthread: add support for TexParameteri and SamplerParameteri
|
||
functions
|
||
- glthread: add support for glFog, glLight, glLightModel, glTexEnv,
|
||
glTexGen
|
||
- glthread: add support for glClearNamedFramebuffer, glMaterial,
|
||
glPointParameter
|
||
- glthread: add support for glCallLists, glPatchParameterfv
|
||
- glthread: add support for glMemoryObjectParameteriv,
|
||
glSemaphoreParameterui64v
|
||
- glthread: don't insert an empty line after (void) cmd;
|
||
- glthread: add marshal_call_after and remove custom glFlush and
|
||
glEnable code
|
||
- glthread: track for each VAO whether the user has set a user pointer
|
||
- glthread: sync instead of disabling glthread for non-VBO pointers
|
||
- glthread: replace custom glBindBuffer marshalling with generated one
|
||
- glthread: merge glBufferData and glNamedBufferData into 1 set of
|
||
functions
|
||
- glthread: merge glBufferSubData and glNamedBufferSubData into 1 set
|
||
of functions
|
||
- glthread: add custom marshalling for glNamedBuffer(Sub)DataEXT
|
||
- glthread: fix a crash with incorrect glShaderSource parameters
|
||
- glthread: fall back if a param size is non-zero and a pointer param
|
||
is NULL
|
||
- radeonsi: add a bug workaround for NGG - LATE_ALLOC_GS
|
||
- ac: add a bug workaround for the 100% NGG culling case
|
||
- radeonsi: determine uses_bindless_samplers correctly
|
||
- st/mesa: flush the bitmap cache before st/dri and vbo flushes
|
||
- st/mesa: fix a possible crash with selection and feedback modes
|
||
- gallium/cso_context: remove cso_delete_xxx_shader helpers to fix the
|
||
live cache
|
||
- st/mesa: keep serialized NIR instead of nir_shader in st_program
|
||
- vbo: use vbo_exec_wrap_upgrade_vertex for glVertex in ATTR_UNION
|
||
- vbo: fix transitions from glVertexN to glVertexM where M < N
|
||
- vbo: fix vbo_copy_vertices for GL_PATCHES and adjacency primitive
|
||
types
|
||
- gallium: add PIPE_CAP_DRAW_INFO_START_WITH_USER_INDICES
|
||
- mesa: don't unroll glMultiDrawElements with user indices for gallium
|
||
- radeonsi/gfx10: cache metadata in L2 on small chips
|
||
- radeonsi: set better tessellation tunables on gfx9 and gfx10
|
||
- radeonsi: tune primitive binning for small chips
|
||
- ac: add radeon_info::use_late_alloc to control LATE_ALLOC globally
|
||
- ac: disable late alloc on small gfx10 chips
|
||
- gallium/u_threaded: don't sync the thread for all unsychronized
|
||
mappings
|
||
- gallium/u_vbuf: simplify the first if statement in
|
||
u_vbuf_upload_buffers
|
||
- ac: unify denorm setting enforcement
|
||
- ac: set new LLVM denormal flags
|
||
- ac: don't set old denormals flags with LLVM >= 11
|
||
- nir: fix clip/cull_distance_array_size in
|
||
nir_lower_clip_cull_distance_arrays
|
||
- mesa: use vbo_attrib_tmp.h to generate display list vertex attrib
|
||
functions
|
||
- mesa: remove redundant api_loopback functions
|
||
- glthread: align the batch buffer to 8 bytes for pointers and doubles
|
||
again
|
||
- glthread: enable display lists
|
||
- glthread: track VAOs created by CreateVertexArrays
|
||
- glthread: don't execute any custom VAO and BindBuffer code in the
|
||
Core profile
|
||
- glthread: remove debug_print_marshal function
|
||
- glthread: clean up debug_print_sync code
|
||
- glthread: don't declare unmarshal functions as inline
|
||
- winsys/radeon: change to 3-space indentation
|
||
- driconf: enable glthread for "From The Depths"
|
||
- glthread: remove \_mesa_post_marshal_hook, because it's not very
|
||
useful
|
||
- glthread: simplify printing safe_mul in gl_marshal.py
|
||
- glthread: autogenerate prototypes for custom-marshalled functions
|
||
- glthread: move buffer functions into glthread_bufferobj.c
|
||
- glthread: rename marshal.h/c to glthread_marshal.h and
|
||
glthread_shaderobj.c
|
||
- mesa: put gl_thread_state inside gl_context to remove pointer
|
||
indirection
|
||
- glthread: handle buffer unbinding via glDeleteBuffers
|
||
- glthread: rename non_vbo helper functions
|
||
- glthread: track which vertex array attribs are enabled
|
||
- glthread: ignore vertex arrays with user pointers if they're disabled
|
||
- glthread: remove the marshal_fail XML attribute
|
||
- vbo,gallium: make glBegin/End buffer size configurable by drivers
|
||
- ac: fix fast division
|
||
- st/mesa: fix use of uninitialized memory due to st_nir_lower_builtin
|
||
- glthread: inline SET_func and add -O1 to build
|
||
\_mesa_create_marshal_table faster
|
||
- glthread: declare marshal and unmarshal functions as non-static
|
||
- glthread: compile marshal_generated.c faster by breaking it up into 8
|
||
files
|
||
- nir: add and gather shader_info::writes_memory
|
||
- glsl_to_tgsi: set shader_info::writes_memory
|
||
- mesa: allow out-of-order drawing to optimize immediate mode if it's
|
||
safe
|
||
- radeonsi: enable full out-of-order drawing when
|
||
allow_draw_out_of_order is set
|
||
- mesa: try to fix the android build
|
||
- Move compiler.h and imports.h/c from src/mesa/main into src/util
|
||
- mesa: don't use <> for including internal headers
|
||
- util: stop including files from mesa/main
|
||
- radv: stop including files from mesa/main
|
||
- util: don't include p_defines.h and u_pointer.h from gallium
|
||
- util: remove duplicated MALLOC_STRUCT and CALLOC_STRUCT
|
||
- radeonsi: remove obsolete TODO comment related to compute-based
|
||
culling
|
||
- radeonsi: fix incorrect ordered_wave_id initilization for
|
||
compute-based culling
|
||
- radeonsi: set amdgpu-gds-size for mode == 2 of compute-based culling
|
||
- radeonsi: always create wait_mem_scratch for compute-based culling
|
||
- radeonsi: add num_vbos_in_user_sgprs into the shader cache key
|
||
- radeonsi/gfx10: don't use NGG culling if compute-based culling is
|
||
used
|
||
- radeonsi/gfx10: fix ds.ordered.add intrinsic for compute-based
|
||
culling
|
||
- radeonsi/gfx10: user correct ACQUIRE_MEM packet for compute-based
|
||
culling
|
||
- radeonsi/gfx10: fix the wave size for compute-based culling
|
||
- radeonsi/gfx10: fix descriptors and compute registers for
|
||
compute-based culling
|
||
- gallium/u_threaded: call the driver to pin threads to L3 immediately
|
||
- st/mesa: add environment variable pin_app_thread for faster glthread
|
||
on AMD Zen
|
||
- driconf: whilelist more games for glthread
|
||
- mesa: optimize initialization of new VAOs
|
||
- mesa: don't ever set NullBufferObj in gl_vertex_array_binding
|
||
- mesa: don't ever bind NullBufferObj for glBindBuffer targets
|
||
- mesa: don't ever bind NullBufferObj to glBindBuffer(Base,Range) slots
|
||
- mesa: remove NullBufferObj
|
||
- mesa: remove no longer needed \_mesa_is_bufferobj function
|
||
- mesa: precompute \_mesa_primitive_restart_index during state changes
|
||
- mesa: split \_mesa_primitive_restart_index into a function without
|
||
gl_context
|
||
- vbo: expose helper function vbo_get_minmax_index_mapped for glthread
|
||
- util: move and adjust the vertex upload heuristic equation from
|
||
u_vbuf
|
||
- st/mesa: fix a crash due to passing a draw vertex shader into the
|
||
driver
|
||
- ac: out-of-order rasterization is not supported on gfx10
|
||
- ac,radeonsi: simplify checking for Navi1x chips
|
||
- radeonsi: use pipe_blend_state::max_rt to update fewer blend
|
||
registers
|
||
- ac: force enable -structurizecfg-skip-uniform-regions for LLVM 11
|
||
- ac: update and document fast math flags used by radeonsi
|
||
- ac: generate FMA for inexact instructions for radeonsi
|
||
- ac: reassociate FP expressions for inexact instructions for radeonsi
|
||
- mesa: replace \_NEW_EVAL with vbo_exec_update_eval_maps
|
||
- mesa: reset primitive restart state in glClientAttribDefaultEXT
|
||
- mesa: remove exec="dynamic" from Draw functions that are not really
|
||
dynamic
|
||
- glthread: use 32-bit align instead of 64-bit ALIGN
|
||
- glthread: reduce dereferences of the next batch
|
||
- glthread: use GLenum16 in batch buffers to save space
|
||
- glthread: sort variables in marshal structures to pack them optimally
|
||
- gallium: add PIPE_CAP_MAP_UNSYNCHRONIZED_THREAD_SAFE for glthread
|
||
- mesa: add Const.BufferCreateMapUnsynchronizedThreadSafe &
|
||
MESA_MAP_THREAD_SAFE
|
||
- mesa: add offset_is_int32 param into \_mesa_bind_vertex_buffer for
|
||
glthread
|
||
- mesa: extend \_mesa_bind_vertex_buffer to take ownership of the
|
||
buffer reference
|
||
- mesa: replace GLenum target with gl_shader_stage in NewProgram
|
||
- ac/surface: rename micro tile mode enums like gfx10 uses them
|
||
- ac/surface: remove RADEON_SURF_TC_COMPATIBLE_HTILE and assume it's
|
||
always set
|
||
- ac/surface: replace RADEON_SURF_OPTIMIZE_FOR_SPACE with
|
||
!FORCE_SWIZZLE_MODE
|
||
- ac/surface: match get_display_flag() with expectations for
|
||
is_displayable
|
||
- ac/surface: don't compute DCC if it's unsupported by DCN on gfx9+
|
||
- ac/surface: move non-displayable DCC to the end of the buffer
|
||
- ac/surface: add code for gfx10 displayable DCC
|
||
- ac/surface: validate that DCC is enabled correctly on gfx9+
|
||
- ac: enable displayable DCC on Navi12 & Navi14
|
||
- mesa: report GL_INVALID_OPERATION for invalid glTextureBuffer target
|
||
- st/mesa: expose more SPIR-V capabilities
|
||
- radeonsi: unify and align down the max SSBO/TBO/UBO buffer binding
|
||
size
|
||
- radeonsi: revert an accidental change in si_clear_buffer
|
||
- Revert "ac/surface: remove RADEON_SURF_TC_COMPATIBLE_HTILE and assume
|
||
it's always set"
|
||
- Revert "ac: reassociate FP expressions for inexact instructions for
|
||
radeonsi"
|
||
- ac/surface: fix MSAA crash with FORCE_SWIZZLE_MODE on gfx9
|
||
- radeonsi: fix compilation of monolithic PS
|
||
- radeonsi: don't expose 16xAA on chips with 1 RB due to an occlusion
|
||
query issue
|
||
|
||
Marek Vasut (4):
|
||
|
||
- etnaviv: Destroy rsc->pending_ctx set in etna_resource_destroy()
|
||
- etnaviv: Emit PE.ALPHA_COLOR_EXT\* on GPUs with half-float support
|
||
- etnaviv: Fix depth stencil ops on GC880/GC2000
|
||
- etnaviv: Disable seamless cube map on GC880
|
||
|
||
Mark Janes (2):
|
||
|
||
- nir: check shader type before writing to shaderinfo.tess union
|
||
- nir: place aligned members after bitfields in shader_info.tess
|
||
|
||
Mark Menzynski (2):
|
||
|
||
- util/blob: Add overwrite function for uint8
|
||
- tgsi/util: Change boolean for bool
|
||
|
||
Martin Fuzzey (3):
|
||
|
||
- freedreno: android: fix build failure on android due to python
|
||
version
|
||
- freedreno: android: add a6xx-pack.xml.h generation to android build
|
||
- freedreno: android: fix build of perfcounters.
|
||
|
||
Mathias Fröhlich (19):
|
||
|
||
- egl: Implement getImage/putImage on pbuffer swrast.
|
||
- mesa: Fix FLUSH_VERTICES in SubpixelPrecisionBiasNV.
|
||
- egl: Fix A2RGB10 platform_{device,surfaceless} PBuffer configs.
|
||
- egl: Factor out dri2_add_pbuffer_configs_for_visuals
|
||
{device,surfaceless}.
|
||
- mesa: Check for OpenGL state change before flushing vertices.
|
||
- mesa: Flush vertices before changing the OpenGL state.
|
||
- i965: Move down genX_upload_sbe in profiles.
|
||
- iris: Move down iris_emit_sbe_swiz in profiles.
|
||
- i965: Use 32 bit u_bit_scan for vertex attribute setup.
|
||
- i965: Use the VAOs binding information in array setup.
|
||
- i965: Test original vertex array pointer to skip array upload.
|
||
- i965: Split merge_inputs and clear_buffers.
|
||
- i965: Reorder workaround flags computation.
|
||
- i965: Remove glbinding from brw_vertex_element.
|
||
- mesa: Remove now unused \_mesa_draw_attrib_and_binding.
|
||
- mesa: Remove now unused \_mesa_draw_attrib.
|
||
- mesa: Provide gl_vertex_format accessors.
|
||
- i965: Make use of the vertex format functions in i965.
|
||
- i965: Use gl_vertex_format in brw_vertex_element.
|
||
|
||
Matt Turner (11):
|
||
|
||
- intel/tools: Do not print type/qualifiers/name for c_literal
|
||
- intel/vec4: Make implied_mrf_writes() a vec4_instruction method
|
||
- intel/compiler: Remove unnecessary local variables
|
||
- intel/compiler: Make instructions_to_schedule a local variable
|
||
- intel/compiler: Mark some methods and parameters const
|
||
- intel/compiler: Mark visitor parameters to scheduler const
|
||
- intel/compiler: Pass backend_shader \* to cfg_t()
|
||
- intel/compiler: Pass shader_stats for each SIMD mode
|
||
- intel/compiler: Discount NOPs from instruction counts
|
||
- isl: Avoid EXPECT_DEATH in unit tests
|
||
- meson: Specify the maximum required libdrm in dri.pc
|
||
|
||
Mauro Rossi (5):
|
||
|
||
- android: gallium/auxiliary: fix "Unused source files" in tesselator
|
||
- android: aco: fix PIPE_FORMAT related building errors
|
||
- android: r600/sfn: fix includes and libmesa_nir dependency
|
||
- android: r600/sfn: Add GDS instructions
|
||
- android: aco: add various compiler statistics
|
||
|
||
Michel Dänzer (33):
|
||
|
||
- gitlab-ci: Update to latest ci-templates HEAD
|
||
- gitlab-ci: Pass -j4 to make
|
||
- gitlab-ci: Merge ccache and libxml2-utils into main apt-get install
|
||
- gitlab-ci: Add ppc64el and s390x cross-build jobs
|
||
- gitlab-ci: Build radeonsi & RADV in the ppc64el job
|
||
- llvmpipe: Bump test timeout to 180 seconds
|
||
- gitlab-ci: Only use gstreamer runners for the s390x job for now
|
||
- gitlab-ci: Sort random failure softpipe skips
|
||
- gitlab-ci: Add three more dEQP-GLES31 tests to softpipe skips
|
||
- st/vdpau: Only call is_video_format_supported hook if needed
|
||
- winsys/amdgpu: Make local variable r signed
|
||
- util: Change os_same_file_description return type from bool to int
|
||
- gitlab-ci: Drop "test-" prefix from llvmpipe/softpipe job names
|
||
- gitlab-ci: Distribute jobs across more stages
|
||
- gitlab-ci: Always name artifacts archive after the job producing it
|
||
- gitlab-ci: Don't restrict ppc64el/s390x build jobs to gstreamer
|
||
runners
|
||
- gitlab-ci: Don't use buster-backports packages by default for
|
||
x86_build
|
||
- gitlab-ci: Fold scons-swr job into scons job
|
||
- gitlab-ci: Move classic driver testing to a new meson-classic job
|
||
- llvmpipe: Use uintptr_t for pointer values
|
||
- gitlab-ci: Enable more Gallium drivers in meson-i386 job
|
||
- gitlab-ci: Restrict s390x/ppc64el jobs to packet runners
|
||
- gitlab-ci: Update to current templates
|
||
- gitlab-ci: Rename "paths" YAML anchor to "all_paths"
|
||
- gitlab-ci/lava: Add needs: for container image to test jobs (again)
|
||
- gitlab-ci: Don't require triggering build/test jobs manually
|
||
- gitlab-ci: Run merge request pipelines automatically only for Marge
|
||
Bot
|
||
- gitlab-ci: Use all_paths in .test-manual rules
|
||
- gbm/dri: Propagate queryDmaBufModifiers return value
|
||
- amd/addrlib: Use enum instead of sparse chars to identify dimensions
|
||
- mesa: Skip 3-byte array formats in \_mesa_array_format_flip_channels
|
||
- Revert "ac,radeonsi: fix compilations issues with LLVM 11"
|
||
- Revert "gallium/gallivm: fix compilation issues with llvm 11"
|
||
|
||
Mike Blumenkrantz (6):
|
||
|
||
- zink: set UBO alignments in nir_intrinsic_load_uniform lowering
|
||
- zink: remove framebuffer cache
|
||
- zink: explicitly unref old fb object when setting new one
|
||
- iris: move iris_vtable to iris_screen
|
||
- gallium: add pipe cap for scissored clears and pass scissor state to
|
||
clear() hook
|
||
- iris: handle PIPE_CAP_CLEAR_SCISSORED
|
||
|
||
Nanley Chery (6):
|
||
|
||
- isl: Add a module which manages aux resolves
|
||
- iris: Use isl_aux_usage_has_fast_clear()
|
||
- iris: Use ISL's access preparation functions
|
||
- iris: Use isl_aux_state_transition_write()
|
||
- i965: Use ISL's access preparation functions
|
||
- i965: Use isl_aux_state_transition_write()
|
||
|
||
Nataraj Deshpande (1):
|
||
|
||
- dri_util: Update internal_format to GL_RGB8 for
|
||
MESA_FORMAT_R8G8B8X8_UNORM
|
||
|
||
Neha Bhende (2):
|
||
|
||
- svga: fix size of format_conversion_table[]
|
||
- svga: Use pipe_shader_state_from_tgsi to set shader state
|
||
|
||
Neil Armstrong (4):
|
||
|
||
- gitlab-ci/lava: fix handling of lava tags
|
||
- Revert "ci: Remove T820 from CI temporarily"
|
||
- gitlab-ci: add FILES_HOST_URL and move FILES_HOST_NAME into jobs
|
||
- gitlab-ci: re-enable mali400/450 and t820 jobs
|
||
|
||
Neil Roberts (17):
|
||
|
||
- nir/opcodes: Add nir_op_f2fmp
|
||
- glsl: Add support for float16 types in the IR tree
|
||
- glsl: Add IR conversion ops for 16-bit float types
|
||
- glsl: Add b2f16 and f162b conversion operations
|
||
- glsl: Add ir_unop_f2fmp
|
||
- glsl/validate: Allow float16 in the expression tree
|
||
- glsl/lower_instructions: Use float16 constants when appropriate
|
||
- glsl/opt_minmax: Add support for float16
|
||
- glsl: Add a method to get precision from a deref instruction
|
||
- glsl/hierarchical_visitor: Call leave_callback on leaf nodes
|
||
- glsl: Add an IR lowering pass to convert mediump operations to 16-bit
|
||
- glsl/standalone: Add an option to lower the precision
|
||
- glsl: Add unit tests for the lower_precision pass
|
||
- freedreno/ir3: Lower bools to bitsize
|
||
- glsl: Inline builtins in a separate pass
|
||
- glsl/lower_precision: Lower builtins depending on arguments
|
||
- glsl/lower_precision: Use vector.back() instead of vector.end()[-1]
|
||
|
||
Paulo Zanoni (8):
|
||
|
||
- intel: fix the gen 11 compute shader scratch IDs
|
||
- intel: fix the gen 12 compute shader scratch IDs
|
||
- intel/device: bdw_gt1 actually has 6 eus per subslice
|
||
- anv: multiply the scratch space by 4 on gen9-10 like iris and i965
|
||
- iris: remove hole from struct iris_bo
|
||
- iris: remove unnecessary forward declaration
|
||
- iris: remove useless bo->gtt_offset assignment
|
||
- iris: make BATCH_SZ smaller by BATCH_RESERVED bytes
|
||
|
||
Peng Huang (1):
|
||
|
||
- radeonsi: make si_fence_server_signal flush pipe without work
|
||
|
||
Pierre Moreau (1):
|
||
|
||
- clover/nir: Check the result of spirv_to_nir
|
||
|
||
Pierre-Eric Pelloux-Prayer (44):
|
||
|
||
- radeonsi/ngg: add VGT_FLUSH when enabling fast launch
|
||
- radeonsi: test subsampled format in testdma
|
||
- format: add format_to_chroma_format
|
||
- gallium/video: remove pipe_video_buffer.chroma_format
|
||
- gallium/vl: add 4:2:2 support
|
||
- radeonsi: fix surf_pitch for subsampled surface
|
||
- st/va: enable 4:2:2 chroma format
|
||
- st/va: add support YUY2
|
||
- radeonsi: remove AMD_DEBUG=sisched option
|
||
- omx: fix build with gcc 10
|
||
- meson: enable -fno-common by default
|
||
- gitlab-ci: rules:changes to test on tested drivers changes
|
||
- vdpau: remove bogus assert
|
||
- st/mesa: disallow deferred flush if there are multiple contexts
|
||
- radeonsi: enable glsl_zero_init for Curse of the Dead Gods
|
||
- radeonsi: clarify the conditions when FLUSH_AND_INV_DB is needed
|
||
- util/os_file: extend os_read_file to return the file size
|
||
- util/u_process: add util_get_process_exec_path
|
||
- util/xmlconfig: add new sha1 application attribute
|
||
- radeonsi: enable workarounds for YoYo engine based games
|
||
- util/u_process: fix Windows build
|
||
- nir: update uses_demote flag in discard_to_demote pass
|
||
- ac: fix ac_build_is_helper_invocation when postponed_kill is null
|
||
- util: fix process_test path
|
||
- ddebug: add missing forward declaration
|
||
- radeon: fix includes
|
||
- radeonsi: switch to 3-spaces style
|
||
- radeon: switch to 3-spaces style
|
||
- gallium/util: let shader live cache users know if a hit occured
|
||
- radeonsi: dump shader stats when hitting the live cache
|
||
- util/xmlconfig: fix sha1 comparison code
|
||
- mesa: update pipeline when re-linking a program in use
|
||
- gallium/u_threaded: flush batch when hitting mapping limit
|
||
- radeonsi: use thread_context::bytes_mapped_limit
|
||
- radeonsi: don't assume ctx is always a threaded_context
|
||
- radeonsi: skip vs output optimizations for some outputs
|
||
- mesa: fix crash in find_value
|
||
- gallium/utils: silence strncpy warning
|
||
- st/omx: fix gcc warnings
|
||
- radeonsi: fix export count
|
||
- mesa: add gl_coontext::ForceIntegerTexNearest
|
||
- driconf: add force_integer_tex_nearest option
|
||
- radeonsi: don't print gs_copy_shader stats for shaderdb
|
||
- amd/addrlib: fix forgotten char -> enum conversions
|
||
|
||
Plamena Manolova (2):
|
||
|
||
- intel/compiler: Add support for variable workgroup size
|
||
- i965: Implement ARB_compute_variable_group_size
|
||
|
||
Qiang Yu (35):
|
||
|
||
- lima: remove definition of lima_is_scanout
|
||
- lima: use util_copy_framebuffer_state
|
||
- lima: always add texture bo to submit
|
||
- lima: remove lima_ctx_buff_va submit flags (v2)
|
||
- lima: pass array as parameter to PLBU and VS command macros
|
||
- lima: delay add plb buffer to submit when flush
|
||
- lima: delay plbu head command generation to flush stage (v2)
|
||
- lima: add render target to submit by dirty buffer flags
|
||
- lima: add missing resolve check for damage and reload
|
||
- lima: move syncobj from lima_submit to lima_context
|
||
- lima: merge gp/pp submit
|
||
- lima: put hardware related info to lima_gpu.h
|
||
- lima: move flush code to lima_submit.c
|
||
- lima: pass submit parameter for functions in lima_submic.c (v2)
|
||
- lima: add lima_submit_create_stream_bo
|
||
- lima: adjust pp_stream to use lima_submit_create_stream_bo
|
||
- lima: use lima_submit_create_stream_bo for plbu/vs_cmd and pp_stack
|
||
- lima: add lima_submit_get
|
||
- lima: make lima_submit one time use drop data (v3)
|
||
- lima: track write submits of context (v3)
|
||
- lima: move plbu/vs_cmd_array into lima_submit
|
||
- lima: move resolve into lima_submit
|
||
- lima: move pp_max_stack_size to lima_submit
|
||
- lima: move damage_rect into lima_submit
|
||
- lima: move clear into submit (v2)
|
||
- lima: move framebuffer info to lima_submit
|
||
- lima: use per submit dump file
|
||
- lima: optinal flush submit in lima_clear
|
||
- lima: enable multi submit optimization
|
||
- lima: move dump check to macro for lima_dump_command_stream_print
|
||
- lima: rename lima_submit to lima_job
|
||
- lima: fix buffer import with offset
|
||
- lima: also check tiled and depth case when import
|
||
- lima: set offset when export resource
|
||
- panfrost: don't always build bifrost_compiler
|
||
|
||
Quentin Glidic (1):
|
||
|
||
- meson: Use dependency.partial_dependency()
|
||
|
||
Rafael Antognolli (18):
|
||
|
||
- intel: Load the driver even if I915_PARAM_REVISION is not found.
|
||
- intel/tools: Update aubinator_error_decode.
|
||
- intel/blorp: Implement GEN:BUG:1605967699.
|
||
- iris: Apply the flushes when switching pipelines.
|
||
- anv: Wait for the GPU to be idle before invalidating the aux table.
|
||
- iris: Split aux map initialization from invalidation.
|
||
- iris: Wait for the GPU to be idle before invalidating the aux table.
|
||
- intel/isl: Implement D16_UNORM workarounds.
|
||
- intel/gen12+: Disable mid thread preemption.
|
||
- iris: Enable EXT_depth_bounds_test extension.
|
||
- drm-uapi: Update headers from Linux 5.7-rc1.
|
||
- i965/bufmgr: Factor out GEM_MMAP ioctl from mmap_cpu and mmap_wc.
|
||
- iris/bufmgr: Factor out GEM_MMAP ioctl from mmap_cpu and mmap_wc.
|
||
- i965/bufmgr: Add support for MMAP_OFFSET ioctl.
|
||
- iris/bufmgr: Add support for MMAP_OFFSET ioctl.
|
||
- anv: Add anv_device parameter to anv_gem_munmap.
|
||
- anv: Add support for new MMAP_OFFSET ioctl.
|
||
- anv: Enable HiZ on multi-layer depth buffers.
|
||
|
||
Rhys Perry (118):
|
||
|
||
- aco: fix gfx10_wave64_bpermute
|
||
- aco: gfx10_wave64_bpermute reduce op to print_ir
|
||
- aco: disable some instruction combining if it could change an exec
|
||
operand
|
||
- aco: improve SCC handling in some SALU combines
|
||
- nir: fix nir_const_value_as_uint bit size in load/store vectorizer
|
||
tests
|
||
- gitlab-ci: remove load_store_vectorizer from expected s390x test
|
||
failures
|
||
- aco: add RegisterFile
|
||
- aco: add some helpers for filling/testing register ranges
|
||
- aco: improve GFX9 1D ddx/ddy assertion
|
||
- spirv: improve creation of memory_barrier
|
||
- spirv: fix memory_barrier_tcs_patch emission
|
||
- aco: keep track of which events are used in a barrier
|
||
- aco: fix carry-out size for wave32 v_add_co_u32_e64
|
||
- aco: handle v_add_co_u32_e64 in parse_base_offset()
|
||
- aco: add new NOP insertion pass for GFX6-9
|
||
- aco: improve get_wait_states()
|
||
- aco: consider non-hazard writes in handle_raw_hazard_internal
|
||
- aco: improve control flow handling in GFX6-9 NOP pass
|
||
- aco: only reserve sgprs for vcc if it's used
|
||
- aco: fix uninitialized data error in waitcnt pass
|
||
- glsl/list: use uintptr_t for exec_node_data()'s subtraction
|
||
- aco: add helpers for moving instructions for scheduling
|
||
- aco: add helpers for ensuring correct ordering while scheduling
|
||
- aco: allow barriers to be skipped during scheduling
|
||
- aco: don't stop scheduling at exports
|
||
- aco: move some register demand helpers into aco_live_var_analysis.cpp
|
||
- aco: add a late kill flag
|
||
- aco: set late kill for v_interp_p1_f32 for some APUs
|
||
- aco: fix instruction encoding for LS VGPR init bug workaround
|
||
- aco: fix operand order for LS VGPR init bug workaround
|
||
- nir/gather_info: handle emit_vertex_with_counter
|
||
- radv: call nir_shader_gather_info again
|
||
- radv/winsys: set has_syncobj_wait_for_submit in the null winsys
|
||
- aco: set has_divergent_branch for discards in loops
|
||
- aco: handle missing second predecessors at merge block phis
|
||
- aco: handle when ACO adds new continue edges
|
||
- aco: skip NIR in unreachable merge blocks
|
||
- aco: improve check for unreachable loop continue blocks
|
||
- aco: emit IR in IF's merge block instead if the other side ends in a
|
||
jump
|
||
- aco: fix boolean undef regclass
|
||
- nir/gather_info: fix per-vertex handling in try_mask_partial_io
|
||
- aco: remove dead code in handle_operands()
|
||
- aco: implement 64-bit VGPR constant copies in handle_operands()
|
||
- aco: look at p_{extract,split}_vector's definitions in
|
||
pred_by_exec_mask()
|
||
- glsl: fix race in instance getters
|
||
- util/u_queue: fix race in total_jobs_size access
|
||
- radv: add code for exposing compiler statistics
|
||
- aco: add various compiler statistics
|
||
- aco: add vmem/smem score statistic
|
||
- radv, aco: collect statistics if requested but executables are not
|
||
- radv: fix null winsys gpu_info array
|
||
- aco: make PhysReg in units of bytes
|
||
- aco: add SDWA_instruction
|
||
- aco: print and validate opsel
|
||
- aco: add emission support for register-allocated sdwa sels
|
||
- aco: remove divergence check in sanitize_if()
|
||
- aco: zero-initialize Temp
|
||
- aco: improve vector optimization with sub-dword vectors
|
||
- aco: fix p_extract_vector validation
|
||
- aco: improve p_create_vector RA for sub-dword operands
|
||
- aco: clear moved operands in get_reg_create_vector()
|
||
- aco: fix 1D textureGrad() on GFX9
|
||
- aco: implement various 8/16-bit conversions
|
||
- aco: add missing scc clobber to nir_op_unpack_32_2x16_split_y
|
||
- aco: fix copy statistic for 64-bit vgpr constant copy
|
||
- aco: add VOP3P_instruction
|
||
- aco: implement sub-dword swaps
|
||
- aco: implement 64-bit sgpr swaps
|
||
- nir/lower_bit_size: fix lowering of shifts
|
||
- nir/lower_bit_size: fix lowering of {imul,umul}_high
|
||
- nir/algebraic: don't undo lowering of 8/16-bit comparisons to 32-bit
|
||
- aco: decrease the uses of other copy operations after
|
||
splitting/removing
|
||
- aco: copy-propagate p_create_vector copies of vectors
|
||
- aco: remove copy in load_input_from_temps()
|
||
- aco: move call to store_output_to_temps in store_ls_or_es_output
|
||
earlier
|
||
- aco: combine VALU and SALU into various VOP3 instructions
|
||
- aco: improve code for 32-bit isign
|
||
- aco: fix v_or(s_lshl) and v_add(s_lshl) optimizations
|
||
- aco: fix outdated label_vec from p_create_vector labelling
|
||
- radv: align buffer descriptor sizes to dword
|
||
- radv: allocate larger shader memory slabs if needed
|
||
- aco: be more careful about using SMEM for load_global
|
||
- aco: add and use RegClass::get() helper
|
||
- aco: add emit_load helper
|
||
- aco: refactor load_lds to use new helpers
|
||
- aco: use emit_load helper for VMEM/SMEM loads
|
||
- aco: add helpers for splitting stores
|
||
- aco: refactor store_lds() to use new helpers
|
||
- aco: refactor store_vmem_mubuf() to use new helpers
|
||
- aco: refactor visit_store_ssbo() to use new helpers
|
||
- aco: refactor visit_store_global() to use new helpers
|
||
- aco: refactor visit_store_scratch() to use new helpers
|
||
- aco: add and use get_buffer_store_op() helper
|
||
- aco: allow 8/16-bit shared loads
|
||
- aco: vectorize global loads/stores
|
||
- aco: handle undef p_create_vector operands in the optimizer
|
||
- aco: clobber scc in s_bfe_u32 in get_alu_src()
|
||
- aco: improve sub-dword emit_split_vector() with sgprs
|
||
- aco: lower 8/16-bit integer arithmetic
|
||
- radv/aco: enable 8/16-bit storage and int8/int16 on GFX8+
|
||
- aco: make RegisterFile::block() take a regclass
|
||
- aco: check alignment of non-subdword registers in get_reg_specified()
|
||
- aco: fix neighboring register check in get_reg_simple()
|
||
- aco: split self-intersecting copies instead of swapping
|
||
- aco: don't recurse in sub-dword get_reg_simple()
|
||
- aco: improve RA for uneven p_split_vector
|
||
- aco: add missing adjust_max_used_regs()
|
||
- aco: fix sub-dword out-of-bounds check in RA validator
|
||
- aco: fix sub-dword overwrite check in RA validator
|
||
- aco: add various GFX10 int16 opcodes
|
||
- aco: improve clamped integer addition disassembly workaround
|
||
- aco: fix vgpr nir_op_vecn with sgpr operands
|
||
- aco: consider blocks unreachable if they are in the logical cfg
|
||
- aco: remove use of f-strings
|
||
- aco: add message to static_assert
|
||
- nir: add missing group_memory_barrier handling
|
||
- nir/opt_if: run opt_peel_loop_initial_if after all other
|
||
optimizations
|
||
- nir: fix lowering to scratch with boolean access
|
||
|
||
Rob Clark (147):
|
||
|
||
- freedreno/drm: readonly cmdstream
|
||
- freedreno/ir3: shuffle a few ir3_register fields
|
||
- freedreno/ir3: cleanup after lower_locals_to_regs
|
||
- freedreno/ir3: fix crash when no non-input instructions
|
||
- freedreno/ir3: split out delay helpers
|
||
- freedreno/ir3: move nop padding to legalize
|
||
- freedreno/ir3: move block-scheduling into legalize
|
||
- freedreno/ir3: move atomic fixup after RA
|
||
- freedreno/ir3: a bit more optmsgs debug
|
||
- freedreno/ir3/ra: make use()/def() functions instead of macros
|
||
- freedreno/ir3: fix kill scheduling
|
||
- freedreno/ir3: post-RA sched pass
|
||
- freedreno/ir3: number instructions from one
|
||
- freedreno/ir3: add is_tex_or_prefetch()
|
||
- freedreno/ir3: don't precolor unused inputs
|
||
- freedreno/ir3: two pass register allocation
|
||
- freedreno/a6xx: fix lrz overflow
|
||
- freedreno/ir3: add RA sanity check
|
||
- freedreno/ir3: remove unused tex arg harder
|
||
- freedreno/ir3: create fragcoord instructions in input block
|
||
- freedreno/ir3: simplify split from collect
|
||
- freedreno/ir3: fix a dirty lie
|
||
- freedreno: allow ctx->batch to be NULL
|
||
- freedreno/ir3: fold const conversion into consumer
|
||
- freedreno: allow INVALID modifier
|
||
- freedreno/registers: teach gen_header.py about a3xx_regid
|
||
- freedreno/a6xx: few register updates
|
||
- freedreno: quiet INFO_MSG
|
||
- freedreno/registers: cleanup CP_SET_MARKER
|
||
- freedreno/computerator: import parser/lexer from fdre-a3xx
|
||
- freedreno/computerator: polish out some of the rust
|
||
- freedreno/computerator: rename prefix asm->ir3
|
||
- freedreno/ir3: allow block->predecessors to be null
|
||
- freedreno/computerator: add computerator
|
||
- freedreno/computerator: fix build dependency
|
||
- freedreno/ir3: remove from_tgsi
|
||
- freedreno/a6xx: remove unused param
|
||
- freedreno/a6xx: emit LRZ clear in sysmem too
|
||
- freedreno/a6xx: whitespace fix
|
||
- freedreno/a6xx: don't emit YIELD packet
|
||
- freedreno/a6xx: enable SKIP_IB2_ENABLE properly
|
||
- freedreno: honor FD_MESA_DEBUG=nogrow
|
||
- freedreno/ir3: remove regmask_set_if_not()
|
||
- freedreno/ir3: rewrite regmask to better support a6xx+
|
||
- freedreno/ir3: don't hide latency when there is none to hide
|
||
- freedreno/ir3: track half-precision live values
|
||
- freedreno/ir3: update SFU delay
|
||
- freedreno/ir3: fix crash with samgq workaround
|
||
- freedreno/ir3: don't precolor unassigned inputs
|
||
- freedreno/ir3: fix assert with getinfo
|
||
- freedreno/ir3: add assert
|
||
- nir/print: show variable precision
|
||
- freedreno/ir3: also lower lowp frag outputs
|
||
- freedreno/computerator: add hrsq/hlog2/hexp2
|
||
- freedreno/ir3: remove extra nops inserted in scheduler
|
||
- freedreno/ir3: add simplified stall estimation
|
||
- freedreno: fix FD_MESA_DEBUG=inorder
|
||
- util/ra: spiff out select_reg_callback
|
||
- util/ra: move NO_REG to header
|
||
- freedreno/ir3: split out has_latency_to_hide()
|
||
- freedreno/ir3: fix has_latency_to_hide
|
||
- freedreno/ir3: track register usage in first RA pass
|
||
- freedreno/ir3: round-robin RA
|
||
- freedreno/ir3: try to avoid syncs
|
||
- freedreno/computerator: add performance counter support
|
||
- freedreno/fdperf: set locale
|
||
- freedreno/a6xx: register update
|
||
- freedreno/ir3: small cleanup and comments
|
||
- freedreno/ir3: add bary_ij as src for meta:tex_prefetch
|
||
- freedreno/ir3: remove unused helper
|
||
- freedreno/ir3: fix bogus register footprint with tess/gs
|
||
- freedreno/ir3: reformat disasm output
|
||
- freedreno/ir3: convert debug bitfield to BITFIELD_BIT()
|
||
- freedreno/ir3/ra: add debug option for RA debug msgs
|
||
- freedreno/ir3/ra: split-up
|
||
- freedreno/ir3/ra: add helper to map name to instruction
|
||
- freedreno/ir3/ra: fix target register calculation
|
||
- freedreno/ir3/ra: add helper to map name to array
|
||
- freedreno/ir3/ra: drop extending output live-ranges
|
||
- freedreno/ir3/ra: add def/use iterators
|
||
- freedreno/ir3/ra: fix array liveranges
|
||
- freedreno/ir3/ra: compute register target from liveranges
|
||
- freedreno/ir3/ra: pick higher numbered scalars in first pass
|
||
- freedreno/ir3/ra: split building regs/classes and conflicts
|
||
- freedreno/ir3/ra: re-work a6xx merged register file conflicts
|
||
- gitlab-ci: disable vs2019 build
|
||
- freedreno: remove some obsolete debug options
|
||
- util: fix u_fifo_pop()
|
||
- freedreno: add logging infrastructure
|
||
- freedreno/a6xx: timestamp logging support
|
||
- freedreno: add some initial fd_log tracepoints
|
||
- freedreno/a6xx: add some more tracepoints
|
||
- freedreno/log: avoid duplicate ts's
|
||
- util: move ALIGN/ROUND_DOWN_TO to u_math.h
|
||
- freedreno/ir3: fix android build
|
||
- freedreno/log: fix build error
|
||
- nir: fix definition of imadsh_mix16 for vectors
|
||
- freedreno/ir3/cf: handle widening too
|
||
- freedreno/ir3: fixup cat3 32b vs 16b
|
||
- freedreno/ir3/cf: skip array load/store
|
||
- freedreno/ir3: add a pass to collect SSA uses
|
||
- freedreno/ir3/cf: use ssa-uses
|
||
- freedreno/a6xx: add some compute logging
|
||
- freedreno: fix missing locking
|
||
- freedreno/ir3: also precompile compute shaders for shaderdb
|
||
- freedreno: limit fp16 to frag and compute
|
||
- glsl: don't limit fp16 lowering to frag
|
||
- nir: add some swizzle helpers
|
||
- nir/lower_amul: fix slot calculation
|
||
- freedreno/log: android support
|
||
- freedreno/log: spiff out parser some more
|
||
- freedreno/log: better decoding for multiple chunks per batch
|
||
- freedreno/ir3: spiff out disasm a bit
|
||
- freedreno/ir3: make falsedep use's optional
|
||
- freedreno/ir3: simplify grouping pass
|
||
- freedreno/ir3: fix location of inserted mov's
|
||
- freedreno/ir3: new pre-RA scheduler
|
||
- freedreno/ir3/sched: awareness of partial liveness
|
||
- freedreno/ir3/postsched: remove some leftovers
|
||
- freedreno/ir3/postsched: avoid moving tex ahead of kill
|
||
- freedreno/ir3: add mov/cov stats
|
||
- freedreno/ir3/ra: handle array case for SFU select_reg opt
|
||
- freedreno/ir3: better cleanup when removing unused instructions
|
||
- freedreno/ir3: rename depth->dce
|
||
- freedreno/ir3/ra: cleanup some leftovers
|
||
- mesa: avoid redundant VBO updates
|
||
- mesa/st: avoid u_vbuf for GLES
|
||
- gallium: add # of MRT to blend state
|
||
- freedreno/computer: add script to test widening/narrowing
|
||
- freedreno/ir3/ra: remove unused variable
|
||
- freedreno/ir3/ra: use ir3_debug_print helper
|
||
- freedreno/ir3/ra: split out helper for array assignment
|
||
- freedreno/ir3/ra: only assign array base in first pass
|
||
- freedreno/a6xx+tu: rename VSC_DATA/VSC_DATA2
|
||
- freedreno: add helper to estimate # of bins per pipe
|
||
- freedreno/a6xx: pre-calculate expected vsc stream sizes
|
||
- freedreno/log-parser: support to read gzip'd logs
|
||
- freedreno: small whitespace fix
|
||
- freedreno: don't realloc idle bo's
|
||
- freedreno: mark more state dirty when rebinding resources
|
||
- freedreno: optimize rebind_resource()
|
||
- freedreno: rebind resource in all contexts
|
||
- freedreno: rebind_resource() \*before\* bo changes
|
||
- freedreno/a6xx: invalidate tex state cache entries on rebind
|
||
- freedreno: fix buffer import
|
||
- freedreno/ir3: fix indirect cb0 load_ubo lowering
|
||
- freedreno: clear last_fence after resource tracking
|
||
|
||
Rohan Garg (5):
|
||
|
||
- ci: Split out radv build-testing on arm64
|
||
- ci: Drop the git dependency in tracie
|
||
- tracie: Switch to using shutil.move for cross filesystem moves
|
||
- tracie: Print results in a machine readable format
|
||
- tracie: Reformat code to fix indentation
|
||
|
||
Roland Scheidegger (7):
|
||
|
||
- gallivm: fix crash with bptc border color sampling
|
||
- gallivm: fix crash in emit_get_buffer_size
|
||
- gallivm: disable rgtc/latc SNORM accellerated fetches
|
||
- gallium/util: Add back (and rename) util_float_to_half implementation
|
||
- gallivm: fix rgtc2 format
|
||
- gallivm: switch the mask6/mask7 cases for signed rgtc formats
|
||
- gallivm: fix stream id fetch
|
||
|
||
Roman Stratiienko (3):
|
||
|
||
- panfrost: Align Android makefiles with recent changes
|
||
- lima: Add missing source file to Android.mk
|
||
- panfrost: Align Android makefiles with recent changes
|
||
|
||
Sagar Ghuge (13):
|
||
|
||
- intel/isl: Move get_format_encoding function to isl
|
||
- intel/isl: Switch to R8_UNORM format for compatiblity
|
||
- intel/tools: Handle illegal instruction
|
||
- intel/tools: Handle STATE_REG in typed source operand
|
||
- intel/tools: Set correct address register file and number in i965_asm
|
||
- intel/tools: Add test for address register as source
|
||
- intel/tools: Add test for state register as source
|
||
- intel/tools: Print c_literals 4 byte wide
|
||
- intel/tools: Allow i965_disasm to disassemble c_literal input type
|
||
- intel/genxml: Add patch count threshold field on gen12
|
||
- intel/compiler: Track patch count threshold
|
||
- anv: Set patch count threshold in 3DSTATE_HS
|
||
- iris: Set patch count threshold in 3DSTATE_HS
|
||
|
||
Samuel Iglesias Gonsálvez (2):
|
||
|
||
- radv: check buffer size in vkCreateBuffer()
|
||
- radv: set sparseAddressSpaceSize to RADV_MAX_MEMORY_ALLOCATION_SIZE
|
||
|
||
Samuel Pitoiset (197):
|
||
|
||
- aco: fix MUBUF VS input loads when expanding vec3 to vec4 on GFX6
|
||
- aco: do not use ds_{read,write}2 on GFX6
|
||
- gitlab-ci: disable a630 tests as mesa-cheza is down (again)
|
||
- aco: fix waiting for scalar stores before "writing back" data on
|
||
GFX8-GFX9
|
||
- radv: make sure to not submit any IBs when RADV_FORCE_FAMILY is set
|
||
- radv: set the chip name to GCN-NOOP when RADV_FORCE_FAMILY is set
|
||
- aco: fix creating v_madak if v_mad_f32 has two sgpr literals
|
||
- nir: do not use De Morgan's Law rules for flt and fge
|
||
- radv: fix line width range and granularity
|
||
- radv: implement VK_EXT_line_rasterization
|
||
- radv: remove LLVM sicheduler enable for The Talos Principle
|
||
- radv: remove RADV_DEBUG=nosisched and RADV_PERFTEST=sisched
|
||
- radv: remove unused RADV_HASH_SHADER_IS_GEOM_COPY_SHADER
|
||
- radv: remove unnecessary RADV_DEBUG=nobatchchain option
|
||
- docs/new_features: empty the feature list for the 20.1 cycle
|
||
- radv: enable shaderStorageImageMultisample on GFX6-GFX7
|
||
- radv: enable VK_EXT_sampler_filter_minmax on GFX6
|
||
- radv: enable VK_NV_compute_shader_derivatives on GFX6-GFX7
|
||
- radv: add a comment about VK_AMD_mixed_attachment_samples on
|
||
GFX6-GFX7
|
||
- docs/envvars: document RADV_TEX_ANISO
|
||
- radv/winsys: add a new flag that requests zerovram allocations
|
||
- radv: use RADEON_FLAG_ZERO_VRAM when creating the trace BO
|
||
- radv: add the trace BO to the BO list at submit time
|
||
- radv: implement a dummy winsys for creating devices without AMDGPU
|
||
- ac,radeonsi: add ac_gpu_info::lds_size_per_cu
|
||
- ac: add more ac_gpu_info related shader fields
|
||
- radv/gfx10: adjust the number of simd per compute unit
|
||
- radv/gfx10: adjust SGPRs/VGPRs related info
|
||
- radv/gfx10: adjust the LDS size used to compute waves
|
||
- radv/gfx10: adjust the number of VGPRs used to compute waves
|
||
- radv: make use of ac_gpu_info::max_wave64_per_simd
|
||
- radv: fix creating null devices if KHR_display is enabled
|
||
- ac/llvm: fix 64-bit fmed3
|
||
- ac/llvm: fix 16-bit fmed3 on GFX8 and older gens
|
||
- ac/llvm: flush denorms for nir_op_fmed3 on GFX8 and older gens
|
||
- ac: add more fields to ac_gpu_info
|
||
- ac/registers: add definitions for thread trace
|
||
- radv: add a small helper that allows to submit internal CS
|
||
- radv: add initial SQ Thread Trace support for GFX9
|
||
- radv: emit thread trace markers after every draw/dispatch call
|
||
- radv: add initial SQTT files generation support
|
||
- radv: allow to capture SQTT traces with
|
||
RADV_THREAD_TRACE=<start_frame>
|
||
- radv: fix 32-bit build failure in radv_queue_internal_submit()
|
||
- radv: fix size of sqtt_file_chunk_asic_info on 32-bit system
|
||
- radv/rgp: adjust trace memory/shader clocks to fix frame duration
|
||
- radv/sqtt: do not assume that the number of shader engines is 4
|
||
- radv/sqtt: update SPI_CONFIG_CNTL.EXP_PRIORITY_ORDER value
|
||
- ac/registers: add definitions for thread trace on GFX10
|
||
- radv/sqtt: add support for GFX10
|
||
- radv: update entrypoints generation from ANV
|
||
- ac: rename lds_size_per_cu to lds_size_per_workgroup
|
||
- ac: rename vgpr_alloc_granularity to wave64_vgpr_alloc_granularity
|
||
- ac: rename min_vgpr_alloc to min_wave64_vgpr_alloc
|
||
- aco: fix image load/store with lod and 1D images
|
||
- gitlab-ci: build Fossilize in the test image for VK
|
||
- gitlab-ci: add Fossilize support to detect compiler regressions
|
||
- gitlab-ci: enable building the test image for VK unconditionally
|
||
- gitlab-ci: add a job that runs Fossilize on RADV/Polaris10
|
||
- radv/winsys: fix missing initializations of shader info in the null
|
||
device
|
||
- radv/sqtt: fix wrong check in radv_is_thread_trace_complete()
|
||
- radv/sqtt: tidy up radv_emit_thread_trace_{start,stop}
|
||
- radv/sqtt: add radv_copy_thread_trace_info_regs() helper
|
||
- ac/registers: adjust some definitions for thread trace on GFX8
|
||
- radv/sqtt: add support for GFX8
|
||
- radv/sqtt: abort if SQTT is used on GFX6-GFX7
|
||
- ac: add ac_gpu_info::cu_mask to store bitmask of compute units
|
||
- radv/rgp: report correct cu_mask info
|
||
- radv/rgp: report correct system ram size
|
||
- nir/lower_input_attachments: remove bogus assert in
|
||
try_lower_input_texop()
|
||
- radv/entrypoints: declare a driver internal layer for SQTT
|
||
- radv: use device entrypoints from the SQTT layer if enabled
|
||
- radv/sqtt: add a helper that emits thread trace userdata markers
|
||
- radv: initial implementation of the driver internal layer SQTT
|
||
- radv/sqtt: describe begin/end command buffers with user markers
|
||
- radv/sqtt: describe draw/dispatch and emit event markers
|
||
- radv/sqtt: describe render pass color/depthstencil clears
|
||
- radv/rgp: bump the instrumentation spec version to 1
|
||
- radv/sqtt: describe pipeline and wait events barriers
|
||
- gitlab-ci: add rules:changes for RADV
|
||
- radv: do not recursively begin/end render pass for meta operations
|
||
- radv: fix 32-bits build (again)
|
||
- gitlab-ci: build RADV in meson-i386 to avoid 32-bit build failures
|
||
- ac/llvm: add missing optimization barrier for 64-bit readlanes
|
||
- radv/sqtt: describe begin/end subpass barriers with user markers
|
||
- radv/sqtt: describe layout transitions with user markers
|
||
- radv/gfx10: cache metadata in L2 on small chips
|
||
- radv: use better tessellation tunables on GFX9+
|
||
- radv: tune primitive binning for small chips
|
||
- radv: rewrite late alloc computation
|
||
- radv: use ac_gpu_info::use_late_alloc
|
||
- radv: cleanup occurences of use_aco everywhere
|
||
- radv: remove radv_shader_variant::aco_used
|
||
- radv: remove unnecessary LLVM includes
|
||
- radv: add llvm_compiler_shader() helper
|
||
- gitlab-ci: remove useless 'patch' package in the VK test image
|
||
- gitlab-ci: allow deqp-runner to use the maximum number of jobs
|
||
- gitlab-ci: do not set the number of deqp-parallel jobs for RADV CTS
|
||
- gitlab-ci: bump Vulkan CTS to 1.2.1.0
|
||
- radv/sqtt: handle thread trace capture in sqtt_QueuePresentKHR()
|
||
- radv: only inject implicit subpass dependencies if necessary
|
||
- radv/gfx10: fix required subgroup size with
|
||
VK_EXT_subgroup_size_control
|
||
- radv/gfx10: fix required ballot size with
|
||
VK_EXT_subgroup_size_control
|
||
- radv: fix random depth range unrestricted failures due to a cache
|
||
issue
|
||
- radv: remove wrong assert that checks compute subgroup size
|
||
- radv: fix optional pSizes parameter when binding streamout buffers
|
||
- radv/winsys: fix wrong PCI ID for Vega10 in the null winsys
|
||
- radv/winsys: spoof some values for num_render_backends in the null
|
||
winsys
|
||
- gitlab-ci: compile fossils with both RADV compiler backends
|
||
(LLVM/ACO)
|
||
- gitlab-ci: compile fossils with more ASICs
|
||
- gitlab-ci: add a new stage for RADV CI
|
||
- gitlab-ci: add a bunch of new fossils from the Sascha Vulkan demos
|
||
- radv/llvm: fix subgroup shuffle for chips without bpermute
|
||
- radv: enable VK_KHR_8bit_storage on GFX6-GFX7
|
||
- ac/nir: use llvm.amdgcn.rcp for nir_op_frcp
|
||
- ac/nir: use llvm.amdgcn.rsq for nir_op_frsq
|
||
- ac/nir: use llvm.amdgcn.rcp in ac_build_fdiv()
|
||
- nir/algebraic: add fexp2(fmul(flog2(a), 0.5) -> fsqrt(a) optimization
|
||
- aco: only break SMEM clauses if XNACK is enabled (mostly APUs)
|
||
- aco: always optimize v_mad to v_madak in presence of literals
|
||
- ac/nir: split 8-bit load/store to global memory on GFX6
|
||
- ac/nir: split 8-bit SSBO stores on GFX6
|
||
- radv/llvm: enable 8-bit storage features on GFX6-GFX7
|
||
- ac/nir: split 16-bit load/store to global memory on GFX6
|
||
- ac/nir: split 16-bit SSBO stores on GFX6
|
||
- radv/llvm: enable 16-bit storage features on GFX6-GFX7
|
||
- radv: rename decompress/resummarize depth/stencil functions
|
||
- radv: rename extra graphics pipeline decompress/resummarize fields
|
||
- radv: cleanup creating the decompress/resummarize pipelines
|
||
- radv: remove radv_layout_has_htile() helper
|
||
- radv: enable lowering of GS intrinsics for the LLVM backend
|
||
- ac,radv: add ac_gpu_info::has_double_rate_fp16
|
||
- radv: only expose shaderFloat16 for chips with double rate fp16
|
||
- radv: only expose storageInputOutput16 for chips with double rate
|
||
fp16
|
||
- radv: only expose fp16 control features for chips with double rate
|
||
fp16
|
||
- radv: only enable TC-compat HTILE for images readable by a shader
|
||
- radv: allow TC-compat HTILE with GENERAL outside of render loops
|
||
- aco: implement 16-bit nir_op_frexp_sig/nir_op_frexp_exp
|
||
- aco: implement 16-bit nir_op_ffract
|
||
- aco: implement 16-bit nir_op_fexp2/nir_op_flog2
|
||
- aco: implement 16-bit nir_op_ftrunc/nir_op_fround_even
|
||
- aco: implement 16-bit nir_op_fsqrt/nir_op_frcp/nir_op_frsq
|
||
- aco: implement 16-bit nir_op_ffloor/nir_op_fceil
|
||
- aco: implement 16-bit nir_op_fmax/nir_op_fmin
|
||
- aco: implement 16-bit nir_op_fabs/nir_op_fneg
|
||
- aco: implement 16-bit nir_op_fsub/nir_op_fadd
|
||
- aco: implement 16-bit nir_op_fcos/nir_op_fsin
|
||
- aco: implement 16-bit nir_op_fmul
|
||
- aco: implement 16-bit nir_op_fsat
|
||
- aco: implement 16-bit nir_op_fsign
|
||
- aco: implement 16-bit nir_op_bcsel
|
||
- aco: implement 16-bit nir_op_f2i32/nir_op_f2u32
|
||
- aco: implement 16-bit nir_op_ldexp
|
||
- aco: implement 16-bit nir_op_fmax3/nir_op_fmin3/nir_op_fmed3
|
||
- aco: implement 16-bit comparisons
|
||
- aco: implement nir_op_b2f16/nir_op_i2f16/nir_op_u2f16
|
||
- aco: fix f2i64/f2u64 with sgprs if the exponent computation overflow
|
||
- aco: implement 16-bit nir_op_f2i64/nir_op_f2u64
|
||
- aco: fix nir_op_pack_32_2x16_split if one operand is a constant
|
||
- radv: add radeon_set_context_reg_rmw() helper
|
||
- radv: use RMW packets for updating the maximum sample distance
|
||
- aco: fix nir_op_frexp_exp with 16-bit floats and negative exponents
|
||
- radv/aco: do not advertise VK_KHR_shader_subgroup_extended_types
|
||
- aco: implement nir_op_f2i8/nir_op_f2u8
|
||
- aco: fix emitting stream output with tess eval shaders
|
||
- radv: do not abort with unknown/unimplemented descriptor types
|
||
- radv: fix geometry shader primitives query with ACO on GFX10
|
||
- radv: set missing SHARED_VGPR_CNT for NGG VS and ACO
|
||
- radv/llvm: fix exporting the viewport index if the fragment shader
|
||
needs it
|
||
- aco: fix exporting the viewport index if the fragment shader needs it
|
||
- nir/lower_int64: lower imin3/imax3/umin3/umax3/imed3/umed3
|
||
- nir/opt_algebraic: lower 64-bit fmin3/fmax3/fmed3
|
||
- gitlab-ci: add a list of excluded tests for RADV
|
||
- radv: make sure to export the viewport index if FS needs it
|
||
- radv: simplify checking for Navi1x chips
|
||
- radv: adjust the supported subgroup stages
|
||
- radv: fix robust_buffer_access if enabled via
|
||
VkPhysicalDeviceFeatures2
|
||
- gitlab-ci: add lists of expected failures for RADV CI
|
||
- ac,radeonsi: fix compilations issues with LLVM 11
|
||
- radv: do not expose GTT as device local memory mostly for APUs
|
||
- radv: enable FMASK for color attachments only
|
||
- radv: remove unused radv_device_memory::map_size field
|
||
- radv: track memory heaps usage if overallocation is explicitly
|
||
disallowed
|
||
- radv: advertise VK_AMD_memory_overallocation_behavior
|
||
- ac/llvm: fix nir_texop_texture_samples with NULL descriptors
|
||
- aco: fix nir_texop_texture_samples with NULL descriptors
|
||
- aco: fix adjusting the sample index with FMASK if value is negative
|
||
- radv: handle NULL descriptors
|
||
- radv: handle NULL vertex bindings
|
||
- radv: advertise VK_EXT_robustness2
|
||
- gitlab-ci: add a list of expected failures for FIJI with ACO
|
||
- ci: fix reporting the number of unexpected/flakes
|
||
- radv: report INITIALIZATION_FAILED when the amdgpu winsys init failed
|
||
- radv: don't report error with other vendor DRM devices
|
||
- aco: fix 64-bit trunc with negative exponents on GFX6
|
||
- radv: limit the Vulkan version to 1.1 for Android
|
||
- radv: handle different Vulkan API versions correctly
|
||
- radv: update the list of allowed Android extensions
|
||
|
||
Satyajit Sahu (1):
|
||
|
||
- st/va: GetConfigAttributes: check profile and entrypoint combination
|
||
|
||
Simon Ser (1):
|
||
|
||
- mesa: add support for NV_pixel_buffer_object
|
||
|
||
Simon Zeni (1):
|
||
|
||
- mesa: enable GL_EXT_draw_instanced for gles2
|
||
|
||
Sonny Jiang (1):
|
||
|
||
- radeonsi: enable EXT_texture_shadow_lod
|
||
|
||
Szymon Andrzejuk (1):
|
||
|
||
- virgl: Use align_free for align_malloc allocated buffer
|
||
|
||
Tapani Pälli (27):
|
||
|
||
- intel/vec4: fix valgrind errors with vf_values array
|
||
- glsl: fix a memory leak with resource_set
|
||
- iris: fix aux buf map failure in 32bits app on Android
|
||
- mesa: introduce boolean toggle for EXT_texture_norm16
|
||
- i965: toggle on EXT_texture_norm16
|
||
- mesa/st: toggle EXT_texture_norm16 based on format support
|
||
- mesa/st: fix formats required for EXT_texture_norm16
|
||
- nir: fix compilation warning on glsl_get_internal_ifc_packing
|
||
- iris: toggle on PIPE_CAP_MIXED_COLOR_DEPTH_BITS
|
||
- nir/glsl: gather bitmask of images used by program
|
||
- iris: use the images_used mask in resolve pass
|
||
- intel/compiler: detect if atomic load store operations are used
|
||
- iris: provide dummy iris_image_view_aux_usage
|
||
- iris: move existing image format fallback as a helper function
|
||
- iris: determine aux usage during predraw and state setup
|
||
- isl: allow compression for storage images on gen12+
|
||
- iris: allow compression conditionally for images on gen12
|
||
- glsl: set error_emitted true if type not ok for assignment
|
||
- mesa/st: unbind shader state before deleting it
|
||
- mesa/st: release variants for active programs before unref
|
||
- mesa: remove redudant check
|
||
- mesa: remove redudant assignment
|
||
- glsl: remove redudant assignment
|
||
- glsl: stop processing function parameters if error happened
|
||
- mesa/st: initialize all winsys_handle fields for memory objects
|
||
- anv: remove assert from GetImageMemoryRequirements[2]
|
||
- st/mesa: destroy only own program variants when program is released
|
||
|
||
Thomas Hellstrom (5):
|
||
|
||
- svga: Fix banded DMA upload
|
||
- svga, winsys/svga: Fix persistent memory discard maps
|
||
- svga: Treat forced coherent maps as maps of persistent memory
|
||
- gallium/pipebuffer: Use persistent maps for slabs
|
||
- winsys/svga: Optionally avoid caching buffer maps
|
||
|
||
Thong Thai (7):
|
||
|
||
- Revert "st/va: Convert interlaced NV12 to progressive"
|
||
- gallium/auxiliary/vl: fix bob compute shaders for deint yuv
|
||
- st/va: remove unneeded code
|
||
- st/va/postproc: reallocate interlaced destination buffer
|
||
- radeonsi: add 10-bit HEVC encode support for VCN2.0 devices
|
||
- radeon: add support for 10-bit HEVC encoding to VCN 2.0
|
||
- st/va: add check for P010 and P016 encode/decode support
|
||
|
||
Timothy Arceri (51):
|
||
|
||
- glsl: fix gl_nir_set_uniform_initializers() for image arrays
|
||
- glsl: fix possible memory leak in nir uniform linker
|
||
- glsl: set the correct number of samplers in a shader
|
||
- glsl: set the correct number of images in a shader
|
||
- glsl: fix resizing of the uniform remap table
|
||
- glsl: reset next_image_index count for each shader stage
|
||
- glsl: fix sampler index calculation in nir linker
|
||
- glsl: add some error checks to the nir uniform linker
|
||
- glsl: move nir link uniforms struct defs earlier
|
||
- glsl: move add_parameter() earlier in nir link uniforms
|
||
- glsl: move get_next_index() earlier in nir link uniforms
|
||
- glsl: add name support to nir uniform linker
|
||
- glsl: correctly find block index when linking glsl with nir linker
|
||
- nir: add glsl_get_internal_ifc_packing() helper
|
||
- nir: add glsl_get_std140_base_alignment() helper
|
||
- nir: add glsl_get_std140_size() helper
|
||
- nir: add glsl_get_std430_base_alignment() helper
|
||
- nir: add glsl_get_std430_size() helper
|
||
- glsl: add std140 and std430 layouts to nir uniform linker
|
||
- glsl: correctly set explicit offsets for struct members
|
||
- glsl: find the base offset for block members from unnamed blocks
|
||
- glsl: nir linker fix setting of ssbo top level array
|
||
- glsl: set ShaderStorageBlocksWriteAccess in the nir linker
|
||
- glsl: add support for builtins to the nir uniform linker
|
||
- glsl: dont try to assign uniform storage for uniform blocks
|
||
- glsl: add subroutine support to nir linker
|
||
- glsl: fix varying packing for 64bit integers
|
||
- nir: fix packing of TCS varyings not read by the TES
|
||
- nir: fix crash in varying packing on interface mismatch
|
||
- glsl_to_nir: remove dead code
|
||
- radeonsi: don't lower constant arrays to uniforms in GLSL IR
|
||
- nir: make opt_if_loop_terminator() less strict
|
||
- nir: add matrix_layout to nir_variable data
|
||
- glsl: fix struct offsets in the nir uniform linker
|
||
- glsl: tidy up uniform storage value count code in NIR linker
|
||
- Revert "glsl: fix resizing of the uniform remap table"
|
||
- glsl: fix explicit locations for the glsl linker
|
||
- glsl: error check max user assignable uniform locations
|
||
- glsl: fix block index in NIR uniform linker
|
||
- glsl: pull mark_array_elements_referenced() out into common helper
|
||
- glsl: only set stage ref when uniforms referenced in stage
|
||
- nir/gcm: allow derivative dependent intrinisics to be moved earlier
|
||
- nir/gcm: be more conservative about moving instructions from loops
|
||
- nir/gcm: dont move movs unless we can replace them later with their
|
||
src
|
||
- glsl: add bindless support to nir uniform linker
|
||
- glsl: fix gl_nir_set_uniform_initializers() for bindless textures
|
||
- st/glsl_to_nir: make use of nir linker for linking uniforms
|
||
- glsl: some nir uniform linker fixes
|
||
- glsl: remove some duplicate code from the nir uniform linker
|
||
- glsl: stop cascading errors if process_parameters() fails
|
||
- glsl: fix slow linking of uniforms in the nir linker
|
||
|
||
Timur Kristóf (90):
|
||
|
||
- aco/optimizer: Don't combine uniform bool s_and to s_andn2.
|
||
- radv: Move some helper functions to the radv_shader.h header file.
|
||
- aco: Extract setup_gs_variables into a separate function.
|
||
- aco: Setup tessellation control shader variables.
|
||
- aco: Implement load_tess_coord.
|
||
- aco: Implement load_primitive_id for tessellation shaders.
|
||
- aco: Implement load_patch_vertices_in.
|
||
- aco: Implement load_invocation_id for tessellation control shaders.
|
||
- aco: Implement control_barrier for tessellation control shaders.
|
||
- aco: Implement memory_barrier_tcs_patch.
|
||
- aco: Implement load_view_index for TCS and TES.
|
||
- aco: Setup correct HW stages when tessellation is used.
|
||
- aco: Use mesa shader stage when loading inputs.
|
||
- aco: Remove vertex_geometry_gs assertion from merged shaders.
|
||
- aco: Extract LDS alignment calculation to a separate function.
|
||
- aco: Remove esgs_itemsize from LDS alignment calculation.
|
||
- aco: Introduce new VMEM load/store helpers.
|
||
- aco: Introduce new helpers for calculating address offsets.
|
||
- aco: Refactor load_per_vertex_input in preparation for tessellation.
|
||
- aco: Refactor VS output stores in preparation for tessellation.
|
||
- aco: Slight fix to lds_store and lds_load.
|
||
- aco: Fix combining DS additions in the optimizer.
|
||
- aco: Implement tessellation control shader input/output.
|
||
- aco: Store VS outputs correctly when tessellation is used.
|
||
- aco: Fix LS VGPR init bug on affected hardware.
|
||
- radv: Enable ACO for tessellation control shaders.
|
||
- aco: Setup tessellation evaluation shader variables.
|
||
- aco: Use TES output info when TES runs on the VS stage.
|
||
- aco: Store TES outputs when TES runs on the HW VS stage.
|
||
- aco: Enable streamout when TES runs on the HW VS stage.
|
||
- aco: Implement loading TES inputs.
|
||
- radv: Enable ACO for TES when there is no GS.
|
||
- aco: Enable running TES as ES, including merged TES+GS.
|
||
- radv: Enable ACO on all stages.
|
||
- aco: Don't generate an if when the first part of a merged HS or GS is
|
||
empty.
|
||
- aco: Store tess factors in VMEM only at the end of the shader.
|
||
- aco: Only write TCS outputs to LDS when they are read by the TCS.
|
||
- aco: Don't store TCS outputs to LDS when we're sure that none are
|
||
read.
|
||
- nir: Add ability to lower non-const quad broadcasts to const ones.
|
||
- radv: Enable lowering dynamic quad broadcasts.
|
||
- radv: Enable subgroup shuffle on GFX10 when ACO is used.
|
||
- aco: Create null exports in instruction selection instead of
|
||
assembler.
|
||
- aco: Extract tcs_driver_location_matches_api_mask to separate
|
||
function.
|
||
- aco: Fix handling of tess factors.
|
||
- aco: Allow combining TCS output VMEM stores.
|
||
- aco: Allow combining LDS loads when loading tess factors.
|
||
- aco: Skip 2nd read of merged wave info when TCS in/out vertices are
|
||
equal.
|
||
- aco: Use more optimal sequence at the beginning of merged shaders.
|
||
- nir: Collect if shader uses cross-invocation or indirect I/O.
|
||
- aco: Treat outputs of the previous stage as inputs of the next stage.
|
||
- aco: Change isel inputs/outputs to a flat array.
|
||
- aco: Zero-fill undefined elements in create_vec_from_array.
|
||
- aco: Extract setup_tcs_info to a separate function.
|
||
- aco: Fix workgroup size calculation.
|
||
- aco: Extract store_output_to_temps into a separate function.
|
||
- aco: When LS and HS invocations are the same, pass LS outputs in
|
||
temps.
|
||
- aco: Don't store LS VS outputs to LDS when TCS doesn't need them.
|
||
- aco: Fix crash in insert_wait_states.
|
||
- aco: Extract uniform if handling to separate functions.
|
||
- aco: Print block_kind_export_end.
|
||
- aco: Extract merged_wave_info_to_mask to its own function.
|
||
- aco: Treat s_setprio as a scheduling barrier.
|
||
- aco/ngg: Add new stage for hw_ngg_gs.
|
||
- aco/ngg: Initialize exec mask for NGG VS and TES.
|
||
- aco/ngg: Fix exports for NGG VS and TES.
|
||
- aco/ngg: Setup NGG VS and TES stages.
|
||
- aco/ngg: Implement NGG VS and TES.
|
||
- aco/ngg: Schedule position exports of NGG VS/TES.
|
||
- aco/ngg: Run GS_ALLOC_REQ on priority 3 for NGG VS and TES.
|
||
- radv: Enable ACO for NGG VS/TES, but disable NGG for ACO GS.
|
||
- aco: Print shader stage in aco_print_program.
|
||
- radv: Print shader stage before disassembly.
|
||
- radv: Add inputs read by TES to radv_shader_info.
|
||
- aco: Only store TCS outputs to VMEM when they are read by TES.
|
||
- aco: Increase barrier_count to 7 to include barrier_barrier.
|
||
- aco: Abort when RA can't find a register.
|
||
- aco: Const correctness for get_barrier_interaction.
|
||
- aco: Const correctness for aco_print_ir.
|
||
- aco: Use 24-bit multiplication in TCS I/O
|
||
- aco: Use 24-bit multiplication for NGG wave id and thread id.
|
||
- aco: Move s_setprio to correct place after the gs_alloc_req.
|
||
- radv: Refactor calculate_tess_lds_size and get_tcs_num_patches.
|
||
- aco: Use context variables instead of calculating TCS inputs/outputs.
|
||
- aco: Remember VS/TCS output driver locations.
|
||
- aco: Calculate workgroup size of legacy GS.
|
||
- aco: Set config->lds_size when TES or VS is running on HW ESGS.
|
||
- nir: Add new linking helper to set linked driver locations.
|
||
- radv: Use new linking helper to set default driver locations.
|
||
- aco: Use new default driver locations.
|
||
- radv: Use smaller esgs_itemsize for ACO.
|
||
|
||
Tobias Jakobi (1):
|
||
|
||
- meson: Link Gallium Nine with ld_args_build_id
|
||
|
||
Tomasz Pyra (1):
|
||
|
||
- gallium/swr: spin-lock performance improvement
|
||
|
||
Tomeu Vizoso (34):
|
||
|
||
- panfrost: Print intended field when decoding
|
||
- panfrost: Add more info to some assertions
|
||
- pan/midgard: Handle nir_intrinsic_load_barycentric_centroid
|
||
- panfrost: Use DBG macro to avoid noise in the console
|
||
- panfrost: Fix decoding of tiled 3D textures
|
||
- panfrost: Only clamp the LOD to disable mipmapping when needed
|
||
- gitlab-ci: Switch kernel for LAVA jobs to 5.5
|
||
- gitlab-ci: Disable the lima job for now
|
||
- gitlab-ci: Run GLES3 tests in dEQP on Panfrost
|
||
- panfrost: Remove some more prints to stdout
|
||
- gitlab-ci: Move to 5.5 kernel plus fixes for Panfrost
|
||
- gitlab-ci: Use PAN_MESA_DEBUG=gles3 for Panfrost
|
||
- gitlab-ci: Remove GLES3 test from Panfrost fails list
|
||
- gitlab-ci: Skip dEQP-GLES3.functional.shaders.derivate.\*
|
||
- gallium: Add forgotten docs for new CAPs related to transform
|
||
feedback
|
||
- gitlab-ci: Update renderdoc
|
||
- gitlab-ci: Use surfaceless platform also for apitrace
|
||
- gitlab-ci: Place files from the Mesa repo into the build tarball
|
||
- gitlab-ci: Serve files for LAVA via separate service
|
||
- gitlab-ci: Disable jobs for Collabora's LAVA lab
|
||
- Revert "gitlab-ci: Disable jobs for Collabora's LAVA lab"
|
||
- panfrost: Remove most usage of midgard_payload_vertex_tiler
|
||
- panfrost: Pass IS_BIFROST to pandecode_jc
|
||
- panfrost: Don't emit write_value jobs on Bifrost
|
||
- panfrost: On Bifrost, set the right tiler descriptor
|
||
- gitlab-ci: Test virgl driver
|
||
- panfrost: Clean up a bit the tiler structs for Bifrost
|
||
- panfrost: Emit sampler descriptor on bifrost
|
||
- panfrost: Emit texture descriptor on bifrost
|
||
- gitlab-ci: Update virglrenderer in the x86_test-gl image
|
||
- gitlab-ci: Allow test jobs to add options to the dEQP invocation
|
||
- gitlab-ci: Test OpenGL ES 3.1 on virgl
|
||
- gitlab-ci: Test Virgl with traces
|
||
- panfrost: Add Bifrost texture trampoline BO to batch
|
||
|
||
Uros Bizjak (1):
|
||
|
||
- doc: Update features.txt for r600 with misc supported features
|
||
|
||
Vasily Khoruzhick (19):
|
||
|
||
- lima: handle early-z and pixel kill better
|
||
- lima: implement PLB PP stream cache
|
||
- lima: add RGBA5551 and RGBA4444 formats
|
||
- lima: don't disable tiling if there's linear modifier in list
|
||
- lima: gpir: enforce instruction limit earlier
|
||
- panfrost: split index cache into shared part
|
||
- lima: enable minmax cache for index buffers
|
||
- lima: print gp uniforms if gp debug is enabled
|
||
- lima/gpir: improve disassembler output
|
||
- lima/gpir: print acc ops even if we have only one source
|
||
- lima/gpir: kill dead writes to regs in DCE
|
||
- lima/gpir: add better lowering for ftrunc
|
||
- lima/gpir: fix crash in schedule_insert_ready_list()
|
||
- lima: disable Z16 format
|
||
- lima: decode depth/stencil write bits in RSW
|
||
- lima: split pixel and texel format tables
|
||
- lima: add support for R and RG formats
|
||
- lima: Implement lima_texture_subdata
|
||
- lima: avoid situations when scissor minx > maxx or miny > maxy
|
||
|
||
Veerabadhran (1):
|
||
|
||
- radeon/vce: Move global function pointer si_get_pic_param to local
|
||
encoder structure Multi gpu use case broken when the function was
|
||
global
|
||
|
||
Vilya Harvey (1):
|
||
|
||
- zink. Don't set incorrect sType in VkImportMemoryFdInfoKHR struct
|
||
|
||
Vinson Lee (16):
|
||
|
||
- swr: Fix build with GCC 10.
|
||
- lima: Fix build with GCC 10.
|
||
- swr: Fix GCC 4.9 checks.
|
||
- panfrost: Remove unused anonymous enum variables.
|
||
- meson: Enable -Wno-deprecated only for bison > 2.3.
|
||
- swr: Fix non-pod-varargs error.
|
||
- st/nine: Fix incompatible-pointer-types-discards-qualifiers errors.
|
||
- panfrost: Fix gnu-empty-initializer error.
|
||
- util/u_process: Add util_get_process_exec_path for macOS.
|
||
- mesa: Change \_mesa_exec_malloc argument type.
|
||
- gallivm: Add missing header for powf.
|
||
- swr/rasterizer: Use private functions for min/max to avoid namespace
|
||
issues.
|
||
- swr: Remove Byte Order Mark.
|
||
- r600/sfn: Initialize VertexStageExportForGS m_num_clip_dist member
|
||
variable.
|
||
- r600/sfn: Use correct setter method.
|
||
- freedreno: Add missing va_end.
|
||
|
||
Yevhenii Kolesnikov (1):
|
||
|
||
- intel/compiler: fix cmod propagation optimisations
|
||
|
||
Zhang, Boyuan (1):
|
||
|
||
- radeonsi: Add support for midstream bitrate change in encoder
|
||
|
||
luc (1):
|
||
|
||
- zink: confused compilation macro usage for zink in target helpers.
|