1Mesa 19.2.0 Release Notes / 2019.09.25 2====================================== 3 4Mesa 19.2.0 is a new development release. People who are concerned with 5stability and reliability should stick with a previous release or wait 6for Mesa 19.2.1. 7 8Mesa 19.2.0 implements the OpenGL 4.5 API, but the version reported by 9glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) / 10glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being 11used. Some drivers don't support all the features required in OpenGL 124.5. OpenGL 4.5 is **only** available if requested at context creation. 13Compatibility contexts may report a lower version depending on each 14driver. 15 16Mesa 19.2.0 implements the Vulkan 1.1 API, but the version reported by 17the apiVersion property of the VkPhysicalDeviceProperties struct depends 18on the particular driver being used. 19 20SHA256 checksums 21---------------- 22 23:: 24 25 b060caa2a00f856431160ff7377d0e8f58f2aa48c16ee5a9e265ebdccb10852a mesa-19.2.0.tar.xz 26 27New features 28------------ 29 30- GL_ARB_post_depth_coverage on radeonsi (Navi) 31- GL_ARB_seamless_cubemap_per_texture on etnaviv (if GPU supports 32 SEAMLESS_CUBE_MAP) 33- GL_EXT_shader_image_load_store on radeonsi (with LLVM >= 10) 34- GL_EXT_shader_samples_identical on iris and radeonsi (if using NIR) 35- GL_EXT_texture_shadow_lod on i965, iris 36- EGL_EXT_platform_device 37- VK_AMD_buffer_marker on radv 38- VK_EXT_index_type_uint8 on radv 39- VK_EXT_post_depth_coverage on radv 40- VK_EXT_queue_family_foreign on radv 41- VK_EXT_sample_locations on radv 42- VK_EXT_shader_demote_to_helper_invocation on Intel. 43- VK_KHR_depth_stencil_resolve on radv 44- VK_KHR_imageless_framebuffer on radv 45- VK_KHR_shader_atomic_int64 on radv 46- VK_KHR_uniform_buffer_standard_layout on radv 47 48Bug fixes 49--------- 50 51- `Bug 103674 <https://bugs.freedesktop.org/show_bug.cgi?id=103674>`__ 52 - u_queue.c:173:7: error: implicit declaration of function 53 'timespec_get' is invalid in C99 54- `Bug 104395 <https://bugs.freedesktop.org/show_bug.cgi?id=104395>`__ 55 - [CTS] GTF-GL46.gtf32.GL3Tests.packed_pixels.packed_pixels tests 56 fail on 32bit Mesa 57- `Bug 110765 <https://bugs.freedesktop.org/show_bug.cgi?id=110765>`__ 58 - ANV regression: Assertion \`pass->attachment_count == 59 framebuffer->attachment_count' failed 60- `Bug 110814 <https://bugs.freedesktop.org/show_bug.cgi?id=110814>`__ 61 - KWin compositor crashes on launch 62- `Bug 111069 <https://bugs.freedesktop.org/show_bug.cgi?id=111069>`__ 63 - Assertion fails in nir_opt_remove_phis.c during compilation of 64 SPIR-V shader 65- `Bug 111213 <https://bugs.freedesktop.org/show_bug.cgi?id=111213>`__ 66 - VA-API nouveau SIGSEGV and asserts 67- `Bug 111241 <https://bugs.freedesktop.org/show_bug.cgi?id=111241>`__ 68 - Shadertoy shader causing hang 69- `Bug 111248 <https://bugs.freedesktop.org/show_bug.cgi?id=111248>`__ 70 - Navi10 Font rendering issue in Overwatch 71- `Bug 111271 <https://bugs.freedesktop.org/show_bug.cgi?id=111271>`__ 72 - Crash in eglMakeCurrent 73- `Bug 111308 <https://bugs.freedesktop.org/show_bug.cgi?id=111308>`__ 74 - [Regression, NIR, bisected] Black squares in Unigine Heaven via 75 DXVK 76- `Bug 111401 <https://bugs.freedesktop.org/show_bug.cgi?id=111401>`__ 77 - Vulkan overlay layer - async compute not supported, making overlay 78 disappear in Doom 79- `Bug 111405 <https://bugs.freedesktop.org/show_bug.cgi?id=111405>`__ 80 - Some infinite 'do{}while' loops lead mesa to an infinite 81 compilation 82- `Bug 111411 <https://bugs.freedesktop.org/show_bug.cgi?id=111411>`__ 83 - SPIR-V shader leads to GPU hang, sometimes making machine unstable 84- `Bug 111414 <https://bugs.freedesktop.org/show_bug.cgi?id=111414>`__ 85 - [REGRESSION] [BISECTED] Segmentation fault in si_bind_blend_state 86 after removal of the blend state NULL check 87- `Bug 111467 <https://bugs.freedesktop.org/show_bug.cgi?id=111467>`__ 88 - WOLF RPG Editor + Gallium Nine Standalone: Rendering issue when 89 using Iris driver 90- `Bug 111490 <https://bugs.freedesktop.org/show_bug.cgi?id=111490>`__ 91 - [REGRESSION] [BISECTED] Shadow Tactics: Blades of the Shogun - 92 problems rendering water 93- `Bug 111493 <https://bugs.freedesktop.org/show_bug.cgi?id=111493>`__ 94 - In the game The Surge (378540) - textures disappear then appear 95 again when I change the camera angle view 96- `Bug 111509 <https://bugs.freedesktop.org/show_bug.cgi?id=111509>`__ 97 - [regression][bisected] 98 piglit.spec.ext_image_dma_buf_import.ext_image_dma_buf_import-export 99 fails on iris 100- `Bug 111522 <https://bugs.freedesktop.org/show_bug.cgi?id=111522>`__ 101 - [bisected] Supraland no longer start 102- `Bug 111529 <https://bugs.freedesktop.org/show_bug.cgi?id=111529>`__ 103 - EGL_PLATFORM=drm doesn't expose MESA_query_driver extension 104- `Bug 111552 <https://bugs.freedesktop.org/show_bug.cgi?id=111552>`__ 105 - Geekbench 5.0 Vulkan compute benchmark fails on Anvil 106- `Bug 111566 <https://bugs.freedesktop.org/show_bug.cgi?id=111566>`__ 107 - [REGRESSION] [BISECTED] Large CS workgroup sizes broken in 108 combination with FP64 on Intel. 109- `Bug 111576 <https://bugs.freedesktop.org/show_bug.cgi?id=111576>`__ 110 - [bisected] Performance regression in X4:Foundations in 19.2 111- `Bug 111676 <https://bugs.freedesktop.org/show_bug.cgi?id=111676>`__ 112 - Tropico 6 apitrace throws error into logs 113- `Bug 111734 <https://bugs.freedesktop.org/show_bug.cgi?id=111734>`__ 114 - Geometry shader with double interpolators fails in LLVM 115 116Changes 117------- 118 119Adam Jackson (1): 120 121- docs: Update bug report URLs for the gitlab migration 122 123Alex Smith (1): 124 125- radv: Change memory type order for GPUs without dedicated VRAM 126 127Alyssa Rosenzweig (1): 128 129- pan/midgard: Fix writeout combining 130 131Andres Gomez (1): 132 133- docs: Add the maximum implemented Vulkan API version in 19.2 rel 134 notes 135 136Andres Rodriguez (1): 137 138- radv: additional query fixes 139 140Arcady Goldmints-Orlov (1): 141 142- anv: fix descriptor limits on gen8 143 144Bas Nieuwenhuizen (6): 145 146- radv: Use correct vgpr_comp_cnt for VS if both prim_id and 147 instance_id are needed. 148- radv: Emit VGT_GS_ONCHIP_CNTL for tess on GFX10. 149- radv: Disable NGG for geometry shaders. 150- Revert "ac/nir: Lower large indirect variables to scratch" 151- tu: Set up glsl types. 152- radv: Add workaround for hang in The Surge 2. 153 154Caio Marcelo de Oliveira Filho (2): 155 156- nir/lower_explicit_io: Handle 1 bit loads and stores 157- glsl/nir: Avoid overflow when setting max_uniform_location 158 159Connor Abbott (1): 160 161- radv: Call nir_propagate_invariant() 162 163Danylo Piliaiev (3): 164 165- nir/loop_unroll: Prepare loop for unrolling in wrapper_unroll 166- nir/loop_analyze: Treat do{}while(false) loops as 0 iterations 167- tgsi_to_nir: Translate TGSI_INTERPOLATE_COLOR as INTERP_MODE_NONE 168 169Dave Airlie (2): 170 171- virgl: fix format conversion for recent gallium changes. 172- gallivm: fix atomic compare-and-swap 173 174Dave Stevenson (1): 175 176- broadcom/v3d: Allow importing linear BOs with arbitrary 177 offset/stride. 178 179Dylan Baker (9): 180 181- bump version to 19.2-rc2 182- nir: Add is_not_negative helper function 183- Bump version for rc3 184- meson: don't generate file into subdirs 185- add patches to be ignored 186- Bump version for 19.2.0-rc4 187- cherry-ignore: Add patches 188- rehardcode from origin/master to upstream/master 189- bin/get-pick-list: use --oneline=pretty instead of --oneline 190 191Emil Velikov (1): 192 193- Update version to 19.2.0-rc1 194 195Eric Engestrom (14): 196 197- ttn: fix 64-bit shift on 32-bit \`1\` 198- egl: fix deadlock in malloc error path 199- util/os_file: fix double-close() 200- anv: fix format string in error message 201- freedreno/drm-shim: fix mem leak 202- nir: fix memleak in error path 203- anv: add support for driconf 204- wsi: add minImageCount override 205- anv: add support for vk_x11_override_min_image_count 206- amd: move adaptive sync to performance section, as it is defined in 207 xmlpool 208- radv: add support for vk_x11_override_min_image_count 209- drirc: override minImageCount=2 for gfxbench 210- gl: drop incorrect pkg-config file for glvnd 211- meson: re-add incorrect pkg-config files with GLVND for backward 212 compatibility 213 214Erik Faye-Lund (2): 215 216- gallium/auxiliary/indices: consistently apply start only to input 217- util: fix SSE-version needed for double opcodes 218 219Haihao Xiang (1): 220 221- i965: support AYUV/XYUV for external import only 222 223Hal Gentz (2): 224 225- glx: Fix SEGV due to dereferencing a NULL ptr from XCB-GLX. 226- gallium/osmesa: Fix the inability to set no context as current. 227 228Iago Toral Quiroga (1): 229 230- v3d: make sure we have enough space in the CL for the primitive 231 counts packet 232 233Ian Romanick (8): 234 235- nir/algrbraic: Don't optimize open-coded bitfield reverse when 236 lowering is enabled 237- intel/compiler: Request bitfield_reverse lowering on pre-Gen7 238 hardware 239- nir/algebraic: Mark some value range analysis-based optimizations 240 imprecise 241- nir/range-analysis: Adjust result range of exp2 to account for 242 flush-to-zero 243- nir/range-analysis: Adjust result range of multiplication to account 244 for flush-to-zero 245- nir/range-analysis: Fix incorrect fadd range result for (ne_zero, 246 ne_zero) 247- nir/range-analysis: Handle constants in nir_op_mov just like 248 nir_op_bcsel 249- nir/algebraic: Do not apply late DPH optimization in vertex 250 processing stages 251 252Ilia Mirkin (1): 253 254- gallium/vl: use compute preference for all multimedia, not just blit 255 256Jason Ekstrand (9): 257 258- anv: Bump maxComputeWorkgroupSize 259- nir: Handle complex derefs in nir_split_array_vars 260- nir: Don't infinitely recurse in lower_ssa_defs_to_regs_block 261- nir: Add a block_is_unreachable helper 262- nir/repair_ssa: Repair dominance for unreachable blocks 263- nir/repair_ssa: Insert deref casts when needed 264- nir/dead_cf: Repair SSA if the pass makes progress 265- intel/fs: Handle UNDEF in split_virtual_grfs 266- nir/repair_ssa: Replace the unreachable check with the phi builder 267 268Jonathan Marek (1): 269 270- freedreno/a2xx: ir2: fix lowering of instructions after float 271 lowering 272 273Jose Maria Casanova Crespo (1): 274 275- mesa: recover target_check before get_current_tex_objects 276 277Juan A. Suarez Romero (1): 278 279- bin/get-pick-list.sh: sha1 commits can be smaller than 8 chars 280 281Kenneth Graunke (20): 282 283- gallium/ddebug: Wrap resource_get_param if available 284- gallium/trace: Wrap resource_get_param if available 285- gallium/rbug: Wrap resource_get_param if available 286- gallium/noop: Implement resource_get_param 287- iris: Replace devinfo->gen with GEN_GEN 288- iris: Fix broken aux.possible/sampler_usages bitmask handling 289- iris: Update fast clear colors on Gen9 with direct immediate writes. 290- iris: Drop copy format hacks from copy region based transfer path. 291- iris: Avoid unnecessary resolves on transfer maps 292- iris: Fix large timeout handling in rel2abs() 293- isl: Drop UnormPathInColorPipe for buffer surfaces. 294- isl: Don't set UnormPathInColorPipe for integer surfaces. 295- util: Add a \_mesa_i64roundevenf() helper. 296- mesa: Fix \_mesa_float_to_unorm() on 32-bit systems. 297- iris: Fix partial fast clear checks to account for miplevel. 298- iris: Report correct number of planes for planar images 299- iris: Fix constant buffer sizes for non-UBOs 300- gallium: Fix util_format_get_depth_only 301- iris: Initialize ice->state.prim_mode to an invalid value 302- intel: Increase Gen11 compute shader scratch IDs to 64. 303 304Lepton Wu (1): 305 306- virgl: Fix pipe_resource leaks under multi-sample. 307 308Lionel Landwerlin (9): 309 310- util/timespec: use unsigned 64 bit integers for nsec values 311- util: fix compilation on macos 312- egl: fix platform selection 313- vulkan/overlay: bounce image back to present layout 314- radv: store engine name 315- driconfig: add a new engine name/version parameter 316- vulkan: add vk_x11_strict_image_count option 317- util/xmlconfig: fix regexp compile failure check 318- drirc: include unreal engine version 0 to 23 319 320Marek Olšák (23): 321 322- radeonsi/gfx10: fix the legacy pipeline by storing as_ngg in the 323 shader cache 324- radeonsi: move some global shader cache flags to per-binary flags 325- radeonsi/gfx10: fix tessellation for the legacy pipeline 326- radeonsi/gfx10: fix the PRIMITIVES_GENERATED query if using legacy 327 streamout 328- radeonsi/gfx10: create the GS copy shader if using legacy streamout 329- radeonsi/gfx10: add as_ngg variant for VS as ES to select Wave32/64 330- radeonsi/gfx10: fix InstanceID for legacy VS+GS 331- radeonsi/gfx10: don't initialize VGT_INSTANCE_STEP_RATE_0 332- radeonsi/gfx10: always use the legacy pipeline for streamout 333- radeonsi/gfx10: finish up Navi14, add PCI ID 334- radeonsi/gfx10: add AMD_DEBUG=nongg 335- winsys/amdgpu+radeon: process AMD_DEBUG in addition to R600_DEBUG 336- radeonsi: add PKT3_CONTEXT_REG_RMW 337- radeonsi/gfx10: remove incorrect ngg/pos_writes_edgeflag variables 338- radeonsi/gfx10: set PA_CL_VS_OUT_CNTL with CONTEXT_REG_RMW to fix 339 edge flags 340- radeonsi: consolidate determining VGPR_COMP_CNT for API VS 341- radeonsi: unbind blend/DSA/rasterizer state correctly in delete 342 functions 343- radeonsi: fix scratch buffer WAVESIZE setting leading to corruption 344- radeonsi/gfx10: don't call gfx10_destroy_query with compute-only 345 contexts 346- radeonsi/gfx10: fix wave occupancy computations 347- radeonsi: add Navi12 PCI ID 348- amd: add more PCI IDs for Navi14 349- ac/addrlib: fix chip identification for Vega10, Arcturus, Raven2, 350 Renoir 351 352Mauro Rossi (2): 353 354- android: mesa: revert "Enable asm unconditionally" 355- android: anv: libmesa_vulkan_common: add libmesa_util static 356 dependency 357 358Paulo Zanoni (2): 359 360- intel/fs: grab fail_msg from v32 instead of v16 when v32->run_cs 361 fails 362- intel/fs: fix SHADER_OPCODE_CLUSTER_BROADCAST for SIMD32 363 364Pierre-Eric Pelloux-Prayer (1): 365 366- glsl: replace 'x + (-x)' with constant 0 367 368Rafael Antognolli (1): 369 370- anv: Only re-emit non-dynamic state that has changed. 371 372Rhys Perry (1): 373 374- radv: always emit a position export in gs copy shaders 375 376Samuel Iglesias Gonsálvez (1): 377 378- intel/nir: do not apply the fsin and fcos trig workarounds for consts 379 380Samuel Pitoiset (11): 381 382- radv: allow to enable VK_AMD_shader_ballot only on GFX8+ 383- radv: add a new debug option called RADV_DEBUG=noshaderballot 384- radv: force enable VK_AMD_shader_ballot for Wolfenstein Youngblood 385- ac: fix exclusive scans on GFX8-GFX9 386- radv/gfx10: don't initialize VGT_INSTANCE_STEP_RATE_0 387- radv/gfx10: do not use NGG with NAVI14 388- radv: fix getting the index type size for uint8_t 389- nir: do not assume that the result of fexp2(a) is always an integral 390- radv: fix allocating number of user sgprs if streamout is used 391- radv: fix loading 64-bit GS inputs 392- radv/gfx10: fix VK_KHR_pipeline_executable_properties with NGG GS 393 394Sergii Romantsov (2): 395 396- intel/dri: finish proper glthread 397- nir/large_constants: more careful data copying 398 399Tapani Pälli (5): 400 401- util: fix os_create_anonymous_file on android 402- iris/android: fix build and link with libmesa_intel_perf 403- egl: reset blob cache set/get functions on terminate 404- iris: close screen fd on iris_destroy_screen 405- egl: check for NULL value like eglGetSyncAttribKHR does 406 407Thong Thai (1): 408 409- Revert "radeonsi: don't emit PKT3_CONTEXT_CONTROL on amdgpu" 410 411Timur Kristóf (1): 412 413- st/nine: Properly initialize GLSL types for NIR shaders. 414 415Vinson Lee (2): 416 417- swr: Fix build with llvm-9.0 again. 418- travis: Fail build if any command in if statement fails. 419