1Revision history for SPIRV-Tools 2 3v2024.4 2024-09-19 4 - General 5 - Add FPEncoding operand type. (#5726) 6 - Support SPV_KHR_untyped_pointers (#5736) 7 - add support for SPV_INTEL_global_variable_host_access (#5786) 8 - Optimizer 9 - Add knowledge of cooperative matrices (#5720) 10 - Add struct-packing pass and unit test. (#5778) 11 - Validator 12 - Validate presence of Stride operand to OpCooperativeMatrix{Load,Store}KHR (#5777) 13 - Update sampled image validation (#5789) 14 - Linker 15 - allow linking functions with different pointer arguments (#5534) 16 17v2024.3 2024-06-20 18 - General 19 - Optimizer 20 - Do not fold mul and adds to generate fmas (#5682) 21 - add OpExtInst forward ref fixup pass (#5708) 22 - Validator 23 - Separate Location check for tess patch (#5654) 24 - Validate MemoryAccessMask of OpCooperativeMatrixStoreKHR (#5668) 25 - OpSampledImage extra validation (#5695) 26 - add support for OpExtInstWithForwardRefs (#5698)A 27 - Disassembler 28 - add decorations to comments (#5675) 29 - Add --nested-indent and --reorder-blocks (#5671) 30 31v2024.2 2024-04-22 32 - General 33 - Add SPIRV_TOOLS_EXPORT to public C++ API (#5591) 34 - Use bazel 7 and bzlmod (#5601) 35 - Optimizer 36 - opt: add GroupNonUniformPartitionedNV capability to trim pass (#5648) 37 - Fix rebuilding types with circular references. (#5637) 38 - Add AliasedPointer decoration (#5635) 39 - add support for vulkan-shader-profiler external passes (#5512) 40 - Validator 41 - A fix to support of SPV_QCOM_image_processing2 (#5646) 42 - spirv-val: Add Vulkan check for Rect Dim in OpTypeImage (#5644) 43 - Validate duplicate decorations and execution modes (#5641) 44 - Validator: Support SPV_NV_raw_access_chains (#5568) 45 46v2024.1 2024-03-06 47 - General 48 - Add tooling support for SPV_KHR_maximal_reconvergence (#5542) 49 - Add support for SPV_KHR_float_controls2 (#5543) 50 - SPV_KHR_quad_control (#5547) 51 - Fold 64-bit int operations (#5561) 52 - update image enum tests to remove Kernel capability (#5562) 53 - Support operand kind for SPV_INTEL_maximum_registers (#5580) 54 - SPV_NV_shader_atomic_fp16_vector (#5581) 55 - Support for SPV_QCOM_image_processing2 (#5582) 56 - Fix access chain struct checks (#5592) 57 - Optimizer 58 - opt: add Int16 and Float16 to capability trim pass (#5519) 59 - Add preserver-interface option to spirv-opt (#5524) 60 - spirv-opt: Fix OpCompositeExtract relaxation with struct operands (#5536) 61 - opt: Add VulkanMemoryModelDeviceScope to trim (#5544) 62 - opt: Add TrimCapabilities pass to spirv-opt tool (#5545) 63 - Add modify-maximal-reconvergence to spirv-opt help (#5546) 64 - opt: add SPV_EXT_mesh_shader to opt allowlist (#5551) 65 - opt: Add OpEntryPoint to DescriptorScalarReplacement pass (#5553) 66 - opt: prevent meld to merge block with MaximalReconvergence (#5557) 67 - [OPT] Use new instruction folder for for all opcodes in spec consti folding (#5569) 68 - [OPT] Identify arrays with unknown length in copy prop arrays (#5570) 69 - [OPT] Add removed unused interface var pass to legalization passes (#5579) 70 - Validator 71 - spirv-val: Re-enable OpControlBarrier VU (#5527) 72 - spirv-val: Add Mesh Primitive Built-In validaiton (#5529) 73 - spirv-val: Validate PhysicalStorageBuffer Stage Interface (#5539) 74 - spirv-val: Multiple interface var with same SC (#5528) 75 - spirv-val: Revert Validate PhysicalStorageBuffer Stage Interface (#5575) 76 - spirv-val: Make Constant evaluation consistent (#5587) 77 78v2023.6 2023-12-18 79 - General 80 - update_build_version.py produce deterministic header. (#5426) 81 - Support missing git in update_build_version.py (#5473) 82 - Optimizer 83 - Add ComputeDerivativeGroup*NV capabilities to trim capabilities pass. (#5430) 84 - Do not crash when tryingto fold unsupported spec constant (#5496) 85 - instrument: Fix handling of gl_InvocationID (#5493) 86 - Fix nullptr argument in MarkInsertChain (#5465) 87 - opt: support 64-bit OpAccessChain index in FixStorageClass (#5446) 88 - opt: add StorageImageReadWithoutFormat to cap trim (#5475) 89 - opt: add PhysicalStorageBufferAddresses to trim (#5476) 90 - Fix array size calculation (#5463 91 - Validator 92 - spirv-val: Loosen restriction on base type of DebugTypePointer and DebugTypeQualifier (#5479) 93 - spirv-val: Add WorkgroupMemoryExplicitLayoutKHR check for Block (#5461) 94 95v2023.5 2023-10-15 96 - General 97 - Support 2 Intel extensions (#5357) 98 - SPV_QCOM_image_processing support (#5223) 99 - Optimizer 100 - opt: fix StorageInputOutput16 trimming. (#5359) 101 - opt: add StoragePushConstant16 to trim pass (#5366) 102 - opt: enable StorageUniform16 (#5371) 103 - opt: add bitmask support for capability trimming (#5372) 104 - opt: Add SwitchDescriptorSetPass (#5375) 105 - opt: add FragmentShader*InterlockEXT to capability trim pass (#5390) 106 - opt: add Int64 capability to trim pass (#5398) 107 - opt: add Float64 capability to trim pass (#5428) 108 - opt: add raytracing/rayquery to trim pass (#5397) 109 - opt: add ImageMSArray capability to trim pass. (#5395) 110 - Add SPV_KHR_physical_storage_buffer to allowlists (#5402) 111 - Add SPV_EXT_fragment_shader_interlock to allow lists (#5393) 112 - Make sure that fragment shader interlock instructions are not removed by DCE (#5400) 113 - instrument: Use Import linkage for instrumentation functions (#5355) 114 - Add a new legalization pass to dedupe invocation interlock instructions (#5409) 115 - instrument: Ensure linking works even of nothing is changed (#5419) 116 - Validator 117 - Move token version/cap/ext checks from parsing to validation (#5370) 118 - val: re-add ImageMSArray validation (#5394) 119 - Linker 120 - linker: Add --use-highest-version option 121 122v2023.4 2023-07-17 123 - General 124 - Set cmake_policy CMP0128 (#5341) 125 - Add python3 requirement for the script (#5326) 126 - Add support for LiteralFloat type (#5323) 127 - SPV_KHR_cooperative_matrix (#5286) 128 - Allow OpTypeBool in UniformConstant (#5237) 129 - Allow physical storage buffer pointer in IO (#5251) 130 - Remove const zero image operands (#5232) 131 - Optimizer 132 - Enable vector constant folding (#4913) (#5272) 133 - Fold negation of integer vectors (#5269) 134 - Add folding rule for OpTranspose (#5241) 135 - Add SPV_NV_bindless_texture to spirv optimizations (#5231) 136 - Fix incorrect half float conversion (#5349) 137 - Add SPV_EXT_shader_atomic_float_add to allow lists (#5348) 138 - Instrument 139 - instrument: Cast gl_VertexIndex and InstanceIndex to uint (#5319) 140 - instrument: Fix buffer address length calculations (#5257) 141 - instrument: Reduce number of inst_bindless_stream_write_6 calls (#5327) 142 - Validator 143 - Validate GroupNonUniform instructions (#5296) 144 - spirv-val: Label SPV_KHR_cooperative_matrix VUID (#5301) 145 - Validate layouts for PhysicalStorageBuffer pointers (#5291) 146 - spirv-val: Remove VUID from 1.3.251 spec (#5244) 147 - Diff 148 - spirv-diff: Update test expectations (#5264) 149 - spirv-diff: Leave undefined ids unpaired. (#5262) 150 - spirv-diff: Properly match SPV_KHR_ray_query types. (#5259) 151 - diff: Don't give up entry point matching too early. (#5224) 152 153v2023.3 2023-05-15 154 - General 155 - Update spirv_headers to include SPV_KHR_ray_tracing_position_fetch (#5205) 156 - spirv-tools: Add support for QNX (#5211) 157 - build: set std=c++17 for BUILD.gn (#5162) 158 - Optimizer 159 - Run ADCE when the printf extension is used. (#5215) 160 - Don't convert struct members to half (#5201) 161 - Apply scalar replacement on vars with Pointer decorations (#5208) 162 - opt: Fix null deref in OpMatrixTimesVector and OpVectorTimesMatrix (#5199) 163 - instrument: Add set and binding to bindless error records (#5204) 164 - instrument: Change descriptor state storage format (#5178) 165 - Fix LICMPass (#5087) 166 - Add Vulkan memory model to allow lists (#5173) 167 - Do not remove control barrier after spv1.3 (#5174) 168 - Validator 169 - spirv-val: Label Interface Location/Component VUIDs (#5221) 170 - Add support for SPV_EXT_shader_tile_image (#5188) 171 - Fix vector OpConstantComposite type validation (#5191) 172 - spirv-val: Label new Vulkan VUID 07951 (#5154) 173 - Fuzz 174 - Do not define GOOGLE_PROTOBUF_INTERNAL_DONATE_STEAL_INLINE if it is already defined. (#5200) 175 176v2023.2 2023-03-10 177 - General 178 - build: move from c++11 to c++17 (#4983) 179 - tools: refactorize tools flags parsing. (#5111) 180 - Add C interface for Optimizer (#5030) 181 - libspirv.cpp: adds c++ api for spvBinaryParse (#5109) 182 - build: change the way we set cxx version for bazel. (#5114) 183 - Optimizer 184 - Fix null pointer in FoldInsertWithConstants. (#5093) 185 - Fix removal of dependent non-semantic instructions (#5122) 186 - Remove duplicate lists of constant and type opcodes (#5106) 187 - opt: fix spirv ABI on Linux again. (#5113) 188 - Validator 189 - Validate decoration of structs with RuntimeArray (#5094) 190 - Validate operand type before operating on it (#5092) 191 - spirv-val: Conditional Branch without an exit is invalid in loop header (#5069) 192 - spirv-val: Initial SPV_EXT_mesh_shader builtins (#5080) 193 194v2023.1 2023-01-17 195 - General 196 - Renamed "master" to "main" (issue#5051) 197 - Validate version 5 of clspv reflection (#5050) 198 - Remove testing support for VS2015 (#5027) 199 - Fix undef behaviour in hex float parsing (#5025) 200 - Require C++11 *or later* (#5020) 201 - Instrument 202 - Instrument: Fix bindless checking for BufferDeviceAddress (#5049) 203 - Optimizer 204 - Optimize allocation of spvtools::opt::Instruction::operands_ (#5024) 205 - spirv-opt: Fix OpCompositeInsert with Null Constant (#5008) 206 - spirv-opt: Handle null CompositeInsert (#4998) 207 - Add option to ADCE to remove output variables from interface. (#4994) 208 - Add support for tesc, tese and geom to EliminateDead*Components (#4990) 209 - Add pass to eliminate dead output components (#4982) 210 - spirv-opt: Add const folding for CompositeInsert (#4943) 211 - Add passes to eliminate dead output stores (#4970) 212 - Prevent eliminating case constructs in block merging (#4976) 213 - Validator 214 - Fix layout validation (#5015) 215 - Fix use of invalid analysis (#5013) 216 - Fix infinite loop in validator (#5006) 217 - Add validation support for SPV_NV_shader_invocation_reorder. (#4979) 218 - Only validate full layout in Vulkan environments (#4972) 219 - spirv-val: Label new Vulkan OpPtrAccessChain VUs (#4975) 220 - spirv-val: Add OpPtrAccessChain Base checks (#4965) 221 222 223v2022.4 2022-10-12 224 - General 225 - Support Narrow Types in BitCast Folding Rule (#4941) 226 - spirv-diff: Allow no SpecId (#4904) 227 - build: cmake: Add support for GNU/Hurd (#4895) 228 - Implement tool changes for SPV_EXT_mesh_shader. (#4915) 229 - Validator 230 - spirv-val: Add SPV_ARM_core_builtins validation (#4958) 231 - spirv-val: Add an option to use friendly names or not (#4951) 232 - spirv-val: Consistently quote ids in messages (#4950) 233 - spirv-val: Add initial SPV_EXT_mesh_shader validation (#4924) 234 - spirv-val: Make it legal to use arrays of ray queries (#4938) 235 - spirv-val: Better message for using OpTypeBool in input/output (#4901) 236 - spirv-val: Add SPV_KHR_ray_tracing storage class (#4868) 237 - Optimizer 238 - spirv-opt: Fix stacked CompositeExtract constant folds (#4932) 239 - Improve time to build dominators (#4916) 240 - Fix ADCE to mark scope and inlined_at of line instructions as live. (#4910) 241 - Improve algorithm to reorder blocks in a function (#4911) 242 - Add structs to eliminate dead input components (#4894) 243 - spirv-opt: fix copy-propagate-arrays index opti on structs. (#4891) 244 - Fix ADCE to not eliminate top level DebugInfo instructions (#4889) 245 - Fix array copy propagation (#4890) 246 247v2022.3 2022-08-08 248 - General 249 - Add SPV_KHR_fragment_shader_barycentric support (#4805) 250 - Add support for SPV_KHR_subgroup_rotate (#4786) 251 - use exec_tools instead of tools for better RBE compatibility (#4837) 252 - Write binary files to stdout in binary on windows. (#4834) 253 - Allow spirv-opt print-all to show pretty IDs (#4888) 254 - Validator 255 - spirv-val: Add PerVertexKHR (#4807) 256 - spirv-opt : Add FixFuncCallArgumentsPass (#4775) 257 - spirv-val: Add CullMaskKHR support (#4792) 258 - Require ColMajor or RowMajor for matrices (#4878) 259 - spirv-val: Add SPV_KHR_ray_query (#4848) 260 - spirv-val: Add SPV_KHR_ray_tracing instructions (#4871) 261 - Implement SPV_NV_bindless_texture related changes (#4847) 262 - spirv-val: Add OpConvertUToAccelerationStructureKHR (#4838) 263 - spirv-val: Add support for SPV_AMD_shader_early_and_late_fragment_tests (#4812) 264 - Optimizer 265 - Fold multiply and subtraction into FMA with negation (#4808) 266 - Add more folding for composite instructions (#4802) 267 - spirv-opt: add pass for interface variable scalar replacement (#4779) 268 - Don't try to unroll loop with step count 0. (#4769) 269 - spirv-opt: SPV_NV_bindless_texture related changes (#4870) 270 - Linker 271 - linker: Recalculate interface variables (#4784) 272 273v2022.2 2022-04-07 274 - General 275 - Add OpModuleProcessed to debug opcode (#4694) 276 - Optimizer 277 - Complete handling of RayQueryKHR type (#4690) 278 - Have scalar replacement use undef instead of null (#4691) 279 - Optimize Instruction::Instruction (#4705) 280 - Handle propagation of arrays with decorations (#4717) 281 - spirv-opt: Add OpExecutionModeId support (#4719) 282 - Optimize Type::HashValue (#4707) 283 - Optimize DefUseManager allocations (#4709) 284 - Add pass to remove DontInline function control (#4747) 285 - Better handling of 0xFFFFFFFF when folding vector shuffle (#4743) 286 - Reset the id bound on the module in compact ids (#4744) 287 - spirv-opt: (WIP) Eliminate Dead Input Component Pass (#4720) 288 - Support SPV_KHR_uniform_group_instructions (#4734) 289 - Handle shaders without execution model in spread-volatile-semantics (#4766) 290 - Validator 291 - Fix handling of Nontemporal image operand (#4692) 292 - [spirv-val] Allow 0 Component Count for DebugTypeArray for Shader (#4706) 293 - spirv-val: Validate DebugTypeMatrix (#4732) 294 - spirv-val: Label Vulkan VUID 04734 (#4739) 295 - spirv-val: Label VUID 06491 (#4745) 296 - spirv-val: Disallow array of push constants (#4742) 297 - spirv-val: Label Vulkan RuntimeArray VUID (#4749) 298 - spirv-val: Add Vulkan Image VUID 06214 (#4750) 299 - spirv-val: Add Vulkan Dref not allowed 3D dim VUID (#4751) 300 - spirv-val: Label and add test for PSB Aligned (#4756) 301 - spirv-val: Add Vulkan 32-bit bit op Base (#4758) 302 - spirv-val: Add more Vulkan VUID labels (#4764) 303 - Diff 304 - Introduce spirv-diff (#4611) 305 - Stabilize the output of spirv-diff (#4698) 306 - spirv-diff: Handle OpSpecConstant array sizes (#4700) 307 - spirv-diff: Match OpSpecConstantComposite correctly (#4704) 308 - spirv-diff: Use GetSingleWord*Operand (#4768) 309 - spirv-diff: Basic support for OpTypeForwardPointer (#4761) 310 - spirv-diff: Fix OpTypeFunction matching w.r.t operand count (#4771) 311 312v2022.1 2022-01-26 313 - General 314 - Add SPIR-V 1.6 support to wasm build (#4674) 315 - Improvements to disassembly within PassManager (#4677) 316 - Basic support for SPIR-V 1.6 (#4663) 317 - reflect debug (#4662) 318 - Fix endianness of string literals (#4622) 319 - Optimizer 320 - spirv-opt: add pass to Spread Volatile semantics (#4667) 321 - Fix constant propagation and folding of FClamp instructions (#4651) 322 - Manually fold floating point division by zero (#4637) 323 - Allow ADCE to remove dead inputs (#4629) 324 - Linker 325 - Linker improvements (#4679) 326 * test/linker: Code factorisation and small tweaks 327 * linker: Do not fail when going over limits 328 - Validator 329 - val: interface struct with builtins must be Block (#4665) 330 - Fuzzer 331 - Avoid id bound errors during opt fuzzing (#4658) 332 - Avoid uninitialised read when parsing hex float (#4646) 333 334v2021.4 2021-11-11 335 - General 336 - Add a WebAssembly build (#3752) 337 - Make cxx exceptions controllable (#4591) 338 - Validator 339 - Improve decoration validation (#4490) 340 - Optimizer 341 - Add spirv-opt pass to replace descriptor accesses based on variable indices (#4574) 342 - Do not fold snegate feeding sdiv (#4600) 343 - Handle overflowing id in merge return (#4606) 344 - Fuzzer 345 - Add libFuzzer target for spirv-fuzz (#4434) 346 - Linter 347 348v2021.3 2021-08-24 349 - General 350 - Initial support for SPV_KHR_integer_dot_product (#4327) 351 - Add non-semantic vulkan extended instruction set (#4362) 352 - Add common enum for debug info instructions from either opencl or vulkan (#4377) 353 - Validator 354 - Add validation for SPV_EXT_shader_atomic_float16_add (#4325) 355 - Disallow loading a runtime-sized array (#4473) 356 - spirv-val: Validate vulkan debug info similarly to opencl debug info (#4466) 357 - Optimizer 358 - spirv-opt: support SPV_EXT_shader_image_int64 (#4379) 359 - spirv-opt: Add dataflow analysis framework (#4402) 360 - Add control dependence analysis to opt (#4380) 361 - Add spirv-opt convert-to-sampled-image pass (#4340) 362 - spirv-opt: Add handling of vulkan debug info to DebugInfoManager (#4423) 363 - Fuzz 364 - spirv-fuzz: support AtomicLoad (#4330) 365 - spirv-fuzz: Support AtomicStore (#4440) 366 - spirv-fuzz: TransformationWrapVectorSynonym that rewrites scalar operations using vectors (#4376) 367 - spirv-fuzz: Add minimal SPIR-V example to test shaders (#4415) 368 - spirv-fuzz: support building using gn (#4365) 369 - Linter 370 - Add new target for spirv-lint (#4446) 371 - spirv-lint: add basic CLI argument handling (#4478) 372 - Add divergence analysis to linter (#4465) 373 374v2021.2 2021-06-18 375 - General 376 - Support SPV_KHR_subgroup_uniform_control_flow (#4318) 377 - Support Intel extensions for fixed point and hls-float (#4321) 378 - Fix crash when optimizing shaders with DebugPrintf (#4280) 379 380 - Validator 381 - Support Vulkan Storage Class for Execution Model (#4212) 382 383 - Optimizer 384 - Handle SPV_KHR_vulkan_memory_model in dead-code elimination (#4320) 385 - Support folding OpBitcast with numeric constants (#4247) 386 387 - Fuzz 388 - Add tests for MaybeGet* functions in fuzzerutil (#4284) 389 - Fix OutlineFunction in presence of unreachable blocks (#4308) 390 - Fix def-use update in PermutePhiOperands (#4309) 391 - Swap positions of two functions in a module (#4236) 392 393v2021.1 2021-04-19 394 - General 395 - Support SPV_KHR_linkonce_odr, SPV_KHR_expect_assume (#4161) 396 - Fixes for the vscode language server extension (#4150) 397 - Validator 398 - Add validation for SPV_EXT_shader_atomic_float_min_max (#4105) 399 - Add Vulkan Execution Scope checks (#4183) 400 - Vulkan 64-bit OpAtomicStore check (#4163) 401 - Optimizer 402 - Add interpolate legalization pass (#4220) 403 - Fuzz 404 - Various performance optimizations 405 - Do not add too many dead blocks (#4217) 406 - Add WGSL compatibility flag to context (#4193) 407 - Add persistent state to the fuzzer (#4137) 408 409v2020.7 2021-02-16 410 - General 411 - Support pending Intel extensions (#4116) 412 - Remove WebGPU support (#4108) 413 - Validator 414 - Vulkan image gather constant component (#4133) 415 - Add Vulkan PSB64 convert VUID (#4122) 416 - Validate SPV_KHR_workgroup_memory_explicit_layout (#4128) 417 - Validate VK_KHR_zero_initialize_workgroup_memory (#4124) 418 - Add Vulkan image gather offset VUID (#4118) 419 - Label Vulkan atomic semantics VUIDs (#4120) 420 - Label VUID 04662 (#4123) 421 - Label VUID 04683 (#4121) 422 - Add Vulkan EXT builtins (#4115) 423 - Validate Sampled=1 for Vulkan ImageQuerySizeLod, ImageQueryLevels, ImageQueryLod (#4103) 424 - Add Vulkan Memory Scope VUs (#4106) 425 - Add Vulkan Addressing Model check (#4107) 426 - Vulkan atomic storage class (#4079) 427 - Label standalone Vulkan VUID (#4091) 428 - Add Vulkan decroation VUID (#4090) 429 - Add Vulkan FP Mode VUID (#4088) 430 - Fix Vulkan image sampled check (#4085) 431 - Add Vulkan ForwardPointer VUID (#4089) 432 - Add Vulkan ImageTexelPointer format check (#4087) 433 - Add Vulkan Group Operation VUID (#4086) 434 - Add first StandAlone VUID 04633 (#4077) 435 - Add Subgroup VUIDs (#4074) 436 - validate return type of OpImageRead (#4072) 437 - tighter validation of multisampled images (#4059) 438 - validate OpTypeImage Sampled values for environemnts (#4064) 439 - validate StorageImageMultisampled capability (#4062) 440 - Add last TessLevelOuter and TessLevelInner VUID (#4055) 441 - Add last ClipDistance and CullDistance VUID (#4054) 442 - Add last ViewportIndex and Layer VUID (#4053) 443 - Add last Position VUID (#4052) 444 - Allow forward pointer to be used in types generally (#4044) 445 - Optimizer 446 - Mark module as modified if convert-to-half removes decorations (#4127) 447 - Fix binding number calculation in desc sroa (#4095) 448 - Run DCE when SPV_KHR_shader_clock is used (#4049) 449 - Debug Info 450 - Set correct scope and line info for DebugValue (#4125) 451 - Avoid integrity check failures caused by propagating line instructions (#4096) 452 - Linker 453 - Linker usability improvements (#4084) 454 - Instrumentation 455 - Generate differentiated error codes for buffer oob checking (#4097) 456 - Fuzz 457 - Fix OpPhi handling in DuplicateRegionWithSelection (#4065) 458 459v2020.6 2020-12-07 460 - General 461 CMake: Add SPIRV_TOOLS_BUILD_STATIC flag (#3910) 462 - Disassembler 463 Add some context comments to disassembly. (#3847) 464 - Optimizer 465 - Take new (raytracing) termination instructions into account. (#4050) 466 - Do run DCE if SPV_KHR_ray_query is used. (#4047) 467 - Handle 8-bit index in elim dead member (#4043) 468 - Add texel buffer out-of-bounds checking instrumentation (#4038) 469 - Update MeshShadingNV dependencies (and land Ray tracing updates) (#4028) 470 - Fix buffer oob instrumentation for matrix refs (#4025) 471 - Fix SSA re-writing in the presence of variable pointers. (#4010) 472 - Add support to prevent functions from being inlined if they have 473 DontInline flag (#3858) 474 - Add SPV_EXT_shader_image_int64 (#3852) 475 - Support SPV_KHR_fragment_shading_rate (#3943) 476 - Fix use-after-move in val/validate.cpp (#3848) 477 - Debug Info 478 - properly preserve DebugValue indexes operand (#4022) 479 - Add DebugValue for invisible store in single_store_elim (#4002) 480 - Propagate OpLine to all applied instructions in spirv-opt (#3951) 481 - Add DebugValue for DebugDecl invisible to value assignment (#3973) 482 - Add DebugValue for function param regardless of scope (#3923) 483 - Debug info preservation in convert-local-access-chains pass (#3835) 484 - Debug info preservation in redundancy-elimination pass (#3839) 485 - Debug info preservation in if-conversion pass (#3861) 486 - Validator 487 - Add validation support for the ray tracing built-in variables (#4041) 488 - Use less stack space when validating Vulkan builtins (#4019) 489 - Fix SPV_KHR_fragment_shading_rate VUID label (#4014) 490 - Label Layer and ViewportIndex VUIDs (#4013) 491 - Allow the ViewportIndex and Layer built-ins on SPIR-V 1.5 (#3986) 492 - Fix validation of OpPhi instructions (#3919) 493 - Fuzz 494 - Fix facts arising from CompositeConstruct (#4034) 495 - Do not flatten conditionals that create synonyms (#4030) 496 - Add support for reining in rogue fuzzer passes (#3987) 497 - Fix assertion failure in FuzzerPassAddCompositeExtract (#3995) 498 - Fix invalid equation facts (#4009) 499 - Fix bugs in TransformationFlattenConditionalBranch (#4006) 500 - Fix bug related to transformation applicability (#3990) 501 - Add expand vector reduction transformation (#3869) 502 - Add FuzzerPassAddCompositeExtract (#3904) 503 - Fix mismatch with shrinker step limit (#3985) 504 - Fix off-by-one error in replayer (#3982) 505 - Get order right for OpSelect arguments (#3974) 506 - Do not add synonym-creating loops in dead blocks (#3975) 507 - Skip OpTypeSampledImage when propagating up (#3976) 508 - Pass OpUndef in function call if needed (#3978) 509 - Fix off-by-one in TransformationCompositeConstruct (#3979) 510 - Tolerate absent ids in data synonym fact management (#3966) 511 - Fix to id availability (#3971) 512 - Fix operand types (#3962) 513 - Don't flatten conditional if condition is irrelevant (#3944) 514 - Do not produce OpPhis of type OpTypeSampledImage (#3964) 515 - Restrict fuzzer pass to reachable blocks (#3970) 516 - Handle more types when extending OpPhi instructions (#3969) 517 - Skip early terminator wrappers when merging returns (#3968) 518 - Avoid irrelevant constants in synonym-creating loops (#3967) 519 - Skip dead blocks in FuzzerPassAddOpPhiSynonyms (#3965) 520 - Avoid the type manager when looking for struct types (#3963) 521 - Fix to TransformationDuplicateRegionWithSelection (#3941) 522 - Skip OpFunction when replacing irrelevant ids (#3932) 523 - Use component-wise selectors when flattening conditional branches (#3921) 524 - Avoid void struct member when outlining functions (#3936) 525 - Do not allow Block-decorated structs when adding parameters (#3931) 526 - Fix to operand id type (#3937) 527 - Handle dead blocks in TransformationEquationInstruction (#3933) 528 - Do not allow sampled image load when flattening conditionals (#3930) 529 - Take care of OpPhi instructions when inlining (#3939) 530 - Fix to TransformationInlineFunction (#3913) 531 - Wrap early terminators before merging returns (#3925) 532 - Lower probability of adding bit instruction synonyms (#3917) 533 - Fix handling of OpPhi in FlattenConditionalBranch (#3916) 534 - Avoid creating blocks without parents (#3908) 535 - Do not allow creation of constants of block-decorated structs (#3903) 536 - Fixes related to irrelevant ids (#3901) 537 - Fix to transformation that adds a synonym via a loop (#3898) 538 - Fix to duplicate region with selection (#3896) 539 - Do not expose synonym facts for non-existent ids (#3891) 540 - Do not add synonyms involving irrelevant ids (#3890) 541 - Do not replace irrelevant ids that are not in blocks (#3892) 542 - Wrap OpKill and similar in function calls (#3884) 543 - Integrate spirv-reduce with shrinker (#3849) 544 - Report fresh ids in transformations (#3856) 545 - Support OpNot bit instruction case (#3841) 546 - Return IR and transformation context after replay (#3846) 547 548v2020.5 2020-09-22 549 - General 550 - Enable building with BUILD_SHARED_LIBS=1 (#3490) 551 - Avoid using /MP4 for clang on windows. (#3662) 552 - Fix compiler error on macOS with XCode12. (#3836) 553 - Optimizer 554 - Preserve OpenCL.DebugInfo.100 through private-to-local pass (#3571) 555 - Preserve debug info in scalar replacement pass (#3461) 556 - Debug info preservation in loop-unroll pass (#3548) 557 - Preserve debug info in dead-insert-elim pass (#3652) 558 - Improve non-semantic instruction handling in the optimizer (#3693) 559 - Let ADCE pass check DebugScope (#3703) 560 - Add undef for inlined void function (#3720) 561 - Fix SSA-rewrite to remove DebugDeclare for variables without loads (#3719) 562 - Handle DebugScope in compact-ids pass (#3724) 563 - Add buffer oob check to bindless instrumentation (#3800) 564 - Validator 565 - Update OpenCL capabilities validation (#3149) 566 - Validator support for non-semantic clspv reflection (#3618) 567 - OpenCL.DebugInfo.100 DebugTypeArray with variable size (#3549) 568 - Only validation locations for appropriate execution models (#3656) 569 - Validate more OpenCL.DebugInfo.100 instructions (#3684) 570 - Allow DebugTypeTemplate for Type operand (#3702) 571 - spirv-val: Add Vulkan VUID labels to BuiltIn (#3756) 572 - Allow SPV_KHR_8bit_storage extension. (#3780) 573 - Validate SPIRV Version number when parsing binary header (#3834) 574 - Reduce 575 - Support reducing a specific function (#3774) 576 - Fuzz 577 - adds TransformationReplaceCopyObjectWithStoreLoad (#3567) 578 - adds TransformationReplaceCopyMemoryWithLoadStore (#3575) 579 - adds TransformationReplaceLoadStoreWithCopyMemory (#3586) 580 - Implement the OpOuterProduct linear algebra case (#3617) 581 - Pass to replace int operands with ints of opposite signedness (#3612) 582 - TransformationMoveInstructionDown (#3477) 583 - Add TransformationMakeVectorOperationDynamic (#3597) 584 - TransformationReplaceAddSubMulWithCarryingExtended (#3598) 585 - FuzzerPassPropagateInstructionsUp (#3478) 586 - add FuzzerPassAddCompositeInserts (#3606) 587 - Add inline function transformation (#3517) 588 - Transformation to replace the use of an irrelevant id (#3697) 589 - Add SPIRV_FUZZ_PROTOC_COMMAND (#3789) 590 - Add TransformationDuplicateRegionWithSelection (#3773) 591 - Transformation to flatten conditional branch (#3667) 592 - Handle OpPhis in TransformationInlineFunction (#3833) 593 - Create synonym of int constant using a loop (#3790) 594 - Support dead blocks in TransformationAddSynonym (#3832) 595 - Linker 596 597v2020.4 2020-07-22 598 - General 599 - Changed variable names to be more descriptive (#3433) 600 - Add support to GPU-AV instrumentation for Task and Mesh shaders (#3512) 601 - Permit Simple and GLSL450 memory model in WEBGPU_0 (#3463) 602 - Support SPV_KHR_terminate_invocation (#3568) 603 - Optimizer 604 - Preserving debug information in optimizations 605 (#3389,#3420,#3425,#3356,#3459,#3444,#3492,#3451,#3497i,#3498,#3542) 606 - Eliminate branches with condition of OpConstantNull (#3438) 607 - Use structured order to unroll loops. (#3443) 608 - Updated desc_sroa to support flattening structures (#3448) 609 - Support OpCompositeExtract pattern in desc_sroa (#3456) 610 - Fix ADCE pass bug for mulitple entries (#3470) 611 - Sink pointer instructions in merge return (#3569) 612 - Validator 613 - Validate location assignments (#3308) 614 - Fix reachability in the validator (#3541) 615 - Reduce 616 - Fuzz 617 - Add support for OpSpecConstant* (#3373) 618 - Add replace linear algebra instruction transformation (#3402) 619 - Implement vector shuffle fuzzer pass (#3412) 620 - Swap operands in OpBranchConditional (#3423) 621 - Permute OpPhi instruction operands (#3421) 622 - Add FuzzerPassAddCopyMemoryInstructions (#3391) 623 - TransformationInvertComparisonOperator (#3475) 624 - Add variables with workgroup storage class (#3485) 625 - Add image sample unused components transformation (#3439) 626 - TransformationReplaceParameterWithGlobal (#3434) 627 - Support adding dead break from back-edge block (#3519) 628 - Fuzzer pass to interchange zero-like constants (#3524) 629 - Linker 630 631v2020.3 2020-05-27 632 - General 633 - Prevent Effcee from installing things when building spirv-tools with testing enabled (#3256) 634 - Update acorn version (#3294) 635 - If SPIRV-Headers is in our tree, include it as subproject (#3299) 636 - allow cross compiling for Windows Store, UWP, etc. (#3330) 637 - Optimizer 638 - Remove deprecated interfaces from instrument passes (#3361) 639 - Preserve debug info in inline pass (#3349) 640 - Handle more cases in dead member elim (#3289) 641 - Preserve debug info in eliminate-dead-functions (#3251) 642 - Fix Struct CFG analysis for single block loop (#3293) 643 - Add tests for recently added command line option (#3297) 644 - Consider sampled images as read-only storage (#3295) 645 - Allow various validation options to be passed to spirv-opt (#3314) 646 - Add debug information analysis (#3305) 647 - Preserve debug info for wrap-opkill (#3331) 648 - refactor inlining pass (#3328) 649 - Add unrolling to performance passes (#3082) 650 - Validator 651 - Add validation support for ImageGatherBiasLodAMD (#3363) 652 - Validate ShaderCallKHR memory scope (#3332) 653 - Validate Buffer and BufferBlock apply only to struct types (#3259) 654 - Reduce 655 - increase default step limit (#3327) 656 - Remove unused uniforms and similar (#3321) 657 - Fuzz 658 - Add support for StorageBuffer (#3348) 659 - Add validator options (#3254) 660 - Limit adding of new variables to 'basic' types (#3257) 661 - Transformation to add OpConstantNull (#3273) 662 - Handling of more fuzzing opportunities (#3277, #3280, #3281, #3290, #3292) 663 - Respect rules for OpSampledImage (#3287) 664 - Do not outline regions that produce pointer outputs (#3291) 665 - Linker 666 667v2020.2 2020-03-26 668 - General: 669 - Support extended instructions in the vscode language server 670 - Make spvOpcodeString part of the public API (#3174) 671 - Added guide to writing a spirv-fuzz fuzzer pass (#3190) 672 - Add support for KHR_ray_{query,tracing} extensions (#3235) 673 - Optimizer 674 - Debug Printf support (#3215) 675 - Add data structure for DebugScope, DebugDeclare in spirv-opt (#3183) 676 - Fix identification of Vulkan images and buffers (#3253) 677 - Validator 678 - Add support for SPV_AMD_shader_image_load_store_lod (#3186) 679 - Add validation rules for OpenCL.DebugInfo.100 extension (#3133) 680 - Adding WebGPU specific Workgroup scope rule (#3204) 681 - Disallow phis of images, samplers and sampled images (#3246) 682 - Reduce 683 - Fuzz 684 - Fuzzer passes to add local and global variables (#3175) 685 - Add fuzzer passes to add loads/stores (#3176) 686 - Fuzzer pass to add function calls (#3178) 687 - Fuzzer pass that adds access chains (#3182) 688 - Fuzzer pass to add equation instructions (#3202) 689 - Add swap commutable operands transformation (#3205) 690 - Add fuzzer pass to permute function parameters (#3212) 691 - Allow OpPhi operand to be replaced with a composite synonym (#3221) 692 - Linker 693 694v2020.1 2020-02-03 695 - General: 696 - Add support for SPV_KHR_non_semantic_info (#3110) 697 - Support OpenCL.DebugInfo.100 extended instruction set (#3080) 698 - Added support for Vulkan 1.2 699 - Add API function to better handle getting the necessary environment (#3142) 700 - Clarify mapping of target env to SPIR-V version (#3150) 701 - Implement constant folding for many transcendentals (#3166) 702 - Optimizer 703 - Change default version for CreatInstBindlessCheckPass to 2 (#3096, #3119) 704 - Better handling of OpLine on merge blocks (#3130) 705 - Use placeholder switch instead of placeholder loop in MergeReturn pass. (#3151) 706 - Handle TimeAMD in AmdExtensionToKhrPass. (#3168) 707 - Validator 708 - Fix structured exit validation (#3141) 709 - Reduce 710 - Fuzz 711 - Fuzzer pass to merge blocks (#3097) 712 - Transformation to add a new function to a module (#3114) 713 - Add fuzzer pass to perform module donation (#3117) 714 - Fuzzer passes to create and branch to new dead blocks (#3135) 715 - Fuzzer pass to add composite types (#3171) 716 - Linker: 717 - Remove names and decorations of imported symbols (#3081) 718 719v2019.5 2019-12-11 720 - General: 721 - Export SPIRV-Tools targets on installation 722 - SPIRV-Tools support for SPIR-V 1.5 (#2865) 723 - Add WebGPU SPIR-V Assembler in JavaScript. (#2876) 724 - Add Bazel build configuration. (#2891) 725 - Add support for building with emscripten (#2948) 726 - Update SPIR-V binary header test for SPIR-V 1.5 (#2967) 727 - Add fuzzer for spirv-as call path (#2976) 728 - Improved CMake install step. (#2963) 729 - Add fuzzer for spirv-dis call path (#2977) 730 - Ensure timestamp does not vary with timezone. (#2982) 731 - Add a vscode extension for SPIR-V disassembly files (#2987) 732 - Add iOS as a supported platform (#3001) 733 - utils/vscode: Add SPIR-V language server support 734 - Respect CMAKE_INSTALL_LIBDIR in installed CMake files (#3054) 735 - Permit the debug instructions in WebGPU SPIR-V (#3063) 736 - Add support for Fuchsia. (#3062) 737 - Optimizer 738 - Add descriptor array scalar replacement (#2742) 739 - Add pass to wrap OpKill in a function call (#2790) 740 - Fold FMix during constant folding. (#2818) 741 - Add pass to replace AMD shader ballot extension (#2811) 742 - Add pass to make Float32 operation relax precision (#2808) 743 - Add pass to make relax precision operation Float16 (#2808) 744 - Add pass to replace uses of 3 AMD extensions (#2814) 745 - Fold Min, Max, and Clamp instructions. (#2836) 746 - Better handling of OpKill in continues (#2842,#2922,#2933) 747 - Enable OpTypeCooperativeMatrix specialization (#2927) 748 - Support constant-folding UConvert and SConvert (#2960) 749 - Update Offset to ConstOffset bitmask if operand is constant. (#3024) 750 - Improve RegisterSizePasses (#3059) 751 - Folding: perform add and sub on mismatched integer types (#3084) 752 - Graphics robust access: use signed clamp (#3073) 753 Fixes: 754 - Instrument: Fix version 2 output record write for tess eval shaders. (#2782) 755 - Instrument: Add support for Buffer Device Address extension (#2792) 756 - Fix check for changed binary in API call. (#2798) 757 - For WebGPU<->Vulkan optimization, set correct execution environment (#2834) 758 - Handle OpConstantNull in copy-prop-arrays. (#2870) 759 - Use OpReturn* in wrap-opkill (#2886) 760 - Validator 761 - Add generic builtin validation of target (#2843) 762 - Extra resource interface validation (#2864) 763 - Adding valilidation checks for OpEntryPoint duplicate names and execution mode (#2862) 764 - Relaxed bitcast with pointers (#2878) 765 - Validate physical storage buffer restrictions (#2930) 766 - Add SPV_KHR_shader_clock validation (#2879, #3013) 767 - Validate that selections are structured (#2962) 768 - Disallow use of OpCompositeExtract/OpCompositeInsert with no indices (#2980) 769 - Check that derivatives operate on 32-bit values (#2983) 770 - Validate array stride does not cause overlap (#3028) 771 - Validate nested constructs (#3068) 772 Fixes: 773 - Fix validation of constant matrices (#2794) 774 - Update "remquor" validation 775 - Only allow previously declared forward refs in structs (#2920) 776 - Reduce 777 - Remove relaxed precision decorations (#2797) 778 - Reduce/fuzz: improve command line args (#2932) 779 - Improve remove unref instr pass (#2945) 780 Fixes: 781 - Fuzz 782 - Fix add-dead-break and add-dead-continue passes to respect dominance (#2838) 783 - Add fuzzer pass to copy objects (#2853) 784 - Add fuzzer pass to replace ids with synonyms (#2857) 785 - Allow validation during spirv-fuzz replay (#2873) 786 - Employ the "swarm testing" idea in spirv-fuzz (#2890) 787 - reduce/fuzz: improve command line args (#2932) 788 - option to convert shader into a form that renders red (#2934) 789 - Add fuzzer pass to change selection controls (#2944) 790 - add transformation and pass to construct composites (#2941) 791 - Add fuzzer pass to change loop controls (#2949) 792 - Add fuzzer pass to change function controls (#2951) 793 - Add fuzzer pass to add NoContraction decorations (#2950) 794 - Add missing functionality for matrix composites (#2974) 795 - Fuzzer pass to adjust memory access operands (#2968) 796 - Transformation to extract from a composite object (#2991) 797 - Vector shuffle transformation (#3015) 798 - Improve debugging facilities (#3074) 799 - Function outlining fuzzer pass (#3078) 800 801 802v2019.4 2019-08-08 803 - General: 804 - Memory model support for SPIR-V 1.4 805 - Add new spirv-fuzz tool 806 - Add option for base branch in check_code_format.sh 807 - Removed MarkV and Stats code. (#2576) 808 - Instrument: Add version 2 of record formats (#2630) 809 - Linker: Better type comparison for OpTypeArray and OpTypeForwardPointer (#2580) 810 - Optimizer 811 - Bindless Validation: Instrument descriptor-based loads and stores (#2583) 812 - Better folding for OpSpecConstantOp (#2585, #2614) 813 - Add in individual flags for Vulkan <-> WebGPU passes (#2615) 814 - Handle nested breaks from switches. (#2624) 815 - Optimizer: Handle array type with OpSpecConstantOp length (#2652) 816 - Perform merge return with single return in loop. (#2714) 817 - Add --preserve-bindings and --preserve-spec-constants (#2693) 818 - Remove Common Uniform Elimination Pass (#2731) 819 - Allow ray tracing shaders in inst bindle check pass. (#2733) 820 - Add pass to inject code for robust-buffer-access semantics (#2771) 821 - Treat access chain indexes as signed in SROA (#2776) 822 - Handle RelaxedPrecision in SROA (#2788) 823 - Add descriptor array scalar replacement (#2742) 824 Fixes: 825 - Handle decorations better in some optimizations (#2716) 826 - Change the order branches are simplified in dead branch elim (#2728) 827 - Fix bug in merge return (#2734) 828 - SSA rewriter: Don't use trivial phis (#2757) 829 - Record correct dominators in merge return (#2760) 830 - Process OpDecorateId in ADCE (#2761) 831 - Fix check for unreachable blocks in merge-return (#2762) 832 - Handle out-of-bounds scalar replacements. (#2767) 833 - Don't move debug or decorations when folding (#2772) 834 - Protect against out-of-bounds references when folding OpCompositeExtract (#2774) 835 - Validator 836 - Validate loop merge (#2579) 837 - Validate construct exits (#2459) 838 - Validate OpenCL memory and addressing model environment rules (#2589) 839 - Validate OpenCL environment rules for OpTypeImage (#2606) 840 - Allow breaks to switch merge from nested construct (#2604) 841 - Validate OpenCL environment rules for OpImageWrite (#2619) 842 - Allow arrays of out per-primitive builtins for mesh shaders (#2617) 843 - Validate OpenCL rules for ImageRead and OpImageSampleExplicitLod (#2643) 844 - Add validation for SPV_EXT_fragment_shader_interlock (#2650) 845 - Add builtin validation for SPV_NV_shader_sm_builtins (#2656) 846 - Add validation for Subgroup builtins (#2637) 847 - Validate variable initializer type (#2668) 848 - Disallow stores to UBOs (#2651)A 849 - Validate Volatile memory semantics bit (#2672) 850 - Basic validation for Component decorations (#2679) 851 - Validate that in OpenGL env block variables have Binding (#2685) 852 - Validate usage of 8- and 16-bit types with only storage capabilities (#2704) 853 - Add validation for SPV_EXT_demote_to_helper_invocation (#2707) 854 - Extra small storage validation (#2732) 855 - For Vulkan, disallow structures containing opaque types (#2546) 856 - Validate storage class OpenCL environment rules for atomics (#2750) 857 - Update OpControlBarriers rules for WebGPU (#2769) 858 - Update OpMemoryBarriers rules for WebGPU (#2775) 859 - Update WebGPU validation rules of OpAtomic*s (#2777) 860 Fixes: 861 - Disallow merge targeting block with OpLoopMerge (#2610) 862 - Update vloadn and vstoren validation to match the OpenCL Extended 863 Instruction Set Specification (#2599) 864 - Update memory scope rules for WebGPU (#2725) 865 - Allow LOD ops in compute shaders with derivative group execution modes (#2752) 866 - Reduce 867 Fixes: 868 869v2019.3 2019-05-14 870 - General: 871 - Require Python 3 since Python 2 will out of service soon. 872 - Add a continuous test that does memory checks using the address sanitizer. 873 - Fix the build files so the SPIRV_USE_SANITIZER=address build works. 874 - Packaging top of tree build artifacts again. 875 - Added support for SPIR-V 1.4. (#2550) 876 - Optimizer 877 - Remove duplicates from list of interface IDs in OpEntryPoint instruction (#2449) 878 - Bindless Validation: Descriptor Initialization Check (#2419) 879 - Add option to validate after each pass (#2462) 880 - Add legalization pass to fix mismatched pointer (#2430, #2535) 881 - Add error messages when the input contains unknown instructions. (#2487) 882 - Add pass to convert from WebGPU Spir-V to Vulkan Spir-V and back. (#2495) 883 Fixes: 884 - #2412: Dead memeber elimination should not change input and output variables. 885 - #2405: Fix OpDot folding of half float vectors. 886 - #2391: Dead branch elim should not fold away back edges. 887 - #2441: Removing decorations when doing constant propagation. 888 - #2455: Maintain inst to block mapping in merge return. 889 - #2453: Fix merge return in the face of breaks. 890 - #2456: Handle dead infinite loops in DCE. 891 - #2458: Handle variable pointer in some optimizations. 892 - #2452: Fix dead branch elimination to handle unreachable blocks better. 893 - #2528: Fix undefined bit shift in sroa. 894 - #2539: Change implementation of post order CFG traversal. 895 - Validator 896 - Add validation of storage classes for WebGPU (#2446) 897 - Add validation for ExecutionMode in WebGPU (#2443) 898 - Implement WebGPU specific CFG validation (#2386) 899 - Allow NonWritable to target struct members. (#2420) 900 - Allow storage type mismatch for parameter in relaxed addressing mode. 901 - Allow non memory objects as parameter in relaxed addressing mode. 902 - Disallow nested Blocks and buffer blocks (#2410). 903 - Add validation for SPV_NV_cooperative_matrix (#2404) 904 - Add --strip-atomic-counter-memory (#2413) 905 - Check OpSampledImage is only passed into valid instructions (#2467) 906 - Handle function decls in Structured CFG analysis (#2474) 907 - Validate that OpUnreacahble is not statically reachable (#2473) 908 - Add pass to generate needed initializers for WebGPU (#2481) 909 - Allow images without format for OpenCL. (#2470) 910 - Remove unreachable block validation (#2525) 911 - Reduce runtime of array layout checks (#2534) 912 - Add validation specific to OpExecutionModeId (#2536) 913 - Validate sign of int types. (#2549) 914 - VK_KHR_uniform_buffer_standard_layout validation (#2562) 915 Fixes: 916 - #2439: Add missing DepthGreater case to Fragment only check. 917 - #2168: Disallow BufferBlock on StorageBuffer variables for Vulkan. 918 - #2408: Restrict and Aliased decorations cannot be applied to the same id. 919 - #2447: Improve function call parameter check. 920 - Reduce 921 - Add Pass to remove unreferenced blocks. (#2398) 922 - Allows passing options to the validator. (#2401) 923 - Improve reducer algorithm and other changes (#2472) 924 - Add Pass to remove selections (#2485) 925 - Add passes to simplify branches (#2507) 926 Fixes: 927 - #2478: fix loop to selection pass for loops with combined header/continue block 928 929v2019.2 2019-02-20 930 - General: 931 - Support SPV_EXT_physical_storage_buffer 932 - A number of memory leak have been fixed. 933 - Removed use of deprecated Google test macro: 934 - Changed the BUILD.gn to only build tests in Chromium. 935 - Optimizer 936 - Upgrade memory model improvments for modf and frexp. 937 - Add a new pass to move loads closer to their uses: code sinking. 938 - Invalidating the type manager now invalidates the constnat manager. 939 - Expand instrumentation pass for bindless bounds checking to runtime-sized descriptor arrays. 940 - Add a new pass that removes members from structs that are not used: dead member elimination. 941 Fixes: 942 - #2292: Remove undefined behaviour when folding bit shifts. 943 - #2294: Fixes for instrumentation code. 944 - #2293: Fix overflow when folding -INT_MIN. 945 - #2374: Don't merge unreachable blocks when merging blocks. 946 - Validator 947 - Support SPV_KHR_no_integer_wrap and related decorations. 948 - Validate Vulkan rules for OpTypeRuntimeArray. 949 - Validate NonWritable decoration. 950 - Many WebGPU specific validation rules were added. 951 - Validate variable pointer related function call rules. 952 - Better error messages. 953 Fixes: 954 - #2307: Check forwards references in OpTypeArray. 955 - #2315, #2303: Fixed the layout check for relaxed layout. 956 - #1628: Emit an error when an OpSwitch target is not an OpLabel. 957 - Reduce 958 - Added more documentation for spirv-reduce. 959 - Add ability to remove OpPhi instructions. 960 - Add ability to merge two basic blocks. 961 - Add ability to remove unused functions and unused basic blocks. 962 Fixes: 963 964v2019.1 2019-01-07 965 - General: 966 - Created a new tool called spirv-reduce. 967 - Add cmake option to turn off SPIRV_TIMER_ENABLED (#2103) 968 - New optimization pass to update the memory model from GLSL450 to VulkanKHR. 969 - Recognize OpTypeAccelerationStructureNV as a type instruction and ray tracing storage classes. 970 - Fix GCC8 build. 971 - Add --target-env flag to spirv-opt. 972 - Add --webgpu-mode flag to run optimizations for webgpu. 973 - The output disassembled line number stead of byte offset in validation errors. (#2091) 974 - Optimizer 975 - Added the instrumentation passes for bindless validation. 976 - Added passes to help preserve OpLine information (#2027) 977 - Add basic support for EXT_fragment_invocation_density (#2100) 978 - Fix invalid OpPhi generated by merge-return. (#2172) 979 - Constant and type manager have been turned into analysies. (#2251) 980 Fixes: 981 - #2018: Don't inline functions with a return in a structured CFG contstruct. 982 - #2047: Fix bug in folding when volatile stores are present. 983 - #2053: Fix check for when folding floating pointer values is allowed. 984 - #2130: Don't inline recursive functions. 985 - #2202: Handle multiple edges between two basic blocks in SSA-rewriter. 986 - #2205: Don't unswitch a latch condition during loop unswitch. 987 - #2245: Don't fold branch in loop unswitch. Run dead branch elimination to fold them. 988 - #2204: Fix eliminate common uniform to place OpPhi instructions correctly. 989 - #2247: Fix type mismatches caused by scalar replacement. 990 - #2248: Fix missing OpPhi after merge return. 991 - #2211: After merge return, fix invalid continue target. 992 - #2210: Fix loop invariant code motion to not place code between merge instruction and branch. 993 - #2258: Handle CompositeInsert with no indices in VDCE. 994 - #2261: Have replace load size handle extact with no index. 995 - Validator 996 - Changed the naming convention of outputing ids with names in diagnostic messages. 997 - Added validation rules for UniformConstant variables in Vulkan. 998 - #1949: Validate uniform variable type in Vulkan 999 - Ensure for OpVariable that result type and storage class operand agree (#2052) 1000 - Validator: Support VK_EXT_scalar_block_layout 1001 - Added Vulkan memory model semantics validation 1002 - Added validation checkes spefic to WebGPU environment. 1003 - Add support for VK_EXT_Transform_feedback capabilities (#2088) 1004 - Add validation for OpArrayLength. (#2117) 1005 - Ensure that function parameter's type is not void (#2118) 1006 - Validate pointer variables (#2111) 1007 - Add check for QueueFamilyKHMR memory scope (#2144) 1008 - Validate PushConstants annotation and type (#2140) 1009 - Allow Float16/Int8 for Vulkan 1.0 (#2153) 1010 - Check binding annotations in resource variables (#2151, #2167) 1011 - Validate OpForwardPointer (#2156) 1012 - Validate operation for OpSpecConstantOp (#2260) 1013 Fixes: 1014 - #2049: Allow InstanceId for NV ray tracing 1015 - Reduce 1016 - Initial commit wit a few passes to reduce test cases. 1017 - Validation is run after each reduction step. 1018 Fixes: 1019 1020 1021v2018.6 2018-11-07 1022 - General: 1023 - Added support for the Nvidia Turing and ray tracing extensions. 1024 - Make C++11 the CXX standard in CMakeLists.txt. 1025 - Enabled a parallel build for MSVC. 1026 - Enable pre-compiled headers for MSVC. 1027 - Added a code of conduct. 1028 - EFFCEE and RE2 are now required when build the tests. 1029 - Optimizer 1030 - Unrolling loops marked for unrolling in the legalization passes. 1031 - Improved the compile time of loop unrolling. 1032 - Changee merge-return to create a placeholder loop around the function. 1033 - Small improvement to merge-blocks to allow it to merge more often. 1034 - Enforce an upper bound for the ids, and add option to set it. 1035 - #1966: Report error if there are unreachable block before running merge return 1036 Fixes: 1037 - #1917: Allow 0 (meaning unlimited) as a parameter to --scalar-replacement 1038 - #1915: Improve handling of group decorations. 1039 - #1942: Fix incorrect uses of the constant manager. Avoids type mismatches in generated code. 1040 - #1997: Fix dead branch elimination when there is a loop in folded selection. 1041 - #1991: Fixes legality check in if-conversion. 1042 - #1987: Add nullptr check to array copy propagation. 1043 - #1984: Better handling of OpUnreachable in ADCE. 1044 - #1983: Run merge return on reachable functions only. 1045 - #1956: Handled atomic operations in ADCE. 1046 - #1963: Fold integer divisions by 0 to 0. 1047 - #2019: Handle MemberDecorateStringGOOGLE in ADCE and strip reflect. 1048 - Validator 1049 - Added validation for OpGroupNonUniformBallotBitCount. 1050 - Added validation for the Vulkan memory model. 1051 - Added support for VK_KHR_shader_atddomic_int64. 1052 - Added validation for execution modes. 1053 - Added validation for runtime array layouts. 1054 - Added validation for 8-bit storage. 1055 - Added validation of OpPhi instructions with pointer result type. 1056 - Added checks for the Vulkan memory model. 1057 - Validate MakeTexelAvailableKHR and MakeTexelVisibleKHR 1058 - Allow atomic function pointer for OpenCL. 1059 - FPRounding mode checks were implemented. 1060 - Added validation for the id bound with an option to set the max id bound. 1061 Fixes: 1062 - #1882: Improve the validation of decorations to reduce memory usage. 1063 - #1891: Fix an potential infinite loop in dead-branch-elimination. 1064 - #1405: Validate the storage class of boolean objects. 1065 - #1880: Identify arrays of type void as invalid. 1066 - #487: Validate OpImageTexelPointer. 1067 - #1922: Validate OpPhi instructions are at the start of a block correctly. 1068 - #1923: Validate function scope variable are at the start of the entry block. 1069 1070v2018.5 2018-09-07 1071 - General: 1072 - Support SPV_KHR_vulkan_memory_model 1073 - Update Dim capabilities, to match SPIR-V 1.3 Rev 4 1074 - Automated build bots no run tests for the VS2013 case 1075 - Support Chromium GN build 1076 - Use Kokoro bots: 1077 - Disable Travis-CI bots 1078 - Disable AppVeyor VisualStudio Release builds. Keep VS 2017 Debug build 1079 - Don't check export symbols on OSX (Darwin): some installations don't have 'objdump' 1080 - Reorganize source files and namespaces 1081 - Fixes for ClangTidy, and whitespace (passes 'git cl presumit --all -uf') 1082 - Fix unused param compile warnings/errors when Effcee not present 1083 - Avoid including time headers when timer functionality is disabled 1084 - Avoid too-stringent warnings flags for Clang on Windows 1085 - Internal refactoring 1086 - Add hooks for automated fuzzing 1087 - Add testing of command line executables 1088 - #1688: Use binary mode on stdin; fixes "spirv-dis <foo.spv" on Windows 1089 - Optimizer 1090 - The optimizer validates the module before it begins 1091 - Add API to register passes by string name 1092 - Fold a vector shuffle feeding a vector shuffle 1093 - Add -combine-access-chains transform 1094 - Refactor how IRContext is handled by passes 1095 - Improve bookkeeping for instruction result type and result id 1096 - Fix over-duplication of decorations 1097 - Fix handling of exits from selections in dead-branch elimination, and dead code 1098 elimination. 1099 - Fix handling of certain kinds of flow control in merge-return 1100 Fixes: 1101 - #1721: Fix size bug when folding vector shuffles 1102 - #1722: Fix size infinite loop when folding vector shuffles 1103 - #1724: Fix finding a constant of a specific type 1104 - #1727: Dead branch elim: Reorder blocks if needed to satisfy dominance rule 1105 - #1729: Handle VariablePointers cases in various optimizations 1106 - #1731: Fix vector shuffle with literal id indicating undef value 1107 - #1736: Fix handling of decorations and phis in merge-return 1108 - #1787: Fix handling of decorations related to access chains 1109 - #1865: Avoid leaking memory for SPIR-V constant values 1110 - Validator 1111 - Improve error messages 1112 - Avoid platform-dependent traversal ordering, to ensure consistent messages 1113 - Use libspirv::Instruction where possible 1114 - Add option to skip all block layout checks 1115 - Validate all type IDs 1116 - Validate uses of OpFunction 1117 - Validate uses of OpTypeFunction 1118 - Disallow a struct containing its own type https://crbug.com/874372 1119 - #1685: Vulkan permits non-monotonic offsets for block members 1120 - #1697: Enforce block layout rules even when using relaxed block layout option 1121 - #1719: Fix line number for vector shuffle valiation error 1122 - #1789: Avoid assertion failure when validating some functions 1123 - #1800: Fix validation of OpCopyMemorySized 1124 - #1822: Stop enforcing struct member offset montonicity 1125 - #1831: Disallow void members in structs 1126 1127v2018.4 2018-07-08 1128 - General: 1129 - Support SPV_KHR_8bit_storage 1130 - Add gclient and presubmit configurations 1131 - Enable Kokoro build bots (#1625) 1132 - Group tests into fewer executables, reduces load on CI 1133 - Port test script to Python 3 1134 - Symbol export tests respect SPIRV_SKIP_TESTS 1135 - #1596: Operand lookup succeeds if enabled by a capability 1136 - #1624: Instruction lookup succeeds if enabled by a capability 1137 - Refactoring namespaces: 1138 - #1678: Change libspirv to spvtools 1139 - Code in source/utils moved into spvtools::utils 1140 - Code in source/comp moved into spvtools::comp 1141 - Optimizer: 1142 - Remove insert-extract-elim pass. Use simplification pass instead. 1143 - Preserve instruction-to-block mapping in most passes, to reduce runtime. 1144 - Small vector optimization for operands 1145 - Add pass to move Private variables to Function. Increase opportunity to optimize. 1146 - Fixes: 1147 #1120: Check static uses of entry point interfaces 1148 #1372: Avoid merging some structs, to preserve names for reflection 1149 #1577: Scalar replacement uses only undecorated types. 1150 #1578: Fix handling of forward-pointer types, and types embedding pointers 1151 to themselves. 1152 #1591: Inliner: Callee variable with initializer should have a store at the call site. 1153 #1634: Fix crash: Use type id in vector type lookup 1154 #1649: Fix assert in compact-ids pass 1155 Fix constant folder: ensure it uses the right type 1156 #1659: Folding rules added to IRContext. Avoids leak. 1157 - Validator 1158 - Add work-in-progress WebGPU environment. Disallows OpUndef 1159 - #670, #1581: Improve error messages; disassemble instruction 1160 - #491: Check structured switches 1161 - #937: Check layout rules for Block and BufferBlock in Uniform, StorageBuffer, PushConstant 1162 - #1281: Check invalid branches into structured constructs 1163 - #1522: Disallow array-of-arrays with DescriptorSets 1164 - #1577: Allow duplicate pointer types. 1165 - #1581: Better messages: output ID names along with numbers in more cases. 1166 - #1597: Check Vulkan 1.1 capabilities 1167 - #1618: Check invalid exit from structured case construct 1168 - #1622: Run IdPass before DataRulesPass 1169 - #1632: Reduce test time by artificially lowering limits in limit test 1170 - #1638: Block-decorated structs member order must respect offset order 1171 - #1657: Improve CFG validation diagnostics 1172 - Khronos SPIR-V #337: GLSL.std.450 Refract instruction Eta param can be any float scalar. 1173 - #1606: PushConstant Blocks follow storage-buffer layout rules 1174 - #1664: Check layout of StorageBuffer variables with Block decoration, using storage buffer 1175 rules 1176 - #1666: Layout validation should permit {vec3; float} packing 1177 - #1637, #1668: Layout validation uses RowMajor, ArrayStride, MatrixStride properly 1178 - Linker 1179 - Avoid buffer overrun when creating OpModuleProcessed 1180 1181v2018.3 2018-05-25 1182 - General: 1183 - Support SPV_EXT_descriptor_indexing 1184 - Support SPV_GOOGLE_decorate_string 1185 - Support SPV_GOOGLE_hlsl_functionality1 1186 - Support SPV_NV_shader_subgroup_partitioned 1187 - Use "unified1" grammar from SPIRV-Headers 1188 - Simplify support for new extensions. Assembler, disassembler, and simple validation 1189 support is automatic if new tokens are introduced with appropriate extension 1190 attributes in the "unified1" SPIR-V core grammar. 1191 - Disassembler: Emit more digits on floating point, to reliably reproduce all 1192 significand bits. (Use std::max_digits10 instead of std::digits10) 1193 - Fix compilation for old XCode versions: Explicit construction required for std::set. 1194 - Optimizer: 1195 - Add --strip-reflect 1196 - Add --time-report 1197 - Add --loop-fission 1198 - Add lop fusion. 1199 - Add loop peeling pass and internal utility. 1200 - Improve optimizer runtime. 1201 - Merge-return now works with structured control flow. 1202 - New (faster) SSA rewriter to convert local loads and stores to SSA IDs and phis. 1203 Can replace load/store elimination passes. 1204 - Fix instruction folding case: insertion that feeds and extract, when the extract 1205 remains. 1206 - Fold OpDot. 1207 - Fold OpFNegate. 1208 - Fold multply and divide of same value. 1209 - Fold FClamp feeding a compare. 1210 - Fold OpLoad feeding an extract, to reduce excessive copying. (#1547) 1211 - Fold Fmix feeding an extract. 1212 - Use simplification pass instead of insert-extract elimination. 1213 - Constant fold OpVectorTimesScalar. 1214 - Copy propagate arrays, in simple cases. 1215 - Aggressive dead code elimination: Can remove more instructions, e.g. derivatives. 1216 - Aggressive dead code elimination: Remove Workgroup variables that are written but not read. 1217 - Better handling of OpImageTexelPointer 1218 - Initial utilities for scalar evolution. 1219 - Add Vector dead code elimination. 1220 - Each pass can only run once. 1221 - Allow code hosting in if-conversion. 1222 - Add external interface for adding a PassToken, so external code can make their own 1223 passes. 1224 - Fixes: 1225 #1404: Don't optimize away the compute compute workgroup size constant. 1226 #1407: Remove a bad assertion 1227 #1456: Fix bug in SSA rewriter related to variables updated in loops. 1228 #1487: Fix long runtime in Dead insertion elimination: Don't revist select phi nodes. 1229 #1492: Aggressive dead code elimination can remove OpDecorateStringGOOGLE. 1230 #1527: Fix inlining of functions having OpKill and OpUnreachable. 1231 #1559: Fix assert failure in reduce-load-size pass. 1232 #1556: Aggressive dead code elimination: Fix handling of OpCopyMemory. 1233 - Validator: 1234 - Check Vulkan built-in variables 1235 - Check Vulkan-specific atomic result type rule. 1236 - Relax control barrier check for SPIR-V 1.3. Fixes #1427 1237 - Check OpPhi. 1238 - Check OpMemoryModel. 1239 - Stop checking sizes derived from spec-constants. 1240 - Re-enable checks for OpUConvert. 1241 - Vulkan: Fix check for PrimitiveId: Permit as Input in fragment shader. 1242 - Validate binary version for the given target environment. 1243 - Add tests for OpBranch checks. 1244 - Vulkan 1.1: Check scope for non-uniform subgroup operations. 1245 - Fix checks for SPV_AMD_gpu_shader_int16. 1246 - Fix logical layout check for OpDecorateId. 1247 - Fix checks for ViewportIndex & Layer for Vulkan and SPV_EXT_shader_viewport_index_layer. 1248 - Fixes: 1249 #1470: Vulkan: Don't restrict WorkgroupSize to Input storage class. 1250 #1469: Vulkan: Permit Subgroup memory scope for Vulkan 1.1. 1251 #1472: Per-vertex variable validation fixes. 1252 #1483: Valdiate barrier execution scopes for Vulkan 1.1. 1253 - Fixes: 1254 #898: Linker properly removes FuncParamAttr from imported symbols. 1255 #924, #1174: Fix handling of decoration groups in optimizer, linker. 1256 1257v2018.2 2018-03-07 1258 - General: 1259 - Support SPIR-V 1.3 and Vulkan 1.1. 1260 - Default target environment is now SPIR-V 1.3. For command-line tools, 1261 use the --target-env option to override the default. Examples: 1262 # Generate a SPIR-V 1.0 binary instead of SPIR-V 1.3 1263 spirv-as --target-env spv1.0 a.spvasm -o a.spv 1264 spirv-as --target-env vulkan1.0 a.spvasm -o a.spv 1265 # Validate as Vulkan 1.0 1266 spirv-val --target-env vulkan1.0 a.spv 1267 - Support SPV_GOOGLE_decorate_string and SPV_GOOGLE_hlsl_functionality1 1268 - Fixes: 1269 - Fix Android.mk build. Compilation was failing due to missing definitions of 1270 SpvCapabilityFloat16ImageAMD and other enumerated values. 1271 - Optimizer: Avoid generating duplicate names when merging types. 1272 - #1375: Validator: SPV_AMD_gpu_shaer_half_float implicitly allows declaration 1273 of the 16-bit floating point type. 1274 - #1376: Optimizer: Avoid folding half-precision float. 1275 1276v2018.1 2018-03-02 1277 - General: 1278 - Support Visual Studio 2013 again. (Continue support for VS 2015 and VS 2017.) 1279 - Support building SPIRV-Tools as a shared library. 1280 - Improve the HLSL legalization optimization recipe. #1311 1281 - Optimizer: 1282 - General speedups. 1283 - Remove generic dead code elimination functionality from transforms: 1284 --eliminate-local-single-block 1285 --eliminate-local-single-store 1286 --eliminate-local-multi-store 1287 To recover the previous behaviour, a recipe using those transforms should now 1288 also invoke the --eliminate-dead-code-aggressive transform. 1289 - Improve folding, including coverage for floating point, OpSelect, and arithmetic 1290 with non-trivial constant operands. 1291 - Add loop-invariant code motion pass. 1292 - Add loop-unrolling pass, for honouring unroll hits. 1293 - Add loop-unswitch pass. 1294 - Add instruction simplification pass. 1295 - Aggressive dead code elimination: Understands capability hierarchy when finding 1296 instructions it can eliminate (combinators). (PR #1268) 1297 - CCP can now fold floating point arithmetic. #1311 1298 - Validator: 1299 - Validate barrier instructions. 1300 - Check Vulkan-specific rules for atomics. 1301 - Check Vulkan prohibition of Location or Component decorations on BuiltIn variables. 1302 - Linker: 1303 - Add --verify-ids option 1304 - Add option to allow a resulting module to be partially linked. 1305 - Handle OpModuleProcessed (instructions in SPIR-V layout section 7c) 1306 - Fixes: 1307 - #1265: Optimizer: Fix use-after free bug in if-conversion. (Fix object lifecycle bug 1308 in type manager.) 1309 - #1282: Fix new warnings found by GCC 8.0.1. 1310 - #1285: Optimizer: Fix random failures during inlining. (Dangling references in DefUseManager) 1311 - #1295: Optimizer: Fix incorrect handling of Phi nodes in CCP. 1312 - #1300: Fix CCP: avoid bad CCP transitions and unsettled values. 1313 - #1304: Avoid static-duration variables of class type (with constructors). 1314 - #1323: Fix folding of an insert composite feeding a composite extract. 1315 - #1339: Fix CCP: Handle OpConstantNull boolean values as conditions. 1316 - #1341: DCEInst: Keep atomic instructions (and some others with side effects). 1317 - #1354: Don't fold integer division. 1318 - #1357: Support OpConstantNull in folding. 1319 - #1361: CCP: Fix handling of non-constant module-scope values 1320 1321v2018.0 2018-02-02 1322 - General 1323 - VisualStudio 2013 is no longer supported. VisualStudio 2015 is supported. 1324 - Use "include/unified1" directory from SPIRV-Headers. Requires recent SPIRV-Headers source. 1325 - Disassembler: spirv-dis adds --color option to force color disassembly. 1326 - Optimizer: 1327 - Add pass to eliminate dead insertions. 1328 - Aggressive dead code elimination now removes OpSwitch constructs. 1329 - Block merging occurs in more cases. 1330 - Add driver workaround transform: replace OpUnreachable with harmless branch to merge. 1331 - Improve instruction folding framework. 1332 - Add loop analysis. 1333 - Add scalar replacement of aggregates to size-optimization recipe. 1334 - Add pass to replace instructions invalid for a shader stage, with a harmless value. 1335 This changes the semantics of the program! Not for general use! 1336 - Rearragne and add passes to performance-optimization recipe, to produce better results. 1337 - Validator: 1338 - Validate OpenCL extended instructions. 1339 - Shaders can't perform atomics on floats. 1340 - Validate memory semantics values in atomics. 1341 - Validate instruction-adjacency constraints, e.g. OpPhi predecessors, merge instructions 1342 immediately precede branches. 1343 - Fixes: 1344 - PR 1198: Optimizer: Fix CCP in presence of matrix constants. 1345 - #1199: Optimizer: Fix CCP: don't propagate spec constants. 1346 - #1203: Optimizer: Fix common uniform elim bug introduced by refactoring. 1347 - #1210: Optimizer: Aggressive dead code elimination: Fix 'break' identification. 1348 - #1212: Optimizer: Aggressive dead code elimination: Was skipping too many instructions. 1349 - #1214: Optimizer: Aggressive dead code elimination: Fix infinite loop. 1350 - #1228: Optimizer: Fix CCP: Handling of varying Phi nodes; was resulting in infinite loop. 1351 - #1245: Optimizer: Dead branch elimination: Avoid a null pointer dereference. 1352 - #1250: Optimizer: Dead branch elimination: Avoid spuriously reporting a change. 1353 1354v2017.3 2018-01-12 1355 - General: 1356 - Support DebugInfo extended instruction set, targeted at OpenCL environments. 1357 See the SPIR-V Registry. 1358 - Generate a SPIRV-Tools.pc file for pkg-config. 1359 - Optimizer: 1360 - Progress for legalization of code generated from HLSL (issue #1118): 1361 - Add --legalize-hlsl option to run transforms used to transform intermediate 1362 code generated by HLSL to SPIR-V for Vulkan compilers. Those compilers 1363 normally run these transforms automatically. This option is used for developing 1364 those transforms. 1365 - Add Private-to-Function variable conversion for modules with logical 1366 addressing. 1367 - Add --ccp: SSA Conditional Constant Propagation (CCP) 1368 - Add --print-all to show disassembly for each optimization pass. 1369 - Internal: Add loop descriptors and post-order tree iterator. 1370 - Generalized dead branch elimination 1371 - Aggressive dead code elimination (ADCE) now removes dead functions and 1372 module-scope variables. 1373 - Vector extract/insert elimination now optimizes through some cases of 1374 VectorShuffle, and GLSL.std.450 Mix extended instruction. 1375 - Validator: 1376 - Add validation for GLSL.std.450 extended instruction set. 1377 - Check out of bounds composite accesses, where that's statically computable. 1378 Fixes #1112. 1379 - Check upper bits of literal numbers that aren't a multiple of 32-bits wide. 1380 - More validation of primitive instructions 1381 - Add optional "relaxed" checking logical addressing mode to permit some 1382 cases of pointer-to-pointer. Contributes to HLSL legalization (issue #1118). 1383 - Fixes: 1384 #1100: Validator: Image operand Sample can be used with OpImageSparseFetch, 1385 OpImageSparseRead. 1386 #1108: Remove duplicates transform was incorrectly removing non-duplicate 1387 decorations. 1388 #1111: Optimizer's type manager could reference deleted memory. 1389 #1112: Fix decoration equality check, e.g. it is now symmetric. 1390 #1129: Validator now disallows Dim=SupbassData for OpImageSparseRead. 1391 #1143: Fix CCP: Was generating incorrect code for loops. 1392 #1153: Fix CCP crash. 1393 #1154: Optimizer's internal instruction-to-block mappings were sometimes 1394 inconsistent. 1395 #1159: Fix CCP infinite loop. 1396 #1168: Fix dead branch elimination intermittently generating incorrect code. 1397 Fixes https://github.com/KhronosGroup/glslang/issues/1205 1398 #1186: Fix validation of PackDouble2x32 and UnpackDouble2x32 1399 1400v2017.2 2017-12-15 1401 - General: 1402 - Support OpenCL 1.2, 2.0 target environments, including embedded profiles 1403 - Add CONTRIBUTING.md 1404 - Fix exit status code for spirv-link 1405 - Disassember: Enable emitting ANSI colour codes to a string 1406 - Library avoids polluting global namespace. The libraries can export C and C++ 1407 symbols starting with "spv", or in a C++ namespace. Add a test for this. 1408 - Linux release builds include debug information, for easier profiling 1409 - Build bots no longer test VisualStudio 2013 1410 - Testing dependency RE2 requires VisualStudio 2015 or later 1411 - Build bots check code formatting 1412 - Optimizer: 1413 - Add --skip-validation to spirv-opt 1414 - Add dominance tree analysis 1415 - Add generic value propagation engine 1416 - Add global redundancy elimination within a function 1417 - Add scalar replacement of function-scope variables of composite type 1418 - Aggressive dead code elimination: Remove empty loops 1419 - Killing an instruction notifies the IRContext 1420 - IRContext::KillInst deletes the instruction 1421 - Move CFG analysis to IRContext 1422 - Add constant manager 1423 - Fix: Don't consider derivative instructions as combinators. 1424 - Fix: Don't delete an instruction twice in local dead-code-elimination 1425 - Fix: Don't consider derivative instructions as combinators. 1426 - Validator: 1427 - Finish checking of image instructions (Section 3.32.10) 1428 - Check sparse image instructions 1429 - Check OpTypeImage, OpTypeSampleImage 1430 - Check composite instructions (Section 3.32.12) 1431 - Check atomic instructions (Section 3.32.18) 1432 - Check OpEmitStreamVertex, OpEndStreamPrimitive instructions 1433 - Re-enable validation of OpCopyObject 1434 - OpKill, image ImplicitLod and QueryLod instructions can only be used in Fragment 1435 shaders. 1436 - Fixes for image instruction validation: 1437 - Lod image operand only usable with ExplicitLod and OpImageFetch 1438 - ExplicitLod Lod image operand must be float scalar 1439 - OpImageFectch Lod image operand must be int scalar 1440 - OpImageGather component operand must be 32-bits (integer scalar) 1441 - OpImageQuerySizeLod Lod must be integer scalar 1442 - Fixes: 1443 #622: Remove names and decorations when inlining 1444 #989: Aggressive dead code elim: Don't optimize away live breaks from a loop 1445 #991: Fix validation of SPV_AMD_shader_ballot 1446 #1004: Use after free of an instruction, in remove-duplicates transform 1447 #1007: OpImageRead not required to return 4-component vector 1448 #1009: OpImageRead can return scalar int/float types 1449 #1011: OpImageWrite should allow scalar int/float texel types 1450 #1012: Fix validat Dref type check 1451 #1017: Load-store elimination considers variable initializations 1452 #1034: Fix Windows debug build: operator< should be a weak ordering 1453 #1083: Inlining: Set parent (function) for each inlined basic block. 1454 #1075: Aggressive dead code elimination: Was leaving dangling references to 1455 removed blocks. 1456 1457v2017.1 2017-11-23 1458 - Update README with details on the public_spirv_tools_dev@khronos.org mailing list. 1459 - General: 1460 - Automatically deploy built artifacts to GitHub Releases 1461 - Add a Linker (module combiner). Under development. 1462 - Add Android.mk for Android NDK builds. 1463 - Add the 'effcee' library as an optional dependency for use in tests. 1464 Eventually it will be a required dependency, once downstream projects have 1465 a chance to adjust. Requires 're2' library. 1466 - Avoid static-duration variables of class type (with constructors). 1467 - Hack around bugs in gcc-4.8.1 template handling 1468 - Faster opcode lookup 1469 - Validator: 1470 - Recognize extensions listed on SPIR-V registry, 1471 through #25 SPV_AMD_shader_fragment_mask 1472 - Validator issues an info message when it sees an unrecognized extension. 1473 - Type check basic arithmetic operations 1474 - Type check carry/extended arithmetic operations 1475 - Type check vector arithmetic operations 1476 - Type check Relational and Logical instructions 1477 - Type check Bit instructions 1478 - Check type uniqueness rules 1479 - Check conversion instructions 1480 - Check image instructions 1481 - Check derivative instructions 1482 - Check OpVectorShuffle 1483 - Check OpBranchConditional 1484 - OpModuleProcessed is only allowed after debug names section and before annotations 1485 section. 1486 - Checks the right kind of return is called for each function (void or non-void). 1487 - Add option to relax type check when storing structs (--relax-store-struct) 1488 - Optimizer: 1489 - Refactoring internal representation of the module, including: 1490 - IRContext: owns a module and manages analyses 1491 - Instructions are owned by intrusive lists, and have unique IDs 1492 - BasicBlock owns its instruction list. 1493 - DefUseManager: change representation of uses, for faster processing 1494 on large modules. 1495 - Add high level recipes: -O, -Os, and -Oconfig 1496 Recipes for -O and -Os are under development. 1497 - Add eliminate-dead-function transform 1498 - Add strength reduction transform: For now, convert multiply by power of 2 1499 to a bit shift. 1500 - Add CFG cleanup transform 1501 - Add removal of dead module-scope variables 1502 - Add merge-return transform for modules without structured control flow 1503 - Add redundancy elimination within a basic block (local value numbering) 1504 - Extract-insert elimination: 1505 - Recognize the case where the first instruction in the sequence is an 1506 OpCompositeConstruct or OpConstantComposite 1507 - Handle some cases of nested structs 1508 - Dead branch elimination now can eliminate entire selection constructs 1509 when all arms are dead. 1510 - Compressing codec: 1511 - Updated algorithm to 1.01, 1.02, 1.03 1512 - Not built by default. Use -DSPIRV_BUILD_COMPRESSION=ON to build. 1513 - Codec can be parameterized by a customized model. 1514 - Fixes: 1515 #728: Fix decoration of inlined functions 1516 #798: spirv-as should fail when given unrecognized long option 1517 #800: Inliner: Fix inlining function into header of multi-block loop 1518 #824: Eliminate-local-multi-store: Fix a crash 1519 #826: Elimiante-local-multi-store: Fix a crash 1520 #827: Fix crash when compact-ids transform runs before another transform. 1521 #834: Add Cmake option to build the compressing codec. Off by default. 1522 #911: Fix classification of Line and NoLine instructions 1523 1524v2017.0 2017-09-01 1525 - Update README to describe that assembler, disassembler, and binary parser support 1526 are based on grammar files from the SPIRV-Headers repository. 1527 1528v2016.7 2017-09-01 1529 - Add SPIR-V 1.2 1530 - OpenCL 2.2 support is now based on SPIR-V 1.2 1531 - Support AMD extensions in assembler, disassembler: 1532 SPV_AMD_gcn_shader 1533 SPV_AMD_shader_ballot 1534 SPV_AMD_shader_explicit_vertex_parameter 1535 SPV_AMD_shader_trinary_minmax 1536 SPV_AMD_gpu_shader_half_float 1537 SPV_AMD_texture_gather_bias_lod 1538 SPV_AMD_gpu_shader_int16 1539 - Optimizer: Add support for: 1540 - Inline all function calls in entry points. 1541 - Flatten decoration groups. Fixes #602 1542 - Id compaction (minimize Id bound). Fixes #624 1543 - Eliminate redundant composite insert followed by extract 1544 - Simplify access chains to local variables 1545 - Eliminate local variables with a single store, if possible 1546 - Eliminate local variables with a several stores, if possible 1547 - Eliminate loads and stores in same block to local variables 1548 - Eliminate redundant insert/extract to composite values 1549 - Aggressive dead instruction elimination 1550 - Eliminate dead branches 1551 - Merge blocks when the second can only be preceded by the first 1552 - Eliminate ommon uniform loads 1553 - Assembler: Add option to preserve numeric ids. Fixes #625 1554 - Add build target spirv-tools-vimsyntax to generate spvasm.vim, a SPIR-V 1555 assembly syntax file for Vim. 1556 - Version string: Allow overriding of wall clock timestamp with contents 1557 of environment variable SOURCE_DATE_EPOCH. 1558 - Validator implements relaxed rules for SPV_KHR_16bit_storage. 1559 - CMake installation rules use GNUInstallDirs. For example, libraries 1560 will be installed into a lib64 directory if that's the norm for the 1561 current system. 1562 - Fixes: 1563 #500: Parameterize validator limit checks 1564 #508: Support compilation under CYGWIN 1565 #517: Fix validation when continue (or case) contstruct is also the head of a 1566 nested control construct. 1567 #551: If a merge block is reachable, it must be *strictly* dominated by its 1568 header. 1569 #548: Validator: Error when the reserved OpImageSparseSampleProj* opcodes 1570 are used. 1571 #611: spvtools::Optimizer was failing to save the module to the output 1572 binary vector when all passes succeded without changes. 1573 #629: The inline-entry-points-all optimization could generate invalidly 1574 structured code when the inlined function had early returns. 1575 #697: Optimizer's Instruction::ForEachInId method was skipping semantics-id 1576 and scope-id. 1577 #755: Inliner: Fix inlining of callee with single Return appearing before 1578 the end of the function. 1579 #776: Fix dead branch elimination in presence of complex but dead control 1580 flow. 1581 #781: SPV_KHR_variable_pointers allows duplicate pointer types 1582 #782: Inliner: Fix remapping of non-label forward references in callee 1583 #787: Inliner: Fix remapping of inlined entry block when called from 1584 single block loop. 1585 #790: Inliner: Fix remapping of inlined entry block when callee has 1586 multiple returns. 1587 1588v2016.6 2016-12-13 1589 - Published the C++ interface for assembling, disassembling, validation, and 1590 optimization. 1591 - Support SPV_KHR_shader_draw_parameters in assembler, disassembler, parser. 1592 - Validator: 1593 - Add validator API accepting raw binary words 1594 - Increased coverage: 1595 - Checks "Data rules" in Universal Validation Rules, section 2.16.1 1596 - WIP: Universal Limits. 1597 - The minimum mandated upper bounds are checked. 1598 - TODO: Parameterize the validator to allow larger limits accepted by 1599 a more than minimally capable implementation. 1600 - OpSampledImage checks 1601 - OpConstantComposite checks 1602 - Id bound check 1603 - Disasssembler: 1604 - Generates friendly GLSL-based names for more builtin variables 1605 - Generates friendly names for numeric OpConstant values 1606 - Vendor tool info extracted from SPIR-V XML registry file. 1607 - Fixes issues: 1608 #429: Validator: Allow OpTypeForwardPointer and OpTypeStruct to reference 1609 undefined IDs 1610 #482: Validator: OpVariable initializer can be an ID of a module-scope variable 1611 1612v2016.5 2016-09-16 1613 - Support SPV_KHR_shader_ballot in assembler, disassembler, parser. 1614 - Disassembler: Generate friendly names for built-in variables. 1615 - Partial fixes: 1616 #359: Add Emacs helper for automatically diassembling/assembling a SPIR-V 1617 binary on file load/save. 1618 - Fixes: 1619 #414: Validator: Allow OpUndef for composite constants 1620 #415: Validator: Phi can use its own value in some cases. 1621 1622v2016.4 2016-09-01 1623 - Relicensed under Apache 2.0 1624 - Add optimization passes (in API and spirv-opt command) 1625 - Fold spec constants defined with OpSpecConstantOp and 1626 OpSpecConstantComposite to normal constants with fixed value(s). 1627 - Fixes issues: 1628 #318: Relicensed under Apache 2.0 1629 1630v2016.3 2016-08-24 1631 - Add target environment enums for OpenCL 2.1, OpenCL 2.2, 1632 OpenGL 4.0, OpenGL 4.1, OpenGL 4.2, OpenGL 4.3, OpenGL 4.5. 1633 - Add spirv-cfg, an experimental tool to dump the control flow graph 1634 as a GraphiViz "dot" graph 1635 - Add optimization pass: Eliminate dead constants. 1636 - Add spirv-lesspipe.sh filter utility 1637 - Fixes issues: 1638 #288: Check def-use dominance rules for OpPhi (variable,parent) operands 1639 #339: Allow OpUndef in types-constants-global-vars section, as required 1640 by SPIR-V 1.0 Rev7, 1.1 Rev 3. 1641 #340: Avoid race on mkdir during build 1642 #365: Relax PointSize, ClipDistance, CullDistance capability check in all 1643 environments not just Vulkan 1.0. 1644 1645v2016.2 2016-08-05 1646 - Validator is incomplete 1647 - Checks ID use block is dominated by definition block 1648 - Add optimization passes (in API and spirv-opt command) 1649 - Strip debug info instructions 1650 - Freeze spec constant to their default values 1651 - Allow INotEqual as operation for OpSpecConstantOp 1652 - Fixes bugs: 1653 #270: validator: crash when continue construct is unreachable 1654 #279: validator: infinite loop when analyzing some degenerate control 1655 flow graphs 1656 #286: validator: don't incorrectly generate def-use error for 1657 (variable,parent) parameters to OpPhi 1658 #290: disassembler: never generate bare % for an identifier 1659 #295: validator: def-use dominance check should ignore unreachable uses 1660 #276: validator: allow unreachable continue constructs 1661 #297: validator: allow an unreachable block to branch to a reachable 1662 merge block 1663 1664v2016.1 2016-07-19 1665 - Fix https://github.com/KhronosGroup/SPIRV-Tools/issues/261 1666 Turn off ClipDistance and CullDistance capability checks for Vulkan. 1667 - The disassembler can emit friendly names based on debug info (OpName 1668 instructions), and will infer somewhat friendly names for most types. 1669 This is turned on by default for the spirv-dis command line tool. 1670 - Updated to support SPIR-V 1.1 rev 2 1671 - Input StorageClass, Sampled1D capability, and SampledBuffer capability 1672 do not require Shader capability anymore. 1673 1674v2016.0 2016-07-04 1675 1676 - Adds v<year>.<index> versioning, with "-dev" indicating 1677 work in progress. The intent is to more easly report 1678 and summarize functionality when SPIRV-Tools is incorporated 1679 in downstream projects. 1680 1681 - Summary of functionality (See the README.md for more): 1682 - Supports SPIR-V 1.1 Rev 1 1683 - Supports SPIR-V 1.0 Rev 5 1684 - Supports GLSL std450 extended instructions 1.0 Rev 3 1685 - Supports OpenCL extended instructions 1.0 Rev 2 1686 - Assembler, disassembler are complete 1687 - Supports floating point widths of 16, 32, 64 bits 1688 - Supports integer widths up to 64 bits 1689 - Validator is incomplete 1690 - Checks capability requirements in most cases 1691 - Checks module layout constraints 1692 - Checks ID use-definition ordering constraints, 1693 ignoring control flow 1694 - Checks some control flow graph rules 1695 - Optimizer is introduced, with few available transforms. 1696 - Supported on Linux, OSX, Android, Windows 1697 1698 - Fixes bugs: 1699 - #143: OpenCL pow and pown arguments 1700