1Revision history for SPIRV-Tools 2 3v2025.1 2025-02-28 4 - General 5 - diff: Fix crash in OpString matching (#5988) 6 - Add SPV_AMDX_shader_enqueue version 2 support (#5838) 7 - add support for SPV_INTEL_subgroup_matrix_multiply_accumulate (#5928) 8 - update cmake_minimum_required to 3.22.1 (#5925) 9 - Add OpImageSampleFootprintNV to IsAllowedSampledImageOperand (#5914) 10 - assembler: ensure progress when seeking the version string (#5910) 11 - Optimizer 12 - opt: keep all OpSource instructions (#5901) 13 - [opt] Fix bug opt::InstructionBuilder::AddVariable (#6007) 14 - [OPT] Add SPV_KHR_ray_tracing to allow list (#5941) 15 - opt: keep all OpSource instructions (#5901) 16 - Validator 17 - spirv-val: Add AllowVulkan32BitBitwise option (#6001) 18 - Fix untyped pointer comparison validation (#6004) 19 - spirv-val: Update VUIDs for 308 header (#5990) 20 - spirv-val: fix env parsing for vk1.1spv1.4 (#5985) 21 - Add validation for SPV_NV_linear_swept_spheres. (#5975) 22 - Add validation SPV_NV_cluster_acceleration_structure. (#5974) 23 - Improve the instruction diagnostic for some access chain errors (#5978) 24 - Update physical storage buffer validation to match SPIR-V 1.6.5 (#5977) 25 - Validate SPV_NV_cooperative_vector (#5972) 26 - Fix layout checks with untyped pointers (#5970) 27 - spirv-val: Update mislabeled VUIDs (#5964) 28 - More explicit layout validation (#5958) 29 - spirv-val: Add VK_KHR_maintenance8 support (#5951) 30 - Add EXT_mesh_shader validation support (#5640) 31 - spirv-val: Remove OpenCL ivec3 req (#5940) 32 - spirv-val: Validate zero product workgroup size (#5407) 33 - Relax DebugLine validation (#5916) 34 - Linker 35 - linker: remove LinkOnceODR decorations when linking executables (#5979) 36 - fix: handle LinkOnceODR correctly (#5938) 37 38v2024.4 2024-12-04 39 - General 40 - Add FPEncoding operand type. (#5726) 41 - Support SPV_KHR_untyped_pointers (#5736) 42 - add support for SPV_INTEL_global_variable_host_access (#5786) 43 - Add support for SPV_KHR_compute_shader_derivative (#5817) 44 - Accept hex representation as binary input (#5870) 45 - Vulkan 1.4 support (#5899) 46 - Optimizer 47 - Add knowledge of cooperative matrices (#5720) 48 - Add struct-packing pass and unit test. (#5778) 49 - Validator 50 - Validate presence of Stride operand to OpCooperativeMatrix{Load,Store}KHR (#5777) 51 - Update sampled image validation (#5789) 52 - Disallow stores according to VUID 06924 (#5368) 53 - Add validation for SPV_NV_tensor_addressing and SPV_NV_cooperative_matrix2 (#5865) 54 - Linker 55 - allow linking functions with different pointer arguments (#5534) 56 57v2024.3 2024-06-20 58 - General 59 - Optimizer 60 - Do not fold mul and adds to generate fmas (#5682) 61 - add OpExtInst forward ref fixup pass (#5708) 62 - Validator 63 - Separate Location check for tess patch (#5654) 64 - Validate MemoryAccessMask of OpCooperativeMatrixStoreKHR (#5668) 65 - OpSampledImage extra validation (#5695) 66 - add support for OpExtInstWithForwardRefs (#5698)A 67 - Disassembler 68 - add decorations to comments (#5675) 69 - Add --nested-indent and --reorder-blocks (#5671) 70 71v2024.2 2024-04-22 72 - General 73 - Add SPIRV_TOOLS_EXPORT to public C++ API (#5591) 74 - Use bazel 7 and bzlmod (#5601) 75 - Optimizer 76 - opt: add GroupNonUniformPartitionedNV capability to trim pass (#5648) 77 - Fix rebuilding types with circular references. (#5637) 78 - Add AliasedPointer decoration (#5635) 79 - add support for vulkan-shader-profiler external passes (#5512) 80 - Validator 81 - A fix to support of SPV_QCOM_image_processing2 (#5646) 82 - spirv-val: Add Vulkan check for Rect Dim in OpTypeImage (#5644) 83 - Validate duplicate decorations and execution modes (#5641) 84 - Validator: Support SPV_NV_raw_access_chains (#5568) 85 86v2024.1 2024-03-06 87 - General 88 - Add tooling support for SPV_KHR_maximal_reconvergence (#5542) 89 - Add support for SPV_KHR_float_controls2 (#5543) 90 - SPV_KHR_quad_control (#5547) 91 - Fold 64-bit int operations (#5561) 92 - update image enum tests to remove Kernel capability (#5562) 93 - Support operand kind for SPV_INTEL_maximum_registers (#5580) 94 - SPV_NV_shader_atomic_fp16_vector (#5581) 95 - Support for SPV_QCOM_image_processing2 (#5582) 96 - Fix access chain struct checks (#5592) 97 - Optimizer 98 - opt: add Int16 and Float16 to capability trim pass (#5519) 99 - Add preserver-interface option to spirv-opt (#5524) 100 - spirv-opt: Fix OpCompositeExtract relaxation with struct operands (#5536) 101 - opt: Add VulkanMemoryModelDeviceScope to trim (#5544) 102 - opt: Add TrimCapabilities pass to spirv-opt tool (#5545) 103 - Add modify-maximal-reconvergence to spirv-opt help (#5546) 104 - opt: add SPV_EXT_mesh_shader to opt allowlist (#5551) 105 - opt: Add OpEntryPoint to DescriptorScalarReplacement pass (#5553) 106 - opt: prevent meld to merge block with MaximalReconvergence (#5557) 107 - [OPT] Use new instruction folder for for all opcodes in spec consti folding (#5569) 108 - [OPT] Identify arrays with unknown length in copy prop arrays (#5570) 109 - [OPT] Add removed unused interface var pass to legalization passes (#5579) 110 - Validator 111 - spirv-val: Re-enable OpControlBarrier VU (#5527) 112 - spirv-val: Add Mesh Primitive Built-In validaiton (#5529) 113 - spirv-val: Validate PhysicalStorageBuffer Stage Interface (#5539) 114 - spirv-val: Multiple interface var with same SC (#5528) 115 - spirv-val: Revert Validate PhysicalStorageBuffer Stage Interface (#5575) 116 - spirv-val: Make Constant evaluation consistent (#5587) 117 118v2023.6 2023-12-18 119 - General 120 - update_build_version.py produce deterministic header. (#5426) 121 - Support missing git in update_build_version.py (#5473) 122 - Optimizer 123 - Add ComputeDerivativeGroup*NV capabilities to trim capabilities pass. (#5430) 124 - Do not crash when tryingto fold unsupported spec constant (#5496) 125 - instrument: Fix handling of gl_InvocationID (#5493) 126 - Fix nullptr argument in MarkInsertChain (#5465) 127 - opt: support 64-bit OpAccessChain index in FixStorageClass (#5446) 128 - opt: add StorageImageReadWithoutFormat to cap trim (#5475) 129 - opt: add PhysicalStorageBufferAddresses to trim (#5476) 130 - Fix array size calculation (#5463 131 - Validator 132 - spirv-val: Loosen restriction on base type of DebugTypePointer and DebugTypeQualifier (#5479) 133 - spirv-val: Add WorkgroupMemoryExplicitLayoutKHR check for Block (#5461) 134 135v2023.5 2023-10-15 136 - General 137 - Support 2 Intel extensions (#5357) 138 - SPV_QCOM_image_processing support (#5223) 139 - Optimizer 140 - opt: fix StorageInputOutput16 trimming. (#5359) 141 - opt: add StoragePushConstant16 to trim pass (#5366) 142 - opt: enable StorageUniform16 (#5371) 143 - opt: add bitmask support for capability trimming (#5372) 144 - opt: Add SwitchDescriptorSetPass (#5375) 145 - opt: add FragmentShader*InterlockEXT to capability trim pass (#5390) 146 - opt: add Int64 capability to trim pass (#5398) 147 - opt: add Float64 capability to trim pass (#5428) 148 - opt: add raytracing/rayquery to trim pass (#5397) 149 - opt: add ImageMSArray capability to trim pass. (#5395) 150 - Add SPV_KHR_physical_storage_buffer to allowlists (#5402) 151 - Add SPV_EXT_fragment_shader_interlock to allow lists (#5393) 152 - Make sure that fragment shader interlock instructions are not removed by DCE (#5400) 153 - instrument: Use Import linkage for instrumentation functions (#5355) 154 - Add a new legalization pass to dedupe invocation interlock instructions (#5409) 155 - instrument: Ensure linking works even of nothing is changed (#5419) 156 - Validator 157 - Move token version/cap/ext checks from parsing to validation (#5370) 158 - val: re-add ImageMSArray validation (#5394) 159 - Linker 160 - linker: Add --use-highest-version option 161 162v2023.4 2023-07-17 163 - General 164 - Set cmake_policy CMP0128 (#5341) 165 - Add python3 requirement for the script (#5326) 166 - Add support for LiteralFloat type (#5323) 167 - SPV_KHR_cooperative_matrix (#5286) 168 - Allow OpTypeBool in UniformConstant (#5237) 169 - Allow physical storage buffer pointer in IO (#5251) 170 - Remove const zero image operands (#5232) 171 - Optimizer 172 - Enable vector constant folding (#4913) (#5272) 173 - Fold negation of integer vectors (#5269) 174 - Add folding rule for OpTranspose (#5241) 175 - Add SPV_NV_bindless_texture to spirv optimizations (#5231) 176 - Fix incorrect half float conversion (#5349) 177 - Add SPV_EXT_shader_atomic_float_add to allow lists (#5348) 178 - Instrument 179 - instrument: Cast gl_VertexIndex and InstanceIndex to uint (#5319) 180 - instrument: Fix buffer address length calculations (#5257) 181 - instrument: Reduce number of inst_bindless_stream_write_6 calls (#5327) 182 - Validator 183 - Validate GroupNonUniform instructions (#5296) 184 - spirv-val: Label SPV_KHR_cooperative_matrix VUID (#5301) 185 - Validate layouts for PhysicalStorageBuffer pointers (#5291) 186 - spirv-val: Remove VUID from 1.3.251 spec (#5244) 187 - Diff 188 - spirv-diff: Update test expectations (#5264) 189 - spirv-diff: Leave undefined ids unpaired. (#5262) 190 - spirv-diff: Properly match SPV_KHR_ray_query types. (#5259) 191 - diff: Don't give up entry point matching too early. (#5224) 192 193v2023.3 2023-05-15 194 - General 195 - Update spirv_headers to include SPV_KHR_ray_tracing_position_fetch (#5205) 196 - spirv-tools: Add support for QNX (#5211) 197 - build: set std=c++17 for BUILD.gn (#5162) 198 - Optimizer 199 - Run ADCE when the printf extension is used. (#5215) 200 - Don't convert struct members to half (#5201) 201 - Apply scalar replacement on vars with Pointer decorations (#5208) 202 - opt: Fix null deref in OpMatrixTimesVector and OpVectorTimesMatrix (#5199) 203 - instrument: Add set and binding to bindless error records (#5204) 204 - instrument: Change descriptor state storage format (#5178) 205 - Fix LICMPass (#5087) 206 - Add Vulkan memory model to allow lists (#5173) 207 - Do not remove control barrier after spv1.3 (#5174) 208 - Validator 209 - spirv-val: Label Interface Location/Component VUIDs (#5221) 210 - Add support for SPV_EXT_shader_tile_image (#5188) 211 - Fix vector OpConstantComposite type validation (#5191) 212 - spirv-val: Label new Vulkan VUID 07951 (#5154) 213 - Fuzz 214 - Do not define GOOGLE_PROTOBUF_INTERNAL_DONATE_STEAL_INLINE if it is already defined. (#5200) 215 216v2023.2 2023-03-10 217 - General 218 - build: move from c++11 to c++17 (#4983) 219 - tools: refactorize tools flags parsing. (#5111) 220 - Add C interface for Optimizer (#5030) 221 - libspirv.cpp: adds c++ api for spvBinaryParse (#5109) 222 - build: change the way we set cxx version for bazel. (#5114) 223 - Optimizer 224 - Fix null pointer in FoldInsertWithConstants. (#5093) 225 - Fix removal of dependent non-semantic instructions (#5122) 226 - Remove duplicate lists of constant and type opcodes (#5106) 227 - opt: fix spirv ABI on Linux again. (#5113) 228 - Validator 229 - Validate decoration of structs with RuntimeArray (#5094) 230 - Validate operand type before operating on it (#5092) 231 - spirv-val: Conditional Branch without an exit is invalid in loop header (#5069) 232 - spirv-val: Initial SPV_EXT_mesh_shader builtins (#5080) 233 234v2023.1 2023-01-17 235 - General 236 - Renamed "master" to "main" (issue#5051) 237 - Validate version 5 of clspv reflection (#5050) 238 - Remove testing support for VS2015 (#5027) 239 - Fix undef behaviour in hex float parsing (#5025) 240 - Require C++11 *or later* (#5020) 241 - Instrument 242 - Instrument: Fix bindless checking for BufferDeviceAddress (#5049) 243 - Optimizer 244 - Optimize allocation of spvtools::opt::Instruction::operands_ (#5024) 245 - spirv-opt: Fix OpCompositeInsert with Null Constant (#5008) 246 - spirv-opt: Handle null CompositeInsert (#4998) 247 - Add option to ADCE to remove output variables from interface. (#4994) 248 - Add support for tesc, tese and geom to EliminateDead*Components (#4990) 249 - Add pass to eliminate dead output components (#4982) 250 - spirv-opt: Add const folding for CompositeInsert (#4943) 251 - Add passes to eliminate dead output stores (#4970) 252 - Prevent eliminating case constructs in block merging (#4976) 253 - Validator 254 - Fix layout validation (#5015) 255 - Fix use of invalid analysis (#5013) 256 - Fix infinite loop in validator (#5006) 257 - Add validation support for SPV_NV_shader_invocation_reorder. (#4979) 258 - Only validate full layout in Vulkan environments (#4972) 259 - spirv-val: Label new Vulkan OpPtrAccessChain VUs (#4975) 260 - spirv-val: Add OpPtrAccessChain Base checks (#4965) 261 262 263v2022.4 2022-10-12 264 - General 265 - Support Narrow Types in BitCast Folding Rule (#4941) 266 - spirv-diff: Allow no SpecId (#4904) 267 - build: cmake: Add support for GNU/Hurd (#4895) 268 - Implement tool changes for SPV_EXT_mesh_shader. (#4915) 269 - Validator 270 - spirv-val: Add SPV_ARM_core_builtins validation (#4958) 271 - spirv-val: Add an option to use friendly names or not (#4951) 272 - spirv-val: Consistently quote ids in messages (#4950) 273 - spirv-val: Add initial SPV_EXT_mesh_shader validation (#4924) 274 - spirv-val: Make it legal to use arrays of ray queries (#4938) 275 - spirv-val: Better message for using OpTypeBool in input/output (#4901) 276 - spirv-val: Add SPV_KHR_ray_tracing storage class (#4868) 277 - Optimizer 278 - spirv-opt: Fix stacked CompositeExtract constant folds (#4932) 279 - Improve time to build dominators (#4916) 280 - Fix ADCE to mark scope and inlined_at of line instructions as live. (#4910) 281 - Improve algorithm to reorder blocks in a function (#4911) 282 - Add structs to eliminate dead input components (#4894) 283 - spirv-opt: fix copy-propagate-arrays index opti on structs. (#4891) 284 - Fix ADCE to not eliminate top level DebugInfo instructions (#4889) 285 - Fix array copy propagation (#4890) 286 287v2022.3 2022-08-08 288 - General 289 - Add SPV_KHR_fragment_shader_barycentric support (#4805) 290 - Add support for SPV_KHR_subgroup_rotate (#4786) 291 - use exec_tools instead of tools for better RBE compatibility (#4837) 292 - Write binary files to stdout in binary on windows. (#4834) 293 - Allow spirv-opt print-all to show pretty IDs (#4888) 294 - Validator 295 - spirv-val: Add PerVertexKHR (#4807) 296 - spirv-opt : Add FixFuncCallArgumentsPass (#4775) 297 - spirv-val: Add CullMaskKHR support (#4792) 298 - Require ColMajor or RowMajor for matrices (#4878) 299 - spirv-val: Add SPV_KHR_ray_query (#4848) 300 - spirv-val: Add SPV_KHR_ray_tracing instructions (#4871) 301 - Implement SPV_NV_bindless_texture related changes (#4847) 302 - spirv-val: Add OpConvertUToAccelerationStructureKHR (#4838) 303 - spirv-val: Add support for SPV_AMD_shader_early_and_late_fragment_tests (#4812) 304 - Optimizer 305 - Fold multiply and subtraction into FMA with negation (#4808) 306 - Add more folding for composite instructions (#4802) 307 - spirv-opt: add pass for interface variable scalar replacement (#4779) 308 - Don't try to unroll loop with step count 0. (#4769) 309 - spirv-opt: SPV_NV_bindless_texture related changes (#4870) 310 - Linker 311 - linker: Recalculate interface variables (#4784) 312 313v2022.2 2022-04-07 314 - General 315 - Add OpModuleProcessed to debug opcode (#4694) 316 - Optimizer 317 - Complete handling of RayQueryKHR type (#4690) 318 - Have scalar replacement use undef instead of null (#4691) 319 - Optimize Instruction::Instruction (#4705) 320 - Handle propagation of arrays with decorations (#4717) 321 - spirv-opt: Add OpExecutionModeId support (#4719) 322 - Optimize Type::HashValue (#4707) 323 - Optimize DefUseManager allocations (#4709) 324 - Add pass to remove DontInline function control (#4747) 325 - Better handling of 0xFFFFFFFF when folding vector shuffle (#4743) 326 - Reset the id bound on the module in compact ids (#4744) 327 - spirv-opt: (WIP) Eliminate Dead Input Component Pass (#4720) 328 - Support SPV_KHR_uniform_group_instructions (#4734) 329 - Handle shaders without execution model in spread-volatile-semantics (#4766) 330 - Validator 331 - Fix handling of Nontemporal image operand (#4692) 332 - [spirv-val] Allow 0 Component Count for DebugTypeArray for Shader (#4706) 333 - spirv-val: Validate DebugTypeMatrix (#4732) 334 - spirv-val: Label Vulkan VUID 04734 (#4739) 335 - spirv-val: Label VUID 06491 (#4745) 336 - spirv-val: Disallow array of push constants (#4742) 337 - spirv-val: Label Vulkan RuntimeArray VUID (#4749) 338 - spirv-val: Add Vulkan Image VUID 06214 (#4750) 339 - spirv-val: Add Vulkan Dref not allowed 3D dim VUID (#4751) 340 - spirv-val: Label and add test for PSB Aligned (#4756) 341 - spirv-val: Add Vulkan 32-bit bit op Base (#4758) 342 - spirv-val: Add more Vulkan VUID labels (#4764) 343 - Diff 344 - Introduce spirv-diff (#4611) 345 - Stabilize the output of spirv-diff (#4698) 346 - spirv-diff: Handle OpSpecConstant array sizes (#4700) 347 - spirv-diff: Match OpSpecConstantComposite correctly (#4704) 348 - spirv-diff: Use GetSingleWord*Operand (#4768) 349 - spirv-diff: Basic support for OpTypeForwardPointer (#4761) 350 - spirv-diff: Fix OpTypeFunction matching w.r.t operand count (#4771) 351 352v2022.1 2022-01-26 353 - General 354 - Add SPIR-V 1.6 support to wasm build (#4674) 355 - Improvements to disassembly within PassManager (#4677) 356 - Basic support for SPIR-V 1.6 (#4663) 357 - reflect debug (#4662) 358 - Fix endianness of string literals (#4622) 359 - Optimizer 360 - spirv-opt: add pass to Spread Volatile semantics (#4667) 361 - Fix constant propagation and folding of FClamp instructions (#4651) 362 - Manually fold floating point division by zero (#4637) 363 - Allow ADCE to remove dead inputs (#4629) 364 - Linker 365 - Linker improvements (#4679) 366 * test/linker: Code factorisation and small tweaks 367 * linker: Do not fail when going over limits 368 - Validator 369 - val: interface struct with builtins must be Block (#4665) 370 - Fuzzer 371 - Avoid id bound errors during opt fuzzing (#4658) 372 - Avoid uninitialised read when parsing hex float (#4646) 373 374v2021.4 2021-11-11 375 - General 376 - Add a WebAssembly build (#3752) 377 - Make cxx exceptions controllable (#4591) 378 - Validator 379 - Improve decoration validation (#4490) 380 - Optimizer 381 - Add spirv-opt pass to replace descriptor accesses based on variable indices (#4574) 382 - Do not fold snegate feeding sdiv (#4600) 383 - Handle overflowing id in merge return (#4606) 384 - Fuzzer 385 - Add libFuzzer target for spirv-fuzz (#4434) 386 - Linter 387 388v2021.3 2021-08-24 389 - General 390 - Initial support for SPV_KHR_integer_dot_product (#4327) 391 - Add non-semantic vulkan extended instruction set (#4362) 392 - Add common enum for debug info instructions from either opencl or vulkan (#4377) 393 - Validator 394 - Add validation for SPV_EXT_shader_atomic_float16_add (#4325) 395 - Disallow loading a runtime-sized array (#4473) 396 - spirv-val: Validate vulkan debug info similarly to opencl debug info (#4466) 397 - Optimizer 398 - spirv-opt: support SPV_EXT_shader_image_int64 (#4379) 399 - spirv-opt: Add dataflow analysis framework (#4402) 400 - Add control dependence analysis to opt (#4380) 401 - Add spirv-opt convert-to-sampled-image pass (#4340) 402 - spirv-opt: Add handling of vulkan debug info to DebugInfoManager (#4423) 403 - Fuzz 404 - spirv-fuzz: support AtomicLoad (#4330) 405 - spirv-fuzz: Support AtomicStore (#4440) 406 - spirv-fuzz: TransformationWrapVectorSynonym that rewrites scalar operations using vectors (#4376) 407 - spirv-fuzz: Add minimal SPIR-V example to test shaders (#4415) 408 - spirv-fuzz: support building using gn (#4365) 409 - Linter 410 - Add new target for spirv-lint (#4446) 411 - spirv-lint: add basic CLI argument handling (#4478) 412 - Add divergence analysis to linter (#4465) 413 414v2021.2 2021-06-18 415 - General 416 - Support SPV_KHR_subgroup_uniform_control_flow (#4318) 417 - Support Intel extensions for fixed point and hls-float (#4321) 418 - Fix crash when optimizing shaders with DebugPrintf (#4280) 419 420 - Validator 421 - Support Vulkan Storage Class for Execution Model (#4212) 422 423 - Optimizer 424 - Handle SPV_KHR_vulkan_memory_model in dead-code elimination (#4320) 425 - Support folding OpBitcast with numeric constants (#4247) 426 427 - Fuzz 428 - Add tests for MaybeGet* functions in fuzzerutil (#4284) 429 - Fix OutlineFunction in presence of unreachable blocks (#4308) 430 - Fix def-use update in PermutePhiOperands (#4309) 431 - Swap positions of two functions in a module (#4236) 432 433v2021.1 2021-04-19 434 - General 435 - Support SPV_KHR_linkonce_odr, SPV_KHR_expect_assume (#4161) 436 - Fixes for the vscode language server extension (#4150) 437 - Validator 438 - Add validation for SPV_EXT_shader_atomic_float_min_max (#4105) 439 - Add Vulkan Execution Scope checks (#4183) 440 - Vulkan 64-bit OpAtomicStore check (#4163) 441 - Optimizer 442 - Add interpolate legalization pass (#4220) 443 - Fuzz 444 - Various performance optimizations 445 - Do not add too many dead blocks (#4217) 446 - Add WGSL compatibility flag to context (#4193) 447 - Add persistent state to the fuzzer (#4137) 448 449v2020.7 2021-02-16 450 - General 451 - Support pending Intel extensions (#4116) 452 - Remove WebGPU support (#4108) 453 - Validator 454 - Vulkan image gather constant component (#4133) 455 - Add Vulkan PSB64 convert VUID (#4122) 456 - Validate SPV_KHR_workgroup_memory_explicit_layout (#4128) 457 - Validate VK_KHR_zero_initialize_workgroup_memory (#4124) 458 - Add Vulkan image gather offset VUID (#4118) 459 - Label Vulkan atomic semantics VUIDs (#4120) 460 - Label VUID 04662 (#4123) 461 - Label VUID 04683 (#4121) 462 - Add Vulkan EXT builtins (#4115) 463 - Validate Sampled=1 for Vulkan ImageQuerySizeLod, ImageQueryLevels, ImageQueryLod (#4103) 464 - Add Vulkan Memory Scope VUs (#4106) 465 - Add Vulkan Addressing Model check (#4107) 466 - Vulkan atomic storage class (#4079) 467 - Label standalone Vulkan VUID (#4091) 468 - Add Vulkan decroation VUID (#4090) 469 - Add Vulkan FP Mode VUID (#4088) 470 - Fix Vulkan image sampled check (#4085) 471 - Add Vulkan ForwardPointer VUID (#4089) 472 - Add Vulkan ImageTexelPointer format check (#4087) 473 - Add Vulkan Group Operation VUID (#4086) 474 - Add first StandAlone VUID 04633 (#4077) 475 - Add Subgroup VUIDs (#4074) 476 - validate return type of OpImageRead (#4072) 477 - tighter validation of multisampled images (#4059) 478 - validate OpTypeImage Sampled values for environemnts (#4064) 479 - validate StorageImageMultisampled capability (#4062) 480 - Add last TessLevelOuter and TessLevelInner VUID (#4055) 481 - Add last ClipDistance and CullDistance VUID (#4054) 482 - Add last ViewportIndex and Layer VUID (#4053) 483 - Add last Position VUID (#4052) 484 - Allow forward pointer to be used in types generally (#4044) 485 - Optimizer 486 - Mark module as modified if convert-to-half removes decorations (#4127) 487 - Fix binding number calculation in desc sroa (#4095) 488 - Run DCE when SPV_KHR_shader_clock is used (#4049) 489 - Debug Info 490 - Set correct scope and line info for DebugValue (#4125) 491 - Avoid integrity check failures caused by propagating line instructions (#4096) 492 - Linker 493 - Linker usability improvements (#4084) 494 - Instrumentation 495 - Generate differentiated error codes for buffer oob checking (#4097) 496 - Fuzz 497 - Fix OpPhi handling in DuplicateRegionWithSelection (#4065) 498 499v2020.6 2020-12-07 500 - General 501 CMake: Add SPIRV_TOOLS_BUILD_STATIC flag (#3910) 502 - Disassembler 503 Add some context comments to disassembly. (#3847) 504 - Optimizer 505 - Take new (raytracing) termination instructions into account. (#4050) 506 - Do run DCE if SPV_KHR_ray_query is used. (#4047) 507 - Handle 8-bit index in elim dead member (#4043) 508 - Add texel buffer out-of-bounds checking instrumentation (#4038) 509 - Update MeshShadingNV dependencies (and land Ray tracing updates) (#4028) 510 - Fix buffer oob instrumentation for matrix refs (#4025) 511 - Fix SSA re-writing in the presence of variable pointers. (#4010) 512 - Add support to prevent functions from being inlined if they have 513 DontInline flag (#3858) 514 - Add SPV_EXT_shader_image_int64 (#3852) 515 - Support SPV_KHR_fragment_shading_rate (#3943) 516 - Fix use-after-move in val/validate.cpp (#3848) 517 - Debug Info 518 - properly preserve DebugValue indexes operand (#4022) 519 - Add DebugValue for invisible store in single_store_elim (#4002) 520 - Propagate OpLine to all applied instructions in spirv-opt (#3951) 521 - Add DebugValue for DebugDecl invisible to value assignment (#3973) 522 - Add DebugValue for function param regardless of scope (#3923) 523 - Debug info preservation in convert-local-access-chains pass (#3835) 524 - Debug info preservation in redundancy-elimination pass (#3839) 525 - Debug info preservation in if-conversion pass (#3861) 526 - Validator 527 - Add validation support for the ray tracing built-in variables (#4041) 528 - Use less stack space when validating Vulkan builtins (#4019) 529 - Fix SPV_KHR_fragment_shading_rate VUID label (#4014) 530 - Label Layer and ViewportIndex VUIDs (#4013) 531 - Allow the ViewportIndex and Layer built-ins on SPIR-V 1.5 (#3986) 532 - Fix validation of OpPhi instructions (#3919) 533 - Fuzz 534 - Fix facts arising from CompositeConstruct (#4034) 535 - Do not flatten conditionals that create synonyms (#4030) 536 - Add support for reining in rogue fuzzer passes (#3987) 537 - Fix assertion failure in FuzzerPassAddCompositeExtract (#3995) 538 - Fix invalid equation facts (#4009) 539 - Fix bugs in TransformationFlattenConditionalBranch (#4006) 540 - Fix bug related to transformation applicability (#3990) 541 - Add expand vector reduction transformation (#3869) 542 - Add FuzzerPassAddCompositeExtract (#3904) 543 - Fix mismatch with shrinker step limit (#3985) 544 - Fix off-by-one error in replayer (#3982) 545 - Get order right for OpSelect arguments (#3974) 546 - Do not add synonym-creating loops in dead blocks (#3975) 547 - Skip OpTypeSampledImage when propagating up (#3976) 548 - Pass OpUndef in function call if needed (#3978) 549 - Fix off-by-one in TransformationCompositeConstruct (#3979) 550 - Tolerate absent ids in data synonym fact management (#3966) 551 - Fix to id availability (#3971) 552 - Fix operand types (#3962) 553 - Don't flatten conditional if condition is irrelevant (#3944) 554 - Do not produce OpPhis of type OpTypeSampledImage (#3964) 555 - Restrict fuzzer pass to reachable blocks (#3970) 556 - Handle more types when extending OpPhi instructions (#3969) 557 - Skip early terminator wrappers when merging returns (#3968) 558 - Avoid irrelevant constants in synonym-creating loops (#3967) 559 - Skip dead blocks in FuzzerPassAddOpPhiSynonyms (#3965) 560 - Avoid the type manager when looking for struct types (#3963) 561 - Fix to TransformationDuplicateRegionWithSelection (#3941) 562 - Skip OpFunction when replacing irrelevant ids (#3932) 563 - Use component-wise selectors when flattening conditional branches (#3921) 564 - Avoid void struct member when outlining functions (#3936) 565 - Do not allow Block-decorated structs when adding parameters (#3931) 566 - Fix to operand id type (#3937) 567 - Handle dead blocks in TransformationEquationInstruction (#3933) 568 - Do not allow sampled image load when flattening conditionals (#3930) 569 - Take care of OpPhi instructions when inlining (#3939) 570 - Fix to TransformationInlineFunction (#3913) 571 - Wrap early terminators before merging returns (#3925) 572 - Lower probability of adding bit instruction synonyms (#3917) 573 - Fix handling of OpPhi in FlattenConditionalBranch (#3916) 574 - Avoid creating blocks without parents (#3908) 575 - Do not allow creation of constants of block-decorated structs (#3903) 576 - Fixes related to irrelevant ids (#3901) 577 - Fix to transformation that adds a synonym via a loop (#3898) 578 - Fix to duplicate region with selection (#3896) 579 - Do not expose synonym facts for non-existent ids (#3891) 580 - Do not add synonyms involving irrelevant ids (#3890) 581 - Do not replace irrelevant ids that are not in blocks (#3892) 582 - Wrap OpKill and similar in function calls (#3884) 583 - Integrate spirv-reduce with shrinker (#3849) 584 - Report fresh ids in transformations (#3856) 585 - Support OpNot bit instruction case (#3841) 586 - Return IR and transformation context after replay (#3846) 587 588v2020.5 2020-09-22 589 - General 590 - Enable building with BUILD_SHARED_LIBS=1 (#3490) 591 - Avoid using /MP4 for clang on windows. (#3662) 592 - Fix compiler error on macOS with XCode12. (#3836) 593 - Optimizer 594 - Preserve OpenCL.DebugInfo.100 through private-to-local pass (#3571) 595 - Preserve debug info in scalar replacement pass (#3461) 596 - Debug info preservation in loop-unroll pass (#3548) 597 - Preserve debug info in dead-insert-elim pass (#3652) 598 - Improve non-semantic instruction handling in the optimizer (#3693) 599 - Let ADCE pass check DebugScope (#3703) 600 - Add undef for inlined void function (#3720) 601 - Fix SSA-rewrite to remove DebugDeclare for variables without loads (#3719) 602 - Handle DebugScope in compact-ids pass (#3724) 603 - Add buffer oob check to bindless instrumentation (#3800) 604 - Validator 605 - Update OpenCL capabilities validation (#3149) 606 - Validator support for non-semantic clspv reflection (#3618) 607 - OpenCL.DebugInfo.100 DebugTypeArray with variable size (#3549) 608 - Only validation locations for appropriate execution models (#3656) 609 - Validate more OpenCL.DebugInfo.100 instructions (#3684) 610 - Allow DebugTypeTemplate for Type operand (#3702) 611 - spirv-val: Add Vulkan VUID labels to BuiltIn (#3756) 612 - Allow SPV_KHR_8bit_storage extension. (#3780) 613 - Validate SPIRV Version number when parsing binary header (#3834) 614 - Reduce 615 - Support reducing a specific function (#3774) 616 - Fuzz 617 - adds TransformationReplaceCopyObjectWithStoreLoad (#3567) 618 - adds TransformationReplaceCopyMemoryWithLoadStore (#3575) 619 - adds TransformationReplaceLoadStoreWithCopyMemory (#3586) 620 - Implement the OpOuterProduct linear algebra case (#3617) 621 - Pass to replace int operands with ints of opposite signedness (#3612) 622 - TransformationMoveInstructionDown (#3477) 623 - Add TransformationMakeVectorOperationDynamic (#3597) 624 - TransformationReplaceAddSubMulWithCarryingExtended (#3598) 625 - FuzzerPassPropagateInstructionsUp (#3478) 626 - add FuzzerPassAddCompositeInserts (#3606) 627 - Add inline function transformation (#3517) 628 - Transformation to replace the use of an irrelevant id (#3697) 629 - Add SPIRV_FUZZ_PROTOC_COMMAND (#3789) 630 - Add TransformationDuplicateRegionWithSelection (#3773) 631 - Transformation to flatten conditional branch (#3667) 632 - Handle OpPhis in TransformationInlineFunction (#3833) 633 - Create synonym of int constant using a loop (#3790) 634 - Support dead blocks in TransformationAddSynonym (#3832) 635 - Linker 636 637v2020.4 2020-07-22 638 - General 639 - Changed variable names to be more descriptive (#3433) 640 - Add support to GPU-AV instrumentation for Task and Mesh shaders (#3512) 641 - Permit Simple and GLSL450 memory model in WEBGPU_0 (#3463) 642 - Support SPV_KHR_terminate_invocation (#3568) 643 - Optimizer 644 - Preserving debug information in optimizations 645 (#3389,#3420,#3425,#3356,#3459,#3444,#3492,#3451,#3497i,#3498,#3542) 646 - Eliminate branches with condition of OpConstantNull (#3438) 647 - Use structured order to unroll loops. (#3443) 648 - Updated desc_sroa to support flattening structures (#3448) 649 - Support OpCompositeExtract pattern in desc_sroa (#3456) 650 - Fix ADCE pass bug for mulitple entries (#3470) 651 - Sink pointer instructions in merge return (#3569) 652 - Validator 653 - Validate location assignments (#3308) 654 - Fix reachability in the validator (#3541) 655 - Reduce 656 - Fuzz 657 - Add support for OpSpecConstant* (#3373) 658 - Add replace linear algebra instruction transformation (#3402) 659 - Implement vector shuffle fuzzer pass (#3412) 660 - Swap operands in OpBranchConditional (#3423) 661 - Permute OpPhi instruction operands (#3421) 662 - Add FuzzerPassAddCopyMemoryInstructions (#3391) 663 - TransformationInvertComparisonOperator (#3475) 664 - Add variables with workgroup storage class (#3485) 665 - Add image sample unused components transformation (#3439) 666 - TransformationReplaceParameterWithGlobal (#3434) 667 - Support adding dead break from back-edge block (#3519) 668 - Fuzzer pass to interchange zero-like constants (#3524) 669 - Linker 670 671v2020.3 2020-05-27 672 - General 673 - Prevent Effcee from installing things when building spirv-tools with testing enabled (#3256) 674 - Update acorn version (#3294) 675 - If SPIRV-Headers is in our tree, include it as subproject (#3299) 676 - allow cross compiling for Windows Store, UWP, etc. (#3330) 677 - Optimizer 678 - Remove deprecated interfaces from instrument passes (#3361) 679 - Preserve debug info in inline pass (#3349) 680 - Handle more cases in dead member elim (#3289) 681 - Preserve debug info in eliminate-dead-functions (#3251) 682 - Fix Struct CFG analysis for single block loop (#3293) 683 - Add tests for recently added command line option (#3297) 684 - Consider sampled images as read-only storage (#3295) 685 - Allow various validation options to be passed to spirv-opt (#3314) 686 - Add debug information analysis (#3305) 687 - Preserve debug info for wrap-opkill (#3331) 688 - refactor inlining pass (#3328) 689 - Add unrolling to performance passes (#3082) 690 - Validator 691 - Add validation support for ImageGatherBiasLodAMD (#3363) 692 - Validate ShaderCallKHR memory scope (#3332) 693 - Validate Buffer and BufferBlock apply only to struct types (#3259) 694 - Reduce 695 - increase default step limit (#3327) 696 - Remove unused uniforms and similar (#3321) 697 - Fuzz 698 - Add support for StorageBuffer (#3348) 699 - Add validator options (#3254) 700 - Limit adding of new variables to 'basic' types (#3257) 701 - Transformation to add OpConstantNull (#3273) 702 - Handling of more fuzzing opportunities (#3277, #3280, #3281, #3290, #3292) 703 - Respect rules for OpSampledImage (#3287) 704 - Do not outline regions that produce pointer outputs (#3291) 705 - Linker 706 707v2020.2 2020-03-26 708 - General: 709 - Support extended instructions in the vscode language server 710 - Make spvOpcodeString part of the public API (#3174) 711 - Added guide to writing a spirv-fuzz fuzzer pass (#3190) 712 - Add support for KHR_ray_{query,tracing} extensions (#3235) 713 - Optimizer 714 - Debug Printf support (#3215) 715 - Add data structure for DebugScope, DebugDeclare in spirv-opt (#3183) 716 - Fix identification of Vulkan images and buffers (#3253) 717 - Validator 718 - Add support for SPV_AMD_shader_image_load_store_lod (#3186) 719 - Add validation rules for OpenCL.DebugInfo.100 extension (#3133) 720 - Adding WebGPU specific Workgroup scope rule (#3204) 721 - Disallow phis of images, samplers and sampled images (#3246) 722 - Reduce 723 - Fuzz 724 - Fuzzer passes to add local and global variables (#3175) 725 - Add fuzzer passes to add loads/stores (#3176) 726 - Fuzzer pass to add function calls (#3178) 727 - Fuzzer pass that adds access chains (#3182) 728 - Fuzzer pass to add equation instructions (#3202) 729 - Add swap commutable operands transformation (#3205) 730 - Add fuzzer pass to permute function parameters (#3212) 731 - Allow OpPhi operand to be replaced with a composite synonym (#3221) 732 - Linker 733 734v2020.1 2020-02-03 735 - General: 736 - Add support for SPV_KHR_non_semantic_info (#3110) 737 - Support OpenCL.DebugInfo.100 extended instruction set (#3080) 738 - Added support for Vulkan 1.2 739 - Add API function to better handle getting the necessary environment (#3142) 740 - Clarify mapping of target env to SPIR-V version (#3150) 741 - Implement constant folding for many transcendentals (#3166) 742 - Optimizer 743 - Change default version for CreatInstBindlessCheckPass to 2 (#3096, #3119) 744 - Better handling of OpLine on merge blocks (#3130) 745 - Use placeholder switch instead of placeholder loop in MergeReturn pass. (#3151) 746 - Handle TimeAMD in AmdExtensionToKhrPass. (#3168) 747 - Validator 748 - Fix structured exit validation (#3141) 749 - Reduce 750 - Fuzz 751 - Fuzzer pass to merge blocks (#3097) 752 - Transformation to add a new function to a module (#3114) 753 - Add fuzzer pass to perform module donation (#3117) 754 - Fuzzer passes to create and branch to new dead blocks (#3135) 755 - Fuzzer pass to add composite types (#3171) 756 - Linker: 757 - Remove names and decorations of imported symbols (#3081) 758 759v2019.5 2019-12-11 760 - General: 761 - Export SPIRV-Tools targets on installation 762 - SPIRV-Tools support for SPIR-V 1.5 (#2865) 763 - Add WebGPU SPIR-V Assembler in JavaScript. (#2876) 764 - Add Bazel build configuration. (#2891) 765 - Add support for building with emscripten (#2948) 766 - Update SPIR-V binary header test for SPIR-V 1.5 (#2967) 767 - Add fuzzer for spirv-as call path (#2976) 768 - Improved CMake install step. (#2963) 769 - Add fuzzer for spirv-dis call path (#2977) 770 - Ensure timestamp does not vary with timezone. (#2982) 771 - Add a vscode extension for SPIR-V disassembly files (#2987) 772 - Add iOS as a supported platform (#3001) 773 - utils/vscode: Add SPIR-V language server support 774 - Respect CMAKE_INSTALL_LIBDIR in installed CMake files (#3054) 775 - Permit the debug instructions in WebGPU SPIR-V (#3063) 776 - Add support for Fuchsia. (#3062) 777 - Optimizer 778 - Add descriptor array scalar replacement (#2742) 779 - Add pass to wrap OpKill in a function call (#2790) 780 - Fold FMix during constant folding. (#2818) 781 - Add pass to replace AMD shader ballot extension (#2811) 782 - Add pass to make Float32 operation relax precision (#2808) 783 - Add pass to make relax precision operation Float16 (#2808) 784 - Add pass to replace uses of 3 AMD extensions (#2814) 785 - Fold Min, Max, and Clamp instructions. (#2836) 786 - Better handling of OpKill in continues (#2842,#2922,#2933) 787 - Enable OpTypeCooperativeMatrix specialization (#2927) 788 - Support constant-folding UConvert and SConvert (#2960) 789 - Update Offset to ConstOffset bitmask if operand is constant. (#3024) 790 - Improve RegisterSizePasses (#3059) 791 - Folding: perform add and sub on mismatched integer types (#3084) 792 - Graphics robust access: use signed clamp (#3073) 793 Fixes: 794 - Instrument: Fix version 2 output record write for tess eval shaders. (#2782) 795 - Instrument: Add support for Buffer Device Address extension (#2792) 796 - Fix check for changed binary in API call. (#2798) 797 - For WebGPU<->Vulkan optimization, set correct execution environment (#2834) 798 - Handle OpConstantNull in copy-prop-arrays. (#2870) 799 - Use OpReturn* in wrap-opkill (#2886) 800 - Validator 801 - Add generic builtin validation of target (#2843) 802 - Extra resource interface validation (#2864) 803 - Adding valilidation checks for OpEntryPoint duplicate names and execution mode (#2862) 804 - Relaxed bitcast with pointers (#2878) 805 - Validate physical storage buffer restrictions (#2930) 806 - Add SPV_KHR_shader_clock validation (#2879, #3013) 807 - Validate that selections are structured (#2962) 808 - Disallow use of OpCompositeExtract/OpCompositeInsert with no indices (#2980) 809 - Check that derivatives operate on 32-bit values (#2983) 810 - Validate array stride does not cause overlap (#3028) 811 - Validate nested constructs (#3068) 812 Fixes: 813 - Fix validation of constant matrices (#2794) 814 - Update "remquor" validation 815 - Only allow previously declared forward refs in structs (#2920) 816 - Reduce 817 - Remove relaxed precision decorations (#2797) 818 - Reduce/fuzz: improve command line args (#2932) 819 - Improve remove unref instr pass (#2945) 820 Fixes: 821 - Fuzz 822 - Fix add-dead-break and add-dead-continue passes to respect dominance (#2838) 823 - Add fuzzer pass to copy objects (#2853) 824 - Add fuzzer pass to replace ids with synonyms (#2857) 825 - Allow validation during spirv-fuzz replay (#2873) 826 - Employ the "swarm testing" idea in spirv-fuzz (#2890) 827 - reduce/fuzz: improve command line args (#2932) 828 - option to convert shader into a form that renders red (#2934) 829 - Add fuzzer pass to change selection controls (#2944) 830 - add transformation and pass to construct composites (#2941) 831 - Add fuzzer pass to change loop controls (#2949) 832 - Add fuzzer pass to change function controls (#2951) 833 - Add fuzzer pass to add NoContraction decorations (#2950) 834 - Add missing functionality for matrix composites (#2974) 835 - Fuzzer pass to adjust memory access operands (#2968) 836 - Transformation to extract from a composite object (#2991) 837 - Vector shuffle transformation (#3015) 838 - Improve debugging facilities (#3074) 839 - Function outlining fuzzer pass (#3078) 840 841 842v2019.4 2019-08-08 843 - General: 844 - Memory model support for SPIR-V 1.4 845 - Add new spirv-fuzz tool 846 - Add option for base branch in check_code_format.sh 847 - Removed MarkV and Stats code. (#2576) 848 - Instrument: Add version 2 of record formats (#2630) 849 - Linker: Better type comparison for OpTypeArray and OpTypeForwardPointer (#2580) 850 - Optimizer 851 - Bindless Validation: Instrument descriptor-based loads and stores (#2583) 852 - Better folding for OpSpecConstantOp (#2585, #2614) 853 - Add in individual flags for Vulkan <-> WebGPU passes (#2615) 854 - Handle nested breaks from switches. (#2624) 855 - Optimizer: Handle array type with OpSpecConstantOp length (#2652) 856 - Perform merge return with single return in loop. (#2714) 857 - Add --preserve-bindings and --preserve-spec-constants (#2693) 858 - Remove Common Uniform Elimination Pass (#2731) 859 - Allow ray tracing shaders in inst bindle check pass. (#2733) 860 - Add pass to inject code for robust-buffer-access semantics (#2771) 861 - Treat access chain indexes as signed in SROA (#2776) 862 - Handle RelaxedPrecision in SROA (#2788) 863 - Add descriptor array scalar replacement (#2742) 864 Fixes: 865 - Handle decorations better in some optimizations (#2716) 866 - Change the order branches are simplified in dead branch elim (#2728) 867 - Fix bug in merge return (#2734) 868 - SSA rewriter: Don't use trivial phis (#2757) 869 - Record correct dominators in merge return (#2760) 870 - Process OpDecorateId in ADCE (#2761) 871 - Fix check for unreachable blocks in merge-return (#2762) 872 - Handle out-of-bounds scalar replacements. (#2767) 873 - Don't move debug or decorations when folding (#2772) 874 - Protect against out-of-bounds references when folding OpCompositeExtract (#2774) 875 - Validator 876 - Validate loop merge (#2579) 877 - Validate construct exits (#2459) 878 - Validate OpenCL memory and addressing model environment rules (#2589) 879 - Validate OpenCL environment rules for OpTypeImage (#2606) 880 - Allow breaks to switch merge from nested construct (#2604) 881 - Validate OpenCL environment rules for OpImageWrite (#2619) 882 - Allow arrays of out per-primitive builtins for mesh shaders (#2617) 883 - Validate OpenCL rules for ImageRead and OpImageSampleExplicitLod (#2643) 884 - Add validation for SPV_EXT_fragment_shader_interlock (#2650) 885 - Add builtin validation for SPV_NV_shader_sm_builtins (#2656) 886 - Add validation for Subgroup builtins (#2637) 887 - Validate variable initializer type (#2668) 888 - Disallow stores to UBOs (#2651)A 889 - Validate Volatile memory semantics bit (#2672) 890 - Basic validation for Component decorations (#2679) 891 - Validate that in OpenGL env block variables have Binding (#2685) 892 - Validate usage of 8- and 16-bit types with only storage capabilities (#2704) 893 - Add validation for SPV_EXT_demote_to_helper_invocation (#2707) 894 - Extra small storage validation (#2732) 895 - For Vulkan, disallow structures containing opaque types (#2546) 896 - Validate storage class OpenCL environment rules for atomics (#2750) 897 - Update OpControlBarriers rules for WebGPU (#2769) 898 - Update OpMemoryBarriers rules for WebGPU (#2775) 899 - Update WebGPU validation rules of OpAtomic*s (#2777) 900 Fixes: 901 - Disallow merge targeting block with OpLoopMerge (#2610) 902 - Update vloadn and vstoren validation to match the OpenCL Extended 903 Instruction Set Specification (#2599) 904 - Update memory scope rules for WebGPU (#2725) 905 - Allow LOD ops in compute shaders with derivative group execution modes (#2752) 906 - Reduce 907 Fixes: 908 909v2019.3 2019-05-14 910 - General: 911 - Require Python 3 since Python 2 will out of service soon. 912 - Add a continuous test that does memory checks using the address sanitizer. 913 - Fix the build files so the SPIRV_USE_SANITIZER=address build works. 914 - Packaging top of tree build artifacts again. 915 - Added support for SPIR-V 1.4. (#2550) 916 - Optimizer 917 - Remove duplicates from list of interface IDs in OpEntryPoint instruction (#2449) 918 - Bindless Validation: Descriptor Initialization Check (#2419) 919 - Add option to validate after each pass (#2462) 920 - Add legalization pass to fix mismatched pointer (#2430, #2535) 921 - Add error messages when the input contains unknown instructions. (#2487) 922 - Add pass to convert from WebGPU Spir-V to Vulkan Spir-V and back. (#2495) 923 Fixes: 924 - #2412: Dead memeber elimination should not change input and output variables. 925 - #2405: Fix OpDot folding of half float vectors. 926 - #2391: Dead branch elim should not fold away back edges. 927 - #2441: Removing decorations when doing constant propagation. 928 - #2455: Maintain inst to block mapping in merge return. 929 - #2453: Fix merge return in the face of breaks. 930 - #2456: Handle dead infinite loops in DCE. 931 - #2458: Handle variable pointer in some optimizations. 932 - #2452: Fix dead branch elimination to handle unreachable blocks better. 933 - #2528: Fix undefined bit shift in sroa. 934 - #2539: Change implementation of post order CFG traversal. 935 - Validator 936 - Add validation of storage classes for WebGPU (#2446) 937 - Add validation for ExecutionMode in WebGPU (#2443) 938 - Implement WebGPU specific CFG validation (#2386) 939 - Allow NonWritable to target struct members. (#2420) 940 - Allow storage type mismatch for parameter in relaxed addressing mode. 941 - Allow non memory objects as parameter in relaxed addressing mode. 942 - Disallow nested Blocks and buffer blocks (#2410). 943 - Add validation for SPV_NV_cooperative_matrix (#2404) 944 - Add --strip-atomic-counter-memory (#2413) 945 - Check OpSampledImage is only passed into valid instructions (#2467) 946 - Handle function decls in Structured CFG analysis (#2474) 947 - Validate that OpUnreacahble is not statically reachable (#2473) 948 - Add pass to generate needed initializers for WebGPU (#2481) 949 - Allow images without format for OpenCL. (#2470) 950 - Remove unreachable block validation (#2525) 951 - Reduce runtime of array layout checks (#2534) 952 - Add validation specific to OpExecutionModeId (#2536) 953 - Validate sign of int types. (#2549) 954 - VK_KHR_uniform_buffer_standard_layout validation (#2562) 955 Fixes: 956 - #2439: Add missing DepthGreater case to Fragment only check. 957 - #2168: Disallow BufferBlock on StorageBuffer variables for Vulkan. 958 - #2408: Restrict and Aliased decorations cannot be applied to the same id. 959 - #2447: Improve function call parameter check. 960 - Reduce 961 - Add Pass to remove unreferenced blocks. (#2398) 962 - Allows passing options to the validator. (#2401) 963 - Improve reducer algorithm and other changes (#2472) 964 - Add Pass to remove selections (#2485) 965 - Add passes to simplify branches (#2507) 966 Fixes: 967 - #2478: fix loop to selection pass for loops with combined header/continue block 968 969v2019.2 2019-02-20 970 - General: 971 - Support SPV_EXT_physical_storage_buffer 972 - A number of memory leak have been fixed. 973 - Removed use of deprecated Google test macro: 974 - Changed the BUILD.gn to only build tests in Chromium. 975 - Optimizer 976 - Upgrade memory model improvments for modf and frexp. 977 - Add a new pass to move loads closer to their uses: code sinking. 978 - Invalidating the type manager now invalidates the constnat manager. 979 - Expand instrumentation pass for bindless bounds checking to runtime-sized descriptor arrays. 980 - Add a new pass that removes members from structs that are not used: dead member elimination. 981 Fixes: 982 - #2292: Remove undefined behaviour when folding bit shifts. 983 - #2294: Fixes for instrumentation code. 984 - #2293: Fix overflow when folding -INT_MIN. 985 - #2374: Don't merge unreachable blocks when merging blocks. 986 - Validator 987 - Support SPV_KHR_no_integer_wrap and related decorations. 988 - Validate Vulkan rules for OpTypeRuntimeArray. 989 - Validate NonWritable decoration. 990 - Many WebGPU specific validation rules were added. 991 - Validate variable pointer related function call rules. 992 - Better error messages. 993 Fixes: 994 - #2307: Check forwards references in OpTypeArray. 995 - #2315, #2303: Fixed the layout check for relaxed layout. 996 - #1628: Emit an error when an OpSwitch target is not an OpLabel. 997 - Reduce 998 - Added more documentation for spirv-reduce. 999 - Add ability to remove OpPhi instructions. 1000 - Add ability to merge two basic blocks. 1001 - Add ability to remove unused functions and unused basic blocks. 1002 Fixes: 1003 1004v2019.1 2019-01-07 1005 - General: 1006 - Created a new tool called spirv-reduce. 1007 - Add cmake option to turn off SPIRV_TIMER_ENABLED (#2103) 1008 - New optimization pass to update the memory model from GLSL450 to VulkanKHR. 1009 - Recognize OpTypeAccelerationStructureNV as a type instruction and ray tracing storage classes. 1010 - Fix GCC8 build. 1011 - Add --target-env flag to spirv-opt. 1012 - Add --webgpu-mode flag to run optimizations for webgpu. 1013 - The output disassembled line number stead of byte offset in validation errors. (#2091) 1014 - Optimizer 1015 - Added the instrumentation passes for bindless validation. 1016 - Added passes to help preserve OpLine information (#2027) 1017 - Add basic support for EXT_fragment_invocation_density (#2100) 1018 - Fix invalid OpPhi generated by merge-return. (#2172) 1019 - Constant and type manager have been turned into analysies. (#2251) 1020 Fixes: 1021 - #2018: Don't inline functions with a return in a structured CFG contstruct. 1022 - #2047: Fix bug in folding when volatile stores are present. 1023 - #2053: Fix check for when folding floating pointer values is allowed. 1024 - #2130: Don't inline recursive functions. 1025 - #2202: Handle multiple edges between two basic blocks in SSA-rewriter. 1026 - #2205: Don't unswitch a latch condition during loop unswitch. 1027 - #2245: Don't fold branch in loop unswitch. Run dead branch elimination to fold them. 1028 - #2204: Fix eliminate common uniform to place OpPhi instructions correctly. 1029 - #2247: Fix type mismatches caused by scalar replacement. 1030 - #2248: Fix missing OpPhi after merge return. 1031 - #2211: After merge return, fix invalid continue target. 1032 - #2210: Fix loop invariant code motion to not place code between merge instruction and branch. 1033 - #2258: Handle CompositeInsert with no indices in VDCE. 1034 - #2261: Have replace load size handle extact with no index. 1035 - Validator 1036 - Changed the naming convention of outputing ids with names in diagnostic messages. 1037 - Added validation rules for UniformConstant variables in Vulkan. 1038 - #1949: Validate uniform variable type in Vulkan 1039 - Ensure for OpVariable that result type and storage class operand agree (#2052) 1040 - Validator: Support VK_EXT_scalar_block_layout 1041 - Added Vulkan memory model semantics validation 1042 - Added validation checkes spefic to WebGPU environment. 1043 - Add support for VK_EXT_Transform_feedback capabilities (#2088) 1044 - Add validation for OpArrayLength. (#2117) 1045 - Ensure that function parameter's type is not void (#2118) 1046 - Validate pointer variables (#2111) 1047 - Add check for QueueFamilyKHMR memory scope (#2144) 1048 - Validate PushConstants annotation and type (#2140) 1049 - Allow Float16/Int8 for Vulkan 1.0 (#2153) 1050 - Check binding annotations in resource variables (#2151, #2167) 1051 - Validate OpForwardPointer (#2156) 1052 - Validate operation for OpSpecConstantOp (#2260) 1053 Fixes: 1054 - #2049: Allow InstanceId for NV ray tracing 1055 - Reduce 1056 - Initial commit wit a few passes to reduce test cases. 1057 - Validation is run after each reduction step. 1058 Fixes: 1059 1060 1061v2018.6 2018-11-07 1062 - General: 1063 - Added support for the Nvidia Turing and ray tracing extensions. 1064 - Make C++11 the CXX standard in CMakeLists.txt. 1065 - Enabled a parallel build for MSVC. 1066 - Enable pre-compiled headers for MSVC. 1067 - Added a code of conduct. 1068 - EFFCEE and RE2 are now required when build the tests. 1069 - Optimizer 1070 - Unrolling loops marked for unrolling in the legalization passes. 1071 - Improved the compile time of loop unrolling. 1072 - Changee merge-return to create a placeholder loop around the function. 1073 - Small improvement to merge-blocks to allow it to merge more often. 1074 - Enforce an upper bound for the ids, and add option to set it. 1075 - #1966: Report error if there are unreachable block before running merge return 1076 Fixes: 1077 - #1917: Allow 0 (meaning unlimited) as a parameter to --scalar-replacement 1078 - #1915: Improve handling of group decorations. 1079 - #1942: Fix incorrect uses of the constant manager. Avoids type mismatches in generated code. 1080 - #1997: Fix dead branch elimination when there is a loop in folded selection. 1081 - #1991: Fixes legality check in if-conversion. 1082 - #1987: Add nullptr check to array copy propagation. 1083 - #1984: Better handling of OpUnreachable in ADCE. 1084 - #1983: Run merge return on reachable functions only. 1085 - #1956: Handled atomic operations in ADCE. 1086 - #1963: Fold integer divisions by 0 to 0. 1087 - #2019: Handle MemberDecorateStringGOOGLE in ADCE and strip reflect. 1088 - Validator 1089 - Added validation for OpGroupNonUniformBallotBitCount. 1090 - Added validation for the Vulkan memory model. 1091 - Added support for VK_KHR_shader_atddomic_int64. 1092 - Added validation for execution modes. 1093 - Added validation for runtime array layouts. 1094 - Added validation for 8-bit storage. 1095 - Added validation of OpPhi instructions with pointer result type. 1096 - Added checks for the Vulkan memory model. 1097 - Validate MakeTexelAvailableKHR and MakeTexelVisibleKHR 1098 - Allow atomic function pointer for OpenCL. 1099 - FPRounding mode checks were implemented. 1100 - Added validation for the id bound with an option to set the max id bound. 1101 Fixes: 1102 - #1882: Improve the validation of decorations to reduce memory usage. 1103 - #1891: Fix an potential infinite loop in dead-branch-elimination. 1104 - #1405: Validate the storage class of boolean objects. 1105 - #1880: Identify arrays of type void as invalid. 1106 - #487: Validate OpImageTexelPointer. 1107 - #1922: Validate OpPhi instructions are at the start of a block correctly. 1108 - #1923: Validate function scope variable are at the start of the entry block. 1109 1110v2018.5 2018-09-07 1111 - General: 1112 - Support SPV_KHR_vulkan_memory_model 1113 - Update Dim capabilities, to match SPIR-V 1.3 Rev 4 1114 - Automated build bots no run tests for the VS2013 case 1115 - Support Chromium GN build 1116 - Use Kokoro bots: 1117 - Disable Travis-CI bots 1118 - Disable AppVeyor VisualStudio Release builds. Keep VS 2017 Debug build 1119 - Don't check export symbols on OSX (Darwin): some installations don't have 'objdump' 1120 - Reorganize source files and namespaces 1121 - Fixes for ClangTidy, and whitespace (passes 'git cl presumit --all -uf') 1122 - Fix unused param compile warnings/errors when Effcee not present 1123 - Avoid including time headers when timer functionality is disabled 1124 - Avoid too-stringent warnings flags for Clang on Windows 1125 - Internal refactoring 1126 - Add hooks for automated fuzzing 1127 - Add testing of command line executables 1128 - #1688: Use binary mode on stdin; fixes "spirv-dis <foo.spv" on Windows 1129 - Optimizer 1130 - The optimizer validates the module before it begins 1131 - Add API to register passes by string name 1132 - Fold a vector shuffle feeding a vector shuffle 1133 - Add -combine-access-chains transform 1134 - Refactor how IRContext is handled by passes 1135 - Improve bookkeeping for instruction result type and result id 1136 - Fix over-duplication of decorations 1137 - Fix handling of exits from selections in dead-branch elimination, and dead code 1138 elimination. 1139 - Fix handling of certain kinds of flow control in merge-return 1140 Fixes: 1141 - #1721: Fix size bug when folding vector shuffles 1142 - #1722: Fix size infinite loop when folding vector shuffles 1143 - #1724: Fix finding a constant of a specific type 1144 - #1727: Dead branch elim: Reorder blocks if needed to satisfy dominance rule 1145 - #1729: Handle VariablePointers cases in various optimizations 1146 - #1731: Fix vector shuffle with literal id indicating undef value 1147 - #1736: Fix handling of decorations and phis in merge-return 1148 - #1787: Fix handling of decorations related to access chains 1149 - #1865: Avoid leaking memory for SPIR-V constant values 1150 - Validator 1151 - Improve error messages 1152 - Avoid platform-dependent traversal ordering, to ensure consistent messages 1153 - Use libspirv::Instruction where possible 1154 - Add option to skip all block layout checks 1155 - Validate all type IDs 1156 - Validate uses of OpFunction 1157 - Validate uses of OpTypeFunction 1158 - Disallow a struct containing its own type https://crbug.com/874372 1159 - #1685: Vulkan permits non-monotonic offsets for block members 1160 - #1697: Enforce block layout rules even when using relaxed block layout option 1161 - #1719: Fix line number for vector shuffle valiation error 1162 - #1789: Avoid assertion failure when validating some functions 1163 - #1800: Fix validation of OpCopyMemorySized 1164 - #1822: Stop enforcing struct member offset montonicity 1165 - #1831: Disallow void members in structs 1166 1167v2018.4 2018-07-08 1168 - General: 1169 - Support SPV_KHR_8bit_storage 1170 - Add gclient and presubmit configurations 1171 - Enable Kokoro build bots (#1625) 1172 - Group tests into fewer executables, reduces load on CI 1173 - Port test script to Python 3 1174 - Symbol export tests respect SPIRV_SKIP_TESTS 1175 - #1596: Operand lookup succeeds if enabled by a capability 1176 - #1624: Instruction lookup succeeds if enabled by a capability 1177 - Refactoring namespaces: 1178 - #1678: Change libspirv to spvtools 1179 - Code in source/utils moved into spvtools::utils 1180 - Code in source/comp moved into spvtools::comp 1181 - Optimizer: 1182 - Remove insert-extract-elim pass. Use simplification pass instead. 1183 - Preserve instruction-to-block mapping in most passes, to reduce runtime. 1184 - Small vector optimization for operands 1185 - Add pass to move Private variables to Function. Increase opportunity to optimize. 1186 - Fixes: 1187 #1120: Check static uses of entry point interfaces 1188 #1372: Avoid merging some structs, to preserve names for reflection 1189 #1577: Scalar replacement uses only undecorated types. 1190 #1578: Fix handling of forward-pointer types, and types embedding pointers 1191 to themselves. 1192 #1591: Inliner: Callee variable with initializer should have a store at the call site. 1193 #1634: Fix crash: Use type id in vector type lookup 1194 #1649: Fix assert in compact-ids pass 1195 Fix constant folder: ensure it uses the right type 1196 #1659: Folding rules added to IRContext. Avoids leak. 1197 - Validator 1198 - Add work-in-progress WebGPU environment. Disallows OpUndef 1199 - #670, #1581: Improve error messages; disassemble instruction 1200 - #491: Check structured switches 1201 - #937: Check layout rules for Block and BufferBlock in Uniform, StorageBuffer, PushConstant 1202 - #1281: Check invalid branches into structured constructs 1203 - #1522: Disallow array-of-arrays with DescriptorSets 1204 - #1577: Allow duplicate pointer types. 1205 - #1581: Better messages: output ID names along with numbers in more cases. 1206 - #1597: Check Vulkan 1.1 capabilities 1207 - #1618: Check invalid exit from structured case construct 1208 - #1622: Run IdPass before DataRulesPass 1209 - #1632: Reduce test time by artificially lowering limits in limit test 1210 - #1638: Block-decorated structs member order must respect offset order 1211 - #1657: Improve CFG validation diagnostics 1212 - Khronos SPIR-V #337: GLSL.std.450 Refract instruction Eta param can be any float scalar. 1213 - #1606: PushConstant Blocks follow storage-buffer layout rules 1214 - #1664: Check layout of StorageBuffer variables with Block decoration, using storage buffer 1215 rules 1216 - #1666: Layout validation should permit {vec3; float} packing 1217 - #1637, #1668: Layout validation uses RowMajor, ArrayStride, MatrixStride properly 1218 - Linker 1219 - Avoid buffer overrun when creating OpModuleProcessed 1220 1221v2018.3 2018-05-25 1222 - General: 1223 - Support SPV_EXT_descriptor_indexing 1224 - Support SPV_GOOGLE_decorate_string 1225 - Support SPV_GOOGLE_hlsl_functionality1 1226 - Support SPV_NV_shader_subgroup_partitioned 1227 - Use "unified1" grammar from SPIRV-Headers 1228 - Simplify support for new extensions. Assembler, disassembler, and simple validation 1229 support is automatic if new tokens are introduced with appropriate extension 1230 attributes in the "unified1" SPIR-V core grammar. 1231 - Disassembler: Emit more digits on floating point, to reliably reproduce all 1232 significand bits. (Use std::max_digits10 instead of std::digits10) 1233 - Fix compilation for old XCode versions: Explicit construction required for std::set. 1234 - Optimizer: 1235 - Add --strip-reflect 1236 - Add --time-report 1237 - Add --loop-fission 1238 - Add lop fusion. 1239 - Add loop peeling pass and internal utility. 1240 - Improve optimizer runtime. 1241 - Merge-return now works with structured control flow. 1242 - New (faster) SSA rewriter to convert local loads and stores to SSA IDs and phis. 1243 Can replace load/store elimination passes. 1244 - Fix instruction folding case: insertion that feeds and extract, when the extract 1245 remains. 1246 - Fold OpDot. 1247 - Fold OpFNegate. 1248 - Fold multply and divide of same value. 1249 - Fold FClamp feeding a compare. 1250 - Fold OpLoad feeding an extract, to reduce excessive copying. (#1547) 1251 - Fold Fmix feeding an extract. 1252 - Use simplification pass instead of insert-extract elimination. 1253 - Constant fold OpVectorTimesScalar. 1254 - Copy propagate arrays, in simple cases. 1255 - Aggressive dead code elimination: Can remove more instructions, e.g. derivatives. 1256 - Aggressive dead code elimination: Remove Workgroup variables that are written but not read. 1257 - Better handling of OpImageTexelPointer 1258 - Initial utilities for scalar evolution. 1259 - Add Vector dead code elimination. 1260 - Each pass can only run once. 1261 - Allow code hosting in if-conversion. 1262 - Add external interface for adding a PassToken, so external code can make their own 1263 passes. 1264 - Fixes: 1265 #1404: Don't optimize away the compute compute workgroup size constant. 1266 #1407: Remove a bad assertion 1267 #1456: Fix bug in SSA rewriter related to variables updated in loops. 1268 #1487: Fix long runtime in Dead insertion elimination: Don't revist select phi nodes. 1269 #1492: Aggressive dead code elimination can remove OpDecorateStringGOOGLE. 1270 #1527: Fix inlining of functions having OpKill and OpUnreachable. 1271 #1559: Fix assert failure in reduce-load-size pass. 1272 #1556: Aggressive dead code elimination: Fix handling of OpCopyMemory. 1273 - Validator: 1274 - Check Vulkan built-in variables 1275 - Check Vulkan-specific atomic result type rule. 1276 - Relax control barrier check for SPIR-V 1.3. Fixes #1427 1277 - Check OpPhi. 1278 - Check OpMemoryModel. 1279 - Stop checking sizes derived from spec-constants. 1280 - Re-enable checks for OpUConvert. 1281 - Vulkan: Fix check for PrimitiveId: Permit as Input in fragment shader. 1282 - Validate binary version for the given target environment. 1283 - Add tests for OpBranch checks. 1284 - Vulkan 1.1: Check scope for non-uniform subgroup operations. 1285 - Fix checks for SPV_AMD_gpu_shader_int16. 1286 - Fix logical layout check for OpDecorateId. 1287 - Fix checks for ViewportIndex & Layer for Vulkan and SPV_EXT_shader_viewport_index_layer. 1288 - Fixes: 1289 #1470: Vulkan: Don't restrict WorkgroupSize to Input storage class. 1290 #1469: Vulkan: Permit Subgroup memory scope for Vulkan 1.1. 1291 #1472: Per-vertex variable validation fixes. 1292 #1483: Valdiate barrier execution scopes for Vulkan 1.1. 1293 - Fixes: 1294 #898: Linker properly removes FuncParamAttr from imported symbols. 1295 #924, #1174: Fix handling of decoration groups in optimizer, linker. 1296 1297v2018.2 2018-03-07 1298 - General: 1299 - Support SPIR-V 1.3 and Vulkan 1.1. 1300 - Default target environment is now SPIR-V 1.3. For command-line tools, 1301 use the --target-env option to override the default. Examples: 1302 # Generate a SPIR-V 1.0 binary instead of SPIR-V 1.3 1303 spirv-as --target-env spv1.0 a.spvasm -o a.spv 1304 spirv-as --target-env vulkan1.0 a.spvasm -o a.spv 1305 # Validate as Vulkan 1.0 1306 spirv-val --target-env vulkan1.0 a.spv 1307 - Support SPV_GOOGLE_decorate_string and SPV_GOOGLE_hlsl_functionality1 1308 - Fixes: 1309 - Fix Android.mk build. Compilation was failing due to missing definitions of 1310 SpvCapabilityFloat16ImageAMD and other enumerated values. 1311 - Optimizer: Avoid generating duplicate names when merging types. 1312 - #1375: Validator: SPV_AMD_gpu_shaer_half_float implicitly allows declaration 1313 of the 16-bit floating point type. 1314 - #1376: Optimizer: Avoid folding half-precision float. 1315 1316v2018.1 2018-03-02 1317 - General: 1318 - Support Visual Studio 2013 again. (Continue support for VS 2015 and VS 2017.) 1319 - Support building SPIRV-Tools as a shared library. 1320 - Improve the HLSL legalization optimization recipe. #1311 1321 - Optimizer: 1322 - General speedups. 1323 - Remove generic dead code elimination functionality from transforms: 1324 --eliminate-local-single-block 1325 --eliminate-local-single-store 1326 --eliminate-local-multi-store 1327 To recover the previous behaviour, a recipe using those transforms should now 1328 also invoke the --eliminate-dead-code-aggressive transform. 1329 - Improve folding, including coverage for floating point, OpSelect, and arithmetic 1330 with non-trivial constant operands. 1331 - Add loop-invariant code motion pass. 1332 - Add loop-unrolling pass, for honouring unroll hits. 1333 - Add loop-unswitch pass. 1334 - Add instruction simplification pass. 1335 - Aggressive dead code elimination: Understands capability hierarchy when finding 1336 instructions it can eliminate (combinators). (PR #1268) 1337 - CCP can now fold floating point arithmetic. #1311 1338 - Validator: 1339 - Validate barrier instructions. 1340 - Check Vulkan-specific rules for atomics. 1341 - Check Vulkan prohibition of Location or Component decorations on BuiltIn variables. 1342 - Linker: 1343 - Add --verify-ids option 1344 - Add option to allow a resulting module to be partially linked. 1345 - Handle OpModuleProcessed (instructions in SPIR-V layout section 7c) 1346 - Fixes: 1347 - #1265: Optimizer: Fix use-after free bug in if-conversion. (Fix object lifecycle bug 1348 in type manager.) 1349 - #1282: Fix new warnings found by GCC 8.0.1. 1350 - #1285: Optimizer: Fix random failures during inlining. (Dangling references in DefUseManager) 1351 - #1295: Optimizer: Fix incorrect handling of Phi nodes in CCP. 1352 - #1300: Fix CCP: avoid bad CCP transitions and unsettled values. 1353 - #1304: Avoid static-duration variables of class type (with constructors). 1354 - #1323: Fix folding of an insert composite feeding a composite extract. 1355 - #1339: Fix CCP: Handle OpConstantNull boolean values as conditions. 1356 - #1341: DCEInst: Keep atomic instructions (and some others with side effects). 1357 - #1354: Don't fold integer division. 1358 - #1357: Support OpConstantNull in folding. 1359 - #1361: CCP: Fix handling of non-constant module-scope values 1360 1361v2018.0 2018-02-02 1362 - General 1363 - VisualStudio 2013 is no longer supported. VisualStudio 2015 is supported. 1364 - Use "include/unified1" directory from SPIRV-Headers. Requires recent SPIRV-Headers source. 1365 - Disassembler: spirv-dis adds --color option to force color disassembly. 1366 - Optimizer: 1367 - Add pass to eliminate dead insertions. 1368 - Aggressive dead code elimination now removes OpSwitch constructs. 1369 - Block merging occurs in more cases. 1370 - Add driver workaround transform: replace OpUnreachable with harmless branch to merge. 1371 - Improve instruction folding framework. 1372 - Add loop analysis. 1373 - Add scalar replacement of aggregates to size-optimization recipe. 1374 - Add pass to replace instructions invalid for a shader stage, with a harmless value. 1375 This changes the semantics of the program! Not for general use! 1376 - Rearragne and add passes to performance-optimization recipe, to produce better results. 1377 - Validator: 1378 - Validate OpenCL extended instructions. 1379 - Shaders can't perform atomics on floats. 1380 - Validate memory semantics values in atomics. 1381 - Validate instruction-adjacency constraints, e.g. OpPhi predecessors, merge instructions 1382 immediately precede branches. 1383 - Fixes: 1384 - PR 1198: Optimizer: Fix CCP in presence of matrix constants. 1385 - #1199: Optimizer: Fix CCP: don't propagate spec constants. 1386 - #1203: Optimizer: Fix common uniform elim bug introduced by refactoring. 1387 - #1210: Optimizer: Aggressive dead code elimination: Fix 'break' identification. 1388 - #1212: Optimizer: Aggressive dead code elimination: Was skipping too many instructions. 1389 - #1214: Optimizer: Aggressive dead code elimination: Fix infinite loop. 1390 - #1228: Optimizer: Fix CCP: Handling of varying Phi nodes; was resulting in infinite loop. 1391 - #1245: Optimizer: Dead branch elimination: Avoid a null pointer dereference. 1392 - #1250: Optimizer: Dead branch elimination: Avoid spuriously reporting a change. 1393 1394v2017.3 2018-01-12 1395 - General: 1396 - Support DebugInfo extended instruction set, targeted at OpenCL environments. 1397 See the SPIR-V Registry. 1398 - Generate a SPIRV-Tools.pc file for pkg-config. 1399 - Optimizer: 1400 - Progress for legalization of code generated from HLSL (issue #1118): 1401 - Add --legalize-hlsl option to run transforms used to transform intermediate 1402 code generated by HLSL to SPIR-V for Vulkan compilers. Those compilers 1403 normally run these transforms automatically. This option is used for developing 1404 those transforms. 1405 - Add Private-to-Function variable conversion for modules with logical 1406 addressing. 1407 - Add --ccp: SSA Conditional Constant Propagation (CCP) 1408 - Add --print-all to show disassembly for each optimization pass. 1409 - Internal: Add loop descriptors and post-order tree iterator. 1410 - Generalized dead branch elimination 1411 - Aggressive dead code elimination (ADCE) now removes dead functions and 1412 module-scope variables. 1413 - Vector extract/insert elimination now optimizes through some cases of 1414 VectorShuffle, and GLSL.std.450 Mix extended instruction. 1415 - Validator: 1416 - Add validation for GLSL.std.450 extended instruction set. 1417 - Check out of bounds composite accesses, where that's statically computable. 1418 Fixes #1112. 1419 - Check upper bits of literal numbers that aren't a multiple of 32-bits wide. 1420 - More validation of primitive instructions 1421 - Add optional "relaxed" checking logical addressing mode to permit some 1422 cases of pointer-to-pointer. Contributes to HLSL legalization (issue #1118). 1423 - Fixes: 1424 #1100: Validator: Image operand Sample can be used with OpImageSparseFetch, 1425 OpImageSparseRead. 1426 #1108: Remove duplicates transform was incorrectly removing non-duplicate 1427 decorations. 1428 #1111: Optimizer's type manager could reference deleted memory. 1429 #1112: Fix decoration equality check, e.g. it is now symmetric. 1430 #1129: Validator now disallows Dim=SupbassData for OpImageSparseRead. 1431 #1143: Fix CCP: Was generating incorrect code for loops. 1432 #1153: Fix CCP crash. 1433 #1154: Optimizer's internal instruction-to-block mappings were sometimes 1434 inconsistent. 1435 #1159: Fix CCP infinite loop. 1436 #1168: Fix dead branch elimination intermittently generating incorrect code. 1437 Fixes https://github.com/KhronosGroup/glslang/issues/1205 1438 #1186: Fix validation of PackDouble2x32 and UnpackDouble2x32 1439 1440v2017.2 2017-12-15 1441 - General: 1442 - Support OpenCL 1.2, 2.0 target environments, including embedded profiles 1443 - Add CONTRIBUTING.md 1444 - Fix exit status code for spirv-link 1445 - Disassember: Enable emitting ANSI colour codes to a string 1446 - Library avoids polluting global namespace. The libraries can export C and C++ 1447 symbols starting with "spv", or in a C++ namespace. Add a test for this. 1448 - Linux release builds include debug information, for easier profiling 1449 - Build bots no longer test VisualStudio 2013 1450 - Testing dependency RE2 requires VisualStudio 2015 or later 1451 - Build bots check code formatting 1452 - Optimizer: 1453 - Add --skip-validation to spirv-opt 1454 - Add dominance tree analysis 1455 - Add generic value propagation engine 1456 - Add global redundancy elimination within a function 1457 - Add scalar replacement of function-scope variables of composite type 1458 - Aggressive dead code elimination: Remove empty loops 1459 - Killing an instruction notifies the IRContext 1460 - IRContext::KillInst deletes the instruction 1461 - Move CFG analysis to IRContext 1462 - Add constant manager 1463 - Fix: Don't consider derivative instructions as combinators. 1464 - Fix: Don't delete an instruction twice in local dead-code-elimination 1465 - Fix: Don't consider derivative instructions as combinators. 1466 - Validator: 1467 - Finish checking of image instructions (Section 3.32.10) 1468 - Check sparse image instructions 1469 - Check OpTypeImage, OpTypeSampleImage 1470 - Check composite instructions (Section 3.32.12) 1471 - Check atomic instructions (Section 3.32.18) 1472 - Check OpEmitStreamVertex, OpEndStreamPrimitive instructions 1473 - Re-enable validation of OpCopyObject 1474 - OpKill, image ImplicitLod and QueryLod instructions can only be used in Fragment 1475 shaders. 1476 - Fixes for image instruction validation: 1477 - Lod image operand only usable with ExplicitLod and OpImageFetch 1478 - ExplicitLod Lod image operand must be float scalar 1479 - OpImageFectch Lod image operand must be int scalar 1480 - OpImageGather component operand must be 32-bits (integer scalar) 1481 - OpImageQuerySizeLod Lod must be integer scalar 1482 - Fixes: 1483 #622: Remove names and decorations when inlining 1484 #989: Aggressive dead code elim: Don't optimize away live breaks from a loop 1485 #991: Fix validation of SPV_AMD_shader_ballot 1486 #1004: Use after free of an instruction, in remove-duplicates transform 1487 #1007: OpImageRead not required to return 4-component vector 1488 #1009: OpImageRead can return scalar int/float types 1489 #1011: OpImageWrite should allow scalar int/float texel types 1490 #1012: Fix validat Dref type check 1491 #1017: Load-store elimination considers variable initializations 1492 #1034: Fix Windows debug build: operator< should be a weak ordering 1493 #1083: Inlining: Set parent (function) for each inlined basic block. 1494 #1075: Aggressive dead code elimination: Was leaving dangling references to 1495 removed blocks. 1496 1497v2017.1 2017-11-23 1498 - Update README with details on the public_spirv_tools_dev@khronos.org mailing list. 1499 - General: 1500 - Automatically deploy built artifacts to GitHub Releases 1501 - Add a Linker (module combiner). Under development. 1502 - Add Android.mk for Android NDK builds. 1503 - Add the 'effcee' library as an optional dependency for use in tests. 1504 Eventually it will be a required dependency, once downstream projects have 1505 a chance to adjust. Requires 're2' library. 1506 - Avoid static-duration variables of class type (with constructors). 1507 - Hack around bugs in gcc-4.8.1 template handling 1508 - Faster opcode lookup 1509 - Validator: 1510 - Recognize extensions listed on SPIR-V registry, 1511 through #25 SPV_AMD_shader_fragment_mask 1512 - Validator issues an info message when it sees an unrecognized extension. 1513 - Type check basic arithmetic operations 1514 - Type check carry/extended arithmetic operations 1515 - Type check vector arithmetic operations 1516 - Type check Relational and Logical instructions 1517 - Type check Bit instructions 1518 - Check type uniqueness rules 1519 - Check conversion instructions 1520 - Check image instructions 1521 - Check derivative instructions 1522 - Check OpVectorShuffle 1523 - Check OpBranchConditional 1524 - OpModuleProcessed is only allowed after debug names section and before annotations 1525 section. 1526 - Checks the right kind of return is called for each function (void or non-void). 1527 - Add option to relax type check when storing structs (--relax-store-struct) 1528 - Optimizer: 1529 - Refactoring internal representation of the module, including: 1530 - IRContext: owns a module and manages analyses 1531 - Instructions are owned by intrusive lists, and have unique IDs 1532 - BasicBlock owns its instruction list. 1533 - DefUseManager: change representation of uses, for faster processing 1534 on large modules. 1535 - Add high level recipes: -O, -Os, and -Oconfig 1536 Recipes for -O and -Os are under development. 1537 - Add eliminate-dead-function transform 1538 - Add strength reduction transform: For now, convert multiply by power of 2 1539 to a bit shift. 1540 - Add CFG cleanup transform 1541 - Add removal of dead module-scope variables 1542 - Add merge-return transform for modules without structured control flow 1543 - Add redundancy elimination within a basic block (local value numbering) 1544 - Extract-insert elimination: 1545 - Recognize the case where the first instruction in the sequence is an 1546 OpCompositeConstruct or OpConstantComposite 1547 - Handle some cases of nested structs 1548 - Dead branch elimination now can eliminate entire selection constructs 1549 when all arms are dead. 1550 - Compressing codec: 1551 - Updated algorithm to 1.01, 1.02, 1.03 1552 - Not built by default. Use -DSPIRV_BUILD_COMPRESSION=ON to build. 1553 - Codec can be parameterized by a customized model. 1554 - Fixes: 1555 #728: Fix decoration of inlined functions 1556 #798: spirv-as should fail when given unrecognized long option 1557 #800: Inliner: Fix inlining function into header of multi-block loop 1558 #824: Eliminate-local-multi-store: Fix a crash 1559 #826: Elimiante-local-multi-store: Fix a crash 1560 #827: Fix crash when compact-ids transform runs before another transform. 1561 #834: Add Cmake option to build the compressing codec. Off by default. 1562 #911: Fix classification of Line and NoLine instructions 1563 1564v2017.0 2017-09-01 1565 - Update README to describe that assembler, disassembler, and binary parser support 1566 are based on grammar files from the SPIRV-Headers repository. 1567 1568v2016.7 2017-09-01 1569 - Add SPIR-V 1.2 1570 - OpenCL 2.2 support is now based on SPIR-V 1.2 1571 - Support AMD extensions in assembler, disassembler: 1572 SPV_AMD_gcn_shader 1573 SPV_AMD_shader_ballot 1574 SPV_AMD_shader_explicit_vertex_parameter 1575 SPV_AMD_shader_trinary_minmax 1576 SPV_AMD_gpu_shader_half_float 1577 SPV_AMD_texture_gather_bias_lod 1578 SPV_AMD_gpu_shader_int16 1579 - Optimizer: Add support for: 1580 - Inline all function calls in entry points. 1581 - Flatten decoration groups. Fixes #602 1582 - Id compaction (minimize Id bound). Fixes #624 1583 - Eliminate redundant composite insert followed by extract 1584 - Simplify access chains to local variables 1585 - Eliminate local variables with a single store, if possible 1586 - Eliminate local variables with a several stores, if possible 1587 - Eliminate loads and stores in same block to local variables 1588 - Eliminate redundant insert/extract to composite values 1589 - Aggressive dead instruction elimination 1590 - Eliminate dead branches 1591 - Merge blocks when the second can only be preceded by the first 1592 - Eliminate ommon uniform loads 1593 - Assembler: Add option to preserve numeric ids. Fixes #625 1594 - Add build target spirv-tools-vimsyntax to generate spvasm.vim, a SPIR-V 1595 assembly syntax file for Vim. 1596 - Version string: Allow overriding of wall clock timestamp with contents 1597 of environment variable SOURCE_DATE_EPOCH. 1598 - Validator implements relaxed rules for SPV_KHR_16bit_storage. 1599 - CMake installation rules use GNUInstallDirs. For example, libraries 1600 will be installed into a lib64 directory if that's the norm for the 1601 current system. 1602 - Fixes: 1603 #500: Parameterize validator limit checks 1604 #508: Support compilation under CYGWIN 1605 #517: Fix validation when continue (or case) contstruct is also the head of a 1606 nested control construct. 1607 #551: If a merge block is reachable, it must be *strictly* dominated by its 1608 header. 1609 #548: Validator: Error when the reserved OpImageSparseSampleProj* opcodes 1610 are used. 1611 #611: spvtools::Optimizer was failing to save the module to the output 1612 binary vector when all passes succeded without changes. 1613 #629: The inline-entry-points-all optimization could generate invalidly 1614 structured code when the inlined function had early returns. 1615 #697: Optimizer's Instruction::ForEachInId method was skipping semantics-id 1616 and scope-id. 1617 #755: Inliner: Fix inlining of callee with single Return appearing before 1618 the end of the function. 1619 #776: Fix dead branch elimination in presence of complex but dead control 1620 flow. 1621 #781: SPV_KHR_variable_pointers allows duplicate pointer types 1622 #782: Inliner: Fix remapping of non-label forward references in callee 1623 #787: Inliner: Fix remapping of inlined entry block when called from 1624 single block loop. 1625 #790: Inliner: Fix remapping of inlined entry block when callee has 1626 multiple returns. 1627 1628v2016.6 2016-12-13 1629 - Published the C++ interface for assembling, disassembling, validation, and 1630 optimization. 1631 - Support SPV_KHR_shader_draw_parameters in assembler, disassembler, parser. 1632 - Validator: 1633 - Add validator API accepting raw binary words 1634 - Increased coverage: 1635 - Checks "Data rules" in Universal Validation Rules, section 2.16.1 1636 - WIP: Universal Limits. 1637 - The minimum mandated upper bounds are checked. 1638 - TODO: Parameterize the validator to allow larger limits accepted by 1639 a more than minimally capable implementation. 1640 - OpSampledImage checks 1641 - OpConstantComposite checks 1642 - Id bound check 1643 - Disasssembler: 1644 - Generates friendly GLSL-based names for more builtin variables 1645 - Generates friendly names for numeric OpConstant values 1646 - Vendor tool info extracted from SPIR-V XML registry file. 1647 - Fixes issues: 1648 #429: Validator: Allow OpTypeForwardPointer and OpTypeStruct to reference 1649 undefined IDs 1650 #482: Validator: OpVariable initializer can be an ID of a module-scope variable 1651 1652v2016.5 2016-09-16 1653 - Support SPV_KHR_shader_ballot in assembler, disassembler, parser. 1654 - Disassembler: Generate friendly names for built-in variables. 1655 - Partial fixes: 1656 #359: Add Emacs helper for automatically diassembling/assembling a SPIR-V 1657 binary on file load/save. 1658 - Fixes: 1659 #414: Validator: Allow OpUndef for composite constants 1660 #415: Validator: Phi can use its own value in some cases. 1661 1662v2016.4 2016-09-01 1663 - Relicensed under Apache 2.0 1664 - Add optimization passes (in API and spirv-opt command) 1665 - Fold spec constants defined with OpSpecConstantOp and 1666 OpSpecConstantComposite to normal constants with fixed value(s). 1667 - Fixes issues: 1668 #318: Relicensed under Apache 2.0 1669 1670v2016.3 2016-08-24 1671 - Add target environment enums for OpenCL 2.1, OpenCL 2.2, 1672 OpenGL 4.0, OpenGL 4.1, OpenGL 4.2, OpenGL 4.3, OpenGL 4.5. 1673 - Add spirv-cfg, an experimental tool to dump the control flow graph 1674 as a GraphiViz "dot" graph 1675 - Add optimization pass: Eliminate dead constants. 1676 - Add spirv-lesspipe.sh filter utility 1677 - Fixes issues: 1678 #288: Check def-use dominance rules for OpPhi (variable,parent) operands 1679 #339: Allow OpUndef in types-constants-global-vars section, as required 1680 by SPIR-V 1.0 Rev7, 1.1 Rev 3. 1681 #340: Avoid race on mkdir during build 1682 #365: Relax PointSize, ClipDistance, CullDistance capability check in all 1683 environments not just Vulkan 1.0. 1684 1685v2016.2 2016-08-05 1686 - Validator is incomplete 1687 - Checks ID use block is dominated by definition block 1688 - Add optimization passes (in API and spirv-opt command) 1689 - Strip debug info instructions 1690 - Freeze spec constant to their default values 1691 - Allow INotEqual as operation for OpSpecConstantOp 1692 - Fixes bugs: 1693 #270: validator: crash when continue construct is unreachable 1694 #279: validator: infinite loop when analyzing some degenerate control 1695 flow graphs 1696 #286: validator: don't incorrectly generate def-use error for 1697 (variable,parent) parameters to OpPhi 1698 #290: disassembler: never generate bare % for an identifier 1699 #295: validator: def-use dominance check should ignore unreachable uses 1700 #276: validator: allow unreachable continue constructs 1701 #297: validator: allow an unreachable block to branch to a reachable 1702 merge block 1703 1704v2016.1 2016-07-19 1705 - Fix https://github.com/KhronosGroup/SPIRV-Tools/issues/261 1706 Turn off ClipDistance and CullDistance capability checks for Vulkan. 1707 - The disassembler can emit friendly names based on debug info (OpName 1708 instructions), and will infer somewhat friendly names for most types. 1709 This is turned on by default for the spirv-dis command line tool. 1710 - Updated to support SPIR-V 1.1 rev 2 1711 - Input StorageClass, Sampled1D capability, and SampledBuffer capability 1712 do not require Shader capability anymore. 1713 1714v2016.0 2016-07-04 1715 1716 - Adds v<year>.<index> versioning, with "-dev" indicating 1717 work in progress. The intent is to more easly report 1718 and summarize functionality when SPIRV-Tools is incorporated 1719 in downstream projects. 1720 1721 - Summary of functionality (See the README.md for more): 1722 - Supports SPIR-V 1.1 Rev 1 1723 - Supports SPIR-V 1.0 Rev 5 1724 - Supports GLSL std450 extended instructions 1.0 Rev 3 1725 - Supports OpenCL extended instructions 1.0 Rev 2 1726 - Assembler, disassembler are complete 1727 - Supports floating point widths of 16, 32, 64 bits 1728 - Supports integer widths up to 64 bits 1729 - Validator is incomplete 1730 - Checks capability requirements in most cases 1731 - Checks module layout constraints 1732 - Checks ID use-definition ordering constraints, 1733 ignoring control flow 1734 - Checks some control flow graph rules 1735 - Optimizer is introduced, with few available transforms. 1736 - Supported on Linux, OSX, Android, Windows 1737 1738 - Fixes bugs: 1739 - #143: OpenCL pow and pown arguments 1740