1Revision history for SPIRV-Tools 2 3v2020.2-dev 2020-02-03 4 - Start v2020.2-dev 5 6v2020.1 2020-02-03 7 - General: 8 - Add support for SPV_KHR_non_semantic_info (#3110) 9 - Support OpenCL.DebugInfo.100 extended instruction set (#3080) 10 - Added support for Vulkan 1.2 11 - Add API function to better handle getting the necessary environment (#3142) 12 - Clarify mapping of target env to SPIR-V version (#3150) 13 - Implement constant folding for many transcendentals (#3166) 14 - Optimizer 15 - Change default version for CreatInstBindlessCheckPass to 2 (#3096, #3119) 16 - Better handling of OpLine on merge blocks (#3130) 17 - Use dummy switch instead of dummy loop in MergeReturn pass. (#3151) 18 - Handle TimeAMD in AmdExtensionToKhrPass. (#3168) 19 - Validator 20 - Fix structured exit validation (#3141) 21 - Reduce 22 - Fuzz 23 - Fuzzer pass to merge blocks (#3097) 24 - Transformation to add a new function to a module (#3114) 25 - Add fuzzer pass to perform module donation (#3117) 26 - Fuzzer passes to create and branch to new dead blocks (#3135) 27 - Fuzzer pass to add composite types (#3171) 28 - Linker: 29 - Remove names and decorations of imported symbols (#3081) 30 31v2019.5 2019-12-11 32 - General: 33 - Export SPIRV-Tools targets on installation 34 - SPIRV-Tools support for SPIR-V 1.5 (#2865) 35 - Add WebGPU SPIR-V Assembler in JavaScript. (#2876) 36 - Add Bazel build configuration. (#2891) 37 - Add support for building with emscripten (#2948) 38 - Update SPIR-V binary header test for SPIR-V 1.5 (#2967) 39 - Add fuzzer for spirv-as call path (#2976) 40 - Improved CMake install step. (#2963) 41 - Add fuzzer for spirv-dis call path (#2977) 42 - Ensure timestamp does not vary with timezone. (#2982) 43 - Add a vscode extension for SPIR-V disassembly files (#2987) 44 - Add iOS as a supported platform (#3001) 45 - utils/vscode: Add SPIR-V language server support 46 - Respect CMAKE_INSTALL_LIBDIR in installed CMake files (#3054) 47 - Permit the debug instructions in WebGPU SPIR-V (#3063) 48 - Add support for Fuchsia. (#3062) 49 - Optimizer 50 - Add descriptor array scalar replacement (#2742) 51 - Add pass to wrap OpKill in a function call (#2790) 52 - Fold FMix during constant folding. (#2818) 53 - Add pass to replace AMD shader ballot extension (#2811) 54 - Add pass to make Float32 operation relax precision (#2808) 55 - Add pass to make relax precision operation Float16 (#2808) 56 - Add pass to replace uses of 3 AMD extensions (#2814) 57 - Fold Min, Max, and Clamp instructions. (#2836) 58 - Better handling of OpKill in continues (#2842,#2922,#2933) 59 - Enable OpTypeCooperativeMatrix specialization (#2927) 60 - Support constant-folding UConvert and SConvert (#2960) 61 - Update Offset to ConstOffset bitmask if operand is constant. (#3024) 62 - Improve RegisterSizePasses (#3059) 63 - Folding: perform add and sub on mismatched integer types (#3084) 64 - Graphics robust access: use signed clamp (#3073) 65 Fixes: 66 - Instrument: Fix version 2 output record write for tess eval shaders. (#2782) 67 - Instrument: Add support for Buffer Device Address extension (#2792) 68 - Fix check for changed binary in API call. (#2798) 69 - For WebGPU<->Vulkan optimization, set correct execution environment (#2834) 70 - Handle OpConstantNull in copy-prop-arrays. (#2870) 71 - Use OpReturn* in wrap-opkill (#2886) 72 - Validator 73 - Add generic builtin validation of target (#2843) 74 - Extra resource interface validation (#2864) 75 - Adding valilidation checks for OpEntryPoint duplicate names and execution mode (#2862) 76 - Relaxed bitcast with pointers (#2878) 77 - Validate physical storage buffer restrictions (#2930) 78 - Add SPV_KHR_shader_clock validation (#2879, #3013) 79 - Validate that selections are structured (#2962) 80 - Disallow use of OpCompositeExtract/OpCompositeInsert with no indices (#2980) 81 - Check that derivatives operate on 32-bit values (#2983) 82 - Validate array stride does not cause overlap (#3028) 83 - Validate nested constructs (#3068) 84 Fixes: 85 - Fix validation of constant matrices (#2794) 86 - Update "remquor" validation 87 - Only allow previously declared forward refs in structs (#2920) 88 - Reduce 89 - Remove relaxed precision decorations (#2797) 90 - Reduce/fuzz: improve command line args (#2932) 91 - Improve remove unref instr pass (#2945) 92 Fixes: 93 - Fuzz 94 - Fix add-dead-break and add-dead-continue passes to respect dominance (#2838) 95 - Add fuzzer pass to copy objects (#2853) 96 - Add fuzzer pass to replace ids with synonyms (#2857) 97 - Allow validation during spirv-fuzz replay (#2873) 98 - Employ the "swarm testing" idea in spirv-fuzz (#2890) 99 - reduce/fuzz: improve command line args (#2932) 100 - option to convert shader into a form that renders red (#2934) 101 - Add fuzzer pass to change selection controls (#2944) 102 - add transformation and pass to construct composites (#2941) 103 - Add fuzzer pass to change loop controls (#2949) 104 - Add fuzzer pass to change function controls (#2951) 105 - Add fuzzer pass to add NoContraction decorations (#2950) 106 - Add missing functionality for matrix composites (#2974) 107 - Fuzzer pass to adjust memory access operands (#2968) 108 - Transformation to extract from a composite object (#2991) 109 - Vector shuffle transformation (#3015) 110 - Improve debugging facilities (#3074) 111 - Function outlining fuzzer pass (#3078) 112 113 114v2019.4 2019-08-08 115 - General: 116 - Memory model support for SPIR-V 1.4 117 - Add new spirv-fuzz tool 118 - Add option for base branch in check_code_format.sh 119 - Removed MarkV and Stats code. (#2576) 120 - Instrument: Add version 2 of record formats (#2630) 121 - Linker: Better type comparison for OpTypeArray and OpTypeForwardPointer (#2580) 122 - Optimizer 123 - Bindless Validation: Instrument descriptor-based loads and stores (#2583) 124 - Better folding for OpSpecConstantOp (#2585, #2614) 125 - Add in individual flags for Vulkan <-> WebGPU passes (#2615) 126 - Handle nested breaks from switches. (#2624) 127 - Optimizer: Handle array type with OpSpecConstantOp length (#2652) 128 - Perform merge return with single return in loop. (#2714) 129 - Add --preserve-bindings and --preserve-spec-constants (#2693) 130 - Remove Common Uniform Elimination Pass (#2731) 131 - Allow ray tracing shaders in inst bindle check pass. (#2733) 132 - Add pass to inject code for robust-buffer-access semantics (#2771) 133 - Treat access chain indexes as signed in SROA (#2776) 134 - Handle RelaxedPrecision in SROA (#2788) 135 - Add descriptor array scalar replacement (#2742) 136 Fixes: 137 - Handle decorations better in some optimizations (#2716) 138 - Change the order branches are simplified in dead branch elim (#2728) 139 - Fix bug in merge return (#2734) 140 - SSA rewriter: Don't use trivial phis (#2757) 141 - Record correct dominators in merge return (#2760) 142 - Process OpDecorateId in ADCE (#2761) 143 - Fix check for unreachable blocks in merge-return (#2762) 144 - Handle out-of-bounds scalar replacements. (#2767) 145 - Don't move debug or decorations when folding (#2772) 146 - Protect against out-of-bounds references when folding OpCompositeExtract (#2774) 147 - Validator 148 - Validate loop merge (#2579) 149 - Validate construct exits (#2459) 150 - Validate OpenCL memory and addressing model environment rules (#2589) 151 - Validate OpenCL environment rules for OpTypeImage (#2606) 152 - Allow breaks to switch merge from nested construct (#2604) 153 - Validate OpenCL environment rules for OpImageWrite (#2619) 154 - Allow arrays of out per-primitive builtins for mesh shaders (#2617) 155 - Validate OpenCL rules for ImageRead and OpImageSampleExplicitLod (#2643) 156 - Add validation for SPV_EXT_fragment_shader_interlock (#2650) 157 - Add builtin validation for SPV_NV_shader_sm_builtins (#2656) 158 - Add validation for Subgroup builtins (#2637) 159 - Validate variable initializer type (#2668) 160 - Disallow stores to UBOs (#2651)A 161 - Validate Volatile memory semantics bit (#2672) 162 - Basic validation for Component decorations (#2679) 163 - Validate that in OpenGL env block variables have Binding (#2685) 164 - Validate usage of 8- and 16-bit types with only storage capabilities (#2704) 165 - Add validation for SPV_EXT_demote_to_helper_invocation (#2707) 166 - Extra small storage validation (#2732) 167 - For Vulkan, disallow structures containing opaque types (#2546) 168 - Validate storage class OpenCL environment rules for atomics (#2750) 169 - Update OpControlBarriers rules for WebGPU (#2769) 170 - Update OpMemoryBarriers rules for WebGPU (#2775) 171 - Update WebGPU validation rules of OpAtomic*s (#2777) 172 Fixes: 173 - Disallow merge targeting block with OpLoopMerge (#2610) 174 - Update vloadn and vstoren validation to match the OpenCL Extended 175 Instruction Set Specification (#2599) 176 - Update memory scope rules for WebGPU (#2725) 177 - Allow LOD ops in compute shaders with derivative group execution modes (#2752) 178 - Reduce 179 Fixes: 180 181v2019.3 2019-05-14 182 - General: 183 - Require Python 3 since Python 2 will out of service soon. 184 - Add a continuous test that does memory checks using the address sanitizer. 185 - Fix the build files so the SPIRV_USE_SANITIZER=address build works. 186 - Packaging top of tree build artifacts again. 187 - Added support for SPIR-V 1.4. (#2550) 188 - Optimizer 189 - Remove duplicates from list of interface IDs in OpEntryPoint instruction (#2449) 190 - Bindless Validation: Descriptor Initialization Check (#2419) 191 - Add option to validate after each pass (#2462) 192 - Add legalization pass to fix mismatched pointer (#2430, #2535) 193 - Add error messages when the input contains unknown instructions. (#2487) 194 - Add pass to convert from WebGPU Spir-V to Vulkan Spir-V and back. (#2495) 195 Fixes: 196 - #2412: Dead memeber elimination should not change input and output variables. 197 - #2405: Fix OpDot folding of half float vectors. 198 - #2391: Dead branch elim should not fold away back edges. 199 - #2441: Removing decorations when doing constant propagation. 200 - #2455: Maintain inst to block mapping in merge return. 201 - #2453: Fix merge return in the face of breaks. 202 - #2456: Handle dead infinite loops in DCE. 203 - #2458: Handle variable pointer in some optimizations. 204 - #2452: Fix dead branch elimination to handle unreachable blocks better. 205 - #2528: Fix undefined bit shift in sroa. 206 - #2539: Change implementation of post order CFG traversal. 207 - Validator 208 - Add validation of storage classes for WebGPU (#2446) 209 - Add validation for ExecutionMode in WebGPU (#2443) 210 - Implement WebGPU specific CFG validation (#2386) 211 - Allow NonWritable to target struct members. (#2420) 212 - Allow storage type mismatch for parameter in relaxed addressing mode. 213 - Allow non memory objects as parameter in relaxed addressing mode. 214 - Disallow nested Blocks and buffer blocks (#2410). 215 - Add validation for SPV_NV_cooperative_matrix (#2404) 216 - Add --strip-atomic-counter-memory (#2413) 217 - Check OpSampledImage is only passed into valid instructions (#2467) 218 - Handle function decls in Structured CFG analysis (#2474) 219 - Validate that OpUnreacahble is not statically reachable (#2473) 220 - Add pass to generate needed initializers for WebGPU (#2481) 221 - Allow images without format for OpenCL. (#2470) 222 - Remove unreachable block validation (#2525) 223 - Reduce runtime of array layout checks (#2534) 224 - Add validation specific to OpExecutionModeId (#2536) 225 - Validate sign of int types. (#2549) 226 - VK_KHR_uniform_buffer_standard_layout validation (#2562) 227 Fixes: 228 - #2439: Add missing DepthGreater case to Fragment only check. 229 - #2168: Disallow BufferBlock on StorageBuffer variables for Vulkan. 230 - #2408: Restrict and Aliased decorations cannot be applied to the same id. 231 - #2447: Improve function call parameter check. 232 - Reduce 233 - Add Pass to remove unreferenced blocks. (#2398) 234 - Allows passing options to the validator. (#2401) 235 - Improve reducer algorithm and other changes (#2472) 236 - Add Pass to remove selections (#2485) 237 - Add passes to simplify branches (#2507) 238 Fixes: 239 - #2478: fix loop to selection pass for loops with combined header/continue block 240 241v2019.2 2019-02-20 242 - General: 243 - Support SPV_EXT_physical_storage_buffer 244 - A number of memory leak have been fixed. 245 - Removed use of deprecated Google test macro: 246 - Changed the BUILD.gn to only build tests in Chromium. 247 - Optimizer 248 - Upgrade memory model improvments for modf and frexp. 249 - Add a new pass to move loads closer to their uses: code sinking. 250 - Invalidating the type manager now invalidates the constnat manager. 251 - Expand instrumentation pass for bindless bounds checking to runtime-sized descriptor arrays. 252 - Add a new pass that removes members from structs that are not used: dead member elimination. 253 Fixes: 254 - #2292: Remove undefined behaviour when folding bit shifts. 255 - #2294: Fixes for instrumentation code. 256 - #2293: Fix overflow when folding -INT_MIN. 257 - #2374: Don't merge unreachable blocks when merging blocks. 258 - Validator 259 - Support SPV_KHR_no_integer_wrap and related decorations. 260 - Validate Vulkan rules for OpTypeRuntimeArray. 261 - Validate NonWritable decoration. 262 - Many WebGPU specific validation rules were added. 263 - Validate variable pointer related function call rules. 264 - Better error messages. 265 Fixes: 266 - #2307: Check forwards references in OpTypeArray. 267 - #2315, #2303: Fixed the layout check for relaxed layout. 268 - #1628: Emit an error when an OpSwitch target is not an OpLabel. 269 - Reduce 270 - Added more documentation for spirv-reduce. 271 - Add ability to remove OpPhi instructions. 272 - Add ability to merge two basic blocks. 273 - Add ability to remove unused functions and unused basic blocks. 274 Fixes: 275 276v2019.1 2019-01-07 277 - General: 278 - Created a new tool called spirv-reduce. 279 - Add cmake option to turn off SPIRV_TIMER_ENABLED (#2103) 280 - New optimization pass to update the memory model from GLSL450 to VulkanKHR. 281 - Recognize OpTypeAccelerationStructureNV as a type instruction and ray tracing storage classes. 282 - Fix GCC8 build. 283 - Add --target-env flag to spirv-opt. 284 - Add --webgpu-mode flag to run optimizations for webgpu. 285 - The output disassembled line number stead of byte offset in validation errors. (#2091) 286 - Optimizer 287 - Added the instrumentation passes for bindless validation. 288 - Added passes to help preserve OpLine information (#2027) 289 - Add basic support for EXT_fragment_invocation_density (#2100) 290 - Fix invalid OpPhi generated by merge-return. (#2172) 291 - Constant and type manager have been turned into analysies. (#2251) 292 Fixes: 293 - #2018: Don't inline functions with a return in a structured CFG contstruct. 294 - #2047: Fix bug in folding when volatile stores are present. 295 - #2053: Fix check for when folding floating pointer values is allowed. 296 - #2130: Don't inline recursive functions. 297 - #2202: Handle multiple edges between two basic blocks in SSA-rewriter. 298 - #2205: Don't unswitch a latch condition during loop unswitch. 299 - #2245: Don't fold branch in loop unswitch. Run dead branch elimination to fold them. 300 - #2204: Fix eliminate common uniform to place OpPhi instructions correctly. 301 - #2247: Fix type mismatches caused by scalar replacement. 302 - #2248: Fix missing OpPhi after merge return. 303 - #2211: After merge return, fix invalid continue target. 304 - #2210: Fix loop invariant code motion to not place code between merge instruction and branch. 305 - #2258: Handle CompositeInsert with no indices in VDCE. 306 - #2261: Have replace load size handle extact with no index. 307 - Validator 308 - Changed the naming convention of outputing ids with names in diagnostic messages. 309 - Added validation rules for UniformConstant variables in Vulkan. 310 - #1949: Validate uniform variable type in Vulkan 311 - Ensure for OpVariable that result type and storage class operand agree (#2052) 312 - Validator: Support VK_EXT_scalar_block_layout 313 - Added Vulkan memory model semantics validation 314 - Added validation checkes spefic to WebGPU environment. 315 - Add support for VK_EXT_Transform_feedback capabilities (#2088) 316 - Add validation for OpArrayLength. (#2117) 317 - Ensure that function parameter's type is not void (#2118) 318 - Validate pointer variables (#2111) 319 - Add check for QueueFamilyKHMR memory scope (#2144) 320 - Validate PushConstants annotation and type (#2140) 321 - Allow Float16/Int8 for Vulkan 1.0 (#2153) 322 - Check binding annotations in resource variables (#2151, #2167) 323 - Validate OpForwardPointer (#2156) 324 - Validate operation for OpSpecConstantOp (#2260) 325 Fixes: 326 - #2049: Allow InstanceId for NV ray tracing 327 - Reduce 328 - Initial commit wit a few passes to reduce test cases. 329 - Validation is run after each reduction step. 330 Fixes: 331 332 333v2018.6 2018-11-07 334 - General: 335 - Added support for the Nvidia Turing and ray tracing extensions. 336 - Make C++11 the CXX standard in CMakeLists.txt. 337 - Enabled a parallel build for MSVC. 338 - Enable pre-compiled headers for MSVC. 339 - Added a code of conduct. 340 - EFFCEE and RE2 are now required when build the tests. 341 - Optimizer 342 - Unrolling loops marked for unrolling in the legalization passes. 343 - Improved the compile time of loop unrolling. 344 - Changee merge-return to create a dummy loop around the function. 345 - Small improvement to merge-blocks to allow it to merge more often. 346 - Enforce an upper bound for the ids, and add option to set it. 347 - #1966: Report error if there are unreachable block before running merge return 348 Fixes: 349 - #1917: Allow 0 (meaning unlimited) as a parameter to --scalar-replacement 350 - #1915: Improve handling of group decorations. 351 - #1942: Fix incorrect uses of the constant manager. Avoids type mismatches in generated code. 352 - #1997: Fix dead branch elimination when there is a loop in folded selection. 353 - #1991: Fixes legality check in if-conversion. 354 - #1987: Add nullptr check to array copy propagation. 355 - #1984: Better handling of OpUnreachable in ADCE. 356 - #1983: Run merge return on reachable functions only. 357 - #1956: Handled atomic operations in ADCE. 358 - #1963: Fold integer divisions by 0 to 0. 359 - #2019: Handle MemberDecorateStringGOOGLE in ADCE and strip reflect. 360 - Validator 361 - Added validation for OpGroupNonUniformBallotBitCount. 362 - Added validation for the Vulkan memory model. 363 - Added support for VK_KHR_shader_atddomic_int64. 364 - Added validation for execution modes. 365 - Added validation for runtime array layouts. 366 - Added validation for 8-bit storage. 367 - Added validation of OpPhi instructions with pointer result type. 368 - Added checks for the Vulkan memory model. 369 - Validate MakeTexelAvailableKHR and MakeTexelVisibleKHR 370 - Allow atomic function pointer for OpenCL. 371 - FPRounding mode checks were implemented. 372 - Added validation for the id bound with an option to set the max id bound. 373 Fixes: 374 - #1882: Improve the validation of decorations to reduce memory usage. 375 - #1891: Fix an potential infinite loop in dead-branch-elimination. 376 - #1405: Validate the storage class of boolean objects. 377 - #1880: Identify arrays of type void as invalid. 378 - #487: Validate OpImageTexelPointer. 379 - #1922: Validate OpPhi instructions are at the start of a block correctly. 380 - #1923: Validate function scope variable are at the start of the entry block. 381 382v2018.5 2018-09-07 383 - General: 384 - Support SPV_KHR_vulkan_memory_model 385 - Update Dim capabilities, to match SPIR-V 1.3 Rev 4 386 - Automated build bots no run tests for the VS2013 case 387 - Support Chromium GN build 388 - Use Kokoro bots: 389 - Disable Travis-CI bots 390 - Disable AppVeyor VisualStudio Release builds. Keep VS 2017 Debug build 391 - Don't check export symbols on OSX (Darwin): some installations don't have 'objdump' 392 - Reorganize source files and namespaces 393 - Fixes for ClangTidy, and whitespace (passes 'git cl presumit --all -uf') 394 - Fix unused param compile warnings/errors when Effcee not present 395 - Avoid including time headers when timer functionality is disabled 396 - Avoid too-stringent warnings flags for Clang on Windows 397 - Internal refactoring 398 - Add hooks for automated fuzzing 399 - Add testing of command line executables 400 - #1688: Use binary mode on stdin; fixes "spirv-dis <foo.spv" on Windows 401 - Optimizer 402 - The optimizer validates the module before it begins 403 - Add API to register passes by string name 404 - Fold a vector shuffle feeding a vector shuffle 405 - Add -combine-access-chains transform 406 - Refactor how IRContext is handled by passes 407 - Improve bookkeeping for instruction result type and result id 408 - Fix over-duplication of decorations 409 - Fix handling of exits from selections in dead-branch elimination, and dead code 410 elimination. 411 - Fix handling of certain kinds of flow control in merge-return 412 Fixes: 413 - #1721: Fix size bug when folding vector shuffles 414 - #1722: Fix size infinite loop when folding vector shuffles 415 - #1724: Fix finding a constant of a specific type 416 - #1727: Dead branch elim: Reorder blocks if needed to satisfy dominance rule 417 - #1729: Handle VariablePointers cases in various optimizations 418 - #1731: Fix vector shuffle with literal id indicating undef value 419 - #1736: Fix handling of decorations and phis in merge-return 420 - #1787: Fix handling of decorations related to access chains 421 - #1865: Avoid leaking memory for SPIR-V constant values 422 - Validator 423 - Improve error messages 424 - Avoid platform-dependent traversal ordering, to ensure consistent messages 425 - Use libspirv::Instruction where possible 426 - Add option to skip all block layout checks 427 - Validate all type IDs 428 - Validate uses of OpFunction 429 - Validate uses of OpTypeFunction 430 - Disallow a struct containing its own type https://crbug.com/874372 431 - #1685: Vulkan permits non-monotonic offsets for block members 432 - #1697: Enforce block layout rules even when using relaxed block layout option 433 - #1719: Fix line number for vector shuffle valiation error 434 - #1789: Avoid assertion failure when validating some functions 435 - #1800: Fix validation of OpCopyMemorySized 436 - #1822: Stop enforcing struct member offset montonicity 437 - #1831: Disallow void members in structs 438 439v2018.4 2018-07-08 440 - General: 441 - Support SPV_KHR_8bit_storage 442 - Add gclient and presubmit configurations 443 - Enable Kokoro build bots (#1625) 444 - Group tests into fewer executables, reduces load on CI 445 - Port test script to Python 3 446 - Symbol export tests respect SPIRV_SKIP_TESTS 447 - #1596: Operand lookup succeeds if enabled by a capability 448 - #1624: Instruction lookup succeeds if enabled by a capability 449 - Refactoring namespaces: 450 - #1678: Change libspirv to spvtools 451 - Code in source/utils moved into spvtools::utils 452 - Code in source/comp moved into spvtools::comp 453 - Optimizer: 454 - Remove insert-extract-elim pass. Use simplification pass instead. 455 - Preserve instruction-to-block mapping in most passes, to reduce runtime. 456 - Small vector optimization for operands 457 - Add pass to move Private variables to Function. Increase opportunity to optimize. 458 - Fixes: 459 #1120: Check static uses of entry point interfaces 460 #1372: Avoid merging some structs, to preserve names for reflection 461 #1577: Scalar replacement uses only undecorated types. 462 #1578: Fix handling of forward-pointer types, and types embedding pointers 463 to themselves. 464 #1591: Inliner: Callee variable with initializer should have a store at the call site. 465 #1634: Fix crash: Use type id in vector type lookup 466 #1649: Fix assert in compact-ids pass 467 Fix constant folder: ensure it uses the right type 468 #1659: Folding rules added to IRContext. Avoids leak. 469 - Validator 470 - Add work-in-progress WebGPU environment. Disallows OpUndef 471 - #670, #1581: Improve error messages; disassemble instruction 472 - #491: Check structured switches 473 - #937: Check layout rules for Block and BufferBlock in Uniform, StorageBuffer, PushConstant 474 - #1281: Check invalid branches into structured constructs 475 - #1522: Disallow array-of-arrays with DescriptorSets 476 - #1577: Allow duplicate pointer types. 477 - #1581: Better messages: output ID names along with numbers in more cases. 478 - #1597: Check Vulkan 1.1 capabilities 479 - #1618: Check invalid exit from structured case construct 480 - #1622: Run IdPass before DataRulesPass 481 - #1632: Reduce test time by artificially lowering limits in limit test 482 - #1638: Block-decorated structs member order must respect offset order 483 - #1657: Improve CFG validation diagnostics 484 - Khronos SPIR-V #337: GLSL.std.450 Refract instruction Eta param can be any float scalar. 485 - #1606: PushConstant Blocks follow storage-buffer layout rules 486 - #1664: Check layout of StorageBuffer variables with Block decoration, using storage buffer 487 rules 488 - #1666: Layout validation should permit {vec3; float} packing 489 - #1637, #1668: Layout validation uses RowMajor, ArrayStride, MatrixStride properly 490 - Linker 491 - Avoid buffer overrun when creating OpModuleProcessed 492 493v2018.3 2018-05-25 494 - General: 495 - Support SPV_EXT_descriptor_indexing 496 - Support SPV_GOOGLE_decorate_string 497 - Support SPV_GOOGLE_hlsl_functionality1 498 - Support SPV_NV_shader_subgroup_partitioned 499 - Use "unified1" grammar from SPIRV-Headers 500 - Simplify support for new extensions. Assembler, disassembler, and simple validation 501 support is automatic if new tokens are introduced with appropriate extension 502 attributes in the "unified1" SPIR-V core grammar. 503 - Disassembler: Emit more digits on floating point, to reliably reproduce all 504 significand bits. (Use std::max_digits10 instead of std::digits10) 505 - Fix compilation for old XCode versions: Explicit construction required for std::set. 506 - Optimizer: 507 - Add --strip-reflect 508 - Add --time-report 509 - Add --loop-fission 510 - Add lop fusion. 511 - Add loop peeling pass and internal utility. 512 - Improve optimizer runtime. 513 - Merge-return now works with structured control flow. 514 - New (faster) SSA rewriter to convert local loads and stores to SSA IDs and phis. 515 Can replace load/store elimination passes. 516 - Fix instruction folding case: insertion that feeds and extract, when the extract 517 remains. 518 - Fold OpDot. 519 - Fold OpFNegate. 520 - Fold multply and divide of same value. 521 - Fold FClamp feeding a compare. 522 - Fold OpLoad feeding an extract, to reduce excessive copying. (#1547) 523 - Fold Fmix feeding an extract. 524 - Use simplification pass instead of insert-extract elimination. 525 - Constant fold OpVectorTimesScalar. 526 - Copy propagate arrays, in simple cases. 527 - Aggressive dead code elimination: Can remove more instructions, e.g. derivatives. 528 - Aggressive dead code elimination: Remove Workgroup variables that are written but not read. 529 - Better handling of OpImageTexelPointer 530 - Initial utilities for scalar evolution. 531 - Add Vector dead code elimination. 532 - Each pass can only run once. 533 - Allow code hosting in if-conversion. 534 - Add external interface for adding a PassToken, so external code can make their own 535 passes. 536 - Fixes: 537 #1404: Don't optimize away the compute compute workgroup size constant. 538 #1407: Remove a bad assertion 539 #1456: Fix bug in SSA rewriter related to variables updated in loops. 540 #1487: Fix long runtime in Dead insertion elimination: Don't revist select phi nodes. 541 #1492: Aggressive dead code elimination can remove OpDecorateStringGOOGLE. 542 #1527: Fix inlining of functions having OpKill and OpUnreachable. 543 #1559: Fix assert failure in reduce-load-size pass. 544 #1556: Aggressive dead code elimination: Fix handling of OpCopyMemory. 545 - Validator: 546 - Check Vulkan built-in variables 547 - Check Vulkan-specific atomic result type rule. 548 - Relax control barrier check for SPIR-V 1.3. Fixes #1427 549 - Check OpPhi. 550 - Check OpMemoryModel. 551 - Stop checking sizes derived from spec-constants. 552 - Re-enable checks for OpUConvert. 553 - Vulkan: Fix check for PrimitiveId: Permit as Input in fragment shader. 554 - Validate binary version for the given target environment. 555 - Add tests for OpBranch checks. 556 - Vulkan 1.1: Check scope for non-uniform subgroup operations. 557 - Fix checks for SPV_AMD_gpu_shader_int16. 558 - Fix logical layout check for OpDecorateId. 559 - Fix checks for ViewportIndex & Layer for Vulkan and SPV_EXT_shader_viewport_index_layer. 560 - Fixes: 561 #1470: Vulkan: Don't restrict WorkgroupSize to Input storage class. 562 #1469: Vulkan: Permit Subgroup memory scope for Vulkan 1.1. 563 #1472: Per-vertex variable validation fixes. 564 #1483: Valdiate barrier execution scopes for Vulkan 1.1. 565 - Fixes: 566 #898: Linker properly removes FuncParamAttr from imported symbols. 567 #924, #1174: Fix handling of decoration groups in optimizer, linker. 568 569v2018.2 2018-03-07 570 - General: 571 - Support SPIR-V 1.3 and Vulkan 1.1. 572 - Default target environment is now SPIR-V 1.3. For command-line tools, 573 use the --target-env option to override the default. Examples: 574 # Generate a SPIR-V 1.0 binary instead of SPIR-V 1.3 575 spirv-as --target-env spv1.0 a.spvasm -o a.spv 576 spirv-as --target-env vulkan1.0 a.spvasm -o a.spv 577 # Validate as Vulkan 1.0 578 spirv-val --target-env vulkan1.0 a.spv 579 - Support SPV_GOOGLE_decorate_string and SPV_GOOGLE_hlsl_functionality1 580 - Fixes: 581 - Fix Android.mk build. Compilation was failing due to missing definitions of 582 SpvCapabilityFloat16ImageAMD and other enumerated values. 583 - Optimizer: Avoid generating duplicate names when merging types. 584 - #1375: Validator: SPV_AMD_gpu_shaer_half_float implicitly allows declaration 585 of the 16-bit floating point type. 586 - #1376: Optimizer: Avoid folding half-precision float. 587 588v2018.1 2018-03-02 589 - General: 590 - Support Visual Studio 2013 again. (Continue support for VS 2015 and VS 2017.) 591 - Support building SPIRV-Tools as a shared library. 592 - Improve the HLSL legalization optimization recipe. #1311 593 - Optimizer: 594 - General speedups. 595 - Remove generic dead code elimination functionality from transforms: 596 --eliminate-local-single-block 597 --eliminate-local-single-store 598 --eliminate-local-multi-store 599 To recover the previous behaviour, a recipe using those transforms should now 600 also invoke the --eliminate-dead-code-aggressive transform. 601 - Improve folding, including coverage for floating point, OpSelect, and arithmetic 602 with non-trivial constant operands. 603 - Add loop-invariant code motion pass. 604 - Add loop-unrolling pass, for honouring unroll hits. 605 - Add loop-unswitch pass. 606 - Add instruction simplification pass. 607 - Aggressive dead code elimination: Understands capability hierarchy when finding 608 instructions it can eliminate (combinators). (PR #1268) 609 - CCP can now fold floating point arithmetic. #1311 610 - Validator: 611 - Validate barrier instructions. 612 - Check Vulkan-specific rules for atomics. 613 - Check Vulkan prohibition of Location or Component decorations on BuiltIn variables. 614 - Linker: 615 - Add --verify-ids option 616 - Add option to allow a resulting module to be partially linked. 617 - Handle OpModuleProcessed (instructions in SPIR-V layout section 7c) 618 - Fixes: 619 - #1265: Optimizer: Fix use-after free bug in if-conversion. (Fix object lifecycle bug 620 in type manager.) 621 - #1282: Fix new warnings found by GCC 8.0.1. 622 - #1285: Optimizer: Fix random failures during inlining. (Dangling references in DefUseManager) 623 - #1295: Optimizer: Fix incorrect handling of Phi nodes in CCP. 624 - #1300: Fix CCP: avoid bad CCP transitions and unsettled values. 625 - #1304: Avoid static-duration variables of class type (with constructors). 626 - #1323: Fix folding of an insert composite feeding a composite extract. 627 - #1339: Fix CCP: Handle OpConstantNull boolean values as conditions. 628 - #1341: DCEInst: Keep atomic instructions (and some others with side effects). 629 - #1354: Don't fold integer division. 630 - #1357: Support OpConstantNull in folding. 631 - #1361: CCP: Fix handling of non-constant module-scope values 632 633v2018.0 2018-02-02 634 - General 635 - VisualStudio 2013 is no longer supported. VisualStudio 2015 is supported. 636 - Use "include/unified1" directory from SPIRV-Headers. Requires recent SPIRV-Headers source. 637 - Disassembler: spirv-dis adds --color option to force color disassembly. 638 - Optimizer: 639 - Add pass to eliminate dead insertions. 640 - Aggressive dead code elimination now removes OpSwitch constructs. 641 - Block merging occurs in more cases. 642 - Add driver workaround transform: replace OpUnreachable with harmless branch to merge. 643 - Improve instruction folding framework. 644 - Add loop analysis. 645 - Add scalar replacement of aggregates to size-optimization recipe. 646 - Add pass to replace instructions invalid for a shader stage, with a harmless value. 647 This changes the semantics of the program! Not for general use! 648 - Rearragne and add passes to performance-optimization recipe, to produce better results. 649 - Validator: 650 - Validate OpenCL extended instructions. 651 - Shaders can't perform atomics on floats. 652 - Validate memory semantics values in atomics. 653 - Validate instruction-adjacency constraints, e.g. OpPhi predecessors, merge instructions 654 immediately precede branches. 655 - Fixes: 656 - PR 1198: Optimizer: Fix CCP in presence of matrix constants. 657 - #1199: Optimizer: Fix CCP: don't propagate spec constants. 658 - #1203: Optimizer: Fix common uniform elim bug introduced by refactoring. 659 - #1210: Optimizer: Aggressive dead code elimination: Fix 'break' identification. 660 - #1212: Optimizer: Aggressive dead code elimination: Was skipping too many instructions. 661 - #1214: Optimizer: Aggressive dead code elimination: Fix infinite loop. 662 - #1228: Optimizer: Fix CCP: Handling of varying Phi nodes; was resulting in infinite loop. 663 - #1245: Optimizer: Dead branch elimination: Avoid a null pointer dereference. 664 - #1250: Optimizer: Dead branch elimination: Avoid spuriously reporting a change. 665 666v2017.3 2018-01-12 667 - General: 668 - Support DebugInfo extended instruction set, targeted at OpenCL environments. 669 See the SPIR-V Registry. 670 - Generate a SPIRV-Tools.pc file for pkg-config. 671 - Optimizer: 672 - Progress for legalization of code generated from HLSL (issue #1118): 673 - Add --legalize-hlsl option to run transforms used to transform intermediate 674 code generated by HLSL to SPIR-V for Vulkan compilers. Those compilers 675 normally run these transforms automatically. This option is used for developing 676 those transforms. 677 - Add Private-to-Function variable conversion for modules with logical 678 addressing. 679 - Add --ccp: SSA Conditional Constant Propagation (CCP) 680 - Add --print-all to show disassembly for each optimization pass. 681 - Internal: Add loop descriptors and post-order tree iterator. 682 - Generalized dead branch elimination 683 - Aggressive dead code elimination (ADCE) now removes dead functions and 684 module-scope variables. 685 - Vector extract/insert elimination now optimizes through some cases of 686 VectorShuffle, and GLSL.std.450 Mix extended instruction. 687 - Validator: 688 - Add validation for GLSL.std.450 extended instruction set. 689 - Check out of bounds composite accesses, where that's statically computable. 690 Fixes #1112. 691 - Check upper bits of literal numbers that aren't a multiple of 32-bits wide. 692 - More validation of primitive instructions 693 - Add optional "relaxed" checking logical addressing mode to permit some 694 cases of pointer-to-pointer. Contributes to HLSL legalization (issue #1118). 695 - Fixes: 696 #1100: Validator: Image operand Sample can be used with OpImageSparseFetch, 697 OpImageSparseRead. 698 #1108: Remove duplicates transform was incorrectly removing non-duplicate 699 decorations. 700 #1111: Optimizer's type manager could reference deleted memory. 701 #1112: Fix decoration equality check, e.g. it is now symmetric. 702 #1129: Validator now disallows Dim=SupbassData for OpImageSparseRead. 703 #1143: Fix CCP: Was generating incorrect code for loops. 704 #1153: Fix CCP crash. 705 #1154: Optimizer's internal instruction-to-block mappings were sometimes 706 inconsistent. 707 #1159: Fix CCP infinite loop. 708 #1168: Fix dead branch elimination intermittently generating incorrect code. 709 Fixes https://github.com/KhronosGroup/glslang/issues/1205 710 #1186: Fix validation of PackDouble2x32 and UnpackDouble2x32 711 712v2017.2 2017-12-15 713 - General: 714 - Support OpenCL 1.2, 2.0 target environments, including embedded profiles 715 - Add CONTRIBUTING.md 716 - Fix exit status code for spirv-link 717 - Disassember: Enable emitting ANSI colour codes to a string 718 - Library avoids polluting global namespace. The libraries can export C and C++ 719 symbols starting with "spv", or in a C++ namespace. Add a test for this. 720 - Linux release builds include debug information, for easier profiling 721 - Build bots no longer test VisualStudio 2013 722 - Testing dependency RE2 requires VisualStudio 2015 or later 723 - Build bots check code formatting 724 - Optimizer: 725 - Add --skip-validation to spirv-opt 726 - Add dominance tree analysis 727 - Add generic value propagation engine 728 - Add global redundancy elimination within a function 729 - Add scalar replacement of function-scope variables of composite type 730 - Aggressive dead code elimination: Remove empty loops 731 - Killing an instruction notifies the IRContext 732 - IRContext::KillInst deletes the instruction 733 - Move CFG analysis to IRContext 734 - Add constant manager 735 - Fix: Don't consider derivative instructions as combinators. 736 - Fix: Don't delete an instruction twice in local dead-code-elimination 737 - Fix: Don't consider derivative instructions as combinators. 738 - Validator: 739 - Finish checking of image instructions (Section 3.32.10) 740 - Check sparse image instructions 741 - Check OpTypeImage, OpTypeSampleImage 742 - Check composite instructions (Section 3.32.12) 743 - Check atomic instructions (Section 3.32.18) 744 - Check OpEmitStreamVertex, OpEndStreamPrimitive instructions 745 - Re-enable validation of OpCopyObject 746 - OpKill, image ImplicitLod and QueryLod instructions can only be used in Fragment 747 shaders. 748 - Fixes for image instruction validation: 749 - Lod image operand only usable with ExplicitLod and OpImageFetch 750 - ExplicitLod Lod image operand must be float scalar 751 - OpImageFectch Lod image operand must be int scalar 752 - OpImageGather component operand must be 32-bits (integer scalar) 753 - OpImageQuerySizeLod Lod must be integer scalar 754 - Fixes: 755 #622: Remove names and decorations when inlining 756 #989: Aggressive dead code elim: Don't optimize away live breaks from a loop 757 #991: Fix validation of SPV_AMD_shader_ballot 758 #1004: Use after free of an instruction, in remove-duplicates transform 759 #1007: OpImageRead not required to return 4-component vector 760 #1009: OpImageRead can return scalar int/float types 761 #1011: OpImageWrite should allow scalar int/float texel types 762 #1012: Fix validat Dref type check 763 #1017: Load-store elimination considers variable initializations 764 #1034: Fix Windows debug build: operator< should be a weak ordering 765 #1083: Inlining: Set parent (function) for each inlined basic block. 766 #1075: Aggressive dead code elimination: Was leaving dangling references to 767 removed blocks. 768 769v2017.1 2017-11-23 770 - Update README with details on the public_spirv_tools_dev@khronos.org mailing list. 771 - General: 772 - Automatically deploy built artifacts to GitHub Releases 773 - Add a Linker (module combiner). Under development. 774 - Add Android.mk for Android NDK builds. 775 - Add the 'effcee' library as an optional dependency for use in tests. 776 Eventually it will be a required dependency, once downstream projects have 777 a chance to adjust. Requires 're2' library. 778 - Avoid static-duration variables of class type (with constructors). 779 - Hack around bugs in gcc-4.8.1 template handling 780 - Faster opcode lookup 781 - Validator: 782 - Recognize extensions listed on SPIR-V registry, 783 through #25 SPV_AMD_shader_fragment_mask 784 - Validator issues an info message when it sees an unrecognized extension. 785 - Type check basic arithmetic operations 786 - Type check carry/extended arithmetic operations 787 - Type check vector arithmetic operations 788 - Type check Relational and Logical instructions 789 - Type check Bit instructions 790 - Check type uniqueness rules 791 - Check conversion instructions 792 - Check image instructions 793 - Check derivative instructions 794 - Check OpVectorShuffle 795 - Check OpBranchConditional 796 - OpModuleProcessed is only allowed after debug names section and before annotations 797 section. 798 - Checks the right kind of return is called for each function (void or non-void). 799 - Add option to relax type check when storing structs (--relax-store-struct) 800 - Optimizer: 801 - Refactoring internal representation of the module, including: 802 - IRContext: owns a module and manages analyses 803 - Instructions are owned by intrusive lists, and have unique IDs 804 - BasicBlock owns its instruction list. 805 - DefUseManager: change representation of uses, for faster processing 806 on large modules. 807 - Add high level recipes: -O, -Os, and -Oconfig 808 Recipes for -O and -Os are under development. 809 - Add eliminate-dead-function transform 810 - Add strength reduction transform: For now, convert multiply by power of 2 811 to a bit shift. 812 - Add CFG cleanup transform 813 - Add removal of dead module-scope variables 814 - Add merge-return transform for modules without structured control flow 815 - Add redundancy elimination within a basic block (local value numbering) 816 - Extract-insert elimination: 817 - Recognize the case where the first instruction in the sequence is an 818 OpCompositeConstruct or OpConstantComposite 819 - Handle some cases of nested structs 820 - Dead branch elimination now can eliminate entire selection constructs 821 when all arms are dead. 822 - Compressing codec: 823 - Updated algorithm to 1.01, 1.02, 1.03 824 - Not built by default. Use -DSPIRV_BUILD_COMPRESSION=ON to build. 825 - Codec can be parameterized by a customized model. 826 - Fixes: 827 #728: Fix decoration of inlined functions 828 #798: spirv-as should fail when given unrecognized long option 829 #800: Inliner: Fix inlining function into header of multi-block loop 830 #824: Eliminate-local-multi-store: Fix a crash 831 #826: Elimiante-local-multi-store: Fix a crash 832 #827: Fix crash when compact-ids transform runs before another transform. 833 #834: Add Cmake option to build the compressing codec. Off by default. 834 #911: Fix classification of Line and NoLine instructions 835 836v2017.0 2017-09-01 837 - Update README to describe that assembler, disassembler, and binary parser support 838 are based on grammar files from the SPIRV-Headers repository. 839 840v2016.7 2017-09-01 841 - Add SPIR-V 1.2 842 - OpenCL 2.2 support is now based on SPIR-V 1.2 843 - Support AMD extensions in assembler, disassembler: 844 SPV_AMD_gcn_shader 845 SPV_AMD_shader_ballot 846 SPV_AMD_shader_explicit_vertex_parameter 847 SPV_AMD_shader_trinary_minmax 848 SPV_AMD_gpu_shader_half_float 849 SPV_AMD_texture_gather_bias_lod 850 SPV_AMD_gpu_shader_int16 851 - Optimizer: Add support for: 852 - Inline all function calls in entry points. 853 - Flatten decoration groups. Fixes #602 854 - Id compaction (minimize Id bound). Fixes #624 855 - Eliminate redundant composite insert followed by extract 856 - Simplify access chains to local variables 857 - Eliminate local variables with a single store, if possible 858 - Eliminate local variables with a several stores, if possible 859 - Eliminate loads and stores in same block to local variables 860 - Eliminate redundant insert/extract to composite values 861 - Aggressive dead instruction elimination 862 - Eliminate dead branches 863 - Merge blocks when the second can only be preceded by the first 864 - Eliminate ommon uniform loads 865 - Assembler: Add option to preserve numeric ids. Fixes #625 866 - Add build target spirv-tools-vimsyntax to generate spvasm.vim, a SPIR-V 867 assembly syntax file for Vim. 868 - Version string: Allow overriding of wall clock timestamp with contents 869 of environment variable SOURCE_DATE_EPOCH. 870 - Validator implements relaxed rules for SPV_KHR_16bit_storage. 871 - CMake installation rules use GNUInstallDirs. For example, libraries 872 will be installed into a lib64 directory if that's the norm for the 873 current system. 874 - Fixes: 875 #500: Parameterize validator limit checks 876 #508: Support compilation under CYGWIN 877 #517: Fix validation when continue (or case) contstruct is also the head of a 878 nested control construct. 879 #551: If a merge block is reachable, it must be *strictly* dominated by its 880 header. 881 #548: Validator: Error when the reserved OpImageSparseSampleProj* opcodes 882 are used. 883 #611: spvtools::Optimizer was failing to save the module to the output 884 binary vector when all passes succeded without changes. 885 #629: The inline-entry-points-all optimization could generate invalidly 886 structured code when the inlined function had early returns. 887 #697: Optimizer's Instruction::ForEachInId method was skipping semantics-id 888 and scope-id. 889 #755: Inliner: Fix inlining of callee with single Return appearing before 890 the end of the function. 891 #776: Fix dead branch elimination in presence of complex but dead control 892 flow. 893 #781: SPV_KHR_variable_pointers allows duplicate pointer types 894 #782: Inliner: Fix remapping of non-label forward references in callee 895 #787: Inliner: Fix remapping of inlined entry block when called from 896 single block loop. 897 #790: Inliner: Fix remapping of inlined entry block when callee has 898 multiple returns. 899 900v2016.6 2016-12-13 901 - Published the C++ interface for assembling, disassembling, validation, and 902 optimization. 903 - Support SPV_KHR_shader_draw_parameters in assembler, disassembler, parser. 904 - Validator: 905 - Add validator API accepting raw binary words 906 - Increased coverage: 907 - Checks "Data rules" in Universal Validation Rules, section 2.16.1 908 - WIP: Universal Limits. 909 - The minimum mandated upper bounds are checked. 910 - TODO: Parameterize the validator to allow larger limits accepted by 911 a more than minimally capable implementation. 912 - OpSampledImage checks 913 - OpConstantComposite checks 914 - Id bound check 915 - Disasssembler: 916 - Generates friendly GLSL-based names for more builtin variables 917 - Generates friendly names for numeric OpConstant values 918 - Vendor tool info extracted from SPIR-V XML registry file. 919 - Fixes issues: 920 #429: Validator: Allow OpTypeForwardPointer and OpTypeStruct to reference 921 undefined IDs 922 #482: Validator: OpVariable initializer can be an ID of a module-scope variable 923 924v2016.5 2016-09-16 925 - Support SPV_KHR_shader_ballot in assembler, disassembler, parser. 926 - Disassembler: Generate friendly names for built-in variables. 927 - Partial fixes: 928 #359: Add Emacs helper for automatically diassembling/assembling a SPIR-V 929 binary on file load/save. 930 - Fixes: 931 #414: Validator: Allow OpUndef for composite constants 932 #415: Validator: Phi can use its own value in some cases. 933 934v2016.4 2016-09-01 935 - Relicensed under Apache 2.0 936 - Add optimization passes (in API and spirv-opt command) 937 - Fold spec constants defined with OpSpecConstantOp and 938 OpSpecConstantComposite to normal constants with fixed value(s). 939 - Fixes issues: 940 #318: Relicensed under Apache 2.0 941 942v2016.3 2016-08-24 943 - Add target environment enums for OpenCL 2.1, OpenCL 2.2, 944 OpenGL 4.0, OpenGL 4.1, OpenGL 4.2, OpenGL 4.3, OpenGL 4.5. 945 - Add spirv-cfg, an experimental tool to dump the control flow graph 946 as a GraphiViz "dot" graph 947 - Add optimization pass: Eliminate dead constants. 948 - Add spirv-lesspipe.sh filter utility 949 - Fixes issues: 950 #288: Check def-use dominance rules for OpPhi (variable,parent) operands 951 #339: Allow OpUndef in types-constants-global-vars section, as required 952 by SPIR-V 1.0 Rev7, 1.1 Rev 3. 953 #340: Avoid race on mkdir during build 954 #365: Relax PointSize, ClipDistance, CullDistance capability check in all 955 environments not just Vulkan 1.0. 956 957v2016.2 2016-08-05 958 - Validator is incomplete 959 - Checks ID use block is dominated by definition block 960 - Add optimization passes (in API and spirv-opt command) 961 - Strip debug info instructions 962 - Freeze spec constant to their default values 963 - Allow INotEqual as operation for OpSpecConstantOp 964 - Fixes bugs: 965 #270: validator: crash when continue construct is unreachable 966 #279: validator: infinite loop when analyzing some degenerate control 967 flow graphs 968 #286: validator: don't incorrectly generate def-use error for 969 (variable,parent) parameters to OpPhi 970 #290: disassembler: never generate bare % for an identifier 971 #295: validator: def-use dominance check should ignore unreachable uses 972 #276: validator: allow unreachable continue constructs 973 #297: validator: allow an unreachable block to branch to a reachable 974 merge block 975 976v2016.1 2016-07-19 977 - Fix https://github.com/KhronosGroup/SPIRV-Tools/issues/261 978 Turn off ClipDistance and CullDistance capability checks for Vulkan. 979 - The disassembler can emit friendly names based on debug info (OpName 980 instructions), and will infer somewhat friendly names for most types. 981 This is turned on by default for the spirv-dis command line tool. 982 - Updated to support SPIR-V 1.1 rev 2 983 - Input StorageClass, Sampled1D capability, and SampledBuffer capability 984 do not require Shader capability anymore. 985 986v2016.0 2016-07-04 987 988 - Adds v<year>.<index> versioning, with "-dev" indicating 989 work in progress. The intent is to more easly report 990 and summarize functionality when SPIRV-Tools is incorporated 991 in downstream projects. 992 993 - Summary of functionality (See the README.md for more): 994 - Supports SPIR-V 1.1 Rev 1 995 - Supports SPIR-V 1.0 Rev 5 996 - Supports GLSL std450 extended instructions 1.0 Rev 3 997 - Supports OpenCL extended instructions 1.0 Rev 2 998 - Assembler, disassembler are complete 999 - Supports floating point widths of 16, 32, 64 bits 1000 - Supports integer widths up to 64 bits 1001 - Validator is incomplete 1002 - Checks capability requirements in most cases 1003 - Checks module layout constraints 1004 - Checks ID use-definition ordering constraints, 1005 ignoring control flow 1006 - Checks some control flow graph rules 1007 - Optimizer is introduced, with few available transforms. 1008 - Supported on Linux, OSX, Android, Windows 1009 1010 - Fixes bugs: 1011 - #143: OpenCL pow and pown arguments 1012