| 1 |
0.4.42 |
| 2 |
====== |
| 3 |
|
| 4 |
- Initial 64-bit RISC-V support (Maksymilian Knust, Filip Wasil) |
| 5 |
- Add 64-bit LoongArch support (Jin Bo, Hecai Yuan) |
| 6 |
- Implement release and reuse of temporary registers for some targets (Jorge Zapata) |
| 7 |
- x86: Implement EVEX encoding and an opcode validation system (Jorge Zapata) |
| 8 |
- x86: Opcode refactor, improved constant handling and various other fixes (Jorge Zapata) |
| 9 |
- x86: add missing rounding operands for AVX and SSE (L. E. Segovia) |
| 10 |
- x86: Implement 64-bit single move constant load (L. E. Segovia) |
| 11 |
- includes: stop exporting the private compiler and OrcTarget definitions (Jorge Zapata) |
| 12 |
- Use hotdoc instead of gtk-doc to generate the documentation (Jorge Zapata) |
| 13 |
- ORC_DEBUG_FATAL environment variable allows abort on log messages of a certain level (Carlos Falgueras García) |
| 14 |
- Error message improvements and NEON backend clean-ups (Alexander Slobodeniuk) |
| 15 |
- Fix a few valgrind issues (Doug Nazar) |
| 16 |
- Build: enable tools such as orcc and orc-bugreport by default (Nirbheek Chauhan) |
| 17 |
- Various build fixes (L. E. Segovia, Sebastian Dröge) |
| 18 |
|
| 19 |
0.4.41 |
| 20 |
====== |
| 21 |
|
| 22 |
- orccodemem: Don't modify the process umask, which caused race conditions |
| 23 |
with other threads (L. E. Segovia) |
| 24 |
- Require glibc >= 2.07 (L. E. Segovia) |
| 25 |
- x86: various SSE and MMX fixes (L. E. Segovia) |
| 26 |
- avx: Fix sqrtps encoding causing an illegal instruction crash (L. E. Segovia) |
| 27 |
- Hide internal symbols from ABI and do not install internal headers |
| 28 |
(L. E. Segovia) |
| 29 |
- Rename backend to target, including orc-backend meson option and ORC_BACKEND |
| 30 |
environment variable (Jorge Zapata) |
| 31 |
- Testsuite, tools: Disambiguate OrcProgram naming conventions (L. E. Segovia) |
| 32 |
- Build: Fix _clear_cache call for Clang and error out on implicit function |
| 33 |
declarations (Sam James) |
| 34 |
- opcodes: Use MIN instead of CLAMP for known unsigned values to fix |
| 35 |
compiler warnings (Edward Hervey) |
| 36 |
- Spelling fix in debug log message (Marc Leeman) |
| 37 |
- ci improvements: Upload the generated .S and .bin and include Windows |
| 38 |
artifacts (Jorge Zapata) |
| 39 |
|
| 40 |
0.4.40 |
| 41 |
====== |
| 42 |
|
| 43 |
- Security: Minor follow-up fixes for CVE-2024-40897 (Sebastian Dröge) |
| 44 |
- powerpc: fix div255w which still used the inexact substitution (L. E. Segovia) |
| 45 |
- x86: work around old GCC versions (pre 9.0) having broken xgetbv |
| 46 |
implementations (L. E. Segovia) |
| 47 |
- x86: consider MSYS2/Cygwin as Windows for ABI purposes only (L. E. Segovia) |
| 48 |
- x86: handle unnatural and misaligned array pointers (L. E. Segovia) |
| 49 |
- orccodemem: Assorted memory mapping fixes (L. E. Segovia) |
| 50 |
- Fix include header use from C++ (Kleis Auke Wolthuizen) |
| 51 |
- Some compatibility fixes for Musl (L. E. Segovia) |
| 52 |
- ppc: Disable VSX and ISA 2.07 for Apple targets (L. E. Segovia) |
| 53 |
- ppc: Allow detection of ppc64 in Mac OS (L. E. Segovia) |
| 54 |
- x86: Fix non-C11 typedefs (L. E. Segovia) |
| 55 |
- meson: Fix detecting XSAVE on older AppleClang (L. E. Segovia) |
| 56 |
- x86: try fixing AVX detection again by adding check for XSAVE (L. E. Segovia) |
| 57 |
- Check return values of malloc() and realloc() (Vladimir D. Seleznev) |
| 58 |
|
| 59 |
0.4.39 |
| 60 |
====== |
| 61 |
|
| 62 |
- Security: Fix error message printing buffer overflow leading to possible |
| 63 |
code execution in orcc with specific input files (CVE-2024-40897). This |
| 64 |
only affects developers and CI environments using orcc, not users of liborc |
| 65 |
(Sebastian Dröge, L. E. Segovia) |
| 66 |
- div255w: fix off-by-one error in the implementations (L. E. Segovia) |
| 67 |
- x86: only run AVX detection if xgetbv is available (L. E. Segovia) |
| 68 |
- x86: fix AVX detection by implementing the check recommended by Intel (L. E. Segovia) |
| 69 |
- Only enable JIT compilation on Apple arm64 if running on macOS, fixes crashes on iOS (Piotr Brzeziński) |
| 70 |
- Fix potential crash in emulation mode if logging is enabled (L. E. Segovia) |
| 71 |
- Handle undefined TARGET_OS_OSX correctly (Joshua Root) |
| 72 |
- orconce: Fix typo in GCC __sync-based implementation (Jakub Vaněk) |
| 73 |
- orconce: Fix usage of __STDC_NO_ATOMICS__ (Alexander Slobodeniuk) |
| 74 |
- Fix build with MSVC 17.10 + C11 (L. E. Segovia) |
| 75 |
- Support stack unwinding on Windows (L. E. Segovia) |
| 76 |
- Major opcode and instruction set code clean-ups and refactoring (Jorge Zapata) |
| 77 |
- Refactor allocation and chunk initialization of code regions (L. E. Segovia) |
| 78 |
- Fall back to emulation on Linux if JIT support is not available, |
| 79 |
e.g. because of SELinux sandboxing or noexec mounting) (L. E. Segovia) |
| 80 |
|
| 81 |
0.4.38 |
| 82 |
====== |
| 83 |
|
| 84 |
- x86: account for XSAVE when checking for AVX support, fixing usage on |
| 85 |
hardened linux kernels where AVX support has been disabled (L. E. Segovia) |
| 86 |
- neon: Use the real intrinsics for divf and sqrtf (L. E. Segovia) |
| 87 |
- orc.m4 for autotools is no longer shipped. If anyone still uses |
| 88 |
it they can copy it into their source tree (Tim-Philipp Müller) |
| 89 |
|
| 90 |
0.4.37 |
| 91 |
====== |
| 92 |
|
| 93 |
- enable neon instructions on Apple ARM64 (Aleix Conchillo Flaqué) |
| 94 |
- orcc: Fix regression, was hard-coded to use "sse" as default target (Sebastian Dröge) |
| 95 |
- MMX backend fixes (L. E. Segovia, Jorge Zapata) |
| 96 |
- testsuite: Build fixes for Clang (L. E. Segovia) |
| 97 |
- testsuite, tools: Fix warning caused by inserting unneeded source operands (L. E. Segovia) |
| 98 |
- orccompiler: call sys_icache_invalidate() to invalidate macos inst cache (Aleix Conchillo Flaqué) |
| 99 |
- macOS/iOS version/target check build fixes (Aleix Conchillo Flaqué) |
| 100 |
|
| 101 |
0.4.36 |
| 102 |
====== |
| 103 |
|
| 104 |
- Only use AVX / AVX2 instructions on CPUs that support both AVX and AVX2 |
| 105 |
(fixes crash on machines that only support AVX) (L. E. Segovia) |
| 106 |
|
| 107 |
0.4.35 |
| 108 |
====== |
| 109 |
|
| 110 |
- Add support for AVX / AVX2 (L. E. Segovia) |
| 111 |
- SSE backend improvements (L. E. Segovia) |
| 112 |
- New `orf` and `andf` opcodes for bitwise AND and OR for single precision floats (Jorge Zapata) |
| 113 |
- Add support for `convwf`, int16 to float conversion (Jorge Zapata) |
| 114 |
- Allow backend selection through ORC_TARGET environment variable (L. E. Segovia) |
| 115 |
- Documentation improvements (Jorge Zapata, L. E. Segovia, Tim-Philipp Müller) |
| 116 |
- orconce: Use Win32 once implementation with MSVC (Seungha Yang, L. E. Segovia) |
| 117 |
- orcc: add --binary option to output raw machine code for functions (L. E. Segovia) |
| 118 |
- orcprofile: Implement Windows high-resolution timestamp for MSVC |
| 119 |
to allow benchmarking on MSVC builds (L. E. Segovia) |
| 120 |
|
| 121 |
0.4.34 |
| 122 |
====== |
| 123 |
|
| 124 |
- Thread-safety improvements around orc codemem allocation/freeing |
| 125 |
(Seungha Yang) |
| 126 |
- Add orc_parse_code() with more detailed error reporting (Francesco Romani) |
| 127 |
- Implement Orc function lazy initialization correctly via atomic operations |
| 128 |
(Sebastian Dröge) |
| 129 |
- orc program parser fixes and improvements (Francesco Romani, Sebastian Dröge) |
| 130 |
- build fixes and compiler warning fixes (Thomas Klausner, Tim-Philipp Müller, |
| 131 |
Sebastian Dröge) |
| 132 |
- coverity and clang scan-build static code analysis fixes (Sebastian Dröge) |
| 133 |
- meson: Do not always generate static library for test library (Jeremy Bicha) |
| 134 |
- ci improvements (Jordan Petridis, Nirbheek Chauhan, Tim-Philipp Müller) |
| 135 |
|
| 136 |
0.4.33 |
| 137 |
====== |
| 138 |
|
| 139 |
- Add support for aarch64 (64-bit ARM) architecture (not yet enabled on Windows though) |
| 140 |
(Marek Vasut, Dongju Chae, Gaetan Bahl) |
| 141 |
- aarch32: Implement loadupdb instruction used e.g. for video pixel |
| 142 |
format packing/unpacking/conversions (Marek Vasut, Gaetan Bahl) |
| 143 |
- neon: Fix unsigned only implementation of loadoffb, loadoffw and loadoffl |
| 144 |
(Daniel Knobe) |
| 145 |
- neon: Fix testsuite not passing on arm CPUs (Gaetan Bahl) |
| 146 |
- orccodemem: Fix use-after-free in error paths (Bastien Nocera) |
| 147 |
- orccpu-powerpc: Fix build with kernel < 4.11 (Fabrice Fontaine) |
| 148 |
- Add support for macOS Hardened Runtime (Doug Nazar) |
| 149 |
- Enable only SSE and MMX backends for Windows (Seungha Yang) |
| 150 |
- Fix ORC_RESTRICT definition for MSVC (Tim-Philipp Müller) |
| 151 |
- pkgconfig: add -DORC_STATIC_COMPILATION flag to .pc file for static-only builds |
| 152 |
(Steve Lhomme) |
| 153 |
|
| 154 |
0.4.32 |
| 155 |
====== |
| 156 |
|
| 157 |
- Add support for JIT code generation in Universal Windows Platform apps |
| 158 |
(Nirbheek Chauhan, Seungha Yang) |
| 159 |
- Minor Meson build system fixes and improvements |
| 160 |
(Jan Alexander Steffens, Tim-Philipp Müller) |
| 161 |
|
| 162 |
0.4.31 |
| 163 |
====== |
| 164 |
|
| 165 |
- Fix OrcTargetPowerPCFlags enum typedef to revert API change on |
| 166 |
macOS/iOS (Pablo Marcos Oltra) |
| 167 |
- Fixes for various PowerPC issues (Doug Nazar) |
| 168 |
- Enable flush-to-zero mode for float programs on ARM/neon (Doug Nazar) |
| 169 |
- Fix some opcodes to support x2/x4 processing on PowerPC (Doug Nazar) |
| 170 |
|
| 171 |
0.4.30 |
| 172 |
====== |
| 173 |
|
| 174 |
- Don't always generate static library but default to shared-only (Xavier Claessens) |
| 175 |
- Work around false positives in Microsoft UWP certification kit (Nirbheek Chauhan) |
| 176 |
- Add endbr32/endbr64 instructions on x86/x86-64 for indirect branch tracking (Wim Taymans) |
| 177 |
- Fix gtk-doc build when orc is used as a meson subproject (Mathieu Duponchelle) |
| 178 |
- Switch float comparison in tests to ULP method to fix spurious failures (Doug Nazar) |
| 179 |
- Fix flushing of ARM icache when using dual map (Doug Nazar) |
| 180 |
- Use float constants/parameters when testing float opcodes (Doug Nazar) |
| 181 |
- Add support for Hygon Dhyana processor (fanjinke) |
| 182 |
- Fix PPC/PPC64 CPU family detection (Doug Nazar) |
| 183 |
- Add little-endian PPC support (Doug Nazar) |
| 184 |
- Fix compiler warnings with clang (Matthew Waters) |
| 185 |
- Mark exec mapping writable in debug mode for allowing breakpoints (Doug Nazar) |
| 186 |
- Various codegen refactorings (Doug Nazar) |
| 187 |
- autotools support has been dropped in favour of Meson as build system (Tim-Philipp Müller) |
| 188 |
- Fix PPC CPU feature detection and add support for VSX/v2.07 (Doug Nazar) |
| 189 |
- Add double/int64 support for PPC (Doug Nazar) |
| 190 |
|
| 191 |
0.4.29 |
| 192 |
====== |
| 193 |
|
| 194 |
- PowerPC: Support ELFv2 ABI (A. Wilcox) and ppc64le (Michel Normand) |
| 195 |
- Mips backend: only enable if the DSPr2 ASE is present (James Cowgill) |
| 196 |
- Windows and MSVC build fixes (Nirbheek Chauhan, Tim-Philipp Müller) |
| 197 |
- orccpu-arm: Allow 'cpuinfo' fallback on non-android (Edward Hervey) |
| 198 |
- pkg-config file for orc-test library (Tim-Philipp Müller) |
| 199 |
- orcc: add --decorator command line argument to add function decorators |
| 200 |
in header files (Tim-Philipp Müller) |
| 201 |
- meson: Make orcc detectable from other subprojects (Seungha Yang) |
| 202 |
- meson: add options to disable tests, docs, benchmarks, examples, |
| 203 |
tools, etc. (Sebastian Dröge, Tim-Philipp Müller) |
| 204 |
- meson: misc. other fixes (James Cowgill, Nirbheek Chauhan, Sebastian Dröge) |
| 205 |
|
| 206 |
0.4.28 |
| 207 |
====== |
| 208 |
|
| 209 |
- Numerous undefined behaviour fixes (Edward Hervey) |
| 210 |
- Ability to disable tests (Edward Hervey) |
| 211 |
- Fix meson dist behaviour (Tim-Philipp Müller) |
| 212 |
|
| 213 |
0.4.27 |
| 214 |
====== |
| 215 |
|
| 216 |
- sse: preserve non volatile sse registers, needed for MSVC (Matej Knopp) |
| 217 |
- x86: don't hard-code register size to zero in orc_x86_emit_*() functions (Igor Rondarev) |
| 218 |
- Fix incorrect asm generation on 64-bit Windows when building with MSVC (Jan Schmidt) |
| 219 |
- Support build using the Meson build system (Nirbheek Chauhan, Tim-Philipp Müller) |
| 220 |
|
| 221 |
0.4.26 |
| 222 |
====== |
| 223 |
|
| 224 |
- Use 64 bit arithmetic to increment the stride if needed (Wim Taymans) |
| 225 |
- Fix generation of ModR/M / SIB bytes for the EBP, R12, R13 registers |
| 226 |
on X86/X86-64 (Sebastian Dröge) |
| 227 |
- Fix test_parse unit test if no executable backend is available (Pascal Terjan) |
| 228 |
- Add orc-test path to the -uninstalled .pc file (Josep Torra) |
| 229 |
- Fix compiler warnings in the tests on OS X (Josep Torra) |
| 230 |
|
| 231 |
0.4.25 |
| 232 |
====== |
| 233 |
|
| 234 |
- compiler: also prefer the backup function when no target, instead |
| 235 |
of trying to use emulation which is usually slower (Wim Taymans) |
| 236 |
- executor: fix load of parameters smaller than 64 bits, fixing crashes |
| 237 |
on ldresnearb and friends in emulated code (Wim Taymans) |
| 238 |
- test-limits: improve test without target (Wim Taymans) |
| 239 |
- Only check for Android's liblog on Android targets, so we don't accidentally |
| 240 |
pick up another liblog that may exist elsewhere (Sebastian Dröge) |
| 241 |
- Don't require libtool for uninstalled setups (-uninstalled pkg-config file) |
| 242 |
(Julien Isorce) |
| 243 |
- Make -Bsymbolic check in configure work with clang (Koop Mast) |
| 244 |
- Coverity code analyser fixes (Luis de Bethencourt) |
| 245 |
- docs: update generated opcode tables |
| 246 |
- add orc_version_string() function and make orcc check the liborc that is |
| 247 |
being picked up to make sure the right lib is being used (Tim-Philipp Müller) |
| 248 |
|
| 249 |
0.4.24 |
| 250 |
====== |
| 251 |
|
| 252 |
- Only reuse constants of the same size and value (Wim Taymans) |
| 253 |
- Fix reading of .orc files with Windows line endings on |
| 254 |
Windows (Tim-Philipp Müller) |
| 255 |
- Fix out of bounds array access in the tests (Luis de Bethencourt) |
| 256 |
- Remove duplicate code path in orcc (Edward Hervey) |
| 257 |
- Put a limit to the memcpy test (Edward Hervey) |
| 258 |
- Fix mmap leak on error path (Vincent Penquerc'h) |
| 259 |
|
| 260 |
0.4.23 |
| 261 |
====== |
| 262 |
|
| 263 |
- Various improvements to the NEON backend to bring it closer to the SSE |
| 264 |
backend (Wim Waymans) |
| 265 |
- Add support for setting a custom backup function (Wim Taymans) |
| 266 |
- Preserve NEON/VFP registers across subroutines (Jerome Laheurte) |
| 267 |
- Fix 64 bit parameter loading on big-endian systems (Tim-Philipp Müller) |
| 268 |
- Improved implementations for various opcodes (Wim Taymans) |
| 269 |
- Various improvements and fixes to constants handling (Wim Taymans) |
| 270 |
- Avoid some undefined operations on signed integers (Wim Taymans) |
| 271 |
- Prefer user specific directories over global ones for intermediate files |
| 272 |
to prevent name collisions (Fabian Deutsch) |
| 273 |
|
| 274 |
0.4.22 |
| 275 |
====== |
| 276 |
|
| 277 |
Maintenance release: |
| 278 |
|
| 279 |
- Handle NOCONFIGURE=1 in autogen.sh (Colin Walters) |
| 280 |
- Some memory leak fixes in the compiler (Sebastian Dröge, Thiago Santos) |
| 281 |
- Fixes for compiler warnings on Win64 (Edward Hervey) |
| 282 |
- Properly detect CPU features on Android in non-debug build (Jan Schmidt) |
| 283 |
- Use Android logging system instead of stderr for debug output (Jan Schmidt) |
| 284 |
|
| 285 |
0.4.21 |
| 286 |
====== |
| 287 |
|
| 288 |
Maintenance release: |
| 289 |
|
| 290 |
- Add libtool versioning to the linker flags again. This was accidentially |
| 291 |
removed in 0.4.20 but should not cause any problems on platforms other |
| 292 |
than OS X (Sebastian Dröge) |
| 293 |
|
| 294 |
|
| 295 |
0.4.20 |
| 296 |
====== |
| 297 |
|
| 298 |
Maintenance release: |
| 299 |
|
| 300 |
- Fix list corruption when splitting code memory chunks, causing crashes |
| 301 |
when allocating a lot of code memory and trying to free it later |
| 302 |
(Tim-Philipp Müller) |
| 303 |
- Add some extra checks for the number of variables used in ORC code to |
| 304 |
prevent overflows and crashes in the compiler (Vincent Penquerc'h) |
| 305 |
- Various compiler warnings, coverity warnings and static code analysis |
| 306 |
fixes (Sebastian Dröge) |
| 307 |
|
| 308 |
0.4.19 |
| 309 |
====== |
| 310 |
|
| 311 |
Maintenance release: |
| 312 |
|
| 313 |
- Fix out-of-tree builds (Edward Hervey) |
| 314 |
- Fix many memory leaks, compiler warnings and coverity warnings (Tim-Philipp Müller, |
| 315 |
Olivier Crête, Todd Agulnick, Sebastian Dröge, Vincent Penquerc'h, Edward Hervey) |
| 316 |
- Documentation fix for mulhsw, mulhuw (William Manley) |
| 317 |
|
| 318 |
0.4.18 |
| 319 |
====== |
| 320 |
|
| 321 |
Maintenance release: |
| 322 |
|
| 323 |
- Important bugfix in reading constants from bytecode. (Tim-Philipp Müller |
| 324 |
and Sebastian Dröge) |
| 325 |
- Documentation and code cleanup (Stefan Sauer) |
| 326 |
- Fix cache flushing on iOS (Andoni Morales Alastruey) |
| 327 |
|
| 328 |
|
| 329 |
0.4.17 |
| 330 |
====== |
| 331 |
|
| 332 |
Maintenance release: |
| 333 |
|
| 334 |
- Merged known distro patches. |
| 335 |
- Added MIPS backend (Guillaume Emont). |
| 336 |
- Disabled ARM backend because of poor coverage. |
| 337 |
- Added bytecode parsing and writing. This can be used instead of |
| 338 |
manual creation of OrcPrograms. |
| 339 |
|
| 340 |
|
| 341 |
0.4.16 |
| 342 |
====== |
| 343 |
|
| 344 |
Fix a few bugs people noticed in 0.4.15. |
| 345 |
|
| 346 |
- orc_init() tried to take the same mutex as generated C code that |
| 347 |
calls (indirectly) orc_init(). |
| 348 |
- sse: Fixes for 64 bit pointers with any of the upper 32 bits set. |
| 349 |
|
| 350 |
|
| 351 |
0.4.15 |
| 352 |
====== |
| 353 |
|
| 354 |
This should have been release much earlier. |
| 355 |
|
| 356 |
- Protect global resources with mutexes. Duh. This solves a bunch |
| 357 |
of bug reports. |
| 358 |
- Restore c64x-c backend. Untested. |
| 359 |
- Convert MMX and SSE backends to a new instruction scheduler. |
| 360 |
- Add alignment and size hints to parser. |
| 361 |
|
| 362 |
|
| 363 |
0.4.14 |
| 364 |
====== |
| 365 |
|
| 366 |
Yet more bug fixing. Altivec should work again, OS/X should |
| 367 |
work again. MMX should work again. Another codegen bug on |
| 368 |
SSE fixed. |
| 369 |
|
| 370 |
|
| 371 |
0.4.13 |
| 372 |
====== |
| 373 |
|
| 374 |
Fixes two serious code generation bugs in 0.4.12 on SSE and |
| 375 |
Altivec. Also added some compatibility code to mitigate |
| 376 |
the previous automatic inclusion of stdint.h. |
| 377 |
|
| 378 |
|
| 379 |
0.4.12 |
| 380 |
====== |
| 381 |
|
| 382 |
This is primarily a bug fixing release. |
| 383 |
|
| 384 |
- Fix gcc-4.6 warnings in generated code |
| 385 |
- Codegen fixes for Altivec. Passes regression tests again. |
| 386 |
- More error checking for code allocation. |
| 387 |
- NEON: floating point improvements |
| 388 |
- Removed stdint.h from API. This could theoretically cause |
| 389 |
breakage if you depended on orc to include stdint.h. |
| 390 |
|
| 391 |
One new feature is the OrcCode structure, which keeps track of |
| 392 |
compiled code. This now allows applications to free unused code. |
| 393 |
|
| 394 |
Internally, x86 code generation was completely refactored to add |
| 395 |
an intermediate stage, which will later be used for instruction |
| 396 |
reordering. None of this is useful yet. |
| 397 |
|
| 398 |
|
| 399 |
0.4.11 |
| 400 |
====== |
| 401 |
|
| 402 |
This is primarily a bug fixing release. |
| 403 |
|
| 404 |
- Fixes for CPUs that don't have backends |
| 405 |
- Fix loading of double parameters |
| 406 |
- mmx: Fix 64-bit parameter loading |
| 407 |
- sse/mmx: Fix x2/x4 with certain opcodes |
| 408 |
|
| 409 |
There are still some issues with the ARM backend on certain |
| 410 |
architecture levels (especially ARMv6). Some assistance from |
| 411 |
a user with access to such hardware would be useful. |
| 412 |
|
| 413 |
|
| 414 |
0.4.10 |
| 415 |
====== |
| 416 |
|
| 417 |
Changes: |
| 418 |
|
| 419 |
- Added several simple 64-bit opcodes |
| 420 |
- Improved debugging by adding ORC_CODE=emulate |
| 421 |
- Allocation of mmap'd areas for code now has several fallback |
| 422 |
methods, in order to placate various SELinux configurations. |
| 423 |
- Various speed improvements in SSE backend |
| 424 |
- Add SSE implementations of ldreslinl and ldresnearl. |
| 425 |
- Update Mersenne Twister example |
| 426 |
|
| 427 |
There was a bug in the calculation of maximum loop shift that, when |
| 428 |
fixed, increases the speed of certain functions by a factor of two. |
| 429 |
However, the fix also triggers a bug in Schroedinger, which is fixed |
| 430 |
in the 1.0.10 release. |
| 431 |
|
| 432 |
|
| 433 |
0.4.9 |
| 434 |
===== |
| 435 |
|
| 436 |
This is primarily a bug fixing release. |
| 437 |
|
| 438 |
Changes: |
| 439 |
|
| 440 |
- Added handling for 64-bit constants |
| 441 |
- Fix building and use of static library |
| 442 |
- Fix register allocation on Win64 (still partly broken, however) |
| 443 |
- Quiet some non-errors printed by orcc in 0.4.8. |
| 444 |
- Fix implementation of several opcodes. |
| 445 |
|
| 446 |
Until this release, the shared libraries all had the same versioning |
| 447 |
information. This should be fixed going forward. |
| 448 |
|
| 449 |
|
| 450 |
0.4.8 |
| 451 |
===== |
| 452 |
|
| 453 |
Changes: |
| 454 |
|
| 455 |
- Fix Windows and OS/X builds |
| 456 |
- Improve behavior in failure cases |
| 457 |
- Major improvements for Altivec backend |
| 458 |
- Significant documentation additions |
| 459 |
|
| 460 |
Memory for executable code storage is now handled in a much more |
| 461 |
controlled manner, and it's now possible to reclaim this memory |
| 462 |
after it's no longer needed. |
| 463 |
|
| 464 |
A few more 64-bit opcodes have been added, mostly related to |
| 465 |
arithmetic on floating point values. |
| 466 |
|
| 467 |
The orcc tool now handles 64-bit and floating point parameters |
| 468 |
and constants. |
| 469 |
|
| 470 |
|
| 471 |
0.4.7 |
| 472 |
===== |
| 473 |
|
| 474 |
Changes: |
| 475 |
|
| 476 |
- Lots of specialized new opcodes and opcode prefixes. |
| 477 |
- Important fixes for ARM backend |
| 478 |
- Improved emulation of programs (much faster) |
| 479 |
- Implemented fallback rules for almost all opcodes for |
| 480 |
SSE and NEON backends |
| 481 |
- Performance improvements for SSE and NEON backends. |
| 482 |
- Many fixes to make larger programs compile properly. |
| 483 |
- 64-bit data types are now fully implemented, although |
| 484 |
there are few operations on them. |
| 485 |
|
| 486 |
Loads and stores are now handled by separate opcodes (loadb, |
| 487 |
storeb, etc). For compatibility, these are automatically |
| 488 |
included where necessary. This allowed new specialized |
| 489 |
loading opcodes, for example, resampling a source array |
| 490 |
for use in scaling images. |
| 491 |
|
| 492 |
Opcodes may now be prefixed by "x2" or "x4", indicating that |
| 493 |
a operation should be done on 2 or 4 parts of a proportionally |
| 494 |
larger value. For example, "x4 addusb" performs 4 saturated |
| 495 |
unsigned additions on each of the four bytes of 32-bit |
| 496 |
quantities. This is useful in pixel operations. |
| 497 |
|
| 498 |
The MMX backend is now (semi-) automatically generated from |
| 499 |
the SSE backend. |
| 500 |
|
| 501 |
The orcc tool has a new option "--inline", which creates inline |
| 502 |
versions of the Orc stub functions. The orcc tool also recognizes |
| 503 |
a new directive '.init', which instructs the compiler to generate |
| 504 |
an initialization function, which when called at application init |
| 505 |
time, compiles all the generated functions. This allows the |
| 506 |
generated stub functions to avoid checking if the function has |
| 507 |
already been compiled. The use of these two features can |
| 508 |
dramatically decrease the cost of calling Orc functions. |
| 509 |
|
| 510 |
Known Bugs: Orc generates code that crashes on 64-bit OS/X. |
| 511 |
|
| 512 |
Plans for 0.4.8: (was 2.5 for 4 this time around, not too bad!) |
| 513 |
Document all the new features in 0.4.7. Instruction scheduler. |
| 514 |
Code and API cleanup. |
| 515 |
|
| 516 |
|
| 517 |
|
| 518 |
0.4.6 |
| 519 |
===== |
| 520 |
|
| 521 |
Changes: |
| 522 |
|
| 523 |
- Various fixes to make Orc more portable |
| 524 |
- Major performance improvements to NEON backend |
| 525 |
- Minor performance improvements to SSE backend |
| 526 |
- Major improvements to ARM backend, now passes regression |
| 527 |
tests. |
| 528 |
|
| 529 |
The defaults for floating point operations have been changed |
| 530 |
somewhat: NANs are handled more like the IEEE 754 standard, |
| 531 |
and denormals in operations are treated as zeros. The NAN |
| 532 |
changes causes certain SSE operations to be slightly slower, |
| 533 |
but produce less surprising results. Treating denormals as |
| 534 |
zero has effects ranging from "slightly faster" to "now possible". |
| 535 |
|
| 536 |
New tool: orc-bugreport. Mainly this is to provide a limited |
| 537 |
testing tool in the field, especially for embedded targets |
| 538 |
which would not have access to the testsuite that is not |
| 539 |
installed. |
| 540 |
|
| 541 |
The environment variable ORC_CODE can now be used to adjust |
| 542 |
some code generation. See orc-bugreport --help for details. |
| 543 |
|
| 544 |
orcc has a new option to generate code that is compatible |
| 545 |
with older versions of Orc. For example, if your software |
| 546 |
package only uses 0.4.5 features, you can use --compat 0.4.5 |
| 547 |
to generate code that run on 0.4.5, otherwise it may generate |
| 548 |
code that requires 0.4.6. Useful for generating source code |
| 549 |
for distribution. |
| 550 |
|
| 551 |
New NEON detection relies on Linux 2.6.29 or later. |
| 552 |
|
| 553 |
Plans for 0.4.7: (not that past predictions have been at all |
| 554 |
accurate) New opcodes for FIR filtering, scaling and compositing |
| 555 |
of images and video. Instruction scheduler, helpful for non-OOO |
| 556 |
CPUs. Minor SSE/NEON improvements. Orcc generation of inline |
| 557 |
macros. |
| 558 |
|
| 559 |
|
| 560 |
0.4.5 |
| 561 |
===== |
| 562 |
|
| 563 |
This release contains many small improvements related to |
| 564 |
converting GStreamer from liboil to Orc. |
| 565 |
|
| 566 |
The major addition in this release is the mainstreaming of |
| 567 |
the NEON backend, made possible by Nokia. |
| 568 |
|
| 569 |
There is a new experimental option to ./configure, |
| 570 |
--enable-backend, which allows you to choose a single code |
| 571 |
generation backend to include in the library. This is mostly |
| 572 |
useful for embedded systems, and is not recommended in general. |
| 573 |
|
| 574 |
The upcoming release will focus on improving code generation |
| 575 |
for the SSE and NEON backends. |
| 576 |
|
| 577 |
|
| 578 |
0.4.4 |
| 579 |
===== |
| 580 |
|
| 581 |
This is almost entirely a cleanup and bug fix release. |
| 582 |
|
| 583 |
- fix register copying on x86-64 |
| 584 |
- better checking for partial test failures |
| 585 |
- fix documention build |
| 586 |
- fix build on many systems I don't personally use |
| 587 |
- various fixes to build/run on Win64 (Ramiro Polla) |
| 588 |
- add performance tests |
| 589 |
|
| 590 |
Next release will merge in the new pixel compositing opcodes |
| 591 |
and the SSE instruction scheduler. |
| 592 |
|
| 593 |
|
| 594 |
0.4.3 |
| 595 |
===== |
| 596 |
|
| 597 |
New opcodes: all the 32-bit float opcodes from the orc-float |
| 598 |
library have been moved into the core library. |
| 599 |
|
| 600 |
New opcodes: splitlw and splitwb, which are equivalent to |
| 601 |
select0lw, select1lw, select0wb, and select1wb, except that |
| 602 |
the new opcodes split a value into two destinations in one |
| 603 |
opcode. |
| 604 |
|
| 605 |
New backend: c64x-c, for the TI C64x+ DSP. This backend only |
| 606 |
produces source code, unlike other backends which can produce |
| 607 |
both source and binary code. Generating code for this backend |
| 608 |
can be done using 'orcc --assembly --target=c64x-c'. |
| 609 |
|
| 610 |
Orc now understands and can generate code for two-dimensional |
| 611 |
arrays. If the size of the array is known at compile time, |
| 612 |
this information can be used to improve generated code. |
| 613 |
|
| 614 |
Various improvements to the ARM backend by Wim Taymans. The |
| 615 |
ARM backend is still experimental. |
| 616 |
|
| 617 |
|
| 618 |
0.4.2 |
| 619 |
===== |
| 620 |
|
| 621 |
Bug fixes to C backend. Turns out this is rather important on |
| 622 |
CPUs that don't have a native backend. |
| 623 |
|
| 624 |
New features have been postponed to 0.4.3. |
| 625 |
|
| 626 |
|
| 627 |
0.4.1 |
| 628 |
===== |
| 629 |
|
| 630 |
This release introduces the orcc program, which parses .orc files and |
| 631 |
outputs C source files to compile into a library or application. The |
| 632 |
main source file implements the functions described by the .orc source |
| 633 |
code, by creating Orc programs and compiling them at runtime. Another |
| 634 |
source file that it outputs is a test program that can be compiled and |
| 635 |
run to determine if Orc is generating code correctly. In future |
| 636 |
releases, the orcc tool will be expanded to output assembly code, as |
| 637 |
well as make it easier to use Orc in a variety of ways. |
| 638 |
|
| 639 |
Much of Schroedinger and GStreamer have been converted to use Orc |
| 640 |
instead of liboil, as well as converting code that wasn't able to |
| 641 |
use liboil. To enable this in Schroedinger, use the --enable-orc |
| 642 |
configure option. The GStreamer changes are in the orc branch in |
| 643 |
the repository at http://cgit.freedesktop.org/~ds/gstreamer |
| 644 |
|
| 645 |
Scheduled changes for 0.4.2 include a 2-D array mode for converting |
| 646 |
the remaining liboil functions used in Schroedinger and GStreamer. |
| 647 |
|
| 648 |
|
| 649 |
Major changes: |
| 650 |
|
| 651 |
- Add the orcc compiler. Generates C code that creates Orc programs |
| 652 |
from .orc source files. |
| 653 |
- Improved testing |
| 654 |
- Fixes in the C backend |
| 655 |
- Fix the MMX backend to emit 'emms' instructions. |
| 656 |
- Add a few rules to the SSE backend. |
| 657 |
|
| 658 |
|
| 659 |
|
| 660 |
0.4.0 |
| 661 |
===== |
| 662 |
|
| 663 |
Stuff happened. |
| 664 |
|