Tags: EmmanuelBoidot/Project-OSRM
Tags
This is the shortened change log: - Windows support: Fetch the latest develop branch build from http://build.project-osrm.org - MinGW compiler support - Parallelization now uses Intel Thread Building blocks, bringing feature parity to all supported platforms - Significant query speed improvements, about twice as fast - The distance table feature does not need any zoom level anymore. Finds best nearest edge automatically - First compressed and succinct data structure to improve memory footprint - Improved and faster heuristics for polyline generalization - Fixes a resource leak on shutdown - Shared-Memory parallel sorting where applicable - Lots of smaller and bigger changes to the code base - Lots of code improvements thanks to C++11 and static analysis tools - Better control the number of used threads during extraction and preprocessing - Updated NodeJS bindings [2] supporting the new distance table API endpoint
This is the shortened change log: - new API call: efficient distance table computation - use 100 locations at max for dist table (for now) - JSON output generation through a dedicated JSON container - deactivate LTO on debug build - fix performance regression in preprocessing - use static assertions for better source code correctness - add some static asserts to guard against memory usage regressions - do less work when compressing geometries - move most geometric calculations to float - replace boost::unordered_map, move hash function for pairs into its own header - refactor route name extraction into its own class, fix name extraction - further includes untangling - introduce range based for loops to traverse graphs - further boost by STL replacements - use lookup tables from some trigonometry functions - fix compile regression in debug build - further code improvements and bug fixes Please not the following (minor) breaking changes: - the JSON formatted output now omits empty, i.e. unused fields - the response file names are now response.js(on) or route.gpx
gain a pretty big release with more than 150 commits: - modernized code base to use C++11 language features - decluttered the code base touching virtually every line of code - cut down on static analyzer warnings - much better linking performance on GCC 4.9 - fixed extraction on 32 bit systems (while officially unsupported)
This is again a pretty big release with more than 170 commits (adding… … about 1500 LOCs). And it is the first release in a series to bring major improvements in memory usage. Depending on the size and location of the extract, you can expect 25--45% improvements. This is the shortened change log: - compress all road geometries (49% on the planet for cars) - make all error messages JSON - activated link time optimizations by default - more test scenarios (25+) - use forward decl's to cut compile times - ongoing effort to make code more legible - several lints fixed that were detected by facebooks flint - several lints fixed that were detected by Coveritys static analyzer - several lints fixed that were detected by OCLint - remove remnants of C-Style - add clang-format style file
The release is the result of hard work of all contributors and comes … …at a hefty count of 278 commits. The shortened changelog is this: - Fix performance regression in JSON generation - Update and enhance test suite - Activate link time optimizations - Update speed profiles to be more realistic - CMake now adds RPATH info to executables - Refactored some code for better readability - Removed pesky function ptr from CRC32 computation - Enhanced feedback on command line options - Escape characters in JSON instead of HTML-entetizing - Decode URI-encoded requests - Fix a possible path quality regression - General support for non-X86(-64) targets, e.g. Raspberry Pi - Remove remaining C-style function calls - Fix minor GPX export issue - Fix resource leak, coverity issue #1121924 - Generate pkgconfig file - Compile and install libOSRM as static lib - Lint removal in CMakeLists.txt - Refactor perpendicular projection to nearest location - Untangling includes to cut down compile time - More precise line generalizations - Make nearest neighbor queries more efficient - And many, many more smaller and bigger changes
PreviousNext