Taichi

Highlights:
- **Examples**
- Add mciso.py and mciso_advanced.py for marching cube (1835) (by **彭于斌**)

Full changelog:
- [cli] Setup pylint: run 'ti lint' to obtain linting results (1846) (by **彭于斌**)
- [async] More SFG optimizations (1877) (by **Yuanming Hu**)
- [Example] Add mciso.py and mciso_advanced.py for marching cube (1835) (by **彭于斌**)

Highlights:
- **Language and syntax**
- Support num_thread limitation and serial execution on CPU (1867) (by **Yuanming Hu**)

Full changelog:
- [refactor] Create class TypePromotionMapping for refactoring function promoted_type in lang_util.cpp (1874) (by **Jiafeng Liu**)
- [async] Draw nodes as record shape, allow embedding states into the nodes. (1876) (by **Ye Kuang**)
- [Lang] Support num_thread limitation and serial excution on CPU (1867) (by **Yuanming Hu**)
- [async] Cache fusion results (1875) (by **xumingkuan**)
- [Async] Fix ClearListStmt metadata (1873) (by **Yuanming Hu**)
- [async] Improve fusion optimization and add StateFlowGraph::verify (1872) (by **xumingkuan**)
- [async] Optimizations on StateFlowGraph (1870) (by **Yuanming Hu**)
- [async] Migrate AsyncEngine::fuse() to StateFlowGraph (1859) (by **xumingkuan**)
- [ir] Add ClearListStmt instead of making clear_list a unique task type (1868) (by **Ye Kuang**)
- [error] [doc] Better error message when X11 environment failed to initialize (1856) (by **彭于斌**)
- [async] Highlight user kernels in the DOT graph, allows configuring rankdir (1866) (by **Ye Kuang**)

Highlights:
- **Examples**
- Make ODOP MGPCG class in mgpcg.py generic and reuse it in stable_fluid.py (1731) (by **彭于斌**)
- **Language and syntax**
- Support ti.zero(x) and ti.one(x) for making zero/one value for the same type with x (1857) (by **彭于斌**)
- **GitHub Actions/Workflows**
- Build and test the CUDA backend using GitHub actions (1863) (by **Yuanming Hu**)

Full changelog:
- [Workflow] Build and test the CUDA backend using GitHub actions (1863) (by **Yuanming Hu**)
- [async] Patch StateFlowGraph (1862) (by **Yuanming Hu**)
- [Example] Make ODOP MGPCG class in mgpcg.py generic and reuse it in stable_fluid.py (1731) (by **彭于斌**)
- [Lang] Support ti.zero(x) and ti.one(x) for making zero/one value for the same type with x (1857) (by **彭于斌**)

Highlights:
- **Error messages**
- Raise TaichiSyntaxError when multiple return path detected (1844) (by **彭于斌**)
- **Examples**
- Physarum simulation example (1850) (by **Alexander Mordvintsev**)
- **Language and syntax**
- Support __len__ and __iter__ on ti.Matrix (1838) (by **彭于斌**)

Full changelog:
- [async] Add TaskLaunchRecord StateFlowGraph::Node::rec (1854) (by **xumingkuan**)
- [Error] Raise TaichiSyntaxError when multiple return path detected (1844) (by **彭于斌**)
- [async] Fix missing edges and nodes in SFG (1853) (by **Yuanming Hu**)
- [pypi] Remove setuptools from hard dependencies (1845) (by **彭于斌**)
- [Example] Physarum simulation example (1850) (by **Alexander Mordvintsev**)
- [workflow] Update issue templates (1794) (by **彭于斌**)
- [async] Generate DOT graph for StateFlowGraph (1852) (by **Ye Kuang**)
- [async] Implement IRBank and IRHandle (1839) (by **xumingkuan**)
- [misc] Fix check previous run by using GITHUB_TOKEN to authenticate APIs (1849) (by **Ye Kuang**)
- [cuda] Fix CUDA driver API signature (1834) (by **Yuanming Hu**)
- [async] Improve StateFlowGraph (1843) (by **Yuanming Hu**)
- [ir] Remove the usage of kernel from type_check() pass (1848) (by **Ye Kuang**)
- [misc] Check the previous Github workflow run status if CI is skipped (1837) (by **Ye Kuang**)
- [Lang] Support __len__ and __iter__ on ti.Matrix (1838) (by **彭于斌**)

Highlights:
- **Documentation**
- Doc the workflow for exporting Taichi kernels to C99 source (1756) (by **彭于斌**)
- Add performance.rst introducing the 'ti.block_dim' API (1691) (by **彭于斌**)
- Add document for XYZW accessors of vector (1797) (by **彭于斌**)
- **GUI**
- Support saving image without GUI showing a window (1817) (by **thinking-tower**)
- **Refactor**
- Refactor Block::parent from Block* to Stmt* (1815) (by **xumingkuan**)

Full changelog:
- [refactor] Use promise/future to retrieve the function compiled asynchronously (1831) (by **Ye Kuang**)
- [GUI] Support saving image without GUI showing a window (1817) (by **thinking-tower**)
- [ir] [bug] Fix demote_operations lacking type_check (1825) (by **彭于斌**)
- [refactor] Rename KernelLaunchRecord to TaskLaunchRecord (1828) (by **Ye Kuang**)
- [Doc] [cc] Doc the workflow for exporting Taichi kernels to C99 source (1756) (by **彭于斌**)
- [Doc] Add performance.rst introducing the 'ti.block_dim' API (1691) (by **彭于斌**)
- [Doc] Add document for XYZW accessors of vector (1797) (by **彭于斌**)
- [ir] [bug] Demote all local atomic operations, not only add (1819) (by **彭于斌**)
- [doc] Document data structure organization (995) (by **Yuanming Hu**)
- [ir] [refactor] Reorder FrontAssertStatement's constructor args (1822) (by **Ye Kuang**)
- [Refactor] [ir] Refactor Block::parent from Block* to Stmt* (1815) (by **xumingkuan**)
- [refactor] Rename 'launched_kernels*' stats to 'launched_tasks*' (1821) (by **Ye Kuang**)

Highlights:
- **Bug fixes**
- Use sys.getfilesystemencoding() to prevent possible locale errors (1775) (by **彭于斌**)
- **C source backend**
- Support AdStack on C backend (1752) (by **彭于斌**)
- **Documentation**
- Add how to set environment variables for fish users (1800) (by **Luyu Cheng**)
- Separate arithmetics.rst from syntax.rst and fill it with more details (1761) (by **彭于斌**)
- **Examples**
- Fix stable_fluid.py boundary broken: int -> ti.floor, and make mpm3d.py 3x faster on CUDA (1784) (by **彭于斌**)
- **Language and syntax**
- Support formatted string with args in assert (1806) (by **Ye Kuang**)
- Support printing lists and tuples in Taichi-scope (1796) (by **彭于斌**)
- Add ti.get_rel_eps() for maximal relative error tolerance on current backend (1798) (by **Ye Kuang**)
- Support querying external array shape by 'x.shape' (1787) (by **thinking-tower**)
- **IR optimization passes**
- Make irpass::replace_all_usages_with bottom-up (1789) (by **xumingkuan**)

Full changelog:
- [misc] Remove unused code/comments in transforms/simpilfy.cpp (1812) (by **Yuanming Hu**)
- [ir] Fix irpass::analysis::clone on TLS/BLS prologue/epilogue (1811) (by **Yuanming Hu**)
- [error] "Better error message when vector used as if's condition variable (1804)" (1810) (by **彭于斌**)
- [Lang] Support formatted string with args in assert (1806) (by **Ye Kuang**)
- [refactor] Replace std::map with switch case statements in lang_util.cpp (1755) (by **Aryan Soman**)
- [misc] Revert "Better error message when vector used as if's condition variable (1804)" (1808) (by **Yuanming Hu**)
- [error] Better error message when vector used as if's condition variable (1804) (by **彭于斌**)
- [Opt] Make irpass::replace_all_usages_with bottom-up (1789) (by **xumingkuan**)
- [Doc] Add how to set environment variables for fish users (1800) (by **Luyu Cheng**)
- [CC] [autodiff] Support AdStack on C backend (1752) (by **彭于斌**)
- [Lang] Support printing lists and tuples in Taichi-scope (1796) (by **彭于斌**)
- [workflow] import missing package sys in code_format.py (1802) (by **Luyu Cheng**)
- [cli] Add path to Python executable to 'ti diagnose' (1801) (by **Luyu Cheng**)
- [refactor] Remove the singleton Program::context (1799) (by **Ye Kuang**)
- [ir] Restore verification for OffloadedStmt (1790) (by **xumingkuan**)
- [Lang] [test] Add ti.get_rel_eps() for maximal relative error tolerance on current backend (1798) (by **Ye Kuang**)
- [Lang] [ir] Support querying external array shape by 'x.shape' (1787) (by **thinking-tower**)
- [workflow] Travis CI only test Python 3.6 unless release tag is used (1793) (by **彭于斌**)
- [refactor] Move Context to taichi/program (1788) (by **Ye Kuang**)
- [doc] Better clarify the difference between two profilers (1759) (by **彭于斌**)
- [ir] [autodiff] Initialize ADStack with a zero (1791) (by **Yuanming Hu**)
- [ir] [opt] Unify ifloordiv implementation over backends in demote_operations (1771) (by **彭于斌**)
- [Bug] [misc] Use sys.getfilesystemencoding() to prevent possible locale errors (1775) (by **彭于斌**)
- [Example] Fix stable_fluid.py boundary broken: int -> ti.floor, and make mpm3d.py 3x faster on CUDA (1784) (by **彭于斌**)
- [test] Expand the tolerance for testing on metal backend (1779) (by **Xudong Feng**)
- [opt] Simplify identical load elimination (1782) (by **xumingkuan**)
- [error] [opengl] Better error message when int64 or sparse SNode not supported for OpenGL (1783) (by **彭于斌**)
- [Doc] Separate arithmetics.rst from syntax.rst and fill it with more details (1761) (by **彭于斌**)
- [ipython] Separate {Expr,Matrix}.{__str__,__repr__} to prevent unexpected materialization when repr (1766) (by **彭于斌**)
- [async] Support constant folding in async mode (1778) (by **Ye Kuang**)
- [workflow] [cli] Improve coverage of 'ti format' and 'ti format_all' (1768) (by **彭于斌**)
- [misc] Add "make_block_local" option to "CompileConfig" (1727) (by **Yuanming Hu**)