v1.8.0

[shi-eric]

Changelog

[1.8.0] - 2025-07-01

Added

• Add wp.map() to map a function over arrays and add math operators for Warp arrays (docs, [REQ] Add a wp.map() host-side API to apply a function to each element of an array #694).
• Add support for dynamic control flow in CUDA graphs, see wp.capture_if() and wp.capture_while() (docs, [REQ] Extend graph capture support to allow conditional branches #597).
• Add wp.capture_debug_dot_print() to write a DOT file describing the structure of a captured CUDA graph ([REQ] Add access to cudaGraphDebugDotPrint to visualize cuda graphs #746).
• Add the Device.sm_count property to get the number of streaming multiprocessors on a CUDA device ([REQ] Query number of SMs in the Device #584).
• Add wp.block_dim() to query the number of threads in the current block inside a kernel ([REQ] Allow block dimension to be retrieved in a kernel with wp.block_dim() #695).
• Add wp.atomic_cas() and wp.atomic_exch() built-ins for atomic compare-and-swap and exchange operations ([REQ] Expose atomicCAS and atomicExch #767).
• Add support for profiling GPU runtime module compilation using the global wp.config.compile_time_trace setting or the module-level "compile_time_trace" option. When used, JSON files in the Trace Event format will be written in the kernel cache, which can be opened in a viewer like chrome://tracing/ (docs, Add support for --fdevice-time-trace #609).
• Add support for returning multiple values from native functions like wp.svd3() and wp.quat_to_axis_angle() ([REQ] Allow calling multi-valued functions from Python (e.g.: wp.quat_to_axis_angle(), wp.svd3()) #503).
• Add support for passing tiles to user wp.func functions ([REQ] Allow tiles to be passed to user functions #682).
• Add wp.tile_squeeze() to remove axes of length one ([REQ] Add wp.tile_squeeze() #662).
• Add wp.tile_reshape() to reshape a tile ([REQ] Add wp.tile_reshape() #663).
• Add wp.tile_astype() to return a new tile with the same data but different data type. ([REQ] Tile API function to convert between data types #683).
• Add support for in-place tile add and subtract operations ([REQ] +=/-= assignment support for tile objects #518).
• Add support for in-place tile-component addition and subtraction ([REQ] Add support for in-place tile component addition and subtraction #659).
• Add support for 2D solves using wp.tile_cholesky_solve() (Update libmathdx, extend tile_cholesky_solve to 2D rhs, and improve trsm based tile operations #773).
• Add wp.tile_scan_inclusive() and wp.tile_scan_exclusive() for performing inclusive and exclusive scans over tiles ([REQ] Add support for tile_inclusive_scan and tile_exclusive_scan #731).
• Support attribute indexing for quaternions on the right-hand side of expressions ([REQ] Support quaternion indexing on the rhs of expressions #625).
• Add wp.transform_compose() and wp.transform_decompose() for converting between transforms and 4x4 matrices with 3D scale information ([REQ] Decompose a 4x4 transformation matrix to get the position, rotation (as quaternion), and scale. #576).
• Add various wp.transform syntax operations for loading and storing ([REQ] Add various wp.transform syntax operations for loading and storing #710).
• Add the as_spheres parameter to UsdRenderer.render_points() in order to choose whether to render the points as USD spheres using a point instancer or as simple USD points ([BUG] render_points does not render different colors #634).
• Add support for animating visibility of objects in the USD renderer ([QUESTION] Render object in certain frames only #598).
• Add wp.sim.VBDIntegrator.rebuild_bvh() to rebuild the BVH used for detecting self-contacts.
• Add damping terms wp.sim.VBDIntegrator collisions, with strength is controlled by Model.soft_contact_kd.
• Improve consistency of the wp.fem.lookup() operator across geometries and add filtering parameters ([QUESTION] How to apply displacement field to update mesh geometry in warp.fem? #618).
• Add two examples demonstrating shape optimization using warp.fem: fem/example_elastic_shape_optimization.py and fem/example_darcy_ls_optimization.py (warp.fem shape optimization example #698).
• Add a py.typed marker file (per PEP 561) to the package to formally support static type checking by downstream users (Add a py.typed file for PEP 561 compliance #780).

Removed

• Remove wp.mlp() (deprecated in v1.6.0). Use tile primitives instead.
• Remove wp.autograd.plot_kernel_jacobians() (deprecated in v1.4.0). Use wp.autograd.jacobian_plot() instead.
• Remove the length and owner keyword arguments from wp.array() constructor (deprecated in v1.6.0). Use the shape and deleter keywords instead.
• Remove the kernel keyword argument from wp.autograd.jacobian() and wp.autograd.jacobian_fd() (deprecated in v1.6.0). Use the function keyword argument instead.
• Remove the outputs keyword argument from wp.autograd.jacobian_plot() (deprecated in v1.6.0).

Changed

• Deprecate the warp.sim module (planned for removal in v1.10). It will be superseded by the upcoming Newton library, a separate package with a new API. Migrating will require code changes; a future guide will be provided (current draft). See the GitHub announcement for details (Deprecation Notice: warp.sim Module #735).
• Deprecate the wp.matrix(pos, quat, scale) built-in function. Use wp.transform_compose() instead ([REQ] Decompose a 4x4 transformation matrix to get the position, rotation (as quaternion), and scale. #576).
• Improve support for tuples in kernels ([REQ] Improved Python Tuple support #506).
• Return a constant value from len() where possible.
• Rename the internal function wp.types.type_length() to wp.types.type_size().
• Rename wp.tile_cholesky_solve() input parameters to align with its docstring ([REQ] Update Tile Cholesky naming scheme #726).
• Change wp.tile_upper_solve() and wp.tile_lower_solve() to use libmathdx 0.2.1 TRSM solver (Update libmathdx, extend tile_cholesky_solve to 2D rhs, and improve trsm based tile operations #773).
• Skip adjoint compilation for wp.tile_matmul() if enable_backward is disabled (Performance optimization: Skip compilation of LTO's for adjoint matmuls #644).
• Allow tile reductions to work with non-scalar tile types ([REQ] Allow tile reductions on non-scalar types #771).
• Permit data-type preservation with preserve_type=True when tiling a value across the block with wp.Tile() ([REQ] preserve type when tiling a value across the block #772).
• Make wp.sparse.bsr_[set_]from_triplets differentiable with respect to the input triplet values ([REQ] Make wp.sparse.bsr_from_triplets differentiable #760).
• Expose new warp.fem operators: node_count, node_index, element_coordinates, element_closest_point.
• Change wp.sim.VBDIntegrator rigid-body-contact handling to use only the shape's friction coefficient, rather than averaging the shape's and the cloth's coefficients.
• Limit usage of the wp.assign_copy() hidden built-in to the kernel scope.
• Describe the distinction between inputs and outputs arguments in the Kernel documentation.
• Reduce the overhead of wp.launch() by avoiding costly native API calls (Reduce cuda api overhead for kernel launch #774).
• Improve error reporting when calling @wp.func-decorated functions from the Python scope (Improve error reporting for calling user-defined functions from the Python scope #521).

Fixed

• Fix missing documentation for geometric structs ([DOCS] Document members of various built-in structs #674).
• Fix the type annotations in various tile functions ([BUG] Accurate tile argument type information in tile primitives documentation #714).
• Fix incorrect stride initialization in tiles returned from functions taking transposed tiles as input ([BUG] Tile copy assignment operator does not properly handle transposed tiles #722).
• Fix adjoint generation for user functions that return a tile ([BUG] Non-contiguous tile with Any typing in wp.func #749).
• Fix tile-based solvers failing to accept and return transposed tiles ([BUG] tile_matmul returns wrong result after tile_lower_solve with transpose #768).
• Fix the Formal parameter space overflowed error during wp.sim.VBDIntegrator kernel compilation for the backward pass in CUDA 11 Warp builds. This was resolved by decoupling collision and elasticity evaluations into separate kernels, increasing parallelism and speeding up the solver ([BUG] Compilation of adjoint kernel in warp.sim.integrator_vbd fails with CUDA 11 #442).
• Fix an issue with graph coloring on an empty graph ([REQ] Rigid body support for VBD Integrator #509).
• Fix an integer overflow bug in the native graph coloring module ([BUG] Occasional corrupted double-linked list crash when running unit tests #718).
• Fix UsdRenderer.render_points() not supporting multiple colors ([BUG] render_points does not render different colors #634).
• Fix an inconsistency in the wp.fem module regarding the orientation of 2D geometry side normals ([QUESTION] How to access vertices of both triangles sharing an edge in a fem.integrand function? #629).
• Fix premature unloading of CUDA modules used in JAX FFI graph captures ([BUG] Warp modules used in JAX FFI graph captures can be unloaded prematurely #782).

---

This discussion was created from the release v1.8.0.