warp-lang 1.4.0__py3-none-win_amd64.whl → 1.4.1__py3-none-win_amd64.whl

warp/builtins.py

@@ -2758,7 +2758,6 @@ add_builtin(
     "rand_init",
     input_types={"seed": int},
     value_type=uint32,
-    export=False,
     group="Random",
     doc="Initialize a new random number generator given a user-defined seed. Returns a 32-bit integer representing the RNG state.",
 )
@@ -2767,7 +2766,6 @@ add_builtin(
     "rand_init",
     input_types={"seed": int, "offset": int},
     value_type=uint32,
-    export=False,
     group="Random",
     doc="""Initialize a new random number generator given a user-defined seed and an offset.
 
@@ -2779,7 +2777,6 @@ add_builtin(
     "randi",
     input_types={"state": uint32},
     value_type=int,
-    export=False,
     group="Random",
     doc="Return a random integer in the range [0, 2^32).",
 )
@@ -2787,7 +2784,6 @@ add_builtin(
     "randi",
     input_types={"state": uint32, "low": int, "high": int},
     value_type=int,
-    export=False,
     group="Random",
     doc="Return a random integer between [low, high).",
 )
@@ -2795,7 +2791,6 @@ add_builtin(
     "randf",
     input_types={"state": uint32},
     value_type=float,
-    export=False,
     group="Random",
     doc="Return a random float between [0.0, 1.0).",
 )
@@ -2803,24 +2798,17 @@ add_builtin(
     "randf",
     input_types={"state": uint32, "low": float, "high": float},
     value_type=float,
-    export=False,
     group="Random",
     doc="Return a random float between [low, high).",
 )
 add_builtin(
-    "randn",
-    input_types={"state": uint32},
-    value_type=float,
-    export=False,
-    group="Random",
-    doc="Sample a normal distribution.",
+    "randn", input_types={"state": uint32}, value_type=float, group="Random", doc="Sample a normal distribution."
 )
 
 add_builtin(
     "sample_cdf",
     input_types={"state": uint32, "cdf": array(dtype=float)},
     value_type=int,
-    export=False,
     group="Random",
     doc="Inverse-transform sample a cumulative distribution function.",
 )
@@ -2828,7 +2816,6 @@ add_builtin(
     "sample_triangle",
     input_types={"state": uint32},
     value_type=vec2,
-    export=False,
     group="Random",
     doc="Uniformly sample a triangle. Returns sample barycentric coordinates.",
 )
@@ -2836,7 +2823,6 @@ add_builtin(
     "sample_unit_ring",
     input_types={"state": uint32},
     value_type=vec2,
-    export=False,
     group="Random",
     doc="Uniformly sample a ring in the xy plane.",
 )
@@ -2844,7 +2830,6 @@ add_builtin(
     "sample_unit_disk",
     input_types={"state": uint32},
     value_type=vec2,
-    export=False,
     group="Random",
     doc="Uniformly sample a disk in the xy plane.",
 )
@@ -2852,7 +2837,6 @@ add_builtin(
     "sample_unit_sphere_surface",
     input_types={"state": uint32},
     value_type=vec3,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit sphere surface.",
 )
@@ -2860,7 +2844,6 @@ add_builtin(
     "sample_unit_sphere",
     input_types={"state": uint32},
     value_type=vec3,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit sphere.",
 )
@@ -2868,7 +2851,6 @@ add_builtin(
     "sample_unit_hemisphere_surface",
     input_types={"state": uint32},
     value_type=vec3,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit hemisphere surface.",
 )
@@ -2876,7 +2858,6 @@ add_builtin(
     "sample_unit_hemisphere",
     input_types={"state": uint32},
     value_type=vec3,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit hemisphere.",
 )
@@ -2884,7 +2865,6 @@ add_builtin(
     "sample_unit_square",
     input_types={"state": uint32},
     value_type=vec2,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit square.",
 )
@@ -2892,7 +2872,6 @@ add_builtin(
     "sample_unit_cube",
     input_types={"state": uint32},
     value_type=vec3,
-    export=False,
     group="Random",
     doc="Uniformly sample a unit cube.",
 )
@@ -2901,7 +2880,6 @@ add_builtin(
     "poisson",
     input_types={"state": uint32, "lam": float},
     value_type=uint32,
-    export=False,
     group="Random",
     doc="""Generate a random sample from a Poisson distribution.
 

warp/codegen.py

@@ -777,6 +777,9 @@ def func_match_args(func, arg_types, kwarg_types):
 
 
 def get_arg_type(arg: Union[Var, Any]):
+    if isinstance(arg, str):
+        return str
+
     if isinstance(arg, Sequence):
         return tuple(get_arg_type(x) for x in arg)
 
@@ -936,7 +939,7 @@ class Adjoint:
 
         adj.return_var = None  # return type for function or kernel
         adj.loop_symbols = []  # symbols at the start of each loop
-        adj.loop_const_iter_symbols = set()  # iteration variables (constant) for static loops
+        adj.loop_const_iter_symbols = []  # iteration variables (constant) for static loops
 
         # blocks
         adj.blocks = [Block()]
@@ -1846,7 +1849,7 @@ class Adjoint:
         # detect symbols with conflicting definitions (assigned inside the for loop)
         for items in symbols.items():
             sym = items[0]
-            if adj.loop_const_iter_symbols is not None and sym in adj.loop_const_iter_symbols:
+            if adj.is_constant_iter_symbol(sym):
                 # ignore constant overwriting in for-loops if it is a loop iterator
                 # (it is no problem to unroll static loops multiple times in sequence)
                 continue
@@ -1998,11 +2001,21 @@ class Adjoint:
         return range_call
 
     def begin_record_constant_iter_symbols(adj):
-        if adj.loop_const_iter_symbols is None:
-            adj.loop_const_iter_symbols = set()
+        if len(adj.loop_const_iter_symbols) > 0:
+            adj.loop_const_iter_symbols.append(adj.loop_const_iter_symbols[-1])
+        else:
+            adj.loop_const_iter_symbols.append(set())
 
     def end_record_constant_iter_symbols(adj):
-        adj.loop_const_iter_symbols = None
+        if len(adj.loop_const_iter_symbols) > 0:
+            adj.loop_const_iter_symbols.pop()
+
+    def record_constant_iter_symbol(adj, sym):
+        if len(adj.loop_const_iter_symbols) > 0:
+            adj.loop_const_iter_symbols[-1].add(sym)
+
+    def is_constant_iter_symbol(adj, sym):
+        return len(adj.loop_const_iter_symbols) > 0 and sym in adj.loop_const_iter_symbols[-1]
 
     def emit_For(adj, node):
         # try and unroll simple range() statements that use constant args
@@ -2010,9 +2023,8 @@ class Adjoint:
 
         if isinstance(unroll_range, range):
             const_iter_sym = node.target.id
-            if adj.loop_const_iter_symbols is not None:
-                # prevent constant conflicts in `materialize_redefinitions()`
-                adj.loop_const_iter_symbols.add(const_iter_sym)
+            # prevent constant conflicts in `materialize_redefinitions()`
+            adj.record_constant_iter_symbol(const_iter_sym)
 
             # unroll static for-loop
             for i in unroll_range:

warp/config.py

@@ -7,7 +7,7 @@
 
 from typing import Optional
 
-version: str = "1.4.0"
+version: str = "1.4.1"
 """Warp version string"""
 
 verify_fp: bool = False

warp/context.py

@@ -6,7 +6,6 @@
 # license agreement from NVIDIA CORPORATION is strictly prohibited.
 
 import ast
-import builtins
 import ctypes
 import functools
 import hashlib
@@ -22,7 +21,6 @@ import typing
 import weakref
 from copy import copy as shallowcopy
 from pathlib import Path
-from struct import pack as struct_pack
 from typing import Any, Callable, Dict, List, Mapping, Optional, Sequence, Tuple, Union
 
 import numpy as np
@@ -1487,30 +1485,16 @@ class ModuleHasher:
         # hash referenced constants
         for name, value in constants.items():
             ch.update(bytes(name, "utf-8"))
-            # hash the referenced object
-            if isinstance(value, builtins.bool):
-                # This needs to come before the check for `int` since all boolean
-                # values are also instances of `int`.
-                ch.update(struct_pack("?", value))
-            elif isinstance(value, int):
-                ch.update(struct_pack("<q", value))
-            elif isinstance(value, float):
-                ch.update(struct_pack("<d", value))
-            elif isinstance(value, warp.types.float16):
-                # float16 is a special case
-                p = ctypes.pointer(ctypes.c_float(value.value))
-                ch.update(p.contents)
-            elif isinstance(value, tuple(warp.types.scalar_types)):
-                p = ctypes.pointer(value._type_(value.value))
-                ch.update(p.contents)
-            elif isinstance(value, ctypes.Array):
-                ch.update(bytes(value))
-            else:
-                raise RuntimeError(f"Invalid constant type: {type(value)}")
+            ch.update(self.get_constant_bytes(value))
 
         # hash wp.static() expressions that were evaluated at declaration time
         for k, v in adj.static_expressions.items():
-            ch.update(bytes(f"{k} = {v}", "utf-8"))
+            ch.update(bytes(k, "utf-8"))
+            if isinstance(v, Function):
+                if v not in self.functions_in_progress:
+                    ch.update(self.hash_function(v))
+            else:
+                ch.update(self.get_constant_bytes(v))
 
         # hash referenced types
         for t in types.keys():
@@ -1523,6 +1507,24 @@ class ModuleHasher:
 
         return ch.digest()
 
+    def get_constant_bytes(self, value):
+        if isinstance(value, int):
+            # this also handles builtins.bool
+            return bytes(ctypes.c_int(value))
+        elif isinstance(value, float):
+            return bytes(ctypes.c_float(value))
+        elif isinstance(value, warp.types.float16):
+            # float16 is a special case
+            return bytes(ctypes.c_float(value.value))
+        elif isinstance(value, tuple(warp.types.scalar_and_bool_types)):
+            return bytes(value._type_(value.value))
+        elif hasattr(value, "_wp_scalar_type_"):
+            return bytes(value)
+        elif isinstance(value, warp.codegen.StructInstance):
+            return bytes(value._ctype)
+        else:
+            raise TypeError(f"Invalid constant type: {type(value)}")
+
     def get_module_hash(self):
         return self.module_hash
 

warp/examples/fem/utils.py

@@ -143,7 +143,7 @@ def gen_hexmesh(res, bounds_lo: Optional[wp.vec3] = None, bounds_hi: Optional[wp
 
     x = np.linspace(bounds_lo[0], bounds_hi[0], Nx + 1)
     y = np.linspace(bounds_lo[1], bounds_hi[1], Ny + 1)
-    z = np.linspace(bounds_lo[1], bounds_hi[1], Nz + 1)
+    z = np.linspace(bounds_lo[2], bounds_hi[2], Nz + 1)
 
     positions = np.transpose(np.meshgrid(x, y, z, indexing="ij"), axes=(1, 2, 3, 0)).reshape(-1, 3)
 
@@ -252,6 +252,7 @@ def bsr_cg(
         check_every=check_every,
         M=M,
         callback=callback,
+        use_cuda_graph=not wp.config.verify_cuda,
     )
 
     if not quiet:

warp/fem/utils.py

@@ -193,9 +193,14 @@ def _givens_rotation(a: Any, b: Any):
     # Givens rotation [[c -s], [s c]] such that sa+cb =0
     zero = type(a)(0.0)
     one = type(a)(1.0)
-    abn_sq = a * a + b * b
-    abn = wp.select(abn_sq == zero, one / wp.sqrt(abn_sq), zero)
-    return a * abn, -b * abn
+
+    b2 = b * b
+    if b2 == zero:
+        # id rotation
+        return one, zero
+
+    scale = one / wp.sqrt(a * a + b2)
+    return a * scale, -b * scale
 
 
 @wp.func
@@ -229,7 +234,7 @@ def tridiagonal_symmetric_eigenvalues_qr(D: Any, L: Any, Q: Any, tol: Any):
     x = D.dtype(0.0)  # coeff atop buldge
 
     for _ in range(32 * m):  # failsafe, usually converges faster than that
-        # Iterate over all idependant (deflated) blocks
+        # Iterate over all independent (deflated) blocks
         end = int(-1)
 
         for k in range(m - 1):

warp/native/bvh.cu

@@ -65,7 +65,7 @@ __global__ void bvh_refit_kernel(int n, const int* __restrict__ parents, int* __
             int finished = atomicAdd(&child_count[parent], 1);
 
             // if we have are the last thread (such that the parent node is now complete)
-            // then update its bounds and move onto the the next parent in the hierarchy
+            // then update its bounds and move onto the next parent in the hierarchy
             if (finished == 1)
             {
                 const int left_child = node_lowers[parent].i;
@@ -273,7 +273,7 @@ __global__ void build_hierarchy(int n, int* root, const int* __restrict__ deltas
             }
 
             // if we have are the last thread (such that the parent node is now complete)
-            // then update its bounds and move onto the the next parent in the hierarchy
+            // then update its bounds and move onto the next parent in the hierarchy
             if (childCount == 1)
             {
                 const int left_child = lowers[parent].i;

warp/native/exports.h

@@ -1013,6 +1013,23 @@ WP_API void builtin_volume_index_to_world_uint64_vec3f(uint64 id, vec3f& uvw, ve
 WP_API void builtin_volume_world_to_index_uint64_vec3f(uint64 id, vec3f& xyz, vec3f* ret) { *ret = wp::volume_world_to_index(id, xyz); }
 WP_API void builtin_volume_index_to_world_dir_uint64_vec3f(uint64 id, vec3f& uvw, vec3f* ret) { *ret = wp::volume_index_to_world_dir(id, uvw); }
 WP_API void builtin_volume_world_to_index_dir_uint64_vec3f(uint64 id, vec3f& xyz, vec3f* ret) { *ret = wp::volume_world_to_index_dir(id, xyz); }
+WP_API void builtin_rand_init_int32(int32 seed, uint32* ret) { *ret = wp::rand_init(seed); }
+WP_API void builtin_rand_init_int32_int32(int32 seed, int32 offset, uint32* ret) { *ret = wp::rand_init(seed, offset); }
+WP_API void builtin_randi_uint32(uint32 state, int* ret) { *ret = wp::randi(state); }
+WP_API void builtin_randi_uint32_int32_int32(uint32 state, int32 low, int32 high, int* ret) { *ret = wp::randi(state, low, high); }
+WP_API void builtin_randf_uint32(uint32 state, float* ret) { *ret = wp::randf(state); }
+WP_API void builtin_randf_uint32_float32_float32(uint32 state, float32 low, float32 high, float* ret) { *ret = wp::randf(state, low, high); }
+WP_API void builtin_randn_uint32(uint32 state, float* ret) { *ret = wp::randn(state); }
+WP_API void builtin_sample_triangle_uint32(uint32 state, vec2f* ret) { *ret = wp::sample_triangle(state); }
+WP_API void builtin_sample_unit_ring_uint32(uint32 state, vec2f* ret) { *ret = wp::sample_unit_ring(state); }
+WP_API void builtin_sample_unit_disk_uint32(uint32 state, vec2f* ret) { *ret = wp::sample_unit_disk(state); }
+WP_API void builtin_sample_unit_sphere_surface_uint32(uint32 state, vec3f* ret) { *ret = wp::sample_unit_sphere_surface(state); }
+WP_API void builtin_sample_unit_sphere_uint32(uint32 state, vec3f* ret) { *ret = wp::sample_unit_sphere(state); }
+WP_API void builtin_sample_unit_hemisphere_surface_uint32(uint32 state, vec3f* ret) { *ret = wp::sample_unit_hemisphere_surface(state); }
+WP_API void builtin_sample_unit_hemisphere_uint32(uint32 state, vec3f* ret) { *ret = wp::sample_unit_hemisphere(state); }
+WP_API void builtin_sample_unit_square_uint32(uint32 state, vec2f* ret) { *ret = wp::sample_unit_square(state); }
+WP_API void builtin_sample_unit_cube_uint32(uint32 state, vec3f* ret) { *ret = wp::sample_unit_cube(state); }
+WP_API void builtin_poisson_uint32_float32(uint32 state, float32 lam, uint32* ret) { *ret = wp::poisson(state, lam); }
 WP_API void builtin_noise_uint32_float32(uint32 state, float32 x, float* ret) { *ret = wp::noise(state, x); }
 WP_API void builtin_noise_uint32_vec2f(uint32 state, vec2f& xy, float* ret) { *ret = wp::noise(state, xy); }
 WP_API void builtin_noise_uint32_vec3f(uint32 state, vec3f& xyz, float* ret) { *ret = wp::noise(state, xyz); }

warp/native/mesh.cu

@@ -101,7 +101,7 @@ __global__ void bvh_refit_with_solid_angle_kernel(int n, const int* __restrict__
             int finished = atomicAdd(&child_count[parent], 1);
 
             // if we have are the last thread (such that the parent node is now complete)
-            // then update its bounds and move onto the the next parent in the hierarchy
+            // then update its bounds and move onto the next parent in the hierarchy
             if (finished == 1)
             {
                 //printf("Compute non-leaf at %d\n", index);
@@ -340,4 +340,4 @@ void mesh_set_velocities_device(uint64_t id, wp::array_t<wp::vec3> velocities)
         fprintf(stderr, "The mesh id provided to mesh_set_velocities_device is not valid!\n");
         return;
     }
-}
+}

warp/native/range.h

@@ -97,8 +97,17 @@ CUDA_CALLABLE inline range_t iter_reverse(const range_t& r)
 {
     // generates a reverse range, equivalent to reversed(range())
     range_t rev;
-    rev.start = r.end-1;
-    rev.end = r.start-1;
+
+    if (r.step > 0)
+    {
+        rev.start = r.start + int((r.end - r.start - 1) / r.step) * r.step;
+    }
+    else
+    {
+        rev.start = r.start + int((r.end - r.start + 1) / r.step) * r.step;
+    }
+
+    rev.end = r.start - r.step;
     rev.step = -r.step;
 
     rev.i = rev.start;

warp/sim/integrator_xpbd.py

@@ -2808,12 +2808,8 @@ class XPBDIntegrator(Integrator):
 
         with wp.ScopedTimer("simulate", False):
             if model.particle_count:
-                if requires_grad:
-                    particle_q = state_out.particle_q
-                    particle_qd = state_out.particle_qd
-                else:
-                    particle_q = state_out.particle_q
-                    particle_qd = state_out.particle_qd
+                particle_q = state_out.particle_q
+                particle_qd = state_out.particle_qd
 
                 self.particle_q_init = wp.clone(state_in.particle_q)
                 if self.enable_restitution:

warp/sim/model.py

@@ -641,7 +641,7 @@ class Model:
         joint_dof_count (int): Total number of velocity degrees of freedom of all joints in the system
         joint_coord_count (int): Total number of position degrees of freedom of all joints in the system
 
-        particle_coloring (list of array): The coloring of all the particles, used for VBD's Gauss-Seidel interation.
+        particle_coloring (list of array): The coloring of all the particles, used for VBD's Gauss-Seidel iteration.
 
         device (wp.Device): Device on which the Model was allocated
 
@@ -1404,9 +1404,8 @@ class ModelBuilder:
             self.joint_X_p.extend(joint_X_p)
             self.joint_q.extend(joint_q)
 
-            self.add_articulation()
-
             # offset the indices
+            self.articulation_start.extend([a + self.joint_count for a in builder.articulation_start])
             self.joint_parent.extend([p + self.joint_count if p != -1 else -1 for p in builder.joint_parent])
             self.joint_child.extend([c + self.joint_count for c in builder.joint_child])
 

warp/sparse.py

@@ -106,7 +106,7 @@ class BsrMatrix(Generic[_BlockType]):
             return
 
         BsrMatrix.__setattr__(
-            self, "_nnz_buf", wp.zeros(dtype=int, shape=(1,), device="cpu", pinned=self.device.is_cuda)
+            self, "_nnz_buf", wp.empty(dtype=int, shape=(1,), device="cpu", pinned=self.device.is_cuda)
         )
         if self.device.is_cuda:
             BsrMatrix.__setattr__(self, "_nnz_event", wp.Event(self.device))
@@ -524,7 +524,7 @@ def _bsr_assign_split_blocks(
     if dest_block >= dest_offsets[dest_row_count]:
         return
 
-    dest_row = wp.lower_bound(dest_offsets, dest_block + 1) - 1
+    dest_row = wp.lower_bound(dest_offsets, 0, dest_row_count + 1, dest_block + 1) - 1
     src_row = dest_row // row_factor
 
     dest_col_in_row = dest_block - dest_offsets[dest_row]
@@ -566,7 +566,7 @@ def _bsr_assign_merge_row_col(
         dest_rows[block] = -1  # invalid
         dest_cols[block] = -1
     else:
-        row = wp.lower_bound(src_offsets, block + 1) - 1
+        row = wp.lower_bound(src_offsets, 0, src_row_count + 1, block + 1) - 1
         dest_rows[block] = row // row_factor
         dest_cols[block] = src_columns[block] // col_factor
 
@@ -589,7 +589,7 @@ def _bsr_assign_merge_blocks(
     if src_block >= src_offsets[src_row_count]:
         return
 
-    src_row = wp.lower_bound(src_offsets, src_block + 1) - 1
+    src_row = wp.lower_bound(src_offsets, 0, src_row_count + 1, src_block + 1) - 1
     src_col = src_columns[src_block]
 
     dest_row = src_row // row_factor
@@ -828,7 +828,7 @@ def bsr_copy(
         block_type=block_type,
         device=A.device,
     )
-    bsr_assign(dest=copy, src=A)
+    bsr_assign(dest=copy, src=A, structure_only=structure_only)
     return copy
 
 
@@ -1190,7 +1190,7 @@ def _bsr_get_block_row(dest_offset: int, row_count: int, bsr_offsets: wp.array(d
     if i >= bsr_offsets[row_count]:
         rows[dest_offset + i] = -1  # invalid
     else:
-        row = wp.lower_bound(bsr_offsets, i + 1) - 1
+        row = wp.lower_bound(bsr_offsets, 0, row_count + 1, i + 1) - 1
         rows[dest_offset + i] = row
 
 
@@ -1461,13 +1461,14 @@ def _bsr_mm_compute_values(
     y_offsets: wp.array(dtype=int),
     y_columns: wp.array(dtype=int),
     y_values: wp.array(dtype=Any),
+    mm_row_count: int,
     mm_offsets: wp.array(dtype=int),
     mm_cols: wp.array(dtype=int),
     mm_values: wp.array(dtype=Any),
 ):
     mm_block = wp.tid()
 
-    row = wp.lower_bound(mm_offsets, mm_block + 1) - 1
+    row = wp.lower_bound(mm_offsets, 0, mm_row_count + 1, mm_block + 1) - 1
     col = mm_cols[mm_block]
 
     mm_val = mm_values.dtype(type(alpha)(0.0))
@@ -1759,6 +1760,7 @@ def bsr_mm(
             work_arrays._old_z_offsets if y == z else y.offsets,
             work_arrays._old_z_columns if y == z else y.columns,
             work_arrays._old_z_values if y == z else y.values,
+            z.nrow,
             z.offsets,
             z.columns,
             mm_values,

warp/tests/test_codegen.py

@@ -503,6 +503,37 @@ def test_error_mutating_constant_in_dynamic_loop(test, device):
     ):
         wp.launch(dynamic_loop_kernel, dim=1, inputs=[3, inputs], device=device)
 
+    # the following nested loop must not raise an error
+    const_a = 7
+    const_b = 5
+
+    @wp.kernel
+    def mixed_dyn_static_loop_kernel(dyn_a: int, dyn_b: int, dyn_c: int, output: wp.array(dtype=float, ndim=2)):
+        tid = wp.tid()
+        for i in range(const_a + 1):
+            for j in range(dyn_a + 1):
+                for k in range(dyn_b + 1):
+                    for l in range(const_b + 1):
+                        for m in range(dyn_c + 1):
+                            coeff = i + j + k + l + m
+                            output[tid, coeff] = 1.0
+
+    dyn_a, dyn_b, dyn_c = 3, 4, 5
+    num_threads = 10
+    output = wp.empty([num_threads, const_a + const_b + dyn_a + dyn_b + dyn_c + 1], dtype=float, device=device)
+    wp.launch(
+        mixed_dyn_static_loop_kernel,
+        num_threads,
+        inputs=[
+            dyn_a,
+            dyn_b,
+            dyn_c,
+        ],
+        outputs=[output],
+        device=device,
+    )
+    assert_np_equal(output.numpy(), np.ones([num_threads, const_a + const_b + dyn_a + dyn_b + dyn_c + 1]))
+
 
 @wp.kernel
 def test_call_syntax():

warp/tests/test_fem.py

@@ -28,6 +28,9 @@ from warp.fem.utils import (
 )
 from warp.tests.unittest_utils import *
 
+vec6f = wp.vec(length=6, dtype=float)
+mat66f = wp.mat(shape=(6, 6), dtype=float)
+
 
 @integrand
 def linear_form(s: Sample, u: Field):
@@ -1507,7 +1510,7 @@ def test_implicit_fields(test, device):
 
 @wp.kernel
 def test_qr_eigenvalues():
-    tol = 1.0e-6
+    tol = 1.0e-8
 
     # zero
     Zero = wp.mat33(0.0)
@@ -1546,6 +1549,19 @@ def test_qr_eigenvalues():
     Err4 = wp.transpose(P4) * wp.diag(D4) * P4 - Rank4
     wp.expect_near(wp.ddot(Err4, Err4), 0.0, tol)
 
+    # test robustness to low requested tolerance
+    Rank6 = mat66f(
+        vec6f(0.00171076, 0.0, 0.0, 0.0, 0.0, 0.0),
+        vec6f(0.0, 0.00169935, 6.14367e-06, -3.52589e-05, 3.02397e-05, -1.53458e-11),
+        vec6f(0.0, 6.14368e-06, 0.00172217, 2.03568e-05, 1.74589e-05, -2.92627e-05),
+        vec6f(0.0, -3.52589e-05, 2.03568e-05, 0.00172178, 2.53422e-05, 3.02397e-05),
+        vec6f(0.0, 3.02397e-05, 1.74589e-05, 2.53422e-05, 0.00171114, 3.52589e-05),
+        vec6f(0.0, 6.42993e-12, -2.92627e-05, 3.02397e-05, 3.52589e-05, 0.00169935),
+    )
+    D6, P6 = symmetric_eigenvalues_qr(Rank6, 0.0)
+    Err6 = wp.transpose(P6) * wp.diag(D6) * P6 - Rank6
+    wp.expect_near(wp.ddot(Err6, Err6), 0.0, 1.0e-13)
+
 
 @wp.kernel
 def test_qr_inverse():

warp/tests/test_model.py

@@ -157,6 +157,19 @@ class TestModel(unittest.TestCase):
         assert builder.body_mass == [1.0, 4.0]
         assert builder.body_inv_mass == [1.0, 0.25]
 
+        # create another builder, test add_builder function
+        builder2 = ModelBuilder()
+        builder2.add_builder(builder)
+        assert builder2.articulation_count == builder.articulation_count
+        assert builder2.joint_count == builder.joint_count
+        assert builder2.body_count == builder.body_count
+        assert builder2.shape_count == builder.shape_count
+        assert builder2.articulation_start == builder.articulation_start
+        # add the same builder again
+        builder2.add_builder(builder)
+        assert builder2.articulation_count == 2 * builder.articulation_count
+        assert builder2.articulation_start == [0, 1, 2, 3]
+
 
 if __name__ == "__main__":
     wp.clear_kernel_cache()

warp/tests/test_static.py

@@ -5,6 +5,8 @@
 # distribution of this software and related documentation without an express
 # license agreement from NVIDIA CORPORATION is strictly prohibited.
 
+import importlib
+import tempfile
 import unittest
 from typing import Dict, List
 
@@ -17,6 +19,23 @@ from warp.tests.unittest_utils import *
 global_variable = 3
 
 
+def load_code_as_module(code, name):
+    file, file_path = tempfile.mkstemp(suffix=".py")
+
+    try:
+        with os.fdopen(file, "w") as f:
+            f.write(code)
+
+        spec = importlib.util.spec_from_file_location(name, file_path)
+        module = importlib.util.module_from_spec(spec)
+        spec.loader.exec_module(module)
+    finally:
+        os.remove(file_path)
+
+    # return Warp module
+    return wp.get_module(module.__name__)
+
+
 @wp.func
 def static_global_variable_func():
     static_var = warp.static(global_variable + 2)
@@ -234,7 +253,7 @@ def test_function_variable(test, device):
             results[0] = wp.static(func)(3, 2)  # noqa: B023
 
         results = wp.zeros(1, dtype=int, device=device)
-        # note that the kernel has to be recompiled everytime the value of func changes
+        # note that the kernel has to be recompiled every time the value of func changes
         wp.launch(function_variable_kernel, 1, [results], device=device)
         assert_np_equal(results.numpy(), np.array([func(3, 2)], dtype=int))
 
@@ -383,6 +402,140 @@ def test_static_if_else_elif(test, device):
         assert_np_equal(counts["else"], 0)
 
 
+static_builtin_constant_template = """
+import warp as wp
+
+# Python builtin literal like 17, 42.0, or True
+C = {value}
+
+@wp.kernel
+def k():
+    print(wp.static(C))
+"""
+
+static_warp_constant_template = """
+import warp as wp
+
+# Warp scalar value like wp.uint8(17)
+C = wp.{dtype}({value})
+
+@wp.kernel
+def k():
+    print(wp.static(C))
+"""
+
+static_struct_constant_template = """
+import warp as wp
+
+@wp.struct
+class SimpleStruct:
+    x: float
+
+C = SimpleStruct()
+C.x = {value}
+
+@wp.kernel
+def k():
+    print(wp.static(C))
+"""
+
+static_func_template = """
+import warp as wp
+
+@wp.func
+def f():
+    # modify the function to verify hashing
+    return {value}
+
+@wp.kernel
+def k():
+    print(wp.static(f)())
+"""
+
+
+def test_static_constant_hash(test, _):
+    # Python literals
+    # (type, value1, value2)
+    literals = [
+        (int, 17, 42),
+        (float, 17.5, 42.5),
+        (bool, True, False),
+    ]
+
+    for builtin_type, value1, value2 in literals:
+        type_name = builtin_type.__name__
+        with test.subTest(msg=f"{type_name}"):
+            source1 = static_builtin_constant_template.format(value=value1)
+            source2 = static_builtin_constant_template.format(value=value2)
+            source3 = static_builtin_constant_template.format(value=value1)
+
+            module1 = load_code_as_module(source1, f"aux_static_constant_builtin_{type_name}_1")
+            module2 = load_code_as_module(source2, f"aux_static_constant_builtin_{type_name}_2")
+            module3 = load_code_as_module(source3, f"aux_static_constant_builtin_{type_name}_3")
+
+            hash1 = module1.hash_module()
+            hash2 = module2.hash_module()
+            hash3 = module3.hash_module()
+
+            test.assertNotEqual(hash1, hash2)
+            test.assertEqual(hash1, hash3)
+
+    # Warp types (scalars, vectors, matrices)
+    for warp_type in [*wp.types.scalar_types, *wp.types.vector_types]:
+        type_name = warp_type.__name__
+        with test.subTest(msg=f"wp.{type_name}"):
+            value1 = ", ".join([str(17)] * warp_type._length_)
+            value2 = ", ".join([str(42)] * warp_type._length_)
+            source1 = static_warp_constant_template.format(dtype=type_name, value=value1)
+            source2 = static_warp_constant_template.format(dtype=type_name, value=value2)
+            source3 = static_warp_constant_template.format(dtype=type_name, value=value1)
+
+            module1 = load_code_as_module(source1, f"aux_static_constant_wp_{type_name}_1")
+            module2 = load_code_as_module(source2, f"aux_static_constant_wp_{type_name}_2")
+            module3 = load_code_as_module(source3, f"aux_static_constant_wp_{type_name}_3")
+
+            hash1 = module1.hash_module()
+            hash2 = module2.hash_module()
+            hash3 = module3.hash_module()
+
+            test.assertNotEqual(hash1, hash2)
+            test.assertEqual(hash1, hash3)
+
+    # structs
+    with test.subTest(msg="struct"):
+        source1 = static_struct_constant_template.format(value=17)
+        source2 = static_struct_constant_template.format(value=42)
+        source3 = static_struct_constant_template.format(value=17)
+
+        module1 = load_code_as_module(source1, "aux_static_constant_struct_1")
+        module2 = load_code_as_module(source2, "aux_static_constant_struct_2")
+        module3 = load_code_as_module(source3, "aux_static_constant_struct_3")
+
+        hash1 = module1.hash_module()
+        hash2 = module2.hash_module()
+        hash3 = module3.hash_module()
+
+        test.assertNotEqual(hash1, hash2)
+        test.assertEqual(hash1, hash3)
+
+
+def test_static_function_hash(test, _):
+    source1 = static_func_template.format(value=17)
+    source2 = static_func_template.format(value=42)
+    source3 = static_func_template.format(value=17)
+
+    module1 = load_code_as_module(source1, "aux_static_func1")
+    module2 = load_code_as_module(source2, "aux_static_func2")
+    module3 = load_code_as_module(source3, "aux_static_func3")
+
+    hash1 = module1.hash_module()
+    hash2 = module2.hash_module()
+    hash3 = module3.hash_module()
+
+    test.assertNotEqual(hash1, hash2)
+    test.assertEqual(hash1, hash3)
+
+
 devices = get_test_devices()
 
 
@@ -406,6 +559,9 @@ add_function_test(
 add_function_test(TestStatic, "test_static_for_loop", test_static_for_loop, devices=devices)
 add_function_test(TestStatic, "test_static_if_else_elif", test_static_if_else_elif, devices=devices)
 
+add_function_test(TestStatic, "test_static_constant_hash", test_static_constant_hash, devices=None)
+add_function_test(TestStatic, "test_static_function_hash", test_static_function_hash, devices=None)
+
 
 if __name__ == "__main__":
     wp.clear_kernel_cache()

warp/types.py

@@ -2261,13 +2261,22 @@ class array(Array):
             self._requires_grad = False
         else:
             # make sure the given gradient array is compatible
-            if (
-                grad.dtype != self.dtype
-                or grad.shape != self.shape
-                or grad.strides != self.strides
-                or grad.device != self.device
-            ):
-                raise ValueError("The given gradient array is incompatible")
+            if grad.dtype != self.dtype:
+                raise ValueError(
+                    f"The given gradient array is incompatible: expected dtype {self.dtype}, got {grad.dtype}"
+                )
+            if grad.shape != self.shape:
+                raise ValueError(
+                    f"The given gradient array is incompatible: expected shape {self.shape}, got {grad.shape}"
+                )
+            if grad.device != self.device:
+                raise ValueError(
+                    f"The given gradient array is incompatible: expected device {self.device}, got {grad.device}"
+                )
+            if grad.strides != self.strides:
+                raise ValueError(
+                    f"The given gradient array is incompatible: expected strides {self.strides}, got {grad.strides}"
+                )
             self._grad = grad
             self._requires_grad = True
 
@@ -3465,7 +3474,7 @@ class Volume:
         )
 
     def feature_array(self, feature_index: int, dtype=None) -> array:
-        """Returns one the the grid's feature data arrays as a Warp array
+        """Returns one the grid's feature data arrays as a Warp array
 
         Args:
             feature_index: Index of the supplemental data array in the grid

{warp_lang-1.4.0.dist-info → warp_lang-1.4.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: warp-lang
-Version: 1.4.0
+Version: 1.4.1
 Summary: A Python framework for high-performance simulation and graphics programming
 Author-email: NVIDIA Corporation <mmacklin@nvidia.com>
 License: NVIDIA Software License
@@ -77,9 +77,9 @@ the `pip install` command, e.g.
 
 | Platform        | Install Command                                                                                                               |
 | --------------- | ----------------------------------------------------------------------------------------------------------------------------- |
-| Linux aarch64   | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.0/warp_lang-1.4.0+cu11-py3-none-manylinux2014_aarch64.whl` |
-| Linux x86-64    | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.0/warp_lang-1.4.0+cu11-py3-none-manylinux2014_x86_64.whl`  |
-| Windows x86-64  | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.0/warp_lang-1.4.0+cu11-py3-none-win_amd64.whl`             |
+| Linux aarch64   | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.1/warp_lang-1.4.1+cu11-py3-none-manylinux2014_aarch64.whl` |
+| Linux x86-64    | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.1/warp_lang-1.4.1+cu11-py3-none-manylinux2014_x86_64.whl`  |
+| Windows x86-64  | `pip install https://github.com/NVIDIA/warp/releases/download/v1.4.1/warp_lang-1.4.1+cu11-py3-none-win_amd64.whl`             |
 
 The `--force-reinstall` option may need to be used to overwrite a previous installation.
 

{warp_lang-1.4.0.dist-info → warp_lang-1.4.1.dist-info}/RECORD

@@ -3,24 +3,24 @@ warp/__init__.pyi,sha256=Vxal9LM-q5ztScfRhrM4INpm5LJV6fN6B-H2NI5pkyU,21
 warp/autograd.py,sha256=5-ClvT3AwDQfUcpgDlVN7D4tFFzqqjBJrEMTM_Ag4gk,35700
 warp/build.py,sha256=N2-SDksxkffUkzV97dtLW2Wjqd5ERfgwpqzVoe8GNRw,3904
 warp/build_dll.py,sha256=jJYepplaVXRdLL_ug4QlR6w7TSjyHB8zr1EyfZROOgA,15801
-warp/builtins.py,sha256=gJndZJZ03HBkEdA_Ri9wfYs5iuzQpR5ULRw39F8TYhM,146096
-warp/codegen.py,sha256=kAhfMF5RfqC1vYhN1EZDYddfiSsiH9GUOR9yrnIEzBA,133509
-warp/config.py,sha256=iNbaVCVfM_C4U6iOxUaNSrEoJg4MwzbkwGyY1hvArQk,2808
+warp/builtins.py,sha256=g1lDmyGZnFU_1jBPMzzYRdJPyPG9kbhbfuDicNBd21w,145755
+warp/codegen.py,sha256=5MLNdPhHGrdxZCDsV1IgmssZ2LEBBd_lB66P4qvWECU,133887
+warp/config.py,sha256=nmywCBQORl5fHziFpCqT17fYbqxDCUtM9hTDEPyQkJ4,2808
 warp/constants.py,sha256=ckfEqwbg109RJDW98SILe_yfOBEFolOATVjt0mU2py4,1321
-warp/context.py,sha256=9zZG0K0zez7K0ThLw3KU5bD_raBQfAtLjJ-l7s56JzI,238090
+warp/context.py,sha256=w0G-nmuzzg9XEina3ovCPijAr-A9PjG9XYXYuZCjT_Q,238071
 warp/dlpack.py,sha256=jyQdL-2c6zM-r5P0DZBCTSkBoRpbmWChvcn3GvGBKIE,16853
 warp/fabric.py,sha256=qWOomvuhM2y3EzScsgaj_Dxvdt0wJPfFu4jUUG1JALg,11439
 warp/jax.py,sha256=QKJXfpz1jjaRD-Yp4C4wC4jUv8jHj0en3MW0VXKq0ZA,5280
 warp/jax_experimental.py,sha256=1WFrXcZk3U9SzqM5NJmpuCd1h29tWfvwZRA7b9JIecs,13930
 warp/paddle.py,sha256=afh5f7n37IRaDiRimFlI4xqW6p7xTPx61wnTkotOUS8,14862
-warp/sparse.py,sha256=APAp2M2FNpmOaQkI-ddxjluEcwJ-Ua2plNwq8IdP_Gw,64640
+warp/sparse.py,sha256=tKF7IdiNZB9WpU9t5rFp5t2W4dVa8kacK6dFNA9RE00,64820
 warp/stubs.py,sha256=vqSKm0q1ptkqhtq1iA1Yepn7Bzrde-tJsnE0P_YZ9wI,67864
 warp/tape.py,sha256=X979fk8oSsJiv-VQyz2AMhAm2udDk2bngx-_zKHIJyo,49015
 warp/torch.py,sha256=jv3SYQHVKsiCzenlWUVmVs5mLYfsWcdTW446IqvEwVA,14505
-warp/types.py,sha256=5zdL67hnDpIT3gzFkFyUMawVjrtgXIj1acnZpk90b_g,180802
+warp/types.py,sha256=HIBjH5N1-n20oOkLUIHJX6rzGbeEFis-REKu2h1VxFY,181341
 warp/utils.py,sha256=T07usjsWbggZTH1jvWd-W4ucWcWOEsLSdv8443LseJ8,35993
-warp/bin/warp-clang.dll,sha256=OLLP1YejFCoD3MtGheOCzVmvoENDQkrnsdFetGpUwcw,47584256
-warp/bin/warp.dll,sha256=cOxkWNbaRuDDCKM9jBQzqSdatETsFBUwC5VkWdPv_q0,101668864
+warp/bin/warp-clang.dll,sha256=3Kk6iFOVupg0F5Tgg1iu1weiEJd2VrxAm290QNl-m0A,47584256
+warp/bin/warp.dll,sha256=werkV2gbHKWhvwxJk3ESa4yIUevU_vmPmoPCAxtjoZQ,101675008
 warp/examples/__init__.py,sha256=ul3vRFtQ17_zIZIWJaCnCY28KZTGOUHQA4wA_IW_qaw,606
 warp/examples/browse.py,sha256=6v7F0vpZz5ZtYGkg8hIpxhbIZ2ECv46cIKAmKVtV1S4,848
 warp/examples/assets/bear.usd,sha256=1SK7s-zpF3Ypzcc8Wt861o174Pe-xMuJpvosACXVNUk,226238
@@ -75,7 +75,7 @@ warp/examples/fem/example_nonconforming_contact.py,sha256=APIbR11KU52zVe6_zgObuk
 warp/examples/fem/example_stokes.py,sha256=7wM-bf1mjfj8r0urpBjH1Vm6bQfc2X8yUr9PnPph_n0,7154
 warp/examples/fem/example_stokes_transfer.py,sha256=UhgR5f2yXVavVYhmuMoOzKApg--MXYCD0r5eeHJ3vDs,9366
 warp/examples/fem/example_streamlines.py,sha256=a-RtZrpqngulgfc8MlJ010qZC7kW5iGKsCYT4HGBRlc,12266
-warp/examples/fem/utils.py,sha256=aZX5_6xWuHhSZKC0S3Qof3-2paUDWA9PieuwCpnaFBo,32112
+warp/examples/fem/utils.py,sha256=jSS_3iQMTrtNP0b-a1wVmw4OBO9P_5X-wAmZNyjtbYo,32162
 warp/examples/optim/example_bounce.py,sha256=V4t9pKXGXv91C-S9-e5YV6ENSgOpK2cyHmPl1q3-zKQ,8470
 warp/examples/optim/example_cloth_throw.py,sha256=mj170UWAhWu4EGQwVgqp08ijoHvwZXlCC4WAa3sl0QE,7213
 warp/examples/optim/example_diffray.py,sha256=GiSmgqlN_j33uFuJ8KChJJwYpABkhUvII-YeQrOmZqo,18545
@@ -107,7 +107,7 @@ warp/fem/integrate.py,sha256=mLKHQMgTu8qbN4TOp7DWlV0kNTZBxhi5R2-XfaWMhCo,62264
 warp/fem/operator.py,sha256=W57PueA32kXd8AFl89ra53LLwO19jsRpFDkclsnwPMI,6383
 warp/fem/polynomial.py,sha256=feVKe3Yv8M53vCC57dhW7aq698DgnAilNOvvi_BnBY0,6565
 warp/fem/types.py,sha256=TA6rsT61qIC0a62FEiNrNBPbjz1rhMqF8SqZgFKKcGk,2331
-warp/fem/utils.py,sha256=ESZ6WWERWZxVB1NTPIhtYW63cAbpb2YtJXaR_--y1_w,22865
+warp/fem/utils.py,sha256=4Io5rTsEt5Bokf3hAupLqXBO0wcqRWi_U6s-nlFQPA8,22899
 warp/fem/field/__init__.py,sha256=WGklY2f5nuXLA9TstvWvHJq9AZRxb4kvPihxb1OZHcM,4020
 warp/fem/field/field.py,sha256=zGA-nrkirNIBYH7umYT1gvI5sK6qudzhZkR6nIAhofc,21124
 warp/fem/field/nodal_field.py,sha256=g-6zUAt3GKRAPBwt1ZcK38_7KblZwzKweSRJQJ-xQQY,12746
@@ -155,7 +155,7 @@ warp/fem/space/shape/triangle_shape_function.py,sha256=XIH1-NFHaMmuFXlzNc07bAMlV
 warp/native/array.h,sha256=vXJLJjo5nA-QW10y7T6TjVnBHM8utXv4OZEDL3cMj98,38984
 warp/native/builtin.h,sha256=D-OzDu5QoCq3p0I_bPy9WVaYupO4VY5IRcB2_3zUAaM,50684
 warp/native/bvh.cpp,sha256=PAETbl9RNUqVmJA25wNrg9_3wnGLkykaRubj2ExdKGY,10474
-warp/native/bvh.cu,sha256=e7ATGMSneNYoBMXcSwtoPOfEisZsEBo_wVAz6XRjWR8,19379
+warp/native/bvh.cu,sha256=G6LuuUVwULtOkhAL2mem8LENFP-dACv35SQu7bScaYY,19371
 warp/native/bvh.h,sha256=GG0aPcT5p7xAJv7dK3MlE-C14ss7iepksAeIwintons,10188
 warp/native/crt.cpp,sha256=_2AwFMh_1Q_XIP9lsbc6q74_3HQOl65YpCejKbZ_0aE,1167
 warp/native/crt.h,sha256=HeMvgbN_8BlITJ2_JkmyGlGra3btP36_rF7K-bnwPuo,10788
@@ -166,7 +166,7 @@ warp/native/cutlass_gemm.cpp,sha256=AlBnMgeO-veCYV-Y5NHEmOLnWKvzTQIPxSzg75Z7Tpk,
 warp/native/cutlass_gemm.cu,sha256=0ReIwBKKp6S2TaPPAFKIg2pBHItF707NoPjXhwvdG54,21615
 warp/native/error.cpp,sha256=vZPoPOWNZxYEhKckC8wTfnrMuZgGcog3uo2j1DprwSA,1907
 warp/native/error.h,sha256=LL74-rb8cackiu7NvHdu8zezVOVsG-ERr8gyqvqIhwQ,934
-warp/native/exports.h,sha256=0hxbQYi0XG61m6aT8uWQCMBG1p7n8imRUdiChUREJug,180658
+warp/native/exports.h,sha256=cyX7gP6hV8lPzUo5Jh932kmhGRkz7Mo4EjxAZAzYv44,182600
 warp/native/fabric.h,sha256=XcTNQx0sbvfSYdVSDPhos9uhH1Z96zylzMFQLuOHSj4,5750
 warp/native/hashgrid.cpp,sha256=xP3KzWKaE_6_5EQ72Abj-1jJ7ocYitQBXXX0zNgL37A,8429
 warp/native/hashgrid.cu,sha256=LSHjbh5ehUBoqR1sFAvwGB0KBB3bBGIrF9c9cZnz89c,2259
@@ -181,12 +181,12 @@ warp/native/marching.h,sha256=NukjDs_VrVC-BXDtIeFhOQFwztYbC9U80UCzr-KGhAg,14
 warp/native/mat.h,sha256=d4YuecbJIjCI8bh6O8ck4-GBYgm_srNL6rW5oQpEPrg,51348
 warp/native/matnn.h,sha256=7PQLOe4cUuL5X7Fi6kVKAqORvz6JMXJHLsDviBMxOZQ,9617
 warp/native/mesh.cpp,sha256=dv1YzFlFxE3HXFWhLgqcHxwAn7JCipJ2jKvai1h90XU,6958
-warp/native/mesh.cu,sha256=1BysYvOoqKbxyuqq_rj-RnCD9jWfaztBYykLo4mM94s,11984
+warp/native/mesh.cu,sha256=hKXK3hj5SqByV5TcTB2ugNpwAqc6CaQiylfagtBlFPk,11981
 warp/native/mesh.h,sha256=zMcdlCop5VrEK7tIpCfYXnV5NqoTRUwz3PBY1x8Ex_Q,59646
 warp/native/noise.h,sha256=fOH_JY0S9h-HSyYim6EdXWGncBIa8Tb0Ngwbkb4HUgw,29762
 warp/native/quat.h,sha256=yuC6qUbtUw96scTAbbk_O55eA1QNFLEq0luyCTFzPwc,42584
 warp/native/rand.h,sha256=oAsX4qPmrgMoqG1ObtpU63leDPX6tIHTiR448aLGyKE,11781
-warp/native/range.h,sha256=D7BLpAA_RXXrWXVcWm6oxZfuaSBCO_U_lFSclI1FQbU,2315
+warp/native/range.h,sha256=qLZFvPygGtClsTh1mtErho6yEXIpyWKN87HliglY2Ac,2504
 warp/native/reduce.cpp,sha256=TrZLQePN8GpoHsWQRbvgnknbD2Bjgl0PuD1JsnmAVMk,4460
 warp/native/reduce.cu,sha256=FkNLEQF8DlP5PbDe0AMavxwFs88634gYKce0OO_xZfc,10667
 warp/native/runlength_encode.cpp,sha256=QSPu8CV-2jJh7Nh-_5Om-WxirBjpfrALvQdlOgpB9F4,1325
@@ -238,8 +238,8 @@ warp/sim/integrator.py,sha256=5UZYDdL6IyQaL6eGc1UQLrLcOrCD0dHmZWgG0K2kCGA,6652
 warp/sim/integrator_euler.py,sha256=OEg0T6aajoX1dEDq0g9OdP1nSOevVCKzv32wNkwQUOI,59787
 warp/sim/integrator_featherstone.py,sha256=f2CS2zoCUaOZSwf2-3w0YeUVuxf3qHCk2nEA-4tQpjg,64873
 warp/sim/integrator_vbd.py,sha256=XZyAmlTeU0GSuv4V_X36ELmtLwYd6zEckvNE_46OuQc,38718
-warp/sim/integrator_xpbd.py,sha256=oJmTNO3UK6NlC7GLmzsoTb0dV44RM4DvqwVqIlUQkGs,115198
-warp/sim/model.py,sha256=3ClBoHbR2aev_Fffv-YZyxRtjuuuccB9sJiK7X99z30,193231
+warp/sim/integrator_xpbd.py,sha256=j11n_X2xBu5CkP66qbybNg-wVH3hw8nhkkJUFvOsqsk,115024
+warp/sim/model.py,sha256=oSwxDY8_nZccrgsoNsZ782CG9YqlOknsaBSFh9Hrsn8,193296
 warp/sim/particles.py,sha256=wVK-LEKZy9TGYyRnmna_p2Q24zZaorZf8EdUizM4VCo,3165
 warp/sim/render.py,sha256=XDNflQIqu3XXzVhwR7wC72-8Z3WgEVuEYoo_sDFaOLg,17874
 warp/sim/utils.py,sha256=WrXAMBi1ufZJpISo4rQfOMGx7u4F3moHDeQW-v8ukyU,12201
@@ -269,7 +269,7 @@ warp/tests/test_bool.py,sha256=f-9kM--pYSg176KD9gRMOdnADfu8HP1ZnZEjvOYPjug,6472
 warp/tests/test_builtins_resolution.py,sha256=ixoOhKgOgbuH-xbqNLx1DB1Nur3PdAtR0qW-TutaHrI,63989
 warp/tests/test_bvh.py,sha256=mr7sRxB1AaQWetCzDIOsW38aWWMwIhfbZBMkjZcKr0M,6123
 warp/tests/test_closest_point_edge_edge.py,sha256=mtiiZOHqgGOHoD6KNK4jVP-9OoCUWwnuOBVB9M3uGlc,9560
-warp/tests/test_codegen.py,sha256=izRRZdVludGcFE3UKq_pDw8_f4LgK0k0GwBIVvWnIdI,17550
+warp/tests/test_codegen.py,sha256=5onVI7Inna0B_OTZKDdTpgEaUc9qE3MlBEek6MzHhT4,18606
 warp/tests/test_codegen_instancing.py,sha256=PRExiuwPpXOwLWnB62eSMgOldZeqPFEKEC537O2N6cI,40664
 warp/tests/test_collision.py,sha256=i5FOrlcY1YIWNxRx5lC1CyVW1QQWtKUYaE68-yoLZag,19762
 warp/tests/test_compile_consts.py,sha256=by_7FTvoC-495n_XS49LBxNReqZ5YXS08EmjhOxs-54,6908
@@ -282,7 +282,7 @@ warp/tests/test_dlpack.py,sha256=H99Md6pJZDv2gdLh4omB6bYTf_pbNOnpHGfvEkpiRv8,238
 warp/tests/test_examples.py,sha256=pZ6FoJCpMLhlSUt1SvXsJdp7PHvTHNsDbAbx4Amj_wE,15612
 warp/tests/test_fabricarray.py,sha256=H186ayJZS95IGUZfzfwiOdEcqczXxN6gNmj9RvZPaPA,32340
 warp/tests/test_fast_math.py,sha256=8jETKu2zSAXLo8EZSnRMi2Ubmrd6JrpHUpppNruqMIw,2228
-warp/tests/test_fem.py,sha256=vn0iUv74HPsdCk7cm52tdO3fglkqK7Vi86_ERZYLFKw,64037
+warp/tests/test_fem.py,sha256=_WEATA9VFA2TojAMM91ovsH4pelM9vE3DwDDOPJSYSo,64831
 warp/tests/test_fp16.py,sha256=g7N5oeKpx3QDnpOQKtyKIYv0eJEy0QwaEHXFCXxTXxE,3818
 warp/tests/test_func.py,sha256=pt3sBVnofb-ihX1WNf5Svektq2lw4OrLVDjuBaF7VXM,11799
 warp/tests/test_future_annotations.py,sha256=eC7VHcAOJ3fAgQsKPnrCq9oqkw3s2RbmcXe4rBaFEpQ,2208
@@ -314,7 +314,7 @@ warp/tests/test_mesh_query_aabb.py,sha256=ttfuEwO6eGyAhUAUt1_cM_0NjFSxKzAy8ZMzf7
 warp/tests/test_mesh_query_point.py,sha256=OBFhIDeA6XnzkAI69JTg4zqD5TAXeZUY0rXlgDsNIuk,27656
 warp/tests/test_mesh_query_ray.py,sha256=vWoPLcREGfVzEAQc60t1rE3gIYBYMDY8xGDMB1Q2RlQ,9193
 warp/tests/test_mlp.py,sha256=nnTMg-Oc_YndafdpyGE2zYZG3V9IdqfYS8LVGk4oUeE,7994
-warp/tests/test_model.py,sha256=78vx9B6xGMGTI1fEGkScHvrUyOmv_VSsUcQ620LT8Ag,7052
+warp/tests/test_model.py,sha256=KZXE8_tx_cvGBZ_KTRwKaJTG3XNIkhavan4zYRaWTD4,7717
 warp/tests/test_module_hashing.py,sha256=ekTFrTEY8xtlOZErIk1bkDjYUg6jEiU0BAo5lOJlZPA,4799
 warp/tests/test_modules_lite.py,sha256=N3dUtTxzluaPF6rXmtYiSf8S27djAvPyj3ak-W_19H8,1152
 warp/tests/test_multigpu.py,sha256=ZPeZb1MA8pqQ_ERRllWbPKY_1Hr9AE9TN-vlka7IQls,5269
@@ -339,7 +339,7 @@ warp/tests/test_snippet.py,sha256=fBbGqlc6VJPmXRSF_J_UtrzCpM5XJHN_ejjgi2CMzOI,94
 warp/tests/test_sparse.py,sha256=jz9-9UtUh4n8r2uCI16Qt1Ju_WdXTAtMaz0Uk7vHcNo,20160
 warp/tests/test_spatial.py,sha256=kVfngbZLi-LDy4ooP1RMF0zb_VYmrRc8UksBXzsFQVU,76435
 warp/tests/test_special_values.py,sha256=5Ga45nJ6S7IPSVmeKUUia7CmWpYN9W-hnLPweUZIts4,14740
-warp/tests/test_static.py,sha256=ekzAeasoRIfiVvSKnGEmIzFGkr7-f5wiWuzQ3iZnUvE,13685
+warp/tests/test_static.py,sha256=0kjwNf-m7uQ-TCpsqUiaXKSGLUsufqxHLBvoh6uQaL0,18489
 warp/tests/test_streams.py,sha256=3jJz2BZ9hIANceRII0OdLHu1rfLMJ1mJnJqEPyuz-h0,19750
 warp/tests/test_struct.py,sha256=vqHTX8vdUIn2adK4bdD9xXEYUtSUKL7hw3RA_IplqhQ,16990
 warp/tests/test_tape.py,sha256=AQ0dE5oWfKv2jDNCMOK5s9up0423uz2_M-WDPVi-fIU,5450
@@ -374,8 +374,8 @@ warp/thirdparty/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 warp/thirdparty/appdirs.py,sha256=2UXPK-AkDGoN5APeqAMMYkkfgf3Q_aUEZxcfkIxlQn0,24254
 warp/thirdparty/dlpack.py,sha256=HYvNWIemrvxLDgQLRseerP8gqM87Ie7SVIX7p5xSM0o,4323
 warp/thirdparty/unittest_parallel.py,sha256=-FgbXrKiQ_RDj7YXYyhpRIPC3OmvJEnjvGkN99cqQA0,20888
-warp_lang-1.4.0.dist-info/LICENSE.md,sha256=Gq_TN0Xfat8Ftq2Rlvc31NgxdpLGVB-34kp-4eoeDV0,19265
-warp_lang-1.4.0.dist-info/METADATA,sha256=Jy807e9CHzSNOv8q94l726kC2p5OFG_-q0dg_c1c7UY,23286
-warp_lang-1.4.0.dist-info/WHEEL,sha256=UZWkB7hdHd8SXLvRSsKYZHaAk7YswVowzA3x4w8Hdww,99
-warp_lang-1.4.0.dist-info/top_level.txt,sha256=8pupHORyKoiN_BYWlTmv5OFBWdhqpppiBYQV5KxgEvg,5
-warp_lang-1.4.0.dist-info/RECORD,,
+warp_lang-1.4.1.dist-info/LICENSE.md,sha256=Gq_TN0Xfat8Ftq2Rlvc31NgxdpLGVB-34kp-4eoeDV0,19265
+warp_lang-1.4.1.dist-info/METADATA,sha256=NT-TgzFD2kfdwrN4Q487Ze9BJ16dOnkjkftjsV3wflI,23286
+warp_lang-1.4.1.dist-info/WHEEL,sha256=UZWkB7hdHd8SXLvRSsKYZHaAk7YswVowzA3x4w8Hdww,99
+warp_lang-1.4.1.dist-info/top_level.txt,sha256=8pupHORyKoiN_BYWlTmv5OFBWdhqpppiBYQV5KxgEvg,5
+warp_lang-1.4.1.dist-info/RECORD,,