warp-lang 0.9.0__py3-none-win_amd64.whl → 0.11.0__py3-none-win_amd64.whl

warp/fem/space/shape/square_shape_function.py

@@ -0,0 +1,611 @@
+import math
+
+import warp as wp
+import numpy as np
+
+from warp.fem.polynomial import Polynomial, quadrature_1d, lagrange_scales, is_closed
+from warp.fem.types import Coords
+from warp.fem import cache
+
+from .triangle_shape_function import Triangle2DPolynomialShapeFunctions
+
+
+class SquareBipolynomialShapeFunctions:
+    def __init__(self, degree: int, family: Polynomial):
+        self.family = family
+
+        self.ORDER = wp.constant(degree)
+        self.NODES_PER_ELEMENT = wp.constant((degree + 1) * (degree + 1))
+        self.NODES_PER_SIDE = wp.constant(degree + 1)
+
+        lobatto_coords, lobatto_weight = quadrature_1d(point_count=degree + 1, family=family)
+        lagrange_scale = lagrange_scales(lobatto_coords)
+
+        NodeVec = wp.types.vector(length=degree + 1, dtype=wp.float32)
+        self.LOBATTO_COORDS = wp.constant(NodeVec(lobatto_coords))
+        self.LOBATTO_WEIGHT = wp.constant(NodeVec(lobatto_weight))
+        self.LAGRANGE_SCALE = wp.constant(NodeVec(lagrange_scale))
+        self.ORDER_PLUS_ONE = wp.constant(self.ORDER + 1)
+
+    @property
+    def name(self) -> str:
+        return f"Square_Q{self.ORDER}_{self.family}"
+
+    def make_node_coords_in_element(self):
+        ORDER = self.ORDER
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_coords_in_element(
+            node_index_in_elt: int,
+        ):
+            node_i = node_index_in_elt // (ORDER + 1)
+            node_j = node_index_in_elt - (ORDER + 1) * node_i
+            return Coords(LOBATTO_COORDS[node_i], LOBATTO_COORDS[node_j], 0.0)
+
+        return node_coords_in_element
+
+    def make_node_quadrature_weight(self):
+        ORDER = self.ORDER
+        LOBATTO_WEIGHT = self.LOBATTO_WEIGHT
+
+        def node_quadrature_weight(
+            node_index_in_elt: int,
+        ):
+            node_i = node_index_in_elt // (ORDER + 1)
+            node_j = node_index_in_elt - (ORDER + 1) * node_i
+            return LOBATTO_WEIGHT[node_i] * LOBATTO_WEIGHT[node_j]
+
+        def node_quadrature_weight_linear(
+            node_index_in_elt: int,
+        ):
+            return 0.25
+
+        if ORDER == 1:
+            return cache.get_func(node_quadrature_weight_linear, self.name)
+
+        return cache.get_func(node_quadrature_weight, self.name)
+
+    @wp.func
+    def _vertex_coords_f(vidx_in_cell: int):
+        x = vidx_in_cell // 2
+        y = vidx_in_cell - 2 * x
+        return wp.vec2(float(x), float(y))
+
+    def make_trace_node_quadrature_weight(self):
+        ORDER = self.ORDER
+        LOBATTO_WEIGHT = self.LOBATTO_WEIGHT
+
+        def trace_node_quadrature_weight(
+            node_index_in_elt: int,
+        ):
+            # We're either on a side interior or at a vertex
+            # I.e., either both indices are at extrema, or only one is
+            # Pick the interior one if possible, if both are at extrema pick any one
+            node_i = node_index_in_elt // (ORDER + 1)
+            if node_i > 0 and node_i < ORDER:
+                return LOBATTO_WEIGHT[node_i]
+
+            node_j = node_index_in_elt - (ORDER + 1) * node_i
+            return LOBATTO_WEIGHT[node_j]
+
+        def trace_node_quadrature_weight_linear(
+            node_index_in_elt: int,
+        ):
+            return 0.5
+
+        def trace_node_quadrature_weight_open(
+            node_index_in_elt: int,
+        ):
+            return 0.0
+
+        if not is_closed(self.family):
+            return cache.get_func(trace_node_quadrature_weight_open, self.name)
+
+        if ORDER == 1:
+            return cache.get_func(trace_node_quadrature_weight_linear, self.name)
+
+        return cache.get_func(trace_node_quadrature_weight, self.name)
+
+    def make_element_inner_weight(self):
+        ORDER_PLUS_ONE = self.ORDER_PLUS_ONE
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+        LAGRANGE_SCALE = self.LAGRANGE_SCALE
+
+        def element_inner_weight(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            node_i = node_index_in_elt // ORDER_PLUS_ONE
+            node_j = node_index_in_elt - ORDER_PLUS_ONE * node_i
+
+            w = float(1.0)
+            for k in range(ORDER_PLUS_ONE):
+                if k != node_i:
+                    w *= coords[0] - LOBATTO_COORDS[k]
+                if k != node_j:
+                    w *= coords[1] - LOBATTO_COORDS[k]
+
+            w *= LAGRANGE_SCALE[node_i] * LAGRANGE_SCALE[node_j]
+
+            return w
+
+        def element_inner_weight_linear(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            v = SquareBipolynomialShapeFunctions._vertex_coords_f(node_index_in_elt)
+
+            wx = (1.0 - coords[0]) * (1.0 - v[0]) + v[0] * coords[0]
+            wy = (1.0 - coords[1]) * (1.0 - v[1]) + v[1] * coords[1]
+            return wx * wy
+
+        if self.ORDER == 1 and is_closed(self.family):
+            return cache.get_func(element_inner_weight_linear, self.name)
+
+        return cache.get_func(element_inner_weight, self.name)
+
+    def make_element_inner_weight_gradient(self):
+        ORDER_PLUS_ONE = self.ORDER_PLUS_ONE
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+        LAGRANGE_SCALE = self.LAGRANGE_SCALE
+
+        def element_inner_weight_gradient(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            node_i = node_index_in_elt // ORDER_PLUS_ONE
+            node_j = node_index_in_elt - ORDER_PLUS_ONE * node_i
+
+            prefix_x = float(1.0)
+            prefix_y = float(1.0)
+            for k in range(ORDER_PLUS_ONE):
+                if k != node_i:
+                    prefix_y *= coords[0] - LOBATTO_COORDS[k]
+                if k != node_j:
+                    prefix_x *= coords[1] - LOBATTO_COORDS[k]
+
+            grad_x = float(0.0)
+            grad_y = float(0.0)
+
+            for k in range(ORDER_PLUS_ONE):
+                if k != node_i:
+                    delta_x = coords[0] - LOBATTO_COORDS[k]
+                    grad_x = grad_x * delta_x + prefix_x
+                    prefix_x *= delta_x
+                if k != node_j:
+                    delta_y = coords[1] - LOBATTO_COORDS[k]
+                    grad_y = grad_y * delta_y + prefix_y
+                    prefix_y *= delta_y
+
+            grad = LAGRANGE_SCALE[node_i] * LAGRANGE_SCALE[node_j] * wp.vec2(grad_x, grad_y)
+
+            return grad
+
+        def element_inner_weight_gradient_linear(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            v = SquareBipolynomialShapeFunctions._vertex_coords_f(node_index_in_elt)
+
+            wx = (1.0 - coords[0]) * (1.0 - v[0]) + v[0] * coords[0]
+            wy = (1.0 - coords[1]) * (1.0 - v[1]) + v[1] * coords[1]
+
+            dx = 2.0 * v[0] - 1.0
+            dy = 2.0 * v[1] - 1.0
+
+            return wp.vec2(dx * wy, dy * wx)
+
+        if self.ORDER == 1 and is_closed(self.family):
+            return cache.get_func(element_inner_weight_gradient_linear, self.name)
+
+        return cache.get_func(element_inner_weight_gradient, self.name)
+
+    def element_node_triangulation(self):
+        from warp.fem.utils import grid_to_tris
+
+        return grid_to_tris(self.ORDER, self.ORDER)
+
+
+class SquareSerendipityShapeFunctions:
+    """
+    Serendipity element ~ tensor product space without interior nodes
+    Side shape functions are usual Lagrange shape functions times a linear function in the normal direction
+    Corner shape functions are bilinear shape functions times a function of (x^{d-1} + y^{d-1})
+    """
+
+    # Node categories
+    VERTEX = wp.constant(0)
+    EDGE_X = wp.constant(1)
+    EDGE_Y = wp.constant(2)
+
+    def __init__(self, degree: int, family: Polynomial):
+        if not is_closed(family):
+            raise ValueError("A closed polynomial family is required to define serendipity elements")
+
+        if degree not in [2, 3]:
+            raise NotImplementedError("Serendipity element only implemented for order 2 or 3")
+
+        self.family = family
+
+        self.ORDER = wp.constant(degree)
+        self.NODES_PER_ELEMENT = wp.constant(4 * degree)
+        self.NODES_PER_SIDE = wp.constant(degree + 1)
+
+        lobatto_coords, lobatto_weight = quadrature_1d(point_count=degree + 1, family=family)
+        lagrange_scale = lagrange_scales(lobatto_coords)
+
+        NodeVec = wp.types.vector(length=degree + 1, dtype=wp.float32)
+        self.LOBATTO_COORDS = wp.constant(NodeVec(lobatto_coords))
+        self.LOBATTO_WEIGHT = wp.constant(NodeVec(lobatto_weight))
+        self.LAGRANGE_SCALE = wp.constant(NodeVec(lagrange_scale))
+        self.ORDER_PLUS_ONE = wp.constant(self.ORDER + 1)
+
+        self.node_type_and_type_index = self._get_node_type_and_type_index()
+        self._node_lobatto_indices = self._get_node_lobatto_indices()
+
+    @property
+    def name(self) -> str:
+        return f"Square_S{self.ORDER}_{self.family}"
+
+    def _get_node_type_and_type_index(self):
+        @cache.dynamic_func(suffix=self.name)
+        def node_type_and_index(
+            node_index_in_elt: int,
+        ):
+            if node_index_in_elt < 4:
+                return SquareSerendipityShapeFunctions.VERTEX, node_index_in_elt
+
+            type_index = (node_index_in_elt - 4) // 2
+            side = node_index_in_elt - 4 - 2 * type_index
+            return SquareSerendipityShapeFunctions.EDGE_X + side, type_index
+
+        return node_type_and_index
+
+    @wp.func
+    def side_offset_and_index(type_index: int):
+        index_in_side = type_index // 2
+        side_offset = type_index - 2 * index_in_side
+
+        return side_offset, index_in_side
+
+    def _get_node_lobatto_indices(self):
+        ORDER = self.ORDER
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_lobatto_indices(node_type: int, type_index: int):
+            if node_type == SquareSerendipityShapeFunctions.VERTEX:
+                node_i = type_index // 2
+                node_j = type_index - 2 * node_i
+                return node_i * ORDER, node_j * ORDER
+
+            side_offset, index_in_side = SquareSerendipityShapeFunctions.side_offset_and_index(type_index)
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_X:
+                node_i = 1 + index_in_side
+                node_j = side_offset * ORDER
+            else:
+                node_j = 1 + index_in_side
+                node_i = side_offset * ORDER
+
+            return node_i, node_j
+
+        return node_lobatto_indices
+
+    def make_node_coords_in_element(self):
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_coords_in_element(
+            node_index_in_elt: int,
+        ):
+            node_type, type_index = self.node_type_and_type_index(node_index_in_elt)
+            node_i, node_j = self._node_lobatto_indices(node_type, type_index)
+            return Coords(LOBATTO_COORDS[node_i], LOBATTO_COORDS[node_j], 0.0)
+
+        return node_coords_in_element
+
+    def make_node_quadrature_weight(self):
+        ORDER = self.ORDER
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_quadrature_weight(
+            node_index_in_elt: int,
+        ):
+            node_type, type_index = self.node_type_and_type_index(node_index_in_elt)
+            if node_type == SquareSerendipityShapeFunctions.VERTEX:
+                return 0.25 / float(ORDER * ORDER)
+
+            return (0.25 - 0.25 / float(ORDER * ORDER)) / float(ORDER - 1)
+
+        return node_quadrature_weight
+
+    def make_trace_node_quadrature_weight(self):
+        LOBATTO_WEIGHT = self.LOBATTO_WEIGHT
+
+        @cache.dynamic_func(suffix=self.name)
+        def trace_node_quadrature_weight(
+            node_index_in_elt: int,
+        ):
+            node_type, type_index = self.node_type_and_type_index(node_index_in_elt)
+            if node_type == SquareSerendipityShapeFunctions.VERTEX:
+                return LOBATTO_WEIGHT[0]
+
+            side_offset, index_in_side = SquareSerendipityShapeFunctions.side_offset_and_index(type_index)
+            return LOBATTO_WEIGHT[1 + index_in_side]
+
+        return trace_node_quadrature_weight
+
+    def make_element_inner_weight(self):
+        ORDER = self.ORDER
+        ORDER_PLUS_ONE = self.ORDER_PLUS_ONE
+
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+        LAGRANGE_SCALE = self.LAGRANGE_SCALE
+
+        DEGREE_3_CIRCLE_RAD = wp.constant(0.5**2 + (0.5 - LOBATTO_COORDS[1]) ** 2)
+        DEGREE_3_CIRCLE_SCALE = 1.0 / (0.5 - DEGREE_3_CIRCLE_RAD)
+
+        @cache.dynamic_func(suffix=self.name)
+        def element_inner_weight(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            node_type, type_index = self.node_type_and_type_index(node_index_in_elt)
+
+            node_i, node_j = self._node_lobatto_indices(node_type, type_index)
+
+            if node_type == SquareSerendipityShapeFunctions.VERTEX:
+                cx = wp.select(node_i == 0, coords[0], 1.0 - coords[0])
+                cy = wp.select(node_j == 0, coords[1], 1.0 - coords[1])
+
+                w = cx * cy
+
+                if ORDER == 2:
+                    w *= cx + cy - 2.0 + LOBATTO_COORDS[1]
+                    return w * LAGRANGE_SCALE[0]
+                if ORDER == 3:
+                    w *= (cx - 0.5) * (cx - 0.5) + (cy - 0.5) * (cy - 0.5) - DEGREE_3_CIRCLE_RAD
+                    return w * DEGREE_3_CIRCLE_SCALE
+
+            w = float(1.0)
+            if node_type == SquareSerendipityShapeFunctions.EDGE_Y:
+                w *= wp.select(node_i == 0, coords[0], 1.0 - coords[0])
+            else:
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_i:
+                        w *= coords[0] - LOBATTO_COORDS[k]
+
+                w *= LAGRANGE_SCALE[node_i]
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_X:
+                w *= wp.select(node_j == 0, coords[1], 1.0 - coords[1])
+            else:
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_j:
+                        w *= coords[1] - LOBATTO_COORDS[k]
+                w *= LAGRANGE_SCALE[node_j]
+
+            return w
+
+        return element_inner_weight
+
+    def make_element_inner_weight_gradient(self):
+        ORDER = self.ORDER
+        ORDER_PLUS_ONE = self.ORDER_PLUS_ONE
+        LOBATTO_COORDS = self.LOBATTO_COORDS
+        LAGRANGE_SCALE = self.LAGRANGE_SCALE
+
+        DEGREE_3_CIRCLE_RAD = wp.constant(0.5**2 + (0.5 - LOBATTO_COORDS[1]) ** 2)
+        DEGREE_3_CIRCLE_SCALE = 1.0 / (0.5 - DEGREE_3_CIRCLE_RAD)
+
+        @cache.dynamic_func(suffix=self.name)
+        def element_inner_weight_gradient(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            node_type, type_index = self.node_type_and_type_index(node_index_in_elt)
+
+            node_i, node_j = self._node_lobatto_indices(node_type, type_index)
+
+            if node_type == SquareSerendipityShapeFunctions.VERTEX:
+                cx = wp.select(node_i == 0, coords[0], 1.0 - coords[0])
+                cy = wp.select(node_j == 0, coords[1], 1.0 - coords[1])
+
+                gx = wp.select(node_i == 0, 1.0, -1.0)
+                gy = wp.select(node_j == 0, 1.0, -1.0)
+
+                if ORDER == 2:
+                    w = cx + cy - 2.0 + LOBATTO_COORDS[1]
+                    grad_x = cy * gx * (w + cx)
+                    grad_y = cx * gy * (w + cy)
+
+                    return wp.vec2(grad_x, grad_y) * LAGRANGE_SCALE[0]
+
+                if ORDER == 3:
+                    w = (cx - 0.5) * (cx - 0.5) + (cy - 0.5) * (cy - 0.5) - DEGREE_3_CIRCLE_RAD
+
+                    dw_dcx = 2.0 * cx - 1.0
+                    dw_dcy = 2.0 * cy - 1.0
+                    grad_x = cy * gx * (w + cx * dw_dcx)
+                    grad_y = cx * gy * (w + cy * dw_dcy)
+
+                    return wp.vec2(grad_x, grad_y) * DEGREE_3_CIRCLE_SCALE
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_X:
+                prefix_x = wp.select(node_j == 0, coords[1], 1.0 - coords[1])
+            else:
+                prefix_x = LAGRANGE_SCALE[node_j]
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_j:
+                        prefix_x *= coords[1] - LOBATTO_COORDS[k]
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_Y:
+                prefix_y = wp.select(node_i == 0, coords[0], 1.0 - coords[0])
+            else:
+                prefix_y = LAGRANGE_SCALE[node_i]
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_i:
+                        prefix_y *= coords[0] - LOBATTO_COORDS[k]
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_X:
+                grad_y = wp.select(node_j == 0, 1.0, -1.0) * prefix_y
+            else:
+                prefix_y *= LAGRANGE_SCALE[node_j]
+                grad_y = float(0.0)
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_j:
+                        delta_y = coords[1] - LOBATTO_COORDS[k]
+                        grad_y = grad_y * delta_y + prefix_y
+                        prefix_y *= delta_y
+
+            if node_type == SquareSerendipityShapeFunctions.EDGE_Y:
+                grad_x = wp.select(node_i == 0, 1.0, -1.0) * prefix_x
+            else:
+                prefix_x *= LAGRANGE_SCALE[node_i]
+                grad_x = float(0.0)
+                for k in range(ORDER_PLUS_ONE):
+                    if k != node_i:
+                        delta_x = coords[0] - LOBATTO_COORDS[k]
+                        grad_x = grad_x * delta_x + prefix_x
+                        prefix_x *= delta_x
+
+            grad = wp.vec2(grad_x, grad_y)
+            return grad
+
+        return element_inner_weight_gradient
+
+    def element_node_triangulation(self):
+        if self.ORDER == 2:
+            element_triangles = [
+                [0, 4, 5],
+                [5, 4, 6],
+                [5, 6, 1],
+                [4, 2, 7],
+                [4, 7, 6],
+                [6, 7, 3],
+            ]
+        else:
+            element_triangles = [
+                [0, 4, 5],
+                [2, 7, 8],
+                [3, 10, 11],
+                [1, 9, 6],
+                [5, 6, 9],
+                [5, 4, 6],
+                [8, 11, 10],
+                [8, 7, 11],
+                [4, 8, 10],
+                [4, 10, 6],
+            ]
+
+        return element_triangles
+
+
+class SquareNonConformingPolynomialShapeFunctions:
+    # embeds the largest equilateral triangle centered at (0.5, 0.5) into the reference square
+    _tri_height = 0.75
+    _tri_side = 2.0 / math.sqrt(3.0) * _tri_height
+    _tri_to_square = np.array([[_tri_side, _tri_side / 2.0], [0.0, _tri_height]])
+
+    _TRI_OFFSET = wp.constant(wp.vec2(0.5 - 0.5 * _tri_side, 0.5 - _tri_height / 3.0))
+
+    def __init__(self, degree: int):
+        self._tri_shape = Triangle2DPolynomialShapeFunctions(degree=degree)
+        self.ORDER = self._tri_shape.ORDER
+        self.NODES_PER_ELEMENT = self._tri_shape.NODES_PER_ELEMENT
+
+        self.element_node_triangulation = self._tri_shape.element_node_triangulation
+
+    @property
+    def name(self) -> str:
+        return f"Square_P{self.ORDER}d"
+
+    def make_node_coords_in_element(self):
+        node_coords_in_tet = self._tri_shape.make_node_coords_in_element()
+
+        TRI_TO_SQUARE = wp.constant(wp.mat22(self._tri_to_square))
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_coords_in_element(
+            node_index_in_elt: int,
+        ):
+            tri_coords = node_coords_in_tet(node_index_in_elt)
+            coords = (
+                TRI_TO_SQUARE * wp.vec2(tri_coords[1], tri_coords[2])
+            ) + SquareNonConformingPolynomialShapeFunctions._TRI_OFFSET
+            return Coords(coords[0], coords[1], 0.0)
+
+        return node_coords_in_element
+
+    def make_node_quadrature_weight(self):
+        NODES_PER_ELEMENT = self.NODES_PER_ELEMENT
+
+        if self.ORDER == 2:
+            # Intrinsic quadrature (order 2)
+            @cache.dynamic_func(suffix=self.name)
+            def node_quadrature_weight_quadratic(
+                node_index_in_elt: int,
+            ):
+                node_type, type_index = self._tri_shape.node_type_and_type_index(node_index_in_elt)
+                if node_type == Triangle2DPolynomialShapeFunctions.VERTEX:
+                    return 0.18518521
+                return 0.14814811
+
+            return node_quadrature_weight_quadratic
+
+        @cache.dynamic_func(suffix=self.name)
+        def node_uniform_quadrature_weight(
+            node_index_in_elt: int,
+        ):
+            return 1.0 / float(NODES_PER_ELEMENT)
+
+        return node_uniform_quadrature_weight
+
+    def make_trace_node_quadrature_weight(self):
+        # Non-conforming, zero measure on sides
+
+        @wp.func
+        def zero(node_index_in_elt: int):
+            return 0.0
+
+        return zero
+
+    def make_element_inner_weight(self):
+        tri_inner_weight = self._tri_shape.make_element_inner_weight()
+
+        SQUARE_TO_TRI = wp.constant(wp.mat22(np.linalg.inv(self._tri_to_square)))
+
+        @cache.dynamic_func(suffix=self.name)
+        def element_inner_weight(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            tri_param = SQUARE_TO_TRI * (
+                wp.vec2(coords[0], coords[1]) - SquareNonConformingPolynomialShapeFunctions._TRI_OFFSET
+            )
+            tri_coords = Coords(1.0 - tri_param[0] - tri_param[1], tri_param[0], tri_param[1])
+
+            return tri_inner_weight(tri_coords, node_index_in_elt)
+
+        return element_inner_weight
+
+    def make_element_inner_weight_gradient(self):
+        tri_inner_weight_gradient = self._tri_shape.make_element_inner_weight_gradient()
+
+        SQUARE_TO_TRI = wp.constant(wp.mat22(np.linalg.inv(self._tri_to_square)))
+
+        @cache.dynamic_func(suffix=self.name)
+        def element_inner_weight_gradient(
+            coords: Coords,
+            node_index_in_elt: int,
+        ):
+            tri_param = SQUARE_TO_TRI * (
+                wp.vec2(coords[0], coords[1]) - SquareNonConformingPolynomialShapeFunctions._TRI_OFFSET
+            )
+            tri_coords = Coords(1.0 - tri_param[0] - tri_param[1], tri_param[0], tri_param[1])
+
+            grad = tri_inner_weight_gradient(tri_coords, node_index_in_elt)
+            return wp.transpose(SQUARE_TO_TRI) * grad
+
+        return element_inner_weight_gradient