NVIDIAGH-439 added 2d SVD function

CHANGELOG.md

@@ -5,6 +5,7 @@
 ### Added
 
 - Added support for vec4f grid construction in `wp.Volume.allocate_by_tiles()`
+- Add 2D SVD `svd2` to support 2d simulations ([GH-436](https://github.com/NVIDIA/warp/issues/436)).
 
 ### Changed
 

docs/modules/functions.rst

@@ -578,6 +578,12 @@ Vector Math
     while the left and right basis vectors are returned in ``U`` and ``V``.
 
 
+.. py:function:: svd2(A: Matrix[2,2,Float], U: Matrix[2,2,Float], sigma: Vector[2,Float], V: Matrix[2,2,Scalar]) -> None
+
+    Compute the SVD of a 2x2 matrix ``A``. The singular values are returned in ``sigma``,
+    while the left and right basis vectors are returned in ``U`` and ``V``.
+
+
 .. py:function:: qr3(A: Matrix[3,3,Float], Q: Matrix[3,3,Float], R: Matrix[3,3,Float]) -> None
 
     Compute the QR decomposition of a 3x3 matrix ``A``. The orthogonal matrix is returned in ``Q``,

warp/builtins.py

@@ -1132,6 +1132,21 @@ def matrix_transform_dispatch_func(input_types: Mapping[str, type], return_type:
     while the left and right basis vectors are returned in ``U`` and ``V``.""",
 )
 
+add_builtin(
+    "svd2",
+    input_types={
+        "A": matrix(shape=(2, 2), dtype=Float),
+        "U": matrix(shape=(2, 2), dtype=Float),
+        "sigma": vector(length=2, dtype=Float),
+        "V": matrix(shape=(2, 2), dtype=Scalar),
+    },
+    value_type=None,
+    group="Vector Math",
+    export=False,
+    doc="""Compute the SVD of a 2x2 matrix ``A``. The singular values are returned in ``sigma``,
+    while the left and right basis vectors are returned in ``U`` and ``V``.""",
+)
+
 add_builtin(
     "qr3",
     input_types={

warp/native/svd.h

@@ -423,6 +423,62 @@ void _svd(// input A
     );
 }
 
+
+template<typename Type>
+inline CUDA_CALLABLE
+void _svd_2(// input A
+        Type a11, Type a12,
+        Type a21, Type a22,
+        // output U
+        Type &u11, Type &u12,
+        Type &u21, Type &u22,
+        // output S
+        Type &s11, Type &s12,
+        Type &s21, Type &s22,
+        // output V
+        Type &v11, Type &v12,
+        Type &v21, Type &v22)
+{
+    // Step 1: Compute ATA
+    Type ATA11 = a11 * a11 + a21 * a21;
+    Type ATA12 = a11 * a12 + a21 * a22;
+    Type ATA22 = a12 * a12 + a22 * a22;
+
+    // Step 2: Eigenanalysis
+    Type trace = ATA11 + ATA22;
+    Type det = ATA11 * ATA22 - ATA12 * ATA12;
+    Type sqrt_term = sqrt(trace * trace - Type(4.0) * det);
+    Type lambda1 = (trace + sqrt_term) * Type(0.5);
+    Type lambda2 = (trace - sqrt_term) * Type(0.5);
+
+    // Step 3: Singular values
+    Type sigma1 = sqrt(lambda1);
+    Type sigma2 = sqrt(lambda2);
+
+    // Step 4: Eigenvectors (find V)
+    Type v1x = ATA12, v1y = lambda1 - ATA11; // For first eigenvector
+    Type v2x = ATA12, v2y = lambda2 - ATA11; // For second eigenvector
+    Type norm1 = sqrt(v1x * v1x + v1y * v1y);
+    Type norm2 = sqrt(v2x * v2x + v2y * v2y);
+
+    v11 = v1x / norm1; v12 = v2x / norm2;
+    v21 = v1y / norm1; v22 = v2y / norm2;
+
+    // Step 5: Compute U
+    Type inv_sigma1 = (sigma1 > Type(1e-6)) ? Type(1.0) / sigma1 : Type(0.0);
+    Type inv_sigma2 = (sigma2 > Type(1e-6)) ? Type(1.0) / sigma2 : Type(0.0);
+
+    u11 = (a11 * v11 + a12 * v21) * inv_sigma1;
+    u12 = (a11 * v12 + a12 * v22) * inv_sigma2;
+    u21 = (a21 * v11 + a22 * v21) * inv_sigma1;
+    u22 = (a21 * v12 + a22 * v22) * inv_sigma2;
+
+    // Step 6: Set S
+    s11 = sigma1; s12 = Type(0.0);
+    s21 = Type(0.0); s22 = sigma2;
+}
+
+
 template<typename Type>
 inline CUDA_CALLABLE void svd3(const mat_t<3,3,Type>& A, mat_t<3,3,Type>& U, vec_t<3,Type>& sigma, mat_t<3,3,Type>& V) {
   Type s12, s13, s21, s23, s31, s32;
@@ -483,6 +539,66 @@ inline CUDA_CALLABLE void adj_svd3(const mat_t<3,3,Type>& A,
   adj_A = adj_A + (u_term + v_term + sigma_term);
 }
 
+template<typename Type>
+inline CUDA_CALLABLE void svd2(const mat_t<2,2,Type>& A, mat_t<2,2,Type>& U, vec_t<2,Type>& sigma, mat_t<2,2,Type>& V) {
+  Type s12, s21;
+  _svd_2(A.data[0][0], A.data[0][1],
+       A.data[1][0], A.data[1][1],
+
+       U.data[0][0], U.data[0][1],
+       U.data[1][0], U.data[1][1],
+
+       sigma[0], s12,
+       s21, sigma[1],
+
+       V.data[0][0], V.data[0][1],
+       V.data[1][0], V.data[1][1]);
+}
+
+template<typename Type>
+inline CUDA_CALLABLE void adj_svd2(const mat_t<2,2,Type>& A,
+                                   const mat_t<2,2,Type>& U,
+                                   const vec_t<2,Type>& sigma,
+                                   const mat_t<2,2,Type>& V,
+                                   mat_t<2,2,Type>& adj_A,
+                                   const mat_t<2,2,Type>& adj_U,
+                                   const vec_t<2,Type>& adj_sigma,
+                                   const mat_t<2,2,Type>& adj_V) {
+    Type s1_squared = sigma[0] * sigma[0];
+    Type s2_squared = sigma[1] * sigma[1];
+
+    // Compute inverse of (s1^2 - s2^2) if possible, use small epsilon to prevent division by zero
+    Type F01 = Type(1) / min(s2_squared - s1_squared, Type(-1e-6f));
+
+    // Construct the matrix F for the adjoint
+    mat_t<2,2,Type> F = mat_t<2,2,Type>(0.0, F01,
+                                        -F01, 0.0);
+
+    // Create a matrix to handle the adjoint of the singular values (diagonal matrix)
+    mat_t<2,2,Type> adj_sigma_mat = mat_t<2,2,Type>(adj_sigma[0], 0.0,
+                                                   0.0, adj_sigma[1]);
+
+    // Matrix for handling singular values (diagonal matrix with sigma values)
+    mat_t<2,2,Type> s_mat = mat_t<2,2,Type>(sigma[0], 0.0,
+                                            0.0, sigma[1]);
+
+    // Compute the transpose of U and V
+    mat_t<2,2,Type> UT = transpose(U);
+    mat_t<2,2,Type> VT = transpose(V);
+
+    // Compute the term for sigma (diagonal matrix of adjoint singular values)
+    mat_t<2,2,Type> sigma_term = mul(U, mul(adj_sigma_mat, VT));
+
+    // Compute the adjoint contributions for U (left singular vectors)
+    mat_t<2,2,Type> u_term = mul(mul(U, mul(cw_mul(F, (mul(UT, adj_U) - mul(transpose(adj_U), U))), s_mat)), VT);
+
+    // Compute the adjoint contributions for V (right singular vectors)
+    mat_t<2,2,Type> v_term = mul(U, mul(s_mat, mul(cw_mul(F, (mul(VT, adj_V) - mul(transpose(adj_V), V))), VT)));
+
+    // Combine the terms to compute the adjoint of A
+    adj_A = adj_A + (u_term + v_term + sigma_term);
+}
+
 
 template<typename Type>
 inline CUDA_CALLABLE void qr3(const mat_t<3,3,Type>& A, mat_t<3,3,Type>& Q, mat_t<3,3,Type>& R) {

warp/stubs.py

@@ -647,6 +647,14 @@ def svd3(A: Matrix[3, 3, Float], U: Matrix[3, 3, Float], sigma: Vector[3, Float]
     ...
 
 
+@over
+def svd2(A: Matrix[2, 2, Float], U: Matrix[2, 2, Float], sigma: Vector[2, Float], V: Matrix[2, 2, Scalar]):
+    """Compute the SVD of a 2x2 matrix ``A``. The singular values are returned in ``sigma``,
+    while the left and right basis vectors are returned in ``U`` and ``V``.
+    """
+    ...
+
+
 @over
 def qr3(A: Matrix[3, 3, Float], Q: Matrix[3, 3, Float], R: Matrix[3, 3, Float]):
     """Compute the QR decomposition of a 3x3 matrix ``A``. The orthogonal matrix is returned in ``Q``,

warp/tests/test_mat.py

@@ -987,6 +987,124 @@ def check_mat_svd(
                 assert_np_equal((plusval - minusval) / (2 * dx), m3grads[ii, jj], tol=fdtol)
 
 
+def test_svd_2D(test, device, dtype, register_kernels=False):
+    rng = np.random.default_rng(123)
+
+    tol = {
+        np.float16: 1.0e-3,
+        np.float32: 1.0e-6,
+        np.float64: 1.0e-6,
+    }.get(dtype, 0)
+
+    wptype = wp.types.np_dtype_to_warp_type[np.dtype(dtype)]
+    vec2 = wp.types.vector(length=2, dtype=wptype)
+    mat22 = wp.types.matrix(shape=(2, 2), dtype=wptype)
+
+    def check_mat_svd2(
+        m2: wp.array(dtype=mat22),
+        Uout: wp.array(dtype=mat22),
+        sigmaout: wp.array(dtype=vec2),
+        Vout: wp.array(dtype=mat22),
+        outcomponents: wp.array(dtype=wptype),
+    ):
+        U = mat22()
+        sigma = vec2()
+        V = mat22()
+
+        wp.svd2(m2[0], U, sigma, V)  # Assuming there's a 2D SVD kernel
+
+        Uout[0] = U
+        sigmaout[0] = sigma
+        Vout[0] = V
+
+        # multiply outputs by 2 so we've got something to backpropagate:
+        idx = 0
+        for i in range(2):
+            for j in range(2):
+                outcomponents[idx] = wptype(2) * U[i, j]
+                idx = idx + 1
+
+        for i in range(2):
+            outcomponents[idx] = wptype(2) * sigma[i]
+            idx = idx + 1
+
+        for i in range(2):
+            for j in range(2):
+                outcomponents[idx] = wptype(2) * V[i, j]
+                idx = idx + 1
+
+    kernel = getkernel(check_mat_svd2, suffix=dtype.__name__)
+
+    output_select_kernel = get_select_kernel(wptype)
+
+    if register_kernels:
+        return
+
+    m2 = wp.array(randvals(rng, [1, 2, 2], dtype) + np.eye(2), dtype=mat22, requires_grad=True, device=device)
+
+    outcomponents = wp.zeros(2 * 2 * 2 + 2, dtype=wptype, requires_grad=True, device=device)
+    Uout = wp.zeros(1, dtype=mat22, requires_grad=True, device=device)
+    sigmaout = wp.zeros(1, dtype=vec2, requires_grad=True, device=device)
+    Vout = wp.zeros(1, dtype=mat22, requires_grad=True, device=device)
+
+    wp.launch(kernel, dim=1, inputs=[m2], outputs=[Uout, sigmaout, Vout, outcomponents], device=device)
+
+    Uout_np = Uout.numpy()[0].astype(np.float64)
+    sigmaout_np = np.diag(sigmaout.numpy()[0].astype(np.float64))
+    Vout_np = Vout.numpy()[0].astype(np.float64)
+
+    assert_np_equal(
+        np.matmul(Uout_np, np.matmul(sigmaout_np, Vout_np.T)), m2.numpy()[0].astype(np.float64), tol=30 * tol
+    )
+
+    if dtype == np.float16:
+        # Skip gradient check for float16 due to rounding errors
+        return
+
+    # Check gradients:
+    out = wp.zeros(1, dtype=wptype, requires_grad=True, device=device)
+    idx = 0
+    for idx in range(2 * 2 + 2 + 2 * 2):
+        tape = wp.Tape()
+        with tape:
+            wp.launch(kernel, dim=1, inputs=[m2], outputs=[Uout, sigmaout, Vout, outcomponents], device=device)
+            wp.launch(output_select_kernel, dim=1, inputs=[outcomponents, idx], outputs=[out], device=device)
+        tape.backward(out)
+        m2grads = 1.0 * tape.gradients[m2].numpy()[0]
+
+        tape.zero()
+
+        dx = 0.0001
+        fdtol = 5.0e-4 if dtype == np.float64 else 2.0e-2
+        for ii in range(2):
+            for jj in range(2):
+                m2test = 1.0 * m2.numpy()
+                m2test[0, ii, jj] += dx
+                wp.launch(
+                    kernel,
+                    dim=1,
+                    inputs=[wp.array(m2test, dtype=mat22, device=device)],
+                    outputs=[Uout, sigmaout, Vout, outcomponents],
+                    device=device,
+                )
+                wp.launch(output_select_kernel, dim=1, inputs=[outcomponents, idx], outputs=[out], device=device)
+                plusval = out.numpy()[0]
+
+                m2test = 1.0 * m2.numpy()
+                m2test[0, ii, jj] -= dx
+                wp.launch(
+                    kernel,
+                    dim=1,
+                    inputs=[wp.array(m2test, dtype=mat22, device=device)],
+                    outputs=[Uout, sigmaout, Vout, outcomponents],
+                    device=device,
+                )
+                wp.launch(output_select_kernel, dim=1, inputs=[outcomponents, idx], outputs=[out], device=device)
+                minusval = out.numpy()[0]
+
+                assert_np_equal((plusval - minusval) / (2 * dx), m2grads[ii, jj], tol=fdtol)
+
+
 def test_qr(test, device, dtype, register_kernels=False):
     rng = np.random.default_rng(123)
 
@@ -1826,6 +1944,9 @@ def test_tpl_ops_with_anon(self):
         TestMat, f"test_inverse_{dtype.__name__}", test_inverse, devices=devices, dtype=dtype
     )
     add_function_test_register_kernel(TestMat, f"test_svd_{dtype.__name__}", test_svd, devices=devices, dtype=dtype)
+    add_function_test_register_kernel(
+        TestMat, f"test_svd_2D{dtype.__name__}", test_svd_2D, devices=devices, dtype=dtype
+    )
     add_function_test_register_kernel(TestMat, f"test_qr_{dtype.__name__}", test_qr, devices=devices, dtype=dtype)
     add_function_test_register_kernel(TestMat, f"test_eig_{dtype.__name__}", test_eig, devices=devices, dtype=dtype)
     add_function_test_register_kernel(