pyfock 0.2.4__tar.gz → 0.2.6__tar.gz

{pyfock-0.2.4/pyfock.egg-info → pyfock-0.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyfock
-Version: 0.2.4
+Version: 0.2.6
 Summary: A simplistic and efficient pure-python quantum chemistry library from Phys Whiz.
 Home-page: https://github.com/manassharma07/pyfock
 Download-URL: 
@@ -202,7 +202,7 @@ Dynamic: summary
 - ✅ **Effective Core Potentials**: Support for evaluation of ECP integrals
 - ✅ **Analytical gradients & forces**: Fast analytical nuclear gradients for density-fitted DFT — one-electron (overlap/kinetic/nuclear), DF Coulomb (3c2e + 2c2e), and XC for LDA, GGA and meta-GGA (native or LibXC) — matching PySCF forces and faster
 - ✅ **ASE Calculator**: Optional ASE interface (geometry optimization and the wider ASE ecosystem), using analytical forces by default
-- ✅ **XC integration grids**: Treutler-Ahlrichs radial and Lebedev angular grids with region-wise angular pruning and Becke partitioning (levels 0-9), built by a parallel Numba kernel or on the GPU (~6x faster, same grid); energies agree with PySCF to the SCF convergence threshold in our benchmarks; numgrid grids remain available as an alternative scheme
+- ✅ **XC integration grids**: Treutler-Ahlrichs radial and Lebedev angular grids with region-wise angular pruning and Becke partitioning (levels 0-9), built by a parallel Numba kernel that screens the partitioning (the grid of a 139-water cluster, 4.7 million points, in 7 s on 4 cores) or on the GPU (same grid); numgrid grids remain available as an alternative scheme
 - ✅ **SANO initial guess**: superposition of atomic natural-orbital densities (ANO-RCC-MB natural orbitals projected onto the calculation basis, no atomic SCF needed) as the default SCF starting point; it roughly halves the number of SCF iterations relative to the core-Hamiltonian guess and prunes the XC grid accurately
 - ✅ **Extrapolated starting densities in geometry optimizations**: `density_guess='extrapolate'` in the ASE calculator carries the converged densities of the last steps along with the atoms and extrapolates them to the new geometry (adapting the density projection/extrapolation code contributed by Prof. Vincenzo Barone), for 16-18 % fewer SCF iterations along the same optimization path
 - ✅ **Cross-Platform**: Works on Linux, macOS, and Windows
@@ -730,6 +730,28 @@ The GPU integral builder supports orbital and auxiliary shells through i (`l=6`)
 See [GPU design and validation](docs/df_algo11_gpu.md) for validation coverage,
 benchmark commands, memory accounting, and limitations of the surrounding driver.
 
+### Exchange-Correlation Term on the CPU
+
+The XC energy, potential matrix and nuclear gradient are evaluated by `XC_algo=3` by default on
+the CPU (`Integrals.eval_xc_3`, `Integrals.eval_xc_grad_3`). The grid is partitioned into compact
+leaves of at most 256 points by recursive bisection, every leaf carries the list of shells whose
+cutoff reaches it, and one Numba parallel region evaluates basis functions, densities, the
+functional and the two matrix products per leaf (BLAS with one thread each), with a static
+longest-processing-time schedule over the threads and a deterministic assembly of the matrix.
+All natively implemented functionals (LDA, GGA, meta-GGA, the global hybrids) are evaluated
+inside the kernel; LibXC functionals are evaluated between two passes. The previous
+batch-parallel algorithm stays available as `XC_algo=2`.
+
+```python
+dft_obj.XC_algo = 3            # default on the CPU
+dft_obj.save_ao_values = True  # optional AO cache, stored leaf by leaf (see docs/xc_algo3.md)
+```
+
+On a 47-water cluster (def2-SVP, 1175 basis functions, 1.28 million grid points, 24 cores) the
+XC term takes 0.29 s instead of 6.3 s (PBE), 0.27 s instead of 3.9 s (LDA) and 0.69 s instead of
+16.5 s (r2SCAN); see [docs/xc_algo3.md](docs/xc_algo3.md) for the design, the validation and the
+single-core, cache-mode and gradient numbers.
+
 ### Analytical Forces & Geometry Optimization
 
 After a converged DFT calculation, analytical nuclear gradients (and forces)

{pyfock-0.2.4 → pyfock-0.2.6}/README.md

@@ -139,7 +139,7 @@
 - ✅ **Effective Core Potentials**: Support for evaluation of ECP integrals
 - ✅ **Analytical gradients & forces**: Fast analytical nuclear gradients for density-fitted DFT — one-electron (overlap/kinetic/nuclear), DF Coulomb (3c2e + 2c2e), and XC for LDA, GGA and meta-GGA (native or LibXC) — matching PySCF forces and faster
 - ✅ **ASE Calculator**: Optional ASE interface (geometry optimization and the wider ASE ecosystem), using analytical forces by default
-- ✅ **XC integration grids**: Treutler-Ahlrichs radial and Lebedev angular grids with region-wise angular pruning and Becke partitioning (levels 0-9), built by a parallel Numba kernel or on the GPU (~6x faster, same grid); energies agree with PySCF to the SCF convergence threshold in our benchmarks; numgrid grids remain available as an alternative scheme
+- ✅ **XC integration grids**: Treutler-Ahlrichs radial and Lebedev angular grids with region-wise angular pruning and Becke partitioning (levels 0-9), built by a parallel Numba kernel that screens the partitioning (the grid of a 139-water cluster, 4.7 million points, in 7 s on 4 cores) or on the GPU (same grid); numgrid grids remain available as an alternative scheme
 - ✅ **SANO initial guess**: superposition of atomic natural-orbital densities (ANO-RCC-MB natural orbitals projected onto the calculation basis, no atomic SCF needed) as the default SCF starting point; it roughly halves the number of SCF iterations relative to the core-Hamiltonian guess and prunes the XC grid accurately
 - ✅ **Extrapolated starting densities in geometry optimizations**: `density_guess='extrapolate'` in the ASE calculator carries the converged densities of the last steps along with the atoms and extrapolates them to the new geometry (adapting the density projection/extrapolation code contributed by Prof. Vincenzo Barone), for 16-18 % fewer SCF iterations along the same optimization path
 - ✅ **Cross-Platform**: Works on Linux, macOS, and Windows
@@ -667,6 +667,28 @@ The GPU integral builder supports orbital and auxiliary shells through i (`l=6`)
 See [GPU design and validation](docs/df_algo11_gpu.md) for validation coverage,
 benchmark commands, memory accounting, and limitations of the surrounding driver.
 
+### Exchange-Correlation Term on the CPU
+
+The XC energy, potential matrix and nuclear gradient are evaluated by `XC_algo=3` by default on
+the CPU (`Integrals.eval_xc_3`, `Integrals.eval_xc_grad_3`). The grid is partitioned into compact
+leaves of at most 256 points by recursive bisection, every leaf carries the list of shells whose
+cutoff reaches it, and one Numba parallel region evaluates basis functions, densities, the
+functional and the two matrix products per leaf (BLAS with one thread each), with a static
+longest-processing-time schedule over the threads and a deterministic assembly of the matrix.
+All natively implemented functionals (LDA, GGA, meta-GGA, the global hybrids) are evaluated
+inside the kernel; LibXC functionals are evaluated between two passes. The previous
+batch-parallel algorithm stays available as `XC_algo=2`.
+
+```python
+dft_obj.XC_algo = 3            # default on the CPU
+dft_obj.save_ao_values = True  # optional AO cache, stored leaf by leaf (see docs/xc_algo3.md)
+```
+
+On a 47-water cluster (def2-SVP, 1175 basis functions, 1.28 million grid points, 24 cores) the
+XC term takes 0.29 s instead of 6.3 s (PBE), 0.27 s instead of 3.9 s (LDA) and 0.69 s instead of
+16.5 s (r2SCAN); see [docs/xc_algo3.md](docs/xc_algo3.md) for the design, the validation and the
+single-core, cache-mode and gradient numbers.
+
 ### Analytical Forces & Geometry Optimization
 
 After a converged DFT calculation, analytical nuclear gradients (and forces)