mod_eos_interp.t File Reference

Go to the source code of this file.

Modules
module	mod_eos_interp
	Interpolation kernels for the EoS tables (pure math; no EoS state).

Functions/Subroutines
pure integer function, public	mod_eos_interp::find_index_guard (nodes, n, val, guard, m, scale_m)
	Fast adaptive index lookup. Returns the smallest 1-based ii such that nodes(ii) >= val. Falls back to binary search if the guard array is not built (M = 0). Equivalent in result to find_index_bsearch but O(1) when the guard is present.
pure double precision function, public	mod_eos_interp::interp_clamped_bilinear_table (vary, varx, table, nx, ny, vmin_y, vmax_y, vmin_x, vmax_x)
pure double precision function, public	mod_eos_interp::bicubic_lookup (var1, var2, tc)
	Dispatch wrapper: bicubic lookup that branches on the table's is_uniform flag. Hides the uniform-vs-adaptive distinction from call sites – they pass (var1, var2, eos%.
pure double precision function, public	mod_eos_interp::bilinear_lookup (var1, var2, tc)
	Dispatch wrapper: bilinear lookup that branches on is_uniform.
subroutine, public	mod_eos_interp::interp_pchip_interleaved (vary, varx, table_il, nq, nx, ny, vmin_y, vmax_y, vmin_x, vmax_x, results)
	Interleaved PCHIP: evaluate N quantities at the same (vary, varx) point. Table layout: table_il(nq, ny, nx) where nq quantities share the same grid. All nq values at each grid point are contiguous in memory (cache-optimal).
subroutine, public	mod_eos_interp::interp_pchip_interleaved_nu (vary, varx, table_il, nq, nx, ny, var1_nodes, var2_nodes, guard_1, m_1, scale_1, guard_2, m_2, scale_2, results)
	Adaptive-grid sibling of interp_pchip_interleaved. Replaces the affine index calculation (x - x_min) * step_inv with binary searches on the explicit node arrays var{1,2}_nodes (Q axis-1 binary searches: O(log_2 N) comparisons each, total ~16 cmps for N = 256). The PCHIP kernel itself is unchanged: it operates on the local 4-tuple of values regardless of grid spacing. The stride-1 access pattern over the q-slot dimension is preserved, so cache behaviour matches the uniform variant for the inner kernel.
subroutine, public	mod_eos_interp::precompute_step_inv (tc)