v0.13.0

SymBoltz v0.13.0

Diff since v0.12.0

Breaking changes

• Support sparse Jacobians (for non-differentiable runs)
• Compute matter power spectrum with adaptive k-refinement
• Toggle multithreading with keyword argument at runtime

Merged pull requests:

• Toggle threading at runtime with keyword argument (#50) (@hersle)
• Support sparse Jacobians (#51) (@hersle)
• Compute matter power spectrum with adaptively refined k-grid (#52) (@hersle)

v0.12.0

SymBoltz v0.12.0

Diff since v0.11.0

Breaking changes

• Use symbolically generated Jacobian by default
• Improve multithreading performance of adaptive source grid (slightly)

Merged pull requests:

• Rewrite multithreaded adaptive source grid to use non-recursive Channel work queue (#48) (@hersle)
• Use symbolic Jacobian by default (#49) (@hersle)

v0.11.0

SymBoltz v0.11.0

Diff since v0.10.1

Breaking changes

• Improve load-balancing of parallelization of perturbation modes over threads
• Line-of-sight integrate several source functions simultaneously
• Only save final times when computing matter power spectrum

Merged pull requests:

• Line-of-sight integrate several source functions simultaneously (#45) (@hersle)
• Save only final times when evaluating matter power spectrum (#46) (@hersle)
• Improve load-balancing of perturbations parallelization (#47) (@hersle)

v0.10.1

SymBoltz v0.10.1

Diff since v0.10.0

Merged pull requests:

• Minor performance improvement and consistency improvements (#44) (@hersle)

v0.10.0

SymBoltz v0.10.0

Diff since v0.9.1

Breaking changes

• Add adaptive k-sampling of given source functions
• Compute line-of-sight integral and spectra of lensing potential

Merged pull requests:

• Add CMB lensing power spectrum (#36) (@hersle)
• Improve sampling of perturbation modes for angular spectra (#41) (@hersle)
• Interpolate smoother polarization source function (#42) (@hersle)

Closed issues:

• Compute lensed CMB power spectrum (#35)
• Optimize sampling of perturbation modes (#40)

v0.9.1

SymBoltz v0.9.1

Diff since v0.9.0

v0.9.0

SymBoltz v0.9.0

Diff since v0.8.2

Improvements

• Improve performance of perturbations integration

Breaking changes

• Cache spherical Bessel function values in line-of-sight integral

Merged pull requests:

• Add paper (#29) (@hersle)
• Tune solver settings for performance (#32) (@hersle)
• Cache spherical Bessel functions in line-of-sight integral (#38) (@hersle)
• Optimize f and use one vector-valued spline for all background unknowns (#39) (@hersle)

Closed issues:

• Spline evaluation is unnecessarily slow (#3)
• Analytical Jacobian does not work with massive neutrinos (#5)
• Tune solver settings for performance (#31)
• Experiment with Magnus methods (#33)

v0.8.2

SymBoltz v0.8.2

Diff since v0.8.1

Merged pull requests:

• Disable precompilation (#28) (@hersle)

v0.8.1

SymBoltz v0.8.1

Diff since v0.8.0

Merged pull requests:

• Add unstructured ΛCDM model example (#27) (@hersle)

v0.8.0

SymBoltz v0.8.0

Diff since v0.7.0

Breaking changes

• Use nonzero initial conformal time and solve for initial radiation-dominated scale factor
• Use proper 1/τ lmax multipole cutoff instead of ℰ
• Use k*τ instead of k/ℰ in initial conditions
• Change variables in massive neutrino momentum quadrature to transform infinite domain into numerically well-defined finite domain
• Default to 4 instead of 5 massive neutrino momenta
• Fix MCMC example in documentation
• Add equal-mass neutrino degeneracy parameter
• Rename E and ℰ to H and ℋ for improved notation consistency

Merged pull requests:

• Use nonzero kτ instead of k/ℰ in initial conditions (#19) (@hersle)
• Change variables in momentum quadrature to get finite interval (#21) (@hersle)
• Initialize scale factor in background with constraint ℰ=1/τ (#23) (@hersle)
• Add neutrino degeneracy parameter (#24) (@hersle)
• Refactor to use only one Hubble function in units where H0 = 1 (#25) (@hersle)

Closed issues:

• Degenerate massive neutrino hierarchy (#14)