unrolling the complicated macros in fdr/teddy.c which are opaque to debugging.
There's reason to believe there is some legacy cruft hiding in parts of this but for now
the functionization has been kept pretty much a direct copy.
While the objective of this PR is maintainability and manageability of the codebase,
there have been a few positive side-effects on performance. specifically performance in SSE2 , AVX512 , and ppc
environments has seen a very slight boost (between 0.1% - 0.5% improvement). performance in other environments is essentially unchanged.
unrolling the complicated macros in fdr/teddy.c which are opaque to debugging. There's reason to believe there is some legacy cruft hiding in parts of this but for now the functionization has been kept pretty much a direct copy. While the objective of this PR is maintainability and manageability of the codebase, there have been a few positive side-effects on performance. specifically performance in SSE2 , AVX512 , and ppc environments has seen a very slight boost (between 0.1% - 0.5% improvement). performance in other environments is essentially unchanged.