cl-slug's functions declare their arguments' full types, which SBCL
checks on entry and warns about for a constant. half-float-bits takes
only a real that rounds to a finite half, and half-float-value only a
finite half's bits, so both signal a type-error where they signalled a
slug-error. The copy functions take a row up to 65535 and a count up
to 65536. The errors page says so and drops half-float-value's row,
and the API reference gives the ranges.
Follow cl-slug 8579836: the pixel shader is cl-slug/clsl, in clsl
The shader pages, the systems, the example that draws a ring on the GPU
and the requirements name clsl and cl-slug/clsl in place of cl-gsl and
cl-slug/gsl; the examples are in clsl's language, the ring's run as
shown. Vertex shaders take Vulkan's clip space, y down, which cl-gfx
and clsl use. A program that calls SLUG-COVERAGE needs the device
feature that keeps NaNs, as PROGRAM-FEATURES reports.
