Is Blender's image input/output system default 32bit?

Hi there, is Blender’s image input/output system default 32bit? By this I mean does Blender operate internally on images and produce renders at a 32bit default?

Choosing an output format determines the bit depth, dynamic range and format of a rendered frame. What I’d like to know is whether the Blender internal operations that produce the data that get written to the chosen format are always at 32bit, or higher, or change according to which output format is chosen.

The Blender 2.42 release log “High Dynamic Range Graphics” seems to suggest that 32bit is the constant minimum at which Blender internally operates. Can somebody confirm or deny this? Here’s a snippet of text from the release log:
Blender’s internal “ImBuf” library, our image input/output system, has been recoded to transparently
support regular 32 bits graphics (4 x 8 bits) or floating point images storing 128 bits per pixel (4 x 32

On reading HDR images, even when they're for example 3 x 10 bits, the pixels are always converted internally to RGBA float values. Optionally, like while displaying the image in the UV Image editor, this then gets converted to regular 32 bits for faster display. When an image has float colors, all imaging functions in Blender default to use that. This includes the Video Sequence Editor, texture mapping, background images, and the Compositor.

Stephen Norrington