float32-filterable

Without this feature, the r32float, rg32float, and rgba32float formats can only be sampled unfiltered (nearest) - shaders that need bilinear lookups must emulate filtering with four taps and manual interpolation. With it, the sampler hardware filters float32 directly.

Typical users are scientific visualization, HDR pipelines that refuse half-float precision, terrain/heightfield rendering, and simulation readbacks. Emulation is straightforward but costs shader instructions and read bandwidth, so knowing real-world availability helps decide whether to ship the fallback path at all.

Desktop support is very common; some mobile GPUs still filter only 16-bit floats, which shows up clearly in the vendor split.