rg11b10ufloat-renderable

The rg11b10ufloat format packs unsigned floating-point red, green, and blue into 32 bits per texel: 11 bits each for red and green, and 10 for blue. The feature permits GPUTextureUsage.RENDER_ATTACHMENT on these textures and enables blending, multisampling, and resolve for the format. Request it at device creation before using such attachments.

This is a useful intermediate target for bloom, lighting, and other HDR passes that need positive color values but no alpha channel. It uses half the texel storage of rgba16float, which can reduce bandwidth in a chain of full-screen passes. Actual performance still depends on the workload and GPU.

The format cannot represent negative values and has less precision than half-float RGB, so it is unsuitable for signed data or algorithms that need more accurate accumulation. Use rgba16float when alpha, signed values, or extra precision matter. Sampling an existing rg11b10ufloat texture and rendering into one are different capabilities; this token specifically expands attachment usage.