Benchmark Measurements and Reference Screenshots for Five Bloom Post-Processing Techniques on a MediaTek Dimensity 8300 Mobile GPU

This dataset contains raw benchmark measurements and reference screenshots from a comparative study of five bloom (glow) post-processing techniques implemented in OpenGL ES 3.1 and executed on a mobile tile-based GPU.

The five techniques evaluated are: separable Gaussian blur, Kawase blur, mipmap-chain downsampling, compute shader blur, and a hybrid Kawase pipeline with HDR tone mapping. All techniques were measured on a MediaTek Dimensity 8300 system-on-chip under controlled thermal conditions.

Contents

Excel spreadsheets containing per-frame GPU timing measurements (captured via GL_TIME_ELAPSED timer queries) and thermal headroom traces sampled through the Android PowerManager API across all five techniques and both test scenes.

Reference screenshots of the two test scenes (animated_wire and filled_dense) rendered with each of the five bloom techniques, used for the visual quality comparison reported in the accompanying paper.

Intended use

The dataset supports reproduction and verification of the timing, thermal, and visual quality results reported in the associated publication, and may be reused for further analysis of mobile GPU post-processing performance, thermal throttling behaviour on tile-based architectures, or visual quality comparisons across blur kernels.