C# / HLSL / D3D12

HLSL Kernel Pipeline

Engine-neutral GPU primitives with explicit external backends, whole-operation measurement and correctness checks.

Public source · Current main; historical results retain their own source identity

Work and recorded results

• Built a D3D12/HLSL primitives SDK with explicit RTS and AMD backends. A 70-process native confirmation found a 1.2404x ratio in one 2^25-pair sort case; six other comparisons favored external implementations.

Scope of the evidence

• The September 9 result is an exact R9700 native workload comparison. The application recording example and Unity deployment remain separately unmeasured.
• The older 5.10 to 2.74 ms (1.86x) result is a synthetic 12-frame, 8.39M-seed GPU pipeline record. Candidate executions are not unique candidate designs or full-game FPS.

Source and evidence

Evidence source revision: b52d18aff221f3f8822ddb54bc679c9e03932e2a

Public source at this revision ↗

• Seven native comparisons and controls
• Historical synthetic showcase

Updated 2026-09-14. CV wording and this page use the same project manifest.