TriadicFrameworks
Обзор

t_Execution — TFT.OpenGPU.Stack.Module

1. Execution Intent#

The execution layer defines the AMD‑friendly GPU primitives that receive substrate intent and convert it into actual GPU work.
This layer is the visible execution surface of the module — the part that interacts with hardware, drivers, firmware, and virtualization systems.

Its purpose is to:

  • provide an open alternative to NVIDIA’s closed execution primitives
  • map TFT substrate primitives into GPU cycles, tiles, warps, and VRAM surfaces
  • support Cloud PC tiering through vGPU slices
  • enable remote reconstruction of dimensional intent
  • maintain RTT coherence across substrate and rendering layers

2. Execution Principles#

2.1 Visible View (Inverted Star Alignment)#

The execution layer is the outer star — the visible machinery that performs actual GPU work.

It receives:

  • dimensional primitives
  • substrate operators
  • primitive streams

And resolves them into:

  • kernels
  • tiles
  • warps
  • VRAM surfaces
  • DMA-safe buffers
  • vGPU slices

2.2 AMD-First Execution Model#

The execution layer is intentionally AMD-first:

  • ROCm compute primitives
  • VCE/NV12 encoding blocks
  • SR‑IOV virtualization
  • DMA-safe surface mapping
  • open driver stack
  • open firmware pathways

This avoids NVIDIA lock-in and aligns with the module’s open substrate philosophy.

2.3 Bounded Drift#

Execution reinterpretation of substrate primitives is allowed only within bounded drift:

  • reinterpretation must preserve dimensional intent
  • reinterpretation must remain structurally consistent
  • reinterpretation must not alter RTT coherence

3. Execution Primitive Families#

3.1 Compute Primitives#

  • Exec.Kernel — execution kernel
  • Exec.Tile — tile-level compute block
  • Exec.Warp — warp-equivalent execution group
  • Exec.Stream — execution stream for parallel work
  • Exec.Dispatch — dispatch primitive for substrate cycles

3.2 Memory Primitives#

  • Exec.VRAM.Surface — GPU surface allocation
  • Exec.VRAM.Region — dimensional region mapping
  • Exec.VRAM.Bind — bind substrate primitive to VRAM
  • Exec.DMA.Buffer — DMA-safe buffer for remote transport

3.3 Virtualization Primitives#

  • Exec.vGPU.Slice — vGPU slice allocation
  • Exec.vGPU.Cycle — execution cycle for Cloud PC tiering
  • Exec.vGPU.Priority — scheduling priority
  • Exec.vGPU.Isolation — SR‑IOV isolation boundary

3.4 Reconstruction Primitives#

  • Exec.Reconstruct.Surface — reconstruct dimensional surface
  • Exec.Reconstruct.Material — apply material intent
  • Exec.Reconstruct.Transform — apply dimensional transform
  • Exec.Reconstruct.Viewport — reconstruct viewport substrate

4. Substrate → Execution Mapping#

Execution primitives resolve substrate primitives through declared coherence:

Substrate Primitive Execution Primitive
Dim.Surface Exec.VRAM.Surface
Dim.Material Exec.Reconstruct.Material
Dim.Transform Exec.Reconstruct.Transform
Dim.Viewport Exec.Reconstruct.Viewport
Op.Cycle Exec.Dispatch / Exec.vGPU.Cycle
Stream.Primitive Exec.DMA.Buffer / Exec.VRAM.Bind

This mapping is declared, not inferred.

5. Execution and RDP Integration#

The execution layer defines the GPU-side reconstruction path for RDP:

  • substrate primitives → encoded as TFT primitive stream
  • stream transported via RDPEGFX or custom virtual channel
  • execution layer reconstructs dimensional intent using GPU primitives
  • rendering layer presents final view

Execution provides:

  • GPU-native decoding
  • VRAM surface allocation
  • DMA-safe buffer mapping
  • vGPU cycle scheduling
  • reconstruction fidelity

RDP retains:

  • TLS
  • CredSSP
  • NLA
  • reliability
  • ordering
  • session setup

6. Execution and Cloud PC Tiering#

Cloud PC tiers map directly to execution primitives:

  • Tier → vGPU slice
  • Tier → VRAM allocation
  • Tier → execution cycles
  • Tier → scheduling priority
  • Tier → reconstruction fidelity

Even minimal tiers achieve smooth remote rendering due to dimensional primitives.

7. AMD Partnership Surface#

The execution layer defines the AMD partnership surface:

  • ROCm alignment
  • VCE encoding integration
  • SR‑IOV virtualization
  • open driver pathways
  • open firmware primitives
  • Cloud PC differentiation
  • alternative to NVIDIA’s closed execution stack

This is the layer AMD can adopt under NDA.

8. Canonical Execution Summary#

The execution layer of the TFT.OpenGPU.Stack.Module:

  • defines AMD-friendly execution primitives
  • resolves substrate intent into GPU work
  • maps coherently into rendering and substrate layers
  • supports RDP integration
  • enables Cloud PC tiering
  • aligns with RTT (rtt=1, coherence declared, drift bounded, paradox structural)
  • establishes an open execution surface independent of NVIDIA’s closed ecosystem

This document captures the execution layer for canonical reference.

Updated

Работает на Docsbook