Update PERFORMANCE_AUDIT.md — P4 done, P5 kernel done (pending integration)
This commit is contained in:
@@ -224,9 +224,9 @@ post-fusion structure.
|
||||
| **KV-1** | FP4 storage for CSA main compressed KV | M | Huge at long context | Next | ✅ Done |
|
||||
| **KV-2** | FP4 storage for HCA compressed KV | M | Same pattern as KV-1 | After KV-1 | ✅ Done |
|
||||
| **KV-3** | FP4 storage for indexer keys (pair with E7) | M | Throughput + paper compliance | After KV-2 |✅ Done |
|
||||
| **P4** | RMSNorm fused into next quantize | S | 122 launches/token | After KV |
|
||||
| **P5** | mHC pre_block + RMSNorm fused | S | ~120 launches/token | After P4 |
|
||||
| **P6** | CUDA graph capture | L | **2–3× total** | After everything above |
|
||||
| **P4** | RMSNorm fused into next quantize | S | 122 launches/token | ✅ Done |
|
||||
| **P5** | mHC pre_block + RMSNorm fused | S | ~120 launches/token | ✅ Done (kernel, pending integration) |
|
||||
| **P6** | CUDA graph capture | L | **2–3× total** | Next |
|
||||
|
||||
|
||||
---
|
||||
|
||||
Reference in New Issue
Block a user