better reports refactor for more code-reuse and patter usage fixes
This commit is contained in:
@@ -81,7 +81,11 @@ ONNX-MLIR ──► Spatial ──► Pim (tensor) ──► Pim (bufferized)
|
||||
standard MLIR `BufferizableOpInterface` machinery
|
||||
(`OpBufferizationInterfaces.*`, `PimBufferization.td`).
|
||||
|
||||
5. **PIM code generation** (`src/PIM/Pass/PimCodegen`):
|
||||
5. **Static memory coalescing** (`src/PIM/Dialect/Pim/Transforms/StaticMemoryCoalescing`).
|
||||
Conservatively reuses same-typed local memref allocations inside PIM cores
|
||||
after bufferization and before code generation.
|
||||
|
||||
6. **PIM code generation** (`src/PIM/Pass/PimCodegen`):
|
||||
- `HostConstantFolding` — folds host-side constants.
|
||||
- `MaterializeHostConstantsPass` — materializes the remaining host
|
||||
constants for emission.
|
||||
|
||||
Reference in New Issue
Block a user