Improved Memory Visualization and Multi-Layer Tiling Profiling (pulp-platform#56)

* Improve memory alloc visualization

* Multi-level profiling + Linting

* profilling string change to const static

* Fix profiling dual loop issue

* Fix README Status Badges

* Update CHANGELOG

* Align comment and type hint

* Refactor profiling methods in TilingPrototype

---------

Co-authored-by: Run Wang <52746141+SamanthaWangdl@users.noreply.github.com>

Author: Victor-Jung

CHANGELOG.md

@@ -7,6 +7,7 @@ This release containing major architectural changes, new platform support, enhan
 ### List of Pull Requests
 - Move PULP SDK to main branch/fork [#88](https://github.com/pulp-platform/Deeploy/pull/88)
 - Finite Lifetime for IO Tensors [#51](https://github.com/pulp-platform/Deeploy/pull/51)
+- Improved Memory Visualization and Multi-Layer Tiling Profiling [#56](https://github.com/pulp-platform/Deeploy/pull/56)
 - Fix Linting in CI and Reformat C Files [#86](https://github.com/pulp-platform/Deeploy/pull/86)
 - Fix Broken CMake Flow For pulp-sdk [#87](https://github.com/pulp-platform/Deeploy/pull/87)
 - Refactor Changelog For Release [#85](https://github.com/pulp-platform/Deeploy/pull/85)
@@ -62,6 +63,7 @@ This release containing major architectural changes, new platform support, enhan
 - Test the correcteness of the memory map generated by the tiler
 - Add attribute to `VariableBuffer` to distinguish I/Os
 - Add proper static memory allocation with finite lifetime for I/Os
+- The memory allocation visualization now displays the allocation for each level used
 - Tutorial section in the documentation
 - Guide on using the debug print topology pass and code transformation
 - VSCode configuration files for improved IDE support
@@ -176,6 +178,8 @@ This release containing major architectural changes, new platform support, enhan
 ### Changed
 - Moved PULP SDK from Victor-Jung/pulp-sdk branch deeploy to pulp-platform/pulp-sdk branch main.
 - Memory arena buffers are now declared at the beginning of the `InitNetwork` function
+- Tiling profiling is now an ON/OFF version where you get the I/O DMA time for each DMA call
+- The profiling strings are const static, such that they are stored in .rodata
 - Adapt the select docker image stage to also select a runner depending on ` github.repository`
 - Adapt the jobs and reusable workflows to use the selected runner.
 - Updated `README.md` description to use a persistent development container

Deeploy/DeeployTypes.py

@@ -69,9 +69,8 @@ class CodeGenVerbosity:
     Encapsulates verbosity options for downstream configuration
     """
 
-    tilingProfiling: Optional[str]  #: str: Specifies the name of the memory level on which to profile tiling
-    untiledProfiling: Optional[
-        bool] = None  #: str: Specifies the name of the memory level on which to profile untiled code
+    tilingProfiling: Optional[bool] = False  # Specifies if we should profile the tiling code
+    untiledProfiling: Optional[bool] = None  #  Specifies if we should profile the untilied code
 
 
 _NoVerbosity = CodeGenVerbosity(None)

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPClusterTiling.py

@@ -45,7 +45,7 @@ def apply(self,
               name: str,
               verbose: CodeGenVerbosity = _NoVerbosity) -> Tuple[NetworkContext, ExecutionBlock]:
 
-        if verbose.tilingProfiling == "L2":
+        if verbose.tilingProfiling:
             ctxt, executionBlock = self.profilingSB.apply(ctxt, executionBlock, name)
             ctxt, executionBlock = self.profilingDB.apply(ctxt, executionBlock, name)
         else:

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPClusterTilingDB.py

@@ -314,7 +314,8 @@ def _tilingLoop(self, ctxt: NetworkContext, executionBlock: ExecutionBlock,
         metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'] + "_L2",
                                   nodeOps = operatorRepresentation['nodeOps'],
                                   numTiles = len(tilingSchedule.outputLoadSchedule),
-                                  tileIdxVar = "TILING_I")
+                                  tileIdxVar = "TILING_I",
+                                  kernelLevelTiling = True)
 
         newExecutionBlock = self.generateAllTilingCode(executionBlock, metaInfo, ingressDMATransferCalls,
                                                        ingressDMAWaitStatements[-1:], ingressDMAUpdates,

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPClusterTilingSB.py

@@ -627,7 +627,8 @@ def _tilingLoop(self, ctxt: NetworkContext, executionBlock: ExecutionBlock,
         metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'] + "_L2",
                                   nodeOps = operatorRepresentation['nodeOps'],
                                   numTiles = len(tilingSchedule.outputLoadSchedule),
-                                  tileIdxVar = "TILING_I")
+                                  tileIdxVar = "TILING_I",
+                                  kernelLevelTiling = True)
 
         newExecutionBlock = self.generateAllTilingCode(executionBlock, metaInfo, ingressDMATransferCalls,
                                                        ingressDMAWaitStatements, ingressDMAUpdates,

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPL3Tiling.py

@@ -45,7 +45,7 @@ def apply(self,
               name: str,
               verbose: CodeGenVerbosity = _NoVerbosity) -> Tuple[NetworkContext, ExecutionBlock]:
 
-        if verbose.tilingProfiling == "L3":
+        if verbose.tilingProfiling:
             ctxt, executionBlock = self.profilingSB.apply(ctxt, executionBlock, name)
             ctxt, executionBlock = self.profilingDB.apply(ctxt, executionBlock, name)
         else:

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPL3TilingDB.py

@@ -284,7 +284,8 @@ def _tilingLoop(self, ctxt: NetworkContext, executionBlock: ExecutionBlock,
         metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'] + "_L3",
                                   nodeOps = operatorRepresentation['nodeOps'],
                                   numTiles = len(tilingSchedule.outputLoadSchedule),
-                                  tileIdxVar = "TILING_I")
+                                  tileIdxVar = "TILING_I",
+                                  kernelLevelTiling = False)
 
         newExecutionBlock = self.generateAllTilingCode(executionBlock, metaInfo, ingressDMATransferCalls,
                                                        ingressDMAWaitStatements, ingressDMAUpdates,

Deeploy/Targets/PULPOpen/CodeTransformationPasses/PULPL3TilingSB.py

@@ -421,10 +421,11 @@ def _tilingLoop(self, ctxt: NetworkContext, executionBlock: ExecutionBlock,
             })
         ]
 
-        metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'],
+        metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'] + "_L3",
                                   nodeOps = operatorRepresentation['nodeOps'],
                                   numTiles = len(tilingSchedule.outputLoadSchedule),
-                                  tileIdxVar = "TILING_I")
+                                  tileIdxVar = "TILING_I",
+                                  kernelLevelTiling = False)
 
         newExecutionBlock = self.generateAllTilingCode(executionBlock, metaInfo, ingressDMATransferCalls,
                                                        ingressDMAWaitStatements, ingressDMAUpdates,

Deeploy/Targets/Snitch/CodeTransformationPasses/SnitchClusterTiling.py

@@ -41,7 +41,7 @@ def apply(self,
               name: str,
               verbose: CodeGenVerbosity = _NoVerbosity) -> Tuple[NetworkContext, ExecutionBlock]:
 
-        if verbose.tilingProfiling == "L2":
+        if verbose.tilingProfiling:
             raise NotImplementedError("Profiling not implemented for L2")
             # ctxt, executionBlock = self.profilingSB.apply(ctxt, executionBlock, name)
         else:

Deeploy/Targets/Snitch/CodeTransformationPasses/SnitchClusterTilingSB.py

@@ -478,7 +478,8 @@ def _tilingLoop(self, ctxt: NetworkContext, executionBlock: ExecutionBlock,
         metaInfo = TilingMetaInfo(nodeName = operatorRepresentation['nodeName'] + "_L2",
                                   nodeOps = operatorRepresentation['nodeOps'],
                                   numTiles = len(tilingSchedule.outputLoadSchedule),
-                                  tileIdxVar = "TILING_I")
+                                  tileIdxVar = "TILING_I",
+                                  kernelLevelTiling = True)
 
         newExecutionBlock = self.generateAllTilingCode(executionBlock, metaInfo, ingressDMATransferCalls,
                                                        ingressDMAWaitStatements, ingressDMAUpdates,