From a3553b57e208f98a2e32ea7ff5110b0e4d717d81 Mon Sep 17 00:00:00 2001
From: Tobi-Alonso <tobi.alonso@gmail.com>
Date: Tue, 27 Oct 2020 14:02:16 +0000
Subject: [PATCH] [fpgadataflow] Add suport for non linear graphs in InsertDWC
transf (#227)
---
.../transformation/fpgadataflow/insert_dwc.py | 94 +++++++++++--------
1 file changed, 53 insertions(+), 41 deletions(-)
diff --git a/src/finn/transformation/fpgadataflow/insert_dwc.py b/src/finn/transformation/fpgadataflow/insert_dwc.py
index 195a005ff..b4b577605 100644
--- a/src/finn/transformation/fpgadataflow/insert_dwc.py
+++ b/src/finn/transformation/fpgadataflow/insert_dwc.py
@@ -4,6 +4,7 @@ from onnx import helper as oh
from finn.custom_op.registry import getCustomOp
from finn.transformation.base import Transformation
from finn.util.fpgadataflow import is_fpgadataflow_node
+import warnings
def _is_dwc_node(node):
@@ -40,48 +41,59 @@ class InsertDWC(Transformation):
for n in graph.node:
node_ind += 1
if _suitable_node(n):
- n_output = n.output[0]
- consumer = model.find_consumer(n_output)
- if _suitable_node(consumer) is True:
- n0 = getCustomOp(n)
- n1 = getCustomOp(consumer)
- n0_out_shape = n0.get_folded_output_shape()
- n1_in_shape = n1.get_folded_input_shape()
- if n0_out_shape[-1] != n1_in_shape[-1]:
- graph_modified = True
- # determine dwc inwidth
- dwc_in_width = n0.get_outstream_width()
- # determine dwc outwidth
- dwc_out_width = n1.get_instream_width()
-
- # determine shape for dwc
- dwc_shape = n0.get_normal_output_shape()
-
- # determine dtype for dwc
- dtype = n0.get_output_datatype()
-
- dwc_output_tensor = oh.make_tensor_value_info(
- model.make_new_valueinfo_name(),
- TensorProto.FLOAT,
- dwc_shape,
+ for n_output in n.output:
+ consumers = model.find_consumers(n_output)
+ if consumers is None:
+ continue
+ if len(consumers) > 1:
+ warnings.warn(
+ n.name
+ + ": HLS node with fan-out higher than 1 cannot be stitched"
)
- graph.value_info.append(dwc_output_tensor)
-
- dwc_node = oh.make_node(
- "StreamingDataWidthConverter_Batch",
- [n_output],
- [dwc_output_tensor.name],
- domain="finn",
- backend="fpgadataflow",
- shape=dwc_shape,
- inWidth=dwc_in_width,
- outWidth=dwc_out_width,
- dataType=str(dtype.name),
- )
- # insert dwc
- graph.node.insert(node_ind + 1, dwc_node)
- # set dwc output tensor as new input tensor of second node
- consumer.input[0] = dwc_output_tensor.name
+ consumer = consumers[0]
+ if _suitable_node(consumer) is True:
+ n0 = getCustomOp(n)
+ n1 = getCustomOp(consumer)
+ n0_out_shape = n0.get_folded_output_shape()
+ n1_in_shape = n1.get_folded_input_shape()
+ if n0_out_shape[-1] != n1_in_shape[-1]:
+ graph_modified = True
+ # determine dwc inwidth
+ dwc_in_width = n0.get_outstream_width()
+ # determine dwc outwidth
+ dwc_out_width = n1.get_instream_width()
+
+ # determine shape for dwc
+ dwc_shape = n0.get_normal_output_shape()
+
+ # determine dtype for dwc
+ dtype = n0.get_output_datatype()
+
+ dwc_output_tensor = oh.make_tensor_value_info(
+ model.make_new_valueinfo_name(),
+ TensorProto.FLOAT,
+ dwc_shape,
+ )
+ graph.value_info.append(dwc_output_tensor)
+
+ dwc_node = oh.make_node(
+ "StreamingDataWidthConverter_Batch",
+ [n_output],
+ [dwc_output_tensor.name],
+ domain="finn",
+ backend="fpgadataflow",
+ shape=dwc_shape,
+ inWidth=dwc_in_width,
+ outWidth=dwc_out_width,
+ dataType=str(dtype.name),
+ )
+ # insert dwc
+ graph.node.insert(node_ind + 1, dwc_node)
+
+ # set dwc output tensor as new input tensor of second node
+ for idx, inp in enumerate(consumer.input):
+ if inp == n_output:
+ consumer.input[idx] = dwc_output_tensor.name
return (model, graph_modified)
--
GitLab