[Test] use binaryXnorMode feature to test bipolar by bipolar

tests/fpgadataflow/test_fpgadataflow_fclayer.py

@@ -27,6 +27,21 @@ def make_single_fclayer_modelwrapper(W, pe, simd, wdt, idt, odt, T=None, tdt=Non
     else:
         tmem = 0
 
+    # there are two ways to implement bipolar weights and inputs for
+    # StreamingFC:
+    # - specify their datatypes as such
+    # - specify their datatypes as BINARY as use binaryXnorMode
+    if wdt == DataType.BIPOLAR and idt == DataType.BIPOLAR:
+        # we'll internally convert weights/inputs to binary and specify the
+        # datatypes as such, and also set the binaryXnorMode attribute to 1
+        export_wdt = DataType.BINARY
+        export_idt = DataType.BINARY
+        binary_xnor_mode = 1
+    else:
+        export_wdt = wdt
+        export_idt = idt
+        binary_xnor_mode = 0
+
     inp = helper.make_tensor_value_info("inp", TensorProto.FLOAT, [1, mw])
     outp = helper.make_tensor_value_info("outp", TensorProto.FLOAT, [1, mh])
     if T is not None:
@@ -52,10 +67,11 @@ def make_single_fclayer_modelwrapper(W, pe, simd, wdt, idt, odt, T=None, tdt=Non
         PE=pe,
         WMEM=wmem,
         TMEM=tmem,
-        inputDataType=idt.name,
-        weightDataType=wdt.name,
+        inputDataType=export_idt.name,
+        weightDataType=export_wdt.name,
         outputDataType=odt.name,
         ActVal=actval,
+        binaryXnorMode=binary_xnor_mode,
     )
     graph = helper.make_graph(
         nodes=[FCLayer_node], name="fclayer_graph", inputs=[inp], outputs=[outp]
@@ -67,15 +83,23 @@ def make_single_fclayer_modelwrapper(W, pe, simd, wdt, idt, odt, T=None, tdt=Non
     model.set_tensor_datatype("inp", idt)
     model.set_tensor_datatype("outp", odt)
     model.set_tensor_datatype("weights", wdt)
-    model.set_initializer("weights", W)
+    if binary_xnor_mode:
+        # convert bipolar to binary
+        model.set_initializer("weights", (W + 1) / 2)
+    else:
+        model.set_initializer("weights", W)
     if T is not None:
         model.set_tensor_datatype("thresh", tdt)
         model.set_initializer("thresh", T)
     return model
 
 
-def prepare_inputs(model, input_tensor, idt):
-    return {"inp": input_tensor}
+def prepare_inputs(input_tensor, idt, wdt):
+    if wdt == DataType.BIPOLAR and idt == DataType.BIPOLAR:
+        # convert bipolar to binary
+        return {"inp": (input_tensor + 1) / 2}
+    else:
+        return {"inp": input_tensor}
 
 
 # activation: None or DataType
@@ -132,7 +156,7 @@ def test_fpgadataflow_fclayer(idt, wdt, act, nf, sf, mw, mh):
     model = model.transform(CodeGen())
     model = model.transform(Compile())
     # prepare input data
-    input_dict = prepare_inputs(model, x, idt)
+    input_dict = prepare_inputs(x, idt, wdt)
     if wdt == DataType.BIPOLAR and idt == DataType.BIPOLAR:
         # convert inputs to binary and use xnorpopcountmatmul
         y = xp.xnorpopcountmatmul((x + 1) / 2, (W + 1) / 2)