Remove ap_ctrl_none

ACCL @ 914a8b52

-Subproject commit 8c7e26f98ba8275ca5294fb7186ee8e9666ac786
+Subproject commit 914a8b520e62adbd4c02a8f50567928fe117d9c9

custom_hls/accl/funcs.hpp

@@ -32,10 +32,6 @@ void accl_out(
     bool leftover = num_bits % accl_width != 0;
     int num_transfer_bits = ((num_bits + accl_width - 1) / accl_width) * accl_width;
 
-    accl.stream_put(num_transfer_bits / 32, 9, destination, 0, false);
-
-    // TODO: Doing it like this is probably not optimal. It seems like we're reinventing a
-    // DWC here.
     send: for (int i = 0; i < num_bits - step + 1; i += step) {
         if (i % stream_width == 0) {
             stream_word = in.read();
@@ -55,6 +51,8 @@ void accl_out(
         data.push(accl_word, 0);
     }
 
+    accl.stream_put(num_transfer_bits / 32, 9, destination, 0, false);
+
 #ifdef CPPSIM
     std::cerr << "accl_out waiting on ack" << std::endl;
 #endif

src/finn/custom_op/fpgadataflow/accl.py

@@ -75,8 +75,12 @@ class ACCLOp(HLSCustomOp):
                 emulator_dir = f"{os.environ['FINN_ROOT']}/ACCL/test/model/emulator"
                 world_size = self.get_nodeattr("worldSize")
 
-                subprocess.run(["/usr/bin/cmake", "."],
-                               cwd=emulator_dir, stdout=subprocess.PIPE)
+                # Make sure the emulator itself is built
+                subprocess.run(
+                    ["/usr/bin/cmake", "."],
+                    cwd=emulator_dir,
+                    stdout=subprocess.PIPE
+                )
 
                 emulator = subprocess.Popen([
                     "python3",
@@ -88,10 +92,7 @@ class ACCLOp(HLSCustomOp):
             executable_path = self.get_nodeattr("executable_path")
             if executable_path == "":
                 raise Exception(
-                    """
-Found no executable for this node, did you run the codegen and
-compilation transformations?
-                """
+                    f"Executable for {self.onnx_node.name} at {executable_path} seems to be missing."
                 )
 
             p = subprocess.Popen(
@@ -225,8 +226,6 @@ class ACCLOut(ACCLOp):
             "#pragma HLS INTERFACE s_axilite port=comm_adr bundle=control",
         ]
 
-        self.code_gen_dict["$PRAGMAS$"].append("#pragma HLS INTERFACE ap_ctrl_none port=return")
-
     def strm_decl(self):
         start_port = self.get_nodeattr("startPort")
         rank = self.get_nodeattr("device_id")