Merge pull request #51 from optimsoc/rework_na

NoC fix and MP update to latest changes

external/lisnoc/rtl/router/lisnoc_router_input_route.v

@@ -95,18 +95,21 @@ module lisnoc_router_input_route (/*AUTOARG*/
    endgenerate
 
    // Some bit selections for better readability below
-   wire [1:0] flit_type;
-   assign flit_type = fifo_flit[flit_width-1:flit_width-2];
-
    wire [flit_data_width-1:0] flit_header;
    assign flit_header = fifo_flit[flit_data_width-1:0];
 
    wire [ph_dest_width-1:0]     flit_dest;
    assign flit_dest = flit_header[flit_data_width-1:flit_data_width-ph_dest_width];
 
+   wire 			is_last;
+   assign is_last = fifo_flit[flit_data_width+1];
+
+   reg                         is_first;
+   wire                        nxt_is_first;
+
    // Generating the current destination selection
    assign nxt_cur_select = ( // If there is no transfer we are waiting for or the active is finished ..
-                             (~active || (active && read)) &&
+                             (~active || (active && read && is_first)) &&
                              // .. and this is valid..
                              (fifo_valid)
                              // .. take selection from the lookup vector
@@ -129,6 +132,8 @@ module lisnoc_router_input_route (/*AUTOARG*/
                // .. the current request is not finished.
                (active & ~read);
 
+   assign nxt_is_first = active && read && is_last;
+
    // Pop from FIFO when active and successfully routed or not active and something waiting in FIFO.
    assign fifo_ready = (active ? read : 1 ) & fifo_valid;
 
@@ -143,10 +148,12 @@ module lisnoc_router_input_route (/*AUTOARG*/
          active         <= 1'b0;
          switch_request <= {directions{1'b0}};
          cur_select     <= {directions{1'b0}};
+         is_first       <= 0;
       end else begin
          cur_select     <= nxt_cur_select;
          switch_request <= nxt_switch_request;
          active         <= nxt_active;
+         is_first       <= nxt_is_first;
 
          // Additionally check whether
          if ((active && read) || !active)

src/soc/sw/baremetal-libs/src/libmp/control.c

@@ -26,6 +26,7 @@ void control_init() {
     optimsoc_mp_simple_addhandler(NOC_CLASS_FIFO, &control_msg_handler);
 
     optimsoc_mp_simple_enable(0);
+    optimsoc_mp_simple_enable(1);
     or1k_interrupts_enable();
 }
 
@@ -51,7 +52,7 @@ void control_msg_handler(uint32_t* buffer,size_t len) {
 
         // Return the get endpoint response to sender
         rbuffer[0] = (src << OPTIMSOC_DEST_LSB) |
-                     (1 << OPTIMSOC_CLASS_LSB) |
+                     (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                      (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                      (CTRL_REQUEST_GETEP_RESP << CTRL_REQUEST_LSB);
 
@@ -73,13 +74,13 @@ void control_msg_handler(uint32_t* buffer,size_t len) {
         }
 
         trace_ep_get_resp_send(src, (struct endpoint*) rbuffer[1]);
-        optimsoc_mp_simple_send(0,2,rbuffer);
+        optimsoc_mp_simple_send(1,2,rbuffer);
         break;
     }
     case CTRL_REQUEST_MSG_ALLOC_REQ:
     {
         rbuffer[0] = (src << OPTIMSOC_DEST_LSB) |
-                (1 << OPTIMSOC_CLASS_LSB) |
+                (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                 (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                 (CTRL_REQUEST_MSG_ALLOC_RESP << CTRL_REQUEST_LSB);
 
@@ -94,12 +95,12 @@ void control_msg_handler(uint32_t* buffer,size_t len) {
             rbuffer[1] = CTRL_REQUEST_ACK;
             rbuffer[2] = ptr;
             trace_msg_alloc_resp_send(src, ep, ptr);
-            optimsoc_mp_simple_send(0,3, rbuffer);
+            optimsoc_mp_simple_send(1,3, rbuffer);
 
         } else {
             rbuffer[1] = CTRL_REQUEST_NACK;
             trace_msg_alloc_resp_send(src, ep, -1);
-            optimsoc_mp_simple_send(0,2,rbuffer);
+            optimsoc_mp_simple_send(1,2,rbuffer);
         }
 
         break;
@@ -134,12 +135,12 @@ void control_msg_handler(uint32_t* buffer,size_t len) {
         ep->remote = (struct endpoint *) buffer[3];
 
         rbuffer[0] = (src << OPTIMSOC_DEST_LSB) |
-                (1 << OPTIMSOC_CLASS_LSB) |
+                (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                 (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                 (CTRL_REQUEST_CHAN_CONNECT_RESP << CTRL_REQUEST_LSB);
 
         rbuffer[1] = endpoint_channel_get_credit(ep);
-        optimsoc_mp_simple_send(0, 2, rbuffer);
+        optimsoc_mp_simple_send(1, 2, rbuffer);
         break;
     }
     case CTRL_REQUEST_CHAN_DATA:
@@ -253,7 +254,7 @@ uint32_t control_msg_alloc(struct endpoint_handle *to_ep, uint32_t size) {
         // Try to retrieve from remote
         // We do this as long as we do not get a valid handle back (-1)
         ctrl_request.buffer[0] = (to_ep->domain << OPTIMSOC_DEST_LSB) |
-                (1 << OPTIMSOC_CLASS_LSB) |
+                (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                 (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                 (CTRL_REQUEST_MSG_ALLOC_REQ << CTRL_REQUEST_LSB);
         ctrl_request.buffer[1] = (unsigned int) to_ep->ep;
@@ -290,7 +291,7 @@ void control_msg_data(struct endpoint_handle *ep, uint32_t address, void* buffer
 
     for (int i=0;i<words;i=i+wordsperpacket) {
         ctrl_request.buffer[0] = (ep->domain << OPTIMSOC_DEST_LSB) |
-                (1 << OPTIMSOC_CLASS_LSB) |
+                (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                 (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                 (CTRL_REQUEST_MSG_DATA << CTRL_REQUEST_LSB);
         ctrl_request.buffer[1] = (unsigned int) ep->ep;
@@ -310,7 +311,7 @@ void control_msg_data(struct endpoint_handle *ep, uint32_t address, void* buffer
     }
 
     ctrl_request.buffer[0] = (ep->domain << OPTIMSOC_DEST_LSB) |
-            (1 << OPTIMSOC_CLASS_LSB) |
+            (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
             (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
             (CTRL_REQUEST_MSG_COMPLETE << CTRL_REQUEST_LSB);
     ctrl_request.buffer[1] = (unsigned int) ep->ep;
@@ -326,7 +327,7 @@ void control_msg_data(struct endpoint_handle *ep, uint32_t address, void* buffer
 uint32_t control_channel_connect(struct endpoint_handle *from,
                                  struct endpoint_handle *to) {
     ctrl_request.buffer[0] = (to->domain << OPTIMSOC_DEST_LSB) |
-            (1 << OPTIMSOC_CLASS_LSB) |
+            (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
             (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
             (CTRL_REQUEST_CHAN_CONNECT_REQ << CTRL_REQUEST_LSB);
     ctrl_request.buffer[1] = (unsigned int) to->ep;
@@ -347,7 +348,7 @@ void control_channel_send(struct endpoint_handle *ep, uint8_t *data, uint32_t si
 
     for (int i=0;i<words;i=i+wordsperpacket) {
         ctrl_request.buffer[0] = (ep->domain << OPTIMSOC_DEST_LSB) |
-                (1 << OPTIMSOC_CLASS_LSB) |
+                (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
                 (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
                 (CTRL_REQUEST_CHAN_DATA << CTRL_REQUEST_LSB);
         ctrl_request.buffer[1] = (unsigned int) ep->ep;
@@ -372,7 +373,7 @@ void control_channel_send(struct endpoint_handle *ep, uint8_t *data, uint32_t si
 
 void control_channel_sendcredit(struct endpoint_handle *ep, int32_t credit) {
     ctrl_request.buffer[0] = (ep->ep->remotedomain << OPTIMSOC_DEST_LSB) |
-            (1 << OPTIMSOC_CLASS_LSB) |
+            (NOC_CLASS_FIFO << OPTIMSOC_CLASS_LSB) |
             (optimsoc_get_tileid() << OPTIMSOC_SRC_LSB) |
             (CTRL_REQUEST_CHAN_CREDIT << CTRL_REQUEST_LSB);
     ctrl_request.buffer[1] = (unsigned int) ep->ep->remote;

src/soc/sw/baremetal-libs/src/libmp/control.h

@@ -5,10 +5,10 @@
 
 #include "endpoints.h"
 
-#define NOC_CLASS_FIFO     1
-#define NOC_CLASS_DMA_REQ  2
-#define NOC_CLASS_DMA_RESP 3
-#define NOC_CLASS_MP       4
+#define NOC_CLASS_FIFO     0
+#define NOC_CLASS_DMA_REQ  1
+#define NOC_CLASS_DMA_RESP 2
+#define NOC_CLASS_MP       3
 
 // The protocol for the MCAPI message passing uses control
 // messages via the NoC.

test/systemtest/conftest.py

@@ -110,3 +110,16 @@ def baremetal_apps_hello_mpsimple(baremetal_apps):
     subprocess.check_output(cmd)
 
     return src_baremetal_apps_hello_mpsimple
+
+@pytest.fixture(scope="module")
+def baremetal_apps_hello_mp(baremetal_apps):
+    """
+    Module-scoped fixture: download and build the hello world mp example
+    from baremetal-apps
+    """
+
+    src_baremetal_apps_hello_mp = baremetal_apps.join('hello_mp')
+    cmd = ['make', '-C', str(src_baremetal_apps_hello_mp)]
+    subprocess.check_output(cmd)
+
+    return src_baremetal_apps_hello_mp

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.000

@@ -0,0 +1,5 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE
+[               66580, 0] Received from 1
+[               73466, 0] Received from 2
+[              175038, 0] Received from 3

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.001

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.002

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.003

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.004

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.005

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.006

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.data/test_tutorial8_hello_mp/stdout.007

@@ -0,0 +1,2 @@
+# OpTiMSoC trace_monitor stdout file
+# [TIME, CORE] MESSAGE

test/systemtest/test_tutorial.py

@@ -293,6 +293,23 @@ def test_tutorial6_hello_mpsimple(self, tmpdir, baremetal_apps_hello_mpsimple,
             assert util.matches_golden_reference(str(tmpdir), f,
                                                  filter_func=util.filter_timestamps)
 
+    def test_tutorial8_hello_mp(self, baremetal_apps_hello_mp, tmpdir):
+        """
+        Tutorial 8: Run hello_mp world application on 2x2 CCCC in Verilator
+        Memory loading is done through Verilator meminit.
+        """
+        # run simulation
+        cmd = ['{}/examples/sim/system_2x2_cccc/system_2x2_cccc_sim_dualcore'.format(os.environ['OPTIMSOC']),
+               '--meminit={}'.format(str(baremetal_apps_hello_mp.join('hello_mp.vmem')))]
+        subprocess.check_output(cmd, cwd=str(tmpdir))
+
+        # check all output files
+        for i in range(0, 7):
+            f = "stdout.{:03d}".format(i)
+            assert tmpdir.join(f).isfile()
+            assert util.matches_golden_reference(str(tmpdir), f,
+                                                 filter_func=util.filter_timestamps)
+
 class TestTutorialFpga:
     """
     Test all FPGA-based tutorial steps in the User Guide