Migrating all tests to new kernel launch API

Author: tinebp

ci/blackbox.sh

@@ -232,11 +232,11 @@ then
         if [ $HAS_ARGS -eq 1 ]
         then
             echo "running: VORTEX_RT_PATH=$TEMPDIR OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1"
-            VORTEX_RT_PATH=$TEMPDIR OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
+            DEBUG=1 VORTEX_RT_PATH=$TEMPDIR OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
             status=$?
         else
             echo "running: VORTEX_RT_PATH=$TEMPDIR make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1"
-            VORTEX_RT_PATH=$TEMPDIR make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
+            DEBUG=1 VORTEX_RT_PATH=$TEMPDIR make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
             status=$?
         fi
 
@@ -257,11 +257,11 @@ then
         if [ $HAS_ARGS -eq 1 ]
         then
             echo "running: OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1"
-            OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
+            DEBUG=1 OPTS=$ARGS make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
             status=$?
         else
             echo "running: make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1"
-            make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
+            DEBUG=1 make -C $APP_PATH run-$DRIVER > $LOGFILE 2>&1
             status=$?
         fi
     fi

ci/regression.sh

@@ -133,6 +133,9 @@ opencl()
     make -C tests/opencl run-simx
     make -C tests/opencl run-rtlsim
 
+    ./ci/blackbox.sh --driver=simx --app=lbm --warps=8
+    ./ci/blackbox.sh --driver=rtlsim --app=lbm --warps=8
+
     echo "opencl tests done!"
 }
 

hw/rtl/afu/opae/vortex_afu.sv

@@ -96,9 +96,8 @@ module vortex_afu import ccip_if_pkg::*; import local_mem_cfg_pkg::*; import VX_
 
     wire [127:0] afu_id = `AFU_ACCEL_UUID;
 
-    wire [63:0] dev_caps = {8'b0,
+    wire [63:0] dev_caps = {16'b0,
                             8'(`LMEM_ENABLED ? `LMEM_LOG_SIZE : 0),
-                            8'(`NUM_BARRIERS),
                             16'(`NUM_CORES * `NUM_CLUSTERS),
                             8'(`NUM_WARPS),
                             8'(`NUM_THREADS),

hw/rtl/afu/xrt/VX_afu_ctrl.sv

@@ -142,9 +142,8 @@ module VX_afu_ctrl #(
         RSTATE_DATA     = 2'd1;
 
     // device caps
-    wire [63:0] dev_caps = {8'b0,
+    wire [63:0] dev_caps = {16'b0,
                             8'(`LMEM_ENABLED ? `LMEM_LOG_SIZE : 0),
-                            8'(`NUM_BARRIERS),
                             16'(`NUM_CORES * `NUM_CLUSTERS),
                             8'(`NUM_WARPS),
                             8'(`NUM_THREADS),

kernel/include/vx_intrinsics.h

@@ -17,9 +17,17 @@
 #ifndef __VX_INTRINSICS_H__
 #define __VX_INTRINSICS_H__
 
-#include <stddef.h>
+#include <stdint.h>
 #include <VX_types.h>
 
+#if __riscv_xlen == 64
+typedef unsigned long size_t;  // 64-bit RISC-V
+#elif __riscv_xlen == 32
+typedef unsigned int size_t;   // 32-bit RISC-V
+#else
+#error "Unknown RISC-V architecture"
+#endif
+
 #if defined(__clang__)
 #define __UNIFORM__   __attribute__((annotate("vortex.uniform")))
 #else

kernel/include/vx_spawn.h

@@ -21,16 +21,6 @@
 extern "C" {
 #endif
 
-typedef void (*vx_spawn_tasks_cb)(uint32_t task_id, void *arg);
-
-typedef void (*vx_serial_cb)(void *arg);
-
-void vx_spawn_tasks(uint32_t num_tasks, vx_spawn_tasks_cb callback, const void * arg);
-
-void vx_serial(vx_serial_cb callback, const void * arg);
-
-///////////////////////////////////////////////////////////////////////////////
-
 typedef union {
   struct {
     uint32_t x;
@@ -46,23 +36,28 @@ extern dim3_t gridDim;
 extern dim3_t blockDim;
 
 extern __thread uint32_t __local_group_id;
-extern uint32_t __groups_per_core;
 extern uint32_t __warps_per_group;
 
 typedef void (*vx_kernel_func_cb)(void *arg);
 
+typedef void (*vx_serial_cb)(void *arg);
+
 #define __local_mem(size) \
   (void*)((int8_t*)csr_read(VX_CSR_LOCAL_MEM_BASE) + __local_group_id * size)
 
 #define __syncthreads() \
-  vx_barrier(__COUNTER__ * __groups_per_core + __local_group_id, __warps_per_group)
+  vx_barrier(__local_group_id, __warps_per_group)
 
+// launch a kernel function with a grid of blocks and block of threads
 int vx_spawn_threads(uint32_t dimension,
                      const uint32_t* grid_dim,
                      const uint32_t* block_dim,
                      vx_kernel_func_cb kernel_func,
                      const void* arg);
 
+// function call serialization
+void vx_serial(vx_serial_cb callback, const void * arg);
+
 #ifdef __cplusplus
 }
 #endif