multicore fix

runtime/intrinsics/vx_intrinsics.S

@@ -35,26 +35,50 @@ vx_join:
 .type vx_warp_id, @function
 .global vx_warp_id
 vx_warp_id:
-	csrr a0, CSR_LWID # read warp index
+	csrr a0, CSR_LWID
 	ret
 
 .type vx_warp_gid, @function
 .global vx_warp_gid
 vx_warp_gid:
-	csrr a0, CSR_GWID # read warp index
+	csrr a0, CSR_GWID
 	ret
 
 .type vx_thread_id, @function
 .global vx_thread_id
 vx_thread_id:
-	csrr a0, CSR_LTID # read thread index
+	csrr a0, CSR_LTID
 	ret
 
 .type vx_thread_gid, @function
 .global vx_thread_gid
 vx_thread_gid:
-	csrr a0, CSR_GTID # read thread index
+	csrr a0, CSR_GTID
 	ret
+
+.type vx_core_id, @function
+.global vx_core_id
+vx_core_id:
+	csrr a0, CSR_GCID
+	ret
+
+.type vx_num_threads, @function
+.global vx_num_threads
+vx_num_threads:
+    csrr a0, CSR_NT
+    ret
+
+.type vx_num_warps, @function
+.global vx_num_warps
+vx_num_warps:
+    csrr a0, CSR_NW
+    ret
+
+.type vx_num_cores, @function
+.global vx_num_cores
+vx_num_cores:
+    csrr a0, CSR_NC
+    ret
 	
 .type vx_num_cycles, @function
 .global vx_num_cycles

runtime/intrinsics/vx_intrinsics.h

@@ -8,37 +8,49 @@ extern "C" {
 #endif
 
 // Spawn warps
-void vx_wspawn(unsigned numWarps, unsigned PC_spawn);
+void vx_wspawn(int numWarps, int PC_spawn);
 
 // Set thread mask
-void vx_tmc(unsigned numThreads);
+void vx_tmc(int numThreads);
 
 // Warp Barrier
-void vx_barrier(unsigned barriedID, unsigned numWarps);
+void vx_barrier(int barriedID, int numWarps);
 
 // Split on a predicate
-void vx_split(unsigned predicate);
+void vx_split(int predicate);
 
 // Join
-void vx_join(void);
+void vx_join();
 
-// Return the warp thread index
-unsigned vx_thread_id(void);
+// Return the warp's unique thread id 
+int vx_thread_id();
 
-// Return the core warp index
-unsigned vx_warp_id(void);
+// Return the core's unique warp id
+int vx_warp_id();
 
-// Return processsor unique thread id
-unsigned vx_thread_gid(void);
+// Return processsor unique core id
+int vx_core_id();
 
-// Return processsor unique warp id
-unsigned vx_warp_gid(void);
+// Return processsor global thread id
+int vx_thread_gid();
 
-// Return number cycles
-unsigned vx_num_cycles(void);
+// Return processsor global warp id
+int vx_warp_gid();
 
-// Return number instructions
-unsigned vx_num_instrs(void);
+// Return the number of threads in a warp
+int vx_num_threads();
+
+// Return the number of warps in a core
+int vx_num_warps();
+
+// Return the number of cores in the processsor
+int vx_num_cores();
+
+// Return the number of cycles
+int vx_num_cycles();
+
+// Return the number of instructions
+int vx_num_instrs();
 
 #define __if(b) vx_split(b); \
                 if (b) 

runtime/startup/vx_start.S

@@ -5,7 +5,7 @@
 .type   _start, @function
 _start:
   la a1, vx_set_sp
-  li a0, NUM_WARPS    # activate all warps
+  csrr a0, CSR_NW  # get num warps
   .word 0x00b5106b # wspawn a0(numWarps), a1(PC SPAWN)
   jal vx_set_sp
   li a0, 1
@@ -32,12 +32,11 @@ _exit:
   li a0, 0
   .word 0x0005006b    # disable all threads
 
-
 .section .text
 .type vx_set_sp, @function
 .global vx_set_sp
 vx_set_sp:
-  li a0, NUM_THREADS
+  csrr a0, CSR_NT     # get num threads
   .word 0x0005006b    # activate all threads
   
   .option push
@@ -46,10 +45,10 @@ vx_set_sp:
     addi  gp, gp, %pcrel_lo(1b)
   .option pop
 
-  csrr a1, CSR_GTID    # get gtid
-  slli a1, a1, 10      # multiply tid by 1024
-  csrr a2, CSR_LTID    # get tid  
-  slli a2, a2, 2       # multiply tid by 4
+  csrr a1, CSR_GTID    # get global thread id
+  slli a1, a1, 10      # multiply by 1024
+  csrr a2, CSR_LTID    # get local thread id  
+  slli a2, a2, 2       # multiply by 4
   lui  sp, STACK_BASE_ADDR # load base sp
   sub  sp, sp, a1      # sub thread block
   add  sp, sp, a2      # reduce addr collision for perf