fixed global obejct sharing between cores

benchmarks/opencl/bfs/Makefile

@@ -52,7 +52,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend 
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/convolution/Makefile

@@ -52,7 +52,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/guassian/Makefile

@@ -55,7 +55,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/kmeans/Makefile

@@ -52,7 +52,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend 
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/nearn/Makefile

@@ -55,7 +55,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend 
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/saxpy/Makefile

@@ -54,7 +54,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/sfilter/Makefile

@@ -54,7 +54,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend 
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/sgemm/Makefile

@@ -54,7 +54,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend
 
 clean-all: clean
-	rm *.pocl *.dump 
+	rm -rf *.pocl *.dump 
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/transpose/Makefile

@@ -52,7 +52,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

benchmarks/opencl/vecadd/Makefile

@@ -54,7 +54,7 @@ clean:
 	rm -rf $(PROJECT) *.o .depend 
 
 clean-all: clean
-	rm *.pocl *.dump
+	rm -rf *.pocl *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

driver/tests/basic/Makefile

@@ -61,8 +61,8 @@ run-simx: $(PROJECT)
 clean:
 	rm -rf $(PROJECT) *.o .depend
 
-clean-all:
-	rm -rf $(PROJECT) *.o *.elf *.bin *.dump .depend
+clean-all: clean
+	rm -rf *.elf *.bin *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

driver/tests/basic/kernel.dump

@@ -390,7 +390,7 @@ Disassembly of section .text:
 80000518:	00492703          	lw	a4,4(s2)
 8000051c:	148a2783          	lw	a5,328(s4)
 80000520:	01871463          	bne	a4,s8,80000528 <__call_exitprocs+0xe4>
-80000524:	f92784e3          	beq	a5,s2,800004ac <__call_exitprocs+0x68>
+80000524:	f8f904e3          	beq	s2,a5,800004ac <__call_exitprocs+0x68>
 80000528:	f80788e3          	beqz	a5,800004b8 <__call_exitprocs+0x74>
 8000052c:	00078913          	mv	s2,a5
 80000530:	f5dff06f          	j	8000048c <__call_exitprocs+0x48>
@@ -450,20 +450,21 @@ Disassembly of section .comment:
    0:	3a434347          	fmsub.d	ft6,ft6,ft4,ft7,rmm
    4:	2820                	fld	fs0,80(s0)
    6:	29554e47          	fmsub.s	ft8,fa0,fs5,ft5,rmm
-   a:	3120                	fld	fs0,96(a0)
-   c:	2e30                	fld	fa2,88(a2)
-   e:	2e32                	fld	ft8,264(sp)
-  10:	0030                	addi	a2,sp,8
+   a:	3920                	fld	fs0,112(a0)
+   c:	322e                	fld	ft4,232(sp)
+   e:	302e                	fld	ft0,232(sp)
+	...
 
 Disassembly of section .riscv.attributes:
 
 00000000 <.riscv.attributes>:
-   0:	2941                	jal	490 <_start-0x7ffffb70>
+   0:	2541                	jal	680 <_start-0x7ffff980>
    2:	0000                	unimp
    4:	7200                	flw	fs0,32(a2)
    6:	7369                	lui	t1,0xffffa
    8:	01007663          	bgeu	zero,a6,14 <_start-0x7fffffec>
-   c:	001f 0000 1004      	0x10040000001f
+   c:	0000001b          	0x1b
+  10:	1004                	addi	s1,sp,32
   12:	7205                	lui	tp,0xfffe1
   14:	3376                	fld	ft6,376(sp)
   16:	6932                	flw	fs2,12(sp)
@@ -472,5 +473,3 @@ Disassembly of section .riscv.attributes:
   1c:	326d                	jal	fffff9c6 <__global_pointer$+0x7fffdc5e>
   1e:	3070                	fld	fa2,224(s0)
   20:	665f 7032 0030      	0x307032665f
-  26:	0108                	addi	a0,sp,128
-  28:	0b0a                	slli	s6,s6,0x2

driver/tests/demo/Makefile

@@ -59,8 +59,8 @@ run-simx: $(PROJECT)
 clean:
 	rm -rf $(PROJECT) *.o .depend
 
-clean-all:
-	rm -rf $(PROJECT) *.o *.elf *.bin *.dump .depend
+clean-all: clean
+	rm -rf *.elf *.bin *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

driver/tests/demo/kernel.dump

@@ -461,7 +461,7 @@ Disassembly of section .text:
 8000061c:	00492703          	lw	a4,4(s2)
 80000620:	148a2783          	lw	a5,328(s4)
 80000624:	01871463          	bne	a4,s8,8000062c <__call_exitprocs+0xe4>
-80000628:	f92784e3          	beq	a5,s2,800005b0 <__call_exitprocs+0x68>
+80000628:	f8f904e3          	beq	s2,a5,800005b0 <__call_exitprocs+0x68>
 8000062c:	f80788e3          	beqz	a5,800005bc <__call_exitprocs+0x74>
 80000630:	00078913          	mv	s2,a5
 80000634:	f5dff06f          	j	80000590 <__call_exitprocs+0x48>
@@ -527,20 +527,21 @@ Disassembly of section .comment:
    0:	3a434347          	fmsub.d	ft6,ft6,ft4,ft7,rmm
    4:	2820                	fld	fs0,80(s0)
    6:	29554e47          	fmsub.s	ft8,fa0,fs5,ft5,rmm
-   a:	3120                	fld	fs0,96(a0)
-   c:	2e30                	fld	fa2,88(a2)
-   e:	2e32                	fld	ft8,264(sp)
-  10:	0030                	addi	a2,sp,8
+   a:	3920                	fld	fs0,112(a0)
+   c:	322e                	fld	ft4,232(sp)
+   e:	302e                	fld	ft0,232(sp)
+	...
 
 Disassembly of section .riscv.attributes:
 
 00000000 <.riscv.attributes>:
-   0:	2941                	jal	490 <_start-0x7ffffb70>
+   0:	2541                	jal	680 <_start-0x7ffff980>
    2:	0000                	unimp
    4:	7200                	flw	fs0,32(a2)
    6:	7369                	lui	t1,0xffffa
    8:	01007663          	bgeu	zero,a6,14 <_start-0x7fffffec>
-   c:	001f 0000 1004      	0x10040000001f
+   c:	0000001b          	0x1b
+  10:	1004                	addi	s1,sp,32
   12:	7205                	lui	tp,0xfffe1
   14:	3376                	fld	ft6,376(sp)
   16:	6932                	flw	fs2,12(sp)
@@ -549,5 +550,3 @@ Disassembly of section .riscv.attributes:
   1c:	326d                	jal	fffff9c6 <__global_pointer$+0x7fffdb56>
   1e:	3070                	fld	fa2,224(s0)
   20:	665f 7032 0030      	0x307032665f
-  26:	0108                	addi	a0,sp,128
-  28:	0b0a                	slli	s6,s6,0x2

driver/tests/dogfood/Makefile

@@ -60,8 +60,8 @@ run-simx: $(PROJECT)
 clean:
 	rm -rf $(PROJECT) *.o .depend
 
-clean-all:
-	rm -rf $(PROJECT) *.o *.elf *.bin *.dump .depend
+clean-all: clean
+	rm -rf *.elf *.bin *.dump
 
 ifneq ($(MAKECMDGOALS),clean)
     -include .depend

runtime/src/vx_spawn.c

@@ -6,35 +6,61 @@
 extern "C" {
 #endif
 
+#define NUM_CORES_MAX 8
+
 typedef struct {
 	func_t function;
 	void * arguments;
 	int    nthreads;
 } spawn_t;
 
-spawn_t* g_spawn = NULL;
+spawn_t* g_spawn[NUM_CORES_MAX];
 
-void spawn_warp_runonce() {
+void spawn_warp_all() {
 	// active all threads
-	vx_tmc(g_spawn->nthreads);
+	int num_threads = vx_num_threads();	
+	vx_tmc(num_threads);
+
+	int core_id = vx_core_id();
+	spawn_t* p_spawn = g_spawn[core_id];
 
 	// call user routine
-	g_spawn->function(g_spawn->arguments);
+	p_spawn->function(p_spawn->arguments);	
 
-	// resume single-thread execution on exit
+	// resume single-warp execution on exit
+	int wid = vx_warp_id();
+	unsigned tmask = (0 == wid) ? 0x1 : 0x0; 
+	vx_tmc(tmask);
+}
+
+void spawn_warp_threads(int num_threads) {
+	// active all threads	
+	vx_tmc(num_threads);
+
+	int core_id = vx_core_id();
+	spawn_t* p_spawn = g_spawn[core_id];
+
+	// call user routine
+	p_spawn->function(p_spawn->arguments);	
+
+	// resume single-warp execution on exit
 	int wid = vx_warp_id();
 	unsigned tmask = (0 == wid) ? 0x1 : 0x0; 
 	vx_tmc(tmask);
 }
 
 void vx_spawn_warps(int num_warps, int num_threads, func_t func_ptr , void * args) {
-	spawn_t spawn = { func_ptr, args, num_threads };
-	g_spawn = &spawn;
+	int core_id = vx_core_id();
+	if (core_id >= NUM_CORES_MAX)
+		return;
+		
+	spawn_t spawn = { func_ptr, args, num_threads };	
+	g_spawn[core_id] = &spawn;
 
 	if (num_warps > 1) {
-		vx_wspawn(num_warps, (unsigned)spawn_warp_runonce);
+		vx_wspawn(num_warps, (unsigned)spawn_warp_all);
 	}
-	spawn_warp_runonce();
+	spawn_warp_threads(num_threads);
 }
 
 #ifdef __cplusplus

runtime/src/vx_start.S

@@ -4,24 +4,37 @@
 .global _start
 .type   _start, @function
 _start:
+  
+  # execute stack initialization on all warps
   la a1, vx_set_sp
   csrr a0, CSR_NW  # get num warps
-  .word 0x00b5106b # wspawn a0(numWarps), a1(PC SPAWN)
+  .word 0x00b5106b # wspawn a0, a1
   jal vx_set_sp
+
+  # return back to single thread execution
   li a0, 1
-  .word 0x0005006b    # back to single thread
-  # Initialize global pointerp
-  # call __cxx_global_var_init
+  .word 0x0005006b # tmc a0
+  
   # Clear the bss segment
   la      a0, _edata
   la      a2, _end
   sub     a2, a2, a0
   li      a1, 0
   call    memset
-  la      a0, __libc_fini_array   # Register global termination functions
-  call    atexit                  # to be called upon exit
-  call    __libc_init_array       # Run global initialization functions
+
+  # Register global termination functions
+  la      a0, __libc_fini_array   
+
+  # to be called upon exit
+  call    atexit                  
+
+  # Run global initialization functions
+  call    __libc_init_array       
+
+  # call main program routine
   call    main
+
+  # call exit routine
   tail    exit
 .size  _start, .-_start
 
@@ -29,34 +42,39 @@ _start:
 .type _exit, @function
 .global _exit
 _exit:
+  # disable all threads in current warp
   li a0, 0
-  .word 0x0005006b    # disable all threads
+  .word 0x0005006b    # tmc a0
 
 .section .text
 .type vx_set_sp, @function
 .global vx_set_sp
 vx_set_sp:
+  # activate all threads
   csrr a0, CSR_NT     # get num threads
-  .word 0x0005006b    # activate all threads
+  .word 0x0005006b    # set thread mask
   
+  # set global pointer register
   .option push
   .option norelax
-  1:auipc gp, %pcrel_hi(__global_pointer$)
-    addi  gp, gp, %pcrel_lo(1b)
+  la gp, __global_pointer$
   .option pop
 
+  # allocate stack region for a threads on the processor 
+  # set stack pointer
   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, (SHARED_MEM_BASE_ADDR>>12) # load base sp
+  la sp, __stack_top$  # load stack base address
   sub  sp, sp, a1      # sub thread block
   add  sp, sp, a2      # reduce addr collision for perf
 
-  csrr a3, CSR_LWID    # get wid
+  # disable active warps except warp0
+  csrr a3, CSR_LWID    # get local wid
   beqz a3, RETURN
   li a0, 0
-  .word 0x0005006b     # tmc 0
+  .word 0x0005006b     # tmc a0
 RETURN:
   ret
 

runtime/tests/dev/vx_dev_main.dump

@@ -271,11 +271,11 @@ Disassembly of section .text:
 80000374:	00008067          	ret
 
 80000378 <vx_num_cycles>:
-80000378:	c0002573          	rdcycle	a0
+80000378:	b0002573          	csrr	a0,mcycle
 8000037c:	00008067          	ret
 
 80000380 <vx_num_instrs>:
-80000380:	c0202573          	rdinstret	a0
+80000380:	b0202573          	csrr	a0,minstret
 80000384:	00008067          	ret
 
 80000388 <vx_vprintf>:

runtime/tests/dev/vx_dev_main.hex

@@ -55,7 +55,7 @@
 :100348007325000267800000732520026780000083
 :100358007325400267800000732550026780000003
 :1003680073256002678000007325700267800000B3
-:10037800732500C067800000732520C067800000D7
+:10037800732500B067800000732520B067800000F7
 :1003880063060520130101F52324810A232E310970
 :100398002326110A2322910A2320210B232C410909
 :1003A800232A510923286109232671099309050085

runtime/tests/nlTest/vx_nl_main.dump

@@ -132,11 +132,11 @@ Disassembly of section .text:
 80000158:	00008067          	ret
 
 8000015c <vx_num_cycles>:
-8000015c:	c0002573          	rdcycle	a0
+8000015c:	b0002573          	csrr	a0,mcycle
 80000160:	00008067          	ret
 
 80000164 <vx_num_instrs>:
-80000164:	c0202573          	rdinstret	a0
+80000164:	b0202573          	csrr	a0,minstret
 80000168:	00008067          	ret
 
 8000016c <vx_vprintf>:

runtime/tests/nlTest/vx_nl_main.hex

@@ -21,7 +21,7 @@
 :1001280067800000732500026780000073252002A5
 :100138006780000073254002678000007325500225
 :1001480067800000732560026780000073257002D5
-:1001580067800000732500C067800000732520C0F9
+:1001580067800000732500B067800000732520B019
 :100168006780000063060520130101F52324810A36
 :10017800232E31092326110A2322910A2320210B39
 :10018800232C4109232A51092328610923267109AF

runtime/tests/simple/vx_simple.dump

@@ -407,11 +407,11 @@ Disassembly of section .text:
 80000574:	00008067          	ret
 
 80000578 <vx_num_cycles>:
-80000578:	c0002573          	rdcycle	a0
+80000578:	b0002573          	csrr	a0,mcycle
 8000057c:	00008067          	ret
 
 80000580 <vx_num_instrs>:
-80000580:	c0202573          	rdinstret	a0
+80000580:	b0202573          	csrr	a0,minstret
 80000584:	00008067          	ret
 
 80000588 <vx_vprintf>:

runtime/tests/simple/vx_simple.hex

@@ -87,7 +87,7 @@
 :100548007325000267800000732520026780000081
 :100558007325400267800000732550026780000001
 :1005680073256002678000007325700267800000B1
-:10057800732500C067800000732520C067800000D5
+:10057800732500B067800000732520B067800000F5
 :1005880063060520130101F52324810A232E31096E
 :100598002326110A2322910A2320210B232C410907
 :1005A800232A510923286109232671099309050083