Added matadd.s and 2warp_matadd.s

src/core.cpp

@@ -4,6 +4,9 @@
 
 #include <iostream>
 
+// #define USE_DEBUG 7
+// #define PRINT_ACTIVE_THREADS
+
 #include "include/types.h"
 #include "include/util.h"
 #include "include/archdef.h"

src/test/2warp_matadd.s

@@ -0,0 +1,82 @@
+/*******************************************************************************
+ Harptools by Chad D. Kersey, Summer 2011
+********************************************************************************
+
+ Sample HARP assmebly program.
+
+*******************************************************************************/
+/* Divergent branch: test immediate postdominator branch divergence support. */
+.def THREADS 8
+
+.align 4096
+.perm x
+.entry
+.global
+entry: ldi %r20, #1
+       ldi %r21, Begin
+       wspawn %r20, %r21, %r20
+       ldi %r20, #0
+Begin: ldi %r0, #1
+       ldi %r1, THREADS
+tc_loop: clone %r0
+       
+       addi %r0, %r0, #1
+       sub %r2, %r1, %r0
+       rtop @p0, %r2
+ @p0 ? jmpi tc_loop
+
+       ldi %r0, #0
+       jalis %r5, %r1, dthread;
+
+       ldi %r25, #55
+       ldi %r26, #1
+       bar %r25, %r26
+
+       subi %r20, %r20, #1
+
+       iszero @p0, %r20
+
+  @p0 ? trap;
+
+       ldi %r0, #0
+       ldi %r1, (__WORD * THREADS)
+       shli %r1, %r1, #1
+
+ploop: ld %r7, %r0, RESULT
+       jali %r5, printdec
+       
+       addi %r0, %r0, __WORD
+       sub %r7, %r1, %r0
+       rtop @p0, %r7
+ @p0 ? jmpi ploop
+
+       trap;
+
+
+dthread: shli  %r15, %r20 , #6
+         shli  %r10, %r0 , #3
+         add   %r10, %r10, %r15
+         ld    %r11, %r10, Array1
+         ld    %r12, %r10, Array2
+
+         subi  %r13, %r0, #4
+         isneg @p0, %r13
+   @p0 ? split
+   @p0 ? jmpi SUBT
+         add %r14, %r11, %r12
+         jmpi after
+SUBT:    sub %r14, %r11, %r12
+after:   join
+
+         st %r14, %r10, RESULT
+
+         jmprt %r5;
+
+.align 4096
+Array1:
+  .word 1 5 10 0 3 1 1 2
+  .word 8 7 8 7 5 7 7 9
+Array2:
+  .word 0 2 2 0 5 0 1 1
+  .word 4 2 2 0 3 2 3 2
+RESULT: .space 512

src/test/Makefile

@@ -5,14 +5,14 @@ HARPDIS = ../harptool -D
 4BARCH  = 4b16/16/2/1
 
 all: simple.bin sieve.bin 2thread.bin simple.4b.bin sieve.4b.bin 2thread.4b.bin bubble.bin bubble.4b.bin dotprod.bin dotprod.4b.bin matmul.bin matmul.4b.bin \
- matmul-mt.s lfsr.bin diverge.bin
+ matmul-mt.s lfsr.bin diverge.bin matadd.bin 2warp_matadd.bin
 
 run: simple.out sieve.out 2thread.out simple.4b.out sieve.4b.out 2thread.4b.out bubble.out bubble.4b.out dotprod.out dotprod.4b.out matmul.out matmul.4b.out\
- matmul-mt.out lfsr.4b.out lfsr.out diverge.out
+ matmul-mt.out lfsr.4b.out lfsr.out diverge.out matadd.out 2warp_matadd.out
 
 disas: simple.d sieve.d 2thread.d simple.4b.d sieve.4b.d 2thread.4b.d bubble.d \
 bubble.4b.d dotprod.d dotprod.4b.d matmul.d matmul.4b.d matmul-mt.d lfsr.d \
-diverge.d
+diverge.d matadd.d 2warp_matadd.d
 
 %.4b.out : %.4b.bin
 	$(HARPEM) -a $(4BARCH) -c $< > $@

src/test/matadd.s

@@ -0,0 +1,66 @@
+/*******************************************************************************
+ Harptools by Chad D. Kersey, Summer 2011
+********************************************************************************
+
+ Sample HARP assmebly program.
+
+*******************************************************************************/
+/* Divergent branch: test immediate postdominator branch divergence support. */
+.def THREADS 8
+
+.align 4096
+.perm x
+.entry
+.global
+entry:
+       ldi %r0, #1
+       ldi %r1, THREADS
+tc_loop: clone %r0
+       
+       addi %r0, %r0, #1
+       sub %r2, %r1, %r0
+       rtop @p0, %r2
+ @p0 ? jmpi tc_loop
+
+       ldi %r0, #0
+       jalis %r5, %r1, dthread;
+
+       ldi %r0, #0
+       ldi %r1, (__WORD * THREADS)
+
+ploop: ld %r7, %r0, RESULT
+       jali %r5, printdec
+       
+       addi %r0, %r0, __WORD
+       sub %r7, %r1, %r0
+       rtop @p0, %r7
+ @p0 ? jmpi ploop
+
+       trap;
+
+
+dthread: shli  %r10, %r0 , #3
+         ld    %r11, %r10, Array1
+         ld    %r12, %r10, Array2
+
+         subi  %r13, %r0, #4
+         isneg @p0, %r13
+   @p0 ? split
+   @p0 ? jmpi SUBT
+         add %r14, %r11, %r12
+         jmpi after
+SUBT:    sub %r14, %r11, %r12
+after:   join
+
+         st %r14, %r10, RESULT
+
+         jmprt %r5;
+
+.align 4096
+Array1:
+  .word 1 5 10 0
+  .word 3 1 1 2
+Array2:
+  .word 0 2 2 0
+  .word 5 0 1 1
+RESULT: .space 512