wu-arch/kernels
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Multi-lane matrix multiplication.

Browse Source
This commit is contained in:
cdkersey
2013-11-11 14:57:55 -05:00
parent 1574f3a60d
commit 3e8d0fc4b0
1 changed files with 12 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
src/test/matmul-mt.s
View File

@@ -66,8 +66,10 @@ matmul: ori %r22, %r5, #0;
add %r10, %r10, %r3; add %r10, %r10, %r3;
ldi %r14, #1; ldi %r14, #1;
shl %r14, %r14, %r10; shl %r14, %r14, %r10;
shl %r17, %r14, %r3;
divi %r17, %r14, THREADS; /* Spawn threads */ divi %r17, %r17, THREADS; /* Spawn threads */
divi %r24, %r4, THREADS;
ori %r18, %r0, #0; ori %r18, %r0, #0;
ori %r19, %r2, #0; ori %r19, %r2, #0;
ldi %r20, #0; ldi %r20, #0;
@@ -76,15 +78,15 @@ sloop: add %r0, %r0, %r17;
addi %r20, %r20, #1; addi %r20, %r20, #1;
subi %r21, %r20, THREADS; subi %r21, %r20, THREADS;
rtop @p0, %r21; rtop @p0, %r21;
notp @p1, @p0; @p0 ? clone %r20;
@p1 ? clone %r20;
@p0 ? jmpi sloop;
ori %r0, %r18, #0; ori %r0, %r18, #0;
ori %r2, %r19, #0; ori %r2, %r19, #0;
clone %r20;
jalis %r5, matmulthd; @p0 ? jmpi sloop;
ldi %r20, THREADS;
jalis %r5, %r20, matmulthd;
jmpr %r22; jmpr %r22;
@@ -92,7 +94,7 @@ sloop: add %r0, %r0, %r17;
* %r0 - matrix a pointer (plus offset) * %r0 - matrix a pointer (plus offset)
* %r1 - matrix b pointer * %r1 - matrix b pointer
* %r2 - destination matrix pointer (plus offset) * %r2 - destination matrix pointer (plus offset)
* %r17 - row count * %r24 - row count
*/ */
matmulthd: ldi %r9, #0; /* result row: %r9 */ matmulthd: ldi %r9, #0; /* result row: %r9 */
rloop: ldi %r6, #0; /* result col: %r6 */ rloop: ldi %r6, #0; /* result col: %r6 */
@@ -105,8 +107,8 @@ cloop: shli %r16, %r6, (`__WORD);
ldi %r8, #0 /* dot prod position: %r8 */ ldi %r8, #0 /* dot prod position: %r8 */
iloop: ld %r7, %r11, #0; iloop: ld %r7, %r11, #0;
ld %r17, %r12, #0; ld %r23, %r12, #0;
fmul %r7, %r7, %r17 fmul %r7, %r7, %r23
fadd %r13, %r13, %r7; fadd %r13, %r13, %r7;
addi %r8, %r8, #1; addi %r8, %r8, #1;
@@ -126,7 +128,7 @@ iloop: ld %r7, %r11, #0;
@p0 ? jmpi cloop; @p0 ? jmpi cloop;
addi %r9, %r9, #1; addi %r9, %r9, #1;
sub %r7, %r9, %r17; sub %r7, %r9, %r24;
rtop @p0, %r7; rtop @p0, %r7;
@p0 ? jmpi rloop; @p0 ? jmpi rloop;