Faster __mulsi routine

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@38736 138bc75d-0d04-0410-961f-82ee72b054a4

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 84f8a29..3b02eb8 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,8 @@
+Fri Jan  5 16:34:18 2001  Nick Clifton  <nickc@redhat.com>
+
+       * config/v850/lib1funcs.asm: Replace __mulsi3 routine with faster
+       version supplied by Matteo Frigo.
+
 2001-01-05  Neil Booth  <neil@daikokuya.demon.co.uk>
 
         * cpp.texi: Update for -MQ.

diff --git a/gcc/config/v850/lib1funcs.asm b/gcc/config/v850/lib1funcs.asm
index a162421..9248acb 100644 (file)
--- a/gcc/config/v850/lib1funcs.asm
+++ b/gcc/config/v850/lib1funcs.asm
@@ -33,57 +33,60 @@ Boston, MA 02111-1307, USA.  */
        .type  ___mulsi3,@function
 
 /*
- * In order to not deal with negative numbers (mulh is a signed multiply
- * and we want an unsigned multiply, code the multiplication as a series
- * of 7 bit multiplies).
+ * #define SHIFT 12
+ * #define MASK ((1 << SHIFT) - 1)
+ *  
+ * #define STEP(i, j)                             \
+ * ({                                             \
+ *     short a_part = (a >> (i)) & MASK;          \
+ *     short b_part = (b >> (j)) & MASK;          \
+ *     int res = (((int)a_part) * ((int)b_part)); \
+ *     res;                                       \
+ * })
  *
- * int __mulsi3 (unsigned a, unsigned b)
+ * int
+ * __mulsi3 (unsigned a, unsigned b)
  * {
- *   int i, j;
- *   int ret = 0;
- *
- *   for (i = 0; i < 32; i += 7)
- *     {
- *       short a_part = a & 0x7f;
- *       unsigned b_tmp = b;
- *       a >>= 7;
- * 
- *       for (j = 0; (i+j) < 32; j += 7)
- *     {
- *       short b_part = b_tmp & 0x7f;
- *       ret += (((int)a_part) * ((int)b_part)) << (i+j);
- *       b_tmp >>= 7;
- *     }
- *    }
- *
- *   return ret;
+ *    return STEP (0, 0) +
+ *        ((STEP (SHIFT, 0) + STEP (0, SHIFT)) << SHIFT) +
+ *        ((STEP (0, 2 * SHIFT) + STEP (SHIFT, SHIFT) + STEP (2 * SHIFT, 0))
+ *         << (2 * SHIFT));
  * }
  */
 
 ___mulsi3:
-       mov 0,r10                       /* total */
-       mov 0,r14                       /* i = 0, index for multiply a's part */
-       movea lo(31),r0,r16             /* upper bounds for loop */
-.L5:
-       mov r7,r13                      /* b_tmp = b */
-       andi 0x7f,r6,r15                /* a_part = (a & 127) */
-       shr 7,r6                        /* a >>= 7 */
-       mov r14,r12                     /* i+j = i */
-.L9:
-       andi 0x7f,r13,r11               /* b_part = (b_tmp & 127) */
-       mulh r15,r11                    /* ((int)a_part) * ((int)b_part) */
-       shr 7,r13                       /* b_tmp >>= 7 */
-       shl r12,r11                     /* (((int)a_part) * ((int)b_part)) << (i+j) */
-       add r11,r10                     /* ret += (((int)a_part) * ((int)b_part)) << (i+j) */
-       add 7,r12                       /* i+j += 7 */
-       cmp r16,r12                     /* i+j < 32 */
-       ble .L9
-
-       add 7,r14                       /* i += 7 */
-       cmp r16,r14                     /* i < 32 */
-       ble .L5
-
-       jmp [r31]                       /* return */
+        mov r6,r13
+        movea lo(4095),r0,r16
+        and r16,r13
+        mov r7,r15
+        and r16,r15
+        mov r13,r10
+        mulh r15,r10
+        shr 12,r6
+        mov r6,r14
+        and r16,r14
+        mov r14,r11
+        mulh r15,r11
+        shr 12,r7
+        mov r7,r12
+        and r16,r12
+        shr 12,r7
+        and r16,r7
+        mulh r13,r7
+        shr 12,r6
+        mulh r12,r13
+        and r16,r6
+        add r13,r11
+        shl 12,r11
+        add r11,r10
+        mov r14,r11
+        mulh r12,r11
+        mulh r15,r6
+        add r11,r7
+        add r6,r7
+        shl 24,r7
+        add r7,r10
+        jmp [r31]
        .size ___mulsi3,.-___mulsi3
 #endif