new version

diff --git a/foo.c b/foo.c
index 4d51f60..ec0b8a5 100644 (file)
--- a/foo.c
+++ b/foo.c
@@ -55,6 +55,110 @@ void mod3(void) {
        RET
        #undef tmp
 }
+void mul(void) { //don't need overhead of function (inline it)
+       // i1.i0 * t -> _.x.t
+       #define a1 x
+       #define a2 _
+       #define a0 t
+       // start MUL -- 92 cycles :( (unrolled and skipping second bit: 76)
+       CLR     (a2)
+       CLR     (a1)
+
+       #define MUL_ADD_ROR \
+               ADD     (a1, i0) \
+               ADC     (a2, i1, carry) \
+               MUL_ROR
+
+       #define MUL_ROR \
+               LSR     (a2) \
+               ROR     (a1) \
+               ROR     (t) //superfluous?
+
+       switch(t) {
+       case 0x58: // 0101 1000
+               MUL_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+               break;
+       case 0x69: // 0110 1001
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+               break;
+       case 0x75: // 0111 0101
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+               break;
+       case 0x84: // 1000 0100
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+
+               MUL_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               break;
+       case 0x8c: // 1000 1100
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+
+               MUL_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               break;
+       case 0x9d: // 1001 1101
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+               break;
+       case 0xb0: // 1011 0000
+               MUL_ROR
+               MUL_ROR
+               MUL_ROR
+               MUL_ROR
+
+               MUL_ADD_ROR
+               MUL_ADD_ROR
+               MUL_ROR
+               MUL_ADD_ROR
+       }
+
+       // end MUL
+       #undef a0
+       #undef a1
+       #undef a2
+       RET
+}
 void g(void) {
        // g(i, t) -> t
        // tempvars: `x` and `_`
@@ -73,30 +177,8 @@ void g(void) {
          ADD X_lo, t
          ADC X_hi, zero
          LD  t, X         */
-       #define a1 x
-       #define a2 _
-       #define a0 t
-       // start MUL -- 92 cycles :( (unrolled and skipping second bit: 76)
-       CLR     (a2)
-       CLR     (a1)
-       LDI     (loop, 8)
-       mul:
-       SBRS    (t, 0)
-       RJMP    (skip2)
-       ADD     (a1, i0)
-       ADC     (a2, i1, carry)
-       skip2:
-       LSR     (a2)
-       ROR     (a1)
-       ROR     (t)
-       DEC     (loop)
-       BRNE    (mul)
-       // end MUL
-       MOV     (t, a1)
-       #undef a0
-       #undef a1
-       #undef a2
-       
+       RCALL   mul(); //stores used value in in x
+       MOV     (t, x)
        RET //TODO: replace CALL/RET with IJMP?
 };