The upper 16 bit differ between signed and unsigned multiplication, so while

for 16x16=16, the unsigned routine can be used, regardless of signedness, this is not true for 16x16=32. git-svn-id: svn://svn.cc65.org/cc65/trunk@4438 b7a2c559-68d2-44c3-8de9-860c34a00d81
2009-11-04 13:45:27 +00:00
parent 1bec071755
commit 7c804513fa
3 changed files with 59 additions and 7 deletions
--- a/libsrc/runtime/imul16x16r32.s
+++ b/libsrc/runtime/imul16x16r32.s
@@ -0,0 +1,50 @@
+;
+; Ullrich von Bassewitz, 2010-11-03
+;
+; CC65 runtime: 16x16 => 32 signed multiplication
+;
+
+        .export         imul16x16r32, _cc65_imul16x16r32
+        .import         popax, negax, umul16x16r32m, negeax
+    	.importzp     	ptr1, ptr3, tmp1
+
+
+;---------------------------------------------------------------------------
+; 16x16 => 32 signed multiplication routine.
+;
+;   lhs         rhs           result          result also in
+; -------------------------------------------------------------
+;   ptr1        ax            ax:sreg
+;
+; There is probably a faster way to do this.
+;
+
+
+_cc65_imul16x16r32:
+        sta     ptr1
+        stx     ptr1+1
+        jsr     popax
+
+imul16x16r32:
+        stx     tmp1
+        cpx     #0
+        bpl     @L1
+        jsr     negax
+@L1:    sta     ptr3
+        stx     ptr3+1
+
+        lda     ptr1+1
+        eor     tmp1
+        sta     tmp1
+        lda     ptr1
+        ldx     ptr1+1
+        bpl     @L2
+        jsr     negax
+        sta     ptr1
+        stx     ptr1+1
+@L2:    jsr     umul16x16r32m
+        ldy     tmp1
+        bpl     @L3
+        jmp     negeax
+@L3:    rts
+