From 7c804513fad888f5ab6b2a36564769ddc95c276b Mon Sep 17 00:00:00 2001
From: uz <uz@b7a2c559-68d2-44c3-8de9-860c34a00d81>
Date: Wed, 4 Nov 2009 13:45:27 +0000
Subject: [PATCH] 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
---
 libsrc/runtime/Makefile                       |  3 +-
 libsrc/runtime/imul16x16r32.s                 | 50 +++++++++++++++++++
 .../runtime/{mul16x16r32.s => umul16x16r32.s} | 13 ++---
 3 files changed, 59 insertions(+), 7 deletions(-)
 create mode 100644 libsrc/runtime/imul16x16r32.s
 rename libsrc/runtime/{mul16x16r32.s => umul16x16r32.s} (77%)

diff --git a/libsrc/runtime/Makefile b/libsrc/runtime/Makefile
index b92863cab..adb35a8f9 100644
--- a/libsrc/runtime/Makefile
+++ b/libsrc/runtime/Makefile
@@ -81,6 +81,7 @@ OBJS = 	add.o		\
        	ge.o		\
        	gt.o		\
        	icmp.o		\
+        imul16x16r32.o  \
 	incax1.o	\
 	incax2.o	\
        	incax3.o	\
@@ -152,7 +153,6 @@ OBJS = 	add.o		\
        	makebool.o	\
        	mod.o  		\
        	mul.o  		\
-        mul16x16r32.o   \
         mul8.o          \
         mul8x8r16.o     \
 	mulax3.o	\
@@ -219,6 +219,7 @@ OBJS = 	add.o		\
        	ule.o		\
        	ult.o		\
        	umod.o		\
+        umul16x16r32.o  \
        	xor.o           \
         zeropage.o
 
diff --git a/libsrc/runtime/imul16x16r32.s b/libsrc/runtime/imul16x16r32.s
new file mode 100644
index 000000000..d105b72a3
--- /dev/null
+++ 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
+
diff --git a/libsrc/runtime/mul16x16r32.s b/libsrc/runtime/umul16x16r32.s
similarity index 77%
rename from libsrc/runtime/mul16x16r32.s
rename to libsrc/runtime/umul16x16r32.s
index 72d9fe35f..d86162c7a 100644
--- a/libsrc/runtime/mul16x16r32.s
+++ b/libsrc/runtime/umul16x16r32.s
@@ -1,17 +1,16 @@
 ;
 ; Ullrich von Bassewitz, 2010-11-03
 ;
-; CC65 runtime: 16x16 => 32 multiplication
+; CC65 runtime: 16x16 => 32 unsigned multiplication
 ;
 
-        .export         umul16x16r32, _cc65_umul16x16r32
-       	.export	       	mul16x16r32 := umul16x16r32
+        .export         _cc65_umul16x16r32, umul16x16r32, umul16x16r32m
         .import         popax
-    	.importzp   	ptr1, ptr2, ptr3, ptr4, sreg
+    	.importzp      	ptr1, ptr3, sreg
 
 
 ;---------------------------------------------------------------------------
-; 16x16 => 32 multiplication routine.
+; 16x16 => 32 unsigned multiplication routine.
 ;
 ;   lhs         rhs           result          result also in
 ; -------------------------------------------------------------
@@ -26,6 +25,8 @@ _cc65_umul16x16r32:
 umul16x16r32:
         sta     ptr3
         stx     ptr3+1
+
+umul16x16r32m:
        	lda	#0
        	sta    	sreg+1
        	ldy    	#16   	       	; Number of bits
@@ -37,7 +38,7 @@ umul16x16r32:
       	clc
       	adc	ptr3
       	pha
-       	txa	      	       	; hi byte of left op
+       	lda     ptr3+1
       	adc	sreg+1
       	sta	sreg+1
       	pla
-- 
2.39.5