From 438b622c5801ebb5ebb055503a045e505f3a5b6a Mon Sep 17 00:00:00 2001
From: Sijmen Schoon <me@sijmenschoon.nl>
Date: Wed, 9 Dec 2020 13:26:53 +0100
Subject: [PATCH] Make 32-64 bit multiplication routine based on the 16-32 bit
 one

---
 c64/main.asm | 20 ++++++++++++++++++
 c64/math.asm | 57 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 c64/math.inc | 31 +++++++++++++++++++---------
 3 files changed, 99 insertions(+), 9 deletions(-)

diff --git a/c64/main.asm b/c64/main.asm
index ab54097..507dfeb 100644
--- a/c64/main.asm
+++ b/c64/main.asm
@@ -23,6 +23,26 @@
     sta dst + 1
 }
 
+.macro u16_u16_move(dst, src) {
+    .for (var i = 0; i < 2; i++) {
+        lda src + i
+        sta dst + i
+    }
+}
+
+.macro u32_u32_move(dst, src) {
+    .for (var i = 0; i < 4; i++) {
+        lda src + i
+        sta dst + i
+    }
+}
+
+.macro u32_u32_move_imm(dst, imm) {
+    .for (var i = 0; i < 4; i++) {
+        lda #((imm >> (i * 8)) & $ff)
+        sta dst + i
+    }
+}
 //
 // main
 //
diff --git a/c64/math.asm b/c64/math.asm
index cae2234..2a54d66 100644
--- a/c64/math.asm
+++ b/c64/math.asm
@@ -203,6 +203,63 @@ _dd:            lda #0
                 rts
 
 
+.const aa = $38
+.const AA = $39
+.const bb = $3a
+.const BB = $3b
+.const dd = $3c
+.const DD = $3d
+.const hh = $3e
+.const HH = $3f
+multiply_32bit_unsigned:
+    //                  BBbb AAaa
+    //             DDdd CCcc
+    //             FFff EEee
+    // +      HHhh GGgg
+    // --------------------------
+    //          3e   3c   3a   38
+    //        HHhh DDdd BBbb AAaa
+    //             FFff CCcc
+    //             GGgg EEee
+
+    u32_u32_move_imm(aa, 0)
+    u32_u32_move_imm(dd, 0)
+
+    // Perform <T1 * <T2 == BBbbAAaa
+    u16_u16_move(T1, $30)
+    u16_u16_move(T2, $34)
+    sec
+    jsr multiply_16bit_unsigned
+    u16_u16_move(aa, PRODUCT + 0) // AAaa
+    u16_u16_move(bb, PRODUCT + 2) // BBbb
+
+    // Perform >T1 * <T2 = DDddCCcc
+    u16_u16_move(T1, $32)
+    u16_u16_move(T2, $34)
+    sec
+    jsr multiply_16bit_unsigned
+    i32_i16_add(bb, PRODUCT) // CCcc
+    i16_i16_add(dd, PRODUCT + 2) // DDdd
+
+    // Perform <T1 * >T2 = FFffEEee
+    u16_u16_move(T1, $30)
+    u16_u16_move(T2, $36)
+    sec
+    jsr multiply_16bit_unsigned
+    i32_i16_add(bb, PRODUCT) // EEee
+    i32_i16_add(dd, PRODUCT + 2) // FFff
+
+    // Perform >T1 * >T2 = HHhhGGgg
+    u16_u16_move(T1, $32)
+    u16_u16_move(T2, $36)
+    sec
+    jsr multiply_16bit_unsigned
+    i32_i16_add(dd, PRODUCT) // GGgg
+    i16_i16_add(hh, PRODUCT + 2) // HHhh
+
+    rts
+
+
 // Description: Signed 16-bit multiplication with signed 32-bit result.
 //
 // Input: 16-bit signed value in T1
diff --git a/c64/math.inc b/c64/math.inc
index 6a2b017..a92df1a 100644
--- a/c64/math.inc
+++ b/c64/math.inc
@@ -96,7 +96,7 @@
     sta $fe
 
     i16_mul4(dst, dst + 1)
-    i16_i16_add(dst, dst + 1, $fd, $fe)
+    i16_i16_add(dst, $fd)
 }
 
 .macro u32_mul2(lsb) {
@@ -124,6 +124,19 @@
     i32_i32_add(lsb, $2c)
 }
 
+.macro i32_i16_add(dst, src) {
+    clc
+    lda dst + 0
+    adc src + 0
+    sta dst + 0
+    lda dst + 1
+    adc src + 1
+    sta dst + 1
+    bcc !+
+    inc dst + 2
+!:
+}
+
 .macro i32_i32_add(dst, src) {
     clc
     lda dst + 0
@@ -150,15 +163,15 @@
 !cc:
 }
 
-// Destroys: a, dst_lo, dst_hi
-.macro i16_i16_add(dst_lo, dst_hi, src_lo, src_hi) {
+// Destroys: a
+.macro i16_i16_add(dst, src) {
     clc
-    lda dst_lo
-    adc src_lo
-    sta dst_lo
-    lda dst_hi
-    adc src_hi
-    sta dst_hi
+    lda dst
+    adc src
+    sta dst
+    lda dst + 1
+    adc src + 1
+    sta dst + 1
 }
 
 // Destroys: a, dst_lo, dst_hi