diff --git a/imul8xe.s b/imul8xe.s
index 15adf64..855e044 100644
--- a/imul8xe.s
+++ b/imul8xe.s
@@ -70,6 +70,106 @@ bankswitch:
 done:
 .endmacro
 
+.macro bank_switch bank
+    lda #((bank << 1) | $c1)
+    sta PORTB
+.endmacro
+
 proc imul8xe_init
+
+    ; go through the input set, in four 16KB chunks
+
+    arg1 = FR1
+    arg2 = FR2
+    result = FR0
+
+    lda #$00
+    sta arg1
+    sta arg2
+
+    ; $00 * $00 -> $3f * $ff
+    bank_switch 0
+    jsr imul8xe_init_section
+
+    ; $40 * $00 -> $7f * $ff
+    bank_switch 1
+    jsr imul8xe_init_section
+
+    ; $80 * $00 -> $bf * $ff
+    bank_switch 2
+    jsr imul8xe_init_section
+
+    ; $c0 * $00 -> $ff * $ff
+    bank_switch 3
+    jsr imul8xe_init_section
+
     rts
 endproc
+
+; Initialize a 16 KB chunk of the table
+; input: multipliers in temp
+; output: new multipliers in temp
+; clobbers: temp, temp2
+proc imul8xe_init_section
+    arg1 = FR1
+    arg2 = FR2
+    result = FR0
+    ptr = temp2
+
+    lda #$00
+    sta ptr
+    lda #$40
+    sta ptr + 1
+
+    ldx #0
+    ldy #0
+
+    ; outer loop: $00 -> $3f
+outer_loop:
+
+    ; reset result to 0
+    lda #0
+    sta result
+    sta result + 1
+
+    ; inner loop: $00 -> $ff
+inner_loop:
+
+    ; copy result to data set
+    lda result
+    sta (ptr),y
+    lda result + 1
+    sta (ptr),y
+
+    ; result += 2 * arg2
+    clc
+    lda arg2
+    adc result
+    sta result
+    lda #0
+    adc result + 1
+    sta result
+    lda arg2
+    adc result
+    sta result
+    lda #0
+    adc result + 1
+    sta result
+
+    ; inner loop check
+    inc arg1
+    inc arg1
+    inc ptr
+    inc ptr
+    bne inner_loop
+
+    ; outer loop check
+    inc arg2
+    inc ptr + 1
+    lda ptr + 1
+    cmp #$40
+    bne outer_loop
+
+    rts
+
+endproc