;;;;;;;;;;;;;;;;;;;;;;
; mul Montgomery
;;;;;;;;;;;;;;;;;;;;;;
mulM:
<%
let r0, r1, r2;
function setR(step) {
    if ((step % 3) == 0) {
        r0 = "r8";
        r1 = "r9";
        r2 = "r10";
    } else if ((step % 3) == 1) {
        r0 = "r9";
        r1 = "r10";
        r2 = "r8";
    } else {
        r0 = "r10";
        r1 = "r8";
        r2 = "r9";
    }
}

const base = bigInt.one.shiftLeft(64);
const np64 = base.minus(q.modInv(base));
%>
        sub     rsp, <%= n64*8 %>  ; Reserve space for ms
        mov     rcx, rdx            ; rdx is needed for multiplications so keep it in cx
        mov     r11, 0x<%= np64.toString(16) %> ; np
        xor     r8,r8
        xor     r9,r9
        xor     r10,r10
<%
// Main loop
for (let i=0; i<n64*2; i++) {
    setR(i);
%>
<%
    // Same Digit
    for (let o1=Math.max(0, i-n64+1); (o1<=i)&&(o1<n64); o1++) {
        const o2= i-o1;
%>
        mov     rax, [rsi + <%= 8*o1 %>]
        mul     qword [rcx + <%= 8*o2 %>]
        add     <%= r0 %>, rax
        adc     <%= r1 %>, rdx
        adc     <%= r2 %>, 0x0
<%
    } // Same digit
%>


<%
    for (let j=i-1; j>=0; j--) { // All ms
        if (((i-j)<n64)&&(j<n64)) {
%>
        mov     rax, [rsp + <%= j*8 %>]
        mul     qword [q + <%= (i-j)*8 %>]
        add     <%= r0 %>, rax
        adc     <%= r1 %>, rdx
        adc     <%= r2 %>, 0x0
<%
        }
    } // ms
%>

<%
    if (i<n64) {
%>
        mov     rax, <%= r0 %>
        mul     r11
        mov     [rsp + <%= i*8 %>], rax
        mul     qword [q]
        add     <%= r0 %>, rax
        adc     <%= r1 %>, rdx
        adc     <%= r2 %>, 0x0
<%
    } else {
%>
        mov     [rdi + <%= (i-n64)*8 %> ], <%= r0 %>
        xor     <%= r0 %>,<%= r0 %>
<%
    }
%>

<%
}   // Main Loop
%>
        cmp     <%= r1 %>, 0x0
        jne     mulM_sq
        ; Compare with q
<%
for (let i=0; i<n64; i++) {
%>
        mov rax, [rdi + <%= (n64-i-1)*8 %>]
        cmp rax, [q + <%= (n64-i-1)*8 %>]
        jg mulM_sq
        jl mulM_done
<%
}
%>
        ; If equal substract q

mulM_sq:
<%
for (let i=0; i<n64; i++) {
%>
        mov rax, [q + <%= i*8 %>]
        <%= i==0 ? "sub" : "sbb" %> [rdi + <%= i*8 %>], rax
<%
}
%>

mulM_done:
        add rsp, <%= n64*8 %>   ; recover rsp
        ret

;;;;;;;;;;;;;;;;;;;;;;
; mul MontgomeryShort
;;;;;;;;;;;;;;;;;;;;;;
mulSM:

;;;;;;;;;;;;;;;;;;;;;;
; mul
;;;;;;;;;;;;;;;;;;;;;;
<%=name%>_mul:
        mov    rax, [rsi]
        bt     rax, 63
        jc     l1
        mov    rcx, [rdx]
        bt     rcx, 63
        jc     s1l2
s1s2:                       ; short first and second
        mul    ecx
        jc     rs2l         ; If if doesn't feed in 32 bits convert the result to long

        ; The shorts multiplication is done. copy the val to destination and return
        mov    [rdi], rax
        ret

rs2l:                       ; The result in the multiplication doen't feed
                            ; we have the result in edx:eax we need to convert it to long
        shl    rdx, 32
        mov    edx, eax     ; pack edx:eax to rdx

        xor    rax, rax     ; Set the format to long
        bts    rax, 63
        mov    [rdi], rax       ; move the first digit

        cmp    rdx, 0       ; check if redx is negative.
        jl     rs2ln

        ; edx is positive.
        mov    [rdi + 8], rdx   ; Set the firs digit

        xor    rax, rax     ; Set the remaining digits to 0
<% for (let i=1; i<n64; i++) { %>
        mov [rdi + <%= (i+1)*8 %>], rax
<% } %>
        ret

        ; edx is negative.
rs2ln:

        add    rdx, [q]         ; Set the firs digit
        mov    [rdi + 8], rdx   ;

        mov    rdx, -1          ; all ones
<% for (let i=1; i<n64; i++) { %>
        mov    rax, rdx                       ; Add to q
        adc    rax, [q + <%= i*8 %> ]
        mov    [rdi + <%= (i+1)*8 %>], rax
<% } %>
        ret

l1:
        mov    rcx, [rdx]
        bt     rcx, 63
        jc     ll

l1s2:
        xor rdx, rdx
        mov edx, ecx
        bt  rax, 62
        jc  lsM
        jmp lsN

s1l2:
        mov rsi, rdx
        xor rdx, rdx
        mov edx, eax
        bt  rcx, 62
        jc  lsM
        jmp lsN


lsN:
        mov byte [rdi + 7], 0xC0    ; set the result to montgomery
        add rsi, 8
        add rdi, 8
        call mulSM
        mov rsi, rdi
        lea rdx, [R3]
        call mulM
        ret

lsM:
        mov byte [rdi + 7], 0x80    ; set the result to long normal
        add rsi, 8
        add rdi, 8
        call mulSM
        ret


ll:

        bt rax, 62
        jc lml
        bt rcx, 62
        jc lnlm

lnln:
        mov byte [rdi + 7], 0xC0    ; set the result to long montgomery
        add rsi, 8
        add rdi, 8
        add rdx, 8
        call mulM
        mov rsi, rdi
        lea rdx, [R3]
        call mulM
        ret

lml:
        bt rcx, 62
        jc lmlm

lnlm:
        mov byte [rdi + 7], 0x80    ; set the result to long normal
        add rsi, 8
        add rdi, 8
        add rdx, 8
        call mulM
        ret

lmlm:
        mov byte [rdi + 7], 0xC0    ; set the result to long montgomery
        add rsi, 8
        add rdi, 8
        add rdx, 8
        call mulM
        ret