From f500408e00aed7b413dc33f02e6cbe1a0942c67a Mon Sep 17 00:00:00 2001
From: Joel <rootjdev@gmail.com>
Date: Sat, 1 Jul 2023 13:48:43 +0930
Subject: [PATCH] replace the asm implementation for mul/square inner

---
 src/field_5x52_asm_impl.h | 826 ++++++++++++++++----------------------
 1 file changed, 336 insertions(+), 490 deletions(-)

diff --git a/src/field_5x52_asm_impl.h b/src/field_5x52_asm_impl.h
index 04a9af2105..08a4890bff 100644
--- a/src/field_5x52_asm_impl.h
+++ b/src/field_5x52_asm_impl.h
@@ -1,14 +1,5 @@
-/***********************************************************************
- * Copyright (c) 2013-2014 Diederik Huys, Pieter Wuille                *
- * Distributed under the MIT software license, see the accompanying    *
- * file COPYING or https://www.opensource.org/licenses/mit-license.php.*
- ***********************************************************************/
-
 /**
- * Changelog:
- * - March 2013, Diederik Huys:    original version
- * - November 2014, Pieter Wuille: updated to use Peter Dettman's parallel multiplication algorithm
- * - December 2014, Pieter Wuille: converted from YASM to GCC inline assembly
+ * Generated by CryptOpt (https://github.com/0xADE1A1DE/CryptOpt)
  */
 
 #ifndef SECP256K1_FIELD_INNER5X52_IMPL_H
@@ -16,489 +7,344 @@
 
 #include "util.h"
 
-SECP256K1_INLINE static void secp256k1_fe_mul_inner(uint64_t *r, const uint64_t *a, const uint64_t * SECP256K1_RESTRICT b) {
-/**
- * Registers: rdx:rax = multiplication accumulator
- *            r9:r8   = c
- *            r15:rcx = d
- *            r10-r14 = a0-a4
- *            rbx     = b
- *            rdi     = r
- *            rsi     = a / t?
- */
-  uint64_t tmp1, tmp2, tmp3;
-__asm__ __volatile__(
-    "movq 0(%%rsi),%%r10\n"
-    "movq 8(%%rsi),%%r11\n"
-    "movq 16(%%rsi),%%r12\n"
-    "movq 24(%%rsi),%%r13\n"
-    "movq 32(%%rsi),%%r14\n"
+SECP256K1_INLINE static void
+secp256k1_fe_mul_inner(uint64_t *r, const uint64_t *a, const uint64_t *b) {
+  uint64_t tmp0, tmp1, tmp2, tmp3;
+  __asm__ __volatile__(
+    "mov    %%rdx,%%rax\n"
+    "mov    0x10(%%rdx),%%rdx\n"
+    "mulx   0x8(%%rsi),%%r10,%%r11\n"
+    "mov    0x20(%%rsi),%%rdx\n"
+    "mulx   0x20(%%rax),%%rcx,%%r8\n"
+    "mov    (%%rax),%%rdx\n"
+    "mulx   0x18(%%rsi),%%r9,%%rbx\n"
+    "mov    0x8(%%rax),%%rdx\n"
+    "mulx   0x10(%%rsi),%%r12,%%r13\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r12,%%r9\n"
+    "adox   %%rbx,%%r13\n"
+    "mov    0x18(%%rax),%%rdx\n"
+    "mulx   (%%rsi),%%rbx,%%r12\n"
+    "adcx   %%r10,%%r9\n"
+    "adcx   %%r13,%%r11\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%rbx,%%r9\n"
+    "adox   %%r11,%%r12\n"
+    "mov    (%%rax),%%rdx\n"
+    "mulx   0x20(%%rsi),%%r10,%%r13\n"
+    "mov    0x18(%%rsi),%%rdx\n"
+    "mulx   0x8(%%rax),%%rbx,%%r11\n"
+    "adcx   %%rbx,%%r10\n"
+    "adcx   %%r13,%%r11\n"
+    "mov    0x8(%%rsi),%%rdx\n"
+    "mulx   (%%rax),%%r13,%%rbx\n"
+    "movabs $0x1000003d10,%%rdx\n"
+    "mulx   %%rcx,%%r14,%%r15\n"
+    "mov    0x18(%%rax),%%rdx\n"
+    "mov    %%rdi,%q0\n"
+    "mulx   0x8(%%rsi),%%rcx,%%rdi\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r9,%%r14\n"
+    "adox   %%r15,%%r12\n"
+    "mov    0x10(%%rax),%%rdx\n"
+    "mulx   0x10(%%rsi),%%r9,%%r15\n"
+    "adcx   %%r9,%%r10\n"
+    "adcx   %%r11,%%r15\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%rcx,%%r10\n"
+    "adox   %%r15,%%rdi\n"
+    "mov    (%%rsi),%%rdx\n"
+    "mulx   0x20(%%rax),%%r11,%%rcx\n"
+    "adcx   %%r11,%%r10\n"
+    "adcx   %%rdi,%%rcx\n"
+    "mov    %%r14,%%rdx\n"
+    "shrd   $0x34,%%r12,%%rdx\n"
+    "add    %%r10,%%rdx\n"
+    "adc    $0x0,%%rcx\n"
+    "mov    %%rdx,%%r12\n"
+    "mov    0x20(%%rsi),%%rdx\n"
+    "mulx   0x8(%%rax),%%r9,%%r15\n"
+    "movabs $0x1000003d10000,%%rdx\n"
+    "mulx   %%r8,%%rdi,%%r11\n"
+    "add    %%r12,%%rdi\n"
+    "adcx   %%r11,%%rcx\n"
+    "mov    %%rdi,%%r10\n"
+    "shrd   $0x34,%%rcx,%%r10\n"
+    "mov    0x20(%%rax),%%rdx\n"
+    "mulx   0x8(%%rsi),%%r12,%%r8\n"
+    "mov    0x10(%%rax),%%rdx\n"
+    "mulx   0x18(%%rsi),%%r11,%%rcx\n"
+    "add    %%r11,%%r9\n"
+    "adcx   %%r15,%%rcx\n"
+    "mov    0x10(%%rsi),%%rdx\n"
+    "mulx   0x18(%%rax),%%r15,%%r11\n"
+    "add    %%r15,%%r9\n"
+    "adcx   %%rcx,%%r11\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r12,%%r9\n"
+    "adox   %%r11,%%r8\n"
+    "adcx   %%r9,%%r10\n"
+    "adc    $0x0,%%r8\n"
+    "movabs $0xfffffffffffff,%%r12\n"
+    "mov    %%r10,%%rcx\n"
+    "and    %%r12,%%rcx\n"
+    "shl    $0x4,%%rcx\n"
+    "mov    0x18(%%rax),%%rdx\n"
+    "mulx   0x18(%%rsi),%%r15,%%r11\n"
+    "and    %%r12,%%rdi\n"
+    "mov    %%rdi,%%rdx\n"
+    "shr    $0x30,%%rdx\n"
+    "lea    (%%rcx,%%rdx,1),%%rcx\n"
+    "shrd   $0x34,%%r8,%%r10\n"
+    "mov    (%%rsi),%%rdx\n"
+    "mulx   (%%rax),%%r9,%%r8\n"
+    "movabs $0x1000003d1,%%rdx\n"
+    "mov    %%rbx,%q1\n"
+    "mulx   %%rcx,%%r12,%%rbx\n"
+    "add    %%r9,%%r12\n"
+    "adcx   %%rbx,%%r8\n"
+    "mov    0x10(%%rax),%%rdx\n"
+    "mulx   0x20(%%rsi),%%rcx,%%r9\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r15,%%rcx\n"
+    "adox   %%r9,%%r11\n"
+    "mov    0x10(%%rsi),%%rdx\n"
+    "mulx   0x20(%%rax),%%r15,%%rbx\n"
+    "mov    %%r12,%%rdx\n"
+    "shrd   $0x34,%%r8,%%rdx\n"
+    "xor    %%r8,%%r8\n"
+    "adox   %%r15,%%rcx\n"
+    "adox   %%r11,%%rbx\n"
+    "adcx   %%rcx,%%r10\n"
+    "adc    $0x0,%%rbx\n"
+    "mov    %%r10,%%r9\n"
+    "shrd   $0x34,%%rbx,%%r9\n"
+    "movabs $0xfffffffffffff,%%r11\n"
+    "and    %%r11,%%r10\n"
+    "mov    %%rdx,%%r15\n"
+    "mov    0x8(%%rax),%%rdx\n"
+    "mulx   (%%rsi),%%rcx,%%rbx\n"
+    "adox   %%rcx,%%r13\n"
+    "adox   %q1,%%rbx\n"
+    "adcx   %%r13,%%r15\n"
+    "adc    $0x0,%%rbx\n"
+    "mov    (%%rax),%%rdx\n"
+    "mulx   0x10(%%rsi),%%rcx,%%r13\n"
+    "mov    0x20(%%rsi),%%rdx\n"
+    "mulx   0x18(%%rax),%%r8,%%r11\n"
+    "mov    0x20(%%rax),%%rdx\n"
+    "mov    %%r13,%q2\n"
+    "mov    %%rcx,%q3\n"
+    "mulx   0x18(%%rsi),%%r13,%%rcx\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r13,%%r8\n"
+    "adox   %%r11,%%rcx\n"
+    "movabs $0x1000003d10,%%r11\n"
+    "mov    %%r11,%%rdx\n"
+    "mulx   %%r10,%%r11,%%r13\n"
+    "adcx   %%r8,%%r9\n"
+    "adc    $0x0,%%rcx\n"
+    "add    %%r15,%%r11\n"
+    "adcx   %%r13,%%rbx\n"
+    "mov    %%r11,%%r10\n"
+    "shrd   $0x34,%%rbx,%%r10\n"
+    "mov    0x8(%%rax),%%rdx\n"
+    "mulx   0x8(%%rsi),%%r15,%%r8\n"
+    "movabs $0xfffffffffffff,%%rdx\n"
+    "and    %%rdx,%%r11\n"
+    "mov    %q0,%%r13\n"
+    "mov    %%r11,0x8(%%r13)\n"
+    "movabs $0x1000003d10000,%%rbx\n"
+    "mov    %%rbx,%%rdx\n"
+    "mulx   %%rcx,%%rbx,%%r11\n"
+    "mov    %%r15,%%rcx\n"
+    "adox   %q3,%%rcx\n"
+    "adox   %q2,%%r8\n"
+    "mov    0x10(%%rax),%%rdx\n"
+    "mulx   (%%rsi),%%r15,%%r13\n"
+    "adcx   %%r15,%%rcx\n"
+    "adcx   %%r8,%%r13\n"
+    "add    %%rcx,%%r10\n"
+    "adc    $0x0,%%r13\n"
+    "movabs $0x1000003d10,%%rdx\n"
+    "mulx   %%r9,%%r8,%%r15\n"
+    "xor    %%r9,%%r9\n"
+    "adox   %%r10,%%r8\n"
+    "adox   %%r15,%%r13\n"
+    "movabs $0xfffffffffffff,%%rcx\n"
+    "mov    %%r8,%%r10\n"
+    "and    %%rcx,%%r10\n"
+    "mov    %q0,%%r15\n"
+    "mov    %%r10,0x10(%%r15)\n"
+    "and    %%rcx,%%r14\n"
+    "shrd   $0x34,%%r13,%%r8\n"
+    "movabs $0xffffffffffff,%%r13\n"
+    "and    %%r13,%%rdi\n"
+    "lea    (%%r14,%%r8,1),%%r14\n"
+    "adox   %%r14,%%rbx\n"
+    "adox   %%r9,%%r11\n"
+    "mov    %%rbx,%%r10\n"
+    "shrd   $0x34,%%r11,%%r10\n"
+    "and    %%rcx,%%rbx\n"
+    "mov    %%rbx,0x18(%%r15)\n"
+    "and    %%rcx,%%r12\n"
+    "mov    %%r12,(%%r15)\n"
+    "lea    (%%rdi,%%r10,1),%%rdi\n"
+    "mov    %%rdi,0x20(%%r15)\n"
 
-    /* d += a3 * b0 */
-    "movq 0(%%rbx),%%rax\n"
-    "mulq %%r13\n"
-    "movq %%rax,%%rcx\n"
-    "movq %%rdx,%%r15\n"
-    /* d += a2 * b1 */
-    "movq 8(%%rbx),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a1 * b2 */
-    "movq 16(%%rbx),%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d = a0 * b3 */
-    "movq 24(%%rbx),%%rax\n"
-    "mulq %%r10\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* c = a4 * b4 */
-    "movq 32(%%rbx),%%rax\n"
-    "mulq %%r14\n"
-    "movq %%rax,%%r8\n"
-    "movq %%rdx,%%r9\n"
-    /* d += (c & M) * R */
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* c >>= 52 (%%r8 only) */
-    "shrdq $52,%%r9,%%r8\n"
-    /* t3 (tmp1) = d & M */
-    "movq %%rcx,%%rsi\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rsi\n"
-    "movq %%rsi,%q1\n"
-    /* d >>= 52 */
-    "shrdq $52,%%r15,%%rcx\n"
-    "xorq %%r15,%%r15\n"
-    /* d += a4 * b0 */
-    "movq 0(%%rbx),%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a3 * b1 */
-    "movq 8(%%rbx),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a2 * b2 */
-    "movq 16(%%rbx),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a1 * b3 */
-    "movq 24(%%rbx),%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a0 * b4 */
-    "movq 32(%%rbx),%%rax\n"
-    "mulq %%r10\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += c * R */
-    "movq %%r8,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* t4 = d & M (%%rsi) */
-    "movq %%rcx,%%rsi\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rsi\n"
-    /* d >>= 52 */
-    "shrdq $52,%%r15,%%rcx\n"
-    "xorq %%r15,%%r15\n"
-    /* tx = t4 >> 48 (tmp3) */
-    "movq %%rsi,%%rax\n"
-    "shrq $48,%%rax\n"
-    "movq %%rax,%q3\n"
-    /* t4 &= (M >> 4) (tmp2) */
-    "movq $0xffffffffffff,%%rax\n"
-    "andq %%rax,%%rsi\n"
-    "movq %%rsi,%q2\n"
-    /* c = a0 * b0 */
-    "movq 0(%%rbx),%%rax\n"
-    "mulq %%r10\n"
-    "movq %%rax,%%r8\n"
-    "movq %%rdx,%%r9\n"
-    /* d += a4 * b1 */
-    "movq 8(%%rbx),%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a3 * b2 */
-    "movq 16(%%rbx),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a2 * b3 */
-    "movq 24(%%rbx),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a1 * b4 */
-    "movq 32(%%rbx),%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* u0 = d & M (%%rsi) */
-    "movq %%rcx,%%rsi\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rsi\n"
-    /* d >>= 52 */
-    "shrdq $52,%%r15,%%rcx\n"
-    "xorq %%r15,%%r15\n"
-    /* u0 = (u0 << 4) | tx (%%rsi) */
-    "shlq $4,%%rsi\n"
-    "movq %q3,%%rax\n"
-    "orq %%rax,%%rsi\n"
-    /* c += u0 * (R >> 4) */
-    "movq $0x1000003d1,%%rax\n"
-    "mulq %%rsi\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* r[0] = c & M */
-    "movq %%r8,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq %%rax,0(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* c += a1 * b0 */
-    "movq 0(%%rbx),%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* c += a0 * b1 */
-    "movq 8(%%rbx),%%rax\n"
-    "mulq %%r10\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d += a4 * b2 */
-    "movq 16(%%rbx),%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a3 * b3 */
-    "movq 24(%%rbx),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a2 * b4 */
-    "movq 32(%%rbx),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* c += (d & M) * R */
-    "movq %%rcx,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d >>= 52 */
-    "shrdq $52,%%r15,%%rcx\n"
-    "xorq %%r15,%%r15\n"
-    /* r[1] = c & M */
-    "movq %%r8,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq %%rax,8(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* c += a2 * b0 */
-    "movq 0(%%rbx),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* c += a1 * b1 */
-    "movq 8(%%rbx),%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* c += a0 * b2 (last use of %%r10 = a0) */
-    "movq 16(%%rbx),%%rax\n"
-    "mulq %%r10\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* fetch t3 (%%r10, overwrites a0), t4 (%%rsi) */
-    "movq %q2,%%rsi\n"
-    "movq %q1,%%r10\n"
-    /* d += a4 * b3 */
-    "movq 24(%%rbx),%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* d += a3 * b4 */
-    "movq 32(%%rbx),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rcx\n"
-    "adcq %%rdx,%%r15\n"
-    /* c += (d & M) * R */
-    "movq %%rcx,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d >>= 52 (%%rcx only) */
-    "shrdq $52,%%r15,%%rcx\n"
-    /* r[2] = c & M */
-    "movq %%r8,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq %%rax,16(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* c += t3 */
-    "addq %%r10,%%r8\n"
-    /* c += d * R */
-    "movq %%rcx,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* r[3] = c & M */
-    "movq %%r8,%%rax\n"
-    "movq $0xfffffffffffff,%%rdx\n"
-    "andq %%rdx,%%rax\n"
-    "movq %%rax,24(%%rdi)\n"
-    /* c >>= 52 (%%r8 only) */
-    "shrdq $52,%%r9,%%r8\n"
-    /* c += t4 (%%r8 only) */
-    "addq %%rsi,%%r8\n"
-    /* r[4] = c */
-    "movq %%r8,32(%%rdi)\n"
-: "+S"(a), "=&m"(tmp1), "=&m"(tmp2), "=&m"(tmp3)
-: "b"(b), "D"(r)
-: "%rax", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15", "cc", "memory"
-);
+      : "=&m"(tmp0), "=&m"(tmp1), "=&m"(tmp2), "=&m"(tmp3), "+D"(r), "+d"(b)
+      : "S"(a)
+      : "rax", "rbx", "rcx",  "r8", "r9", "r10", "r11", "r12", "r13",
+        "r14", "r15", "cc", "memory");
 }
 
 SECP256K1_INLINE static void secp256k1_fe_sqr_inner(uint64_t *r, const uint64_t *a) {
-/**
- * Registers: rdx:rax = multiplication accumulator
- *            r9:r8   = c
- *            rcx:rbx = d
- *            r10-r14 = a0-a4
- *            r15     = M (0xfffffffffffff)
- *            rdi     = r
- *            rsi     = a / t?
- */
-  uint64_t tmp1, tmp2, tmp3;
-__asm__ __volatile__(
-    "movq 0(%%rsi),%%r10\n"
-    "movq 8(%%rsi),%%r11\n"
-    "movq 16(%%rsi),%%r12\n"
-    "movq 24(%%rsi),%%r13\n"
-    "movq 32(%%rsi),%%r14\n"
-    "movq $0xfffffffffffff,%%r15\n"
+  uint64_t tmp0, tmp1, tmp2;
 
-    /* d = (a0*2) * a3 */
-    "leaq (%%r10,%%r10,1),%%rax\n"
-    "mulq %%r13\n"
-    "movq %%rax,%%rbx\n"
-    "movq %%rdx,%%rcx\n"
-    /* d += (a1*2) * a2 */
-    "leaq (%%r11,%%r11,1),%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* c = a4 * a4 */
-    "movq %%r14,%%rax\n"
-    "mulq %%r14\n"
-    "movq %%rax,%%r8\n"
-    "movq %%rdx,%%r9\n"
-    /* d += (c & M) * R */
-    "andq %%r15,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* c >>= 52 (%%r8 only) */
-    "shrdq $52,%%r9,%%r8\n"
-    /* t3 (tmp1) = d & M */
-    "movq %%rbx,%%rsi\n"
-    "andq %%r15,%%rsi\n"
-    "movq %%rsi,%q1\n"
-    /* d >>= 52 */
-    "shrdq $52,%%rcx,%%rbx\n"
-    "xorq %%rcx,%%rcx\n"
-    /* a4 *= 2 */
-    "addq %%r14,%%r14\n"
-    /* d += a0 * a4 */
-    "movq %%r10,%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* d+= (a1*2) * a3 */
-    "leaq (%%r11,%%r11,1),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* d += a2 * a2 */
-    "movq %%r12,%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* d += c * R */
-    "movq %%r8,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* t4 = d & M (%%rsi) */
-    "movq %%rbx,%%rsi\n"
-    "andq %%r15,%%rsi\n"
-    /* d >>= 52 */
-    "shrdq $52,%%rcx,%%rbx\n"
-    "xorq %%rcx,%%rcx\n"
-    /* tx = t4 >> 48 (tmp3) */
-    "movq %%rsi,%%rax\n"
-    "shrq $48,%%rax\n"
-    "movq %%rax,%q3\n"
-    /* t4 &= (M >> 4) (tmp2) */
-    "movq $0xffffffffffff,%%rax\n"
-    "andq %%rax,%%rsi\n"
-    "movq %%rsi,%q2\n"
-    /* c = a0 * a0 */
-    "movq %%r10,%%rax\n"
-    "mulq %%r10\n"
-    "movq %%rax,%%r8\n"
-    "movq %%rdx,%%r9\n"
-    /* d += a1 * a4 */
-    "movq %%r11,%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* d += (a2*2) * a3 */
-    "leaq (%%r12,%%r12,1),%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* u0 = d & M (%%rsi) */
-    "movq %%rbx,%%rsi\n"
-    "andq %%r15,%%rsi\n"
-    /* d >>= 52 */
-    "shrdq $52,%%rcx,%%rbx\n"
-    "xorq %%rcx,%%rcx\n"
-    /* u0 = (u0 << 4) | tx (%%rsi) */
-    "shlq $4,%%rsi\n"
-    "movq %q3,%%rax\n"
-    "orq %%rax,%%rsi\n"
-    /* c += u0 * (R >> 4) */
-    "movq $0x1000003d1,%%rax\n"
-    "mulq %%rsi\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* r[0] = c & M */
-    "movq %%r8,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq %%rax,0(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* a0 *= 2 */
-    "addq %%r10,%%r10\n"
-    /* c += a0 * a1 */
-    "movq %%r10,%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d += a2 * a4 */
-    "movq %%r12,%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* d += a3 * a3 */
-    "movq %%r13,%%rax\n"
-    "mulq %%r13\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* c += (d & M) * R */
-    "movq %%rbx,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d >>= 52 */
-    "shrdq $52,%%rcx,%%rbx\n"
-    "xorq %%rcx,%%rcx\n"
-    /* r[1] = c & M */
-    "movq %%r8,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq %%rax,8(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* c += a0 * a2 (last use of %%r10) */
-    "movq %%r10,%%rax\n"
-    "mulq %%r12\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* fetch t3 (%%r10, overwrites a0),t4 (%%rsi) */
-    "movq %q2,%%rsi\n"
-    "movq %q1,%%r10\n"
-    /* c += a1 * a1 */
-    "movq %%r11,%%rax\n"
-    "mulq %%r11\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d += a3 * a4 */
-    "movq %%r13,%%rax\n"
-    "mulq %%r14\n"
-    "addq %%rax,%%rbx\n"
-    "adcq %%rdx,%%rcx\n"
-    /* c += (d & M) * R */
-    "movq %%rbx,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* d >>= 52 (%%rbx only) */
-    "shrdq $52,%%rcx,%%rbx\n"
-    /* r[2] = c & M */
-    "movq %%r8,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq %%rax,16(%%rdi)\n"
-    /* c >>= 52 */
-    "shrdq $52,%%r9,%%r8\n"
-    "xorq %%r9,%%r9\n"
-    /* c += t3 */
-    "addq %%r10,%%r8\n"
-    /* c += d * R */
-    "movq %%rbx,%%rax\n"
-    "movq $0x1000003d10,%%rdx\n"
-    "mulq %%rdx\n"
-    "addq %%rax,%%r8\n"
-    "adcq %%rdx,%%r9\n"
-    /* r[3] = c & M */
-    "movq %%r8,%%rax\n"
-    "andq %%r15,%%rax\n"
-    "movq %%rax,24(%%rdi)\n"
-    /* c >>= 52 (%%r8 only) */
-    "shrdq $52,%%r9,%%r8\n"
-    /* c += t4 (%%r8 only) */
-    "addq %%rsi,%%r8\n"
-    /* r[4] = c */
-    "movq %%r8,32(%%rdi)\n"
-: "+S"(a), "=&m"(tmp1), "=&m"(tmp2), "=&m"(tmp3)
-: "D"(r)
-: "%rax", "%rbx", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r12", "%r13", "%r14", "%r15", "cc", "memory"
-);
-}
+  __asm__ __volatile__(
+    "mov    0x8(%%rsi),%%rax\n"
+    "lea    (%%rax,%%rax,1),%%r10\n"
+    "mov    0x20(%%rsi),%%rdx\n"
+    "mulx   %%rdx,%%rax,%%r11\n"
+    "mov    (%%rsi),%%rdx\n"
+    "lea    (%%rdx,%%rdx,1),%%rcx\n"
+    "mov    %%rcx,%%rdx\n"
+    "mulx   0x18(%%rsi),%%rcx,%%r8\n"
+    "mov    %%rdx,%%r9\n"
+    "mov    0x10(%%rsi),%%rdx\n"
+    "mulx   %%r10,%%rbx,%%r12\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%rcx,%%rbx\n"
+    "adox   %%r12,%%r8\n"
+    "movabs $0x1000003d10,%%rcx\n"
+    "mov    %%rax,%%rdx\n"
+    "mulx   %%rcx,%%rax,%%r12\n"
+    "adcx   %%rbx,%%rax\n"
+    "adcx   %%r12,%%r8\n"
+    "mov    %%rax,%%rdx\n"
+    "shrd   $0x34,%%r8,%%rdx\n"
+    "mov    %%rdx,%%rbx\n"
+    "mov    0x10(%%rsi),%%rdx\n"
+    "mulx   %%rdx,%%r12,%%r8\n"
+    "mov    %%r10,%%rdx\n"
+    "mulx   0x18(%%rsi),%%r10,%%r13\n"
+    "xor    %%r14,%%r14\n"
+    "adox   %%r10,%%r12\n"
+    "adox   %%r8,%%r13\n"
+    "mov    %%rdx,%%r8\n"
+    "mov    0x20(%%rsi),%%rdx\n"
+    "mulx   %%r9,%%r10,%%r14\n"
+    "adcx   %%r10,%%r12\n"
+    "adcx   %%r13,%%r14\n"
+    "add    %%r12,%%rbx\n"
+    "adc    $0x0,%%r14\n"
+    "movabs $0x1000003d10000,%%rdx\n"
+    "mulx   %%r11,%%r13,%%r10\n"
+    "mov    0x10(%%rsi),%%r11\n"
+    "mov    %%r11,%%r12\n"
+    "shl    %%r12\n"
+    "mov    %%r8,%%rdx\n"
+    "mulx   0x20(%%rsi),%%r11,%%r8\n"
+    "mov    %%r12,%%rdx\n"
+    "mulx   0x18(%%rsi),%%r12,%%r15\n"
+    "add    %%rbx,%%r13\n"
+    "adcx   %%r10,%%r14\n"
+    "mov    %%r13,%%rbx\n"
+    "shrd   $0x34,%%r14,%%rbx\n"
+    "movabs $0xfffffffffffff,%%r10\n"
+    "and    %%r10,%%r13\n"
+    "adox   %%r11,%%r12\n"
+    "adox   %%r15,%%r8\n"
+    "adcx   %%r12,%%rbx\n"
+    "adc    $0x0,%%r8\n"
+    "mov    %%rdx,%%r11\n"
+    "mov    0x18(%%rsi),%%rdx\n"
+    "mulx   %%rdx,%%r15,%%r14\n"
+    "mov    %%r11,%%rdx\n"
+    "mulx   0x20(%%rsi),%%r11,%%r12\n"
+    "mov    %%rbx,%%rdx\n"
+    "and    %%r10,%%rdx\n"
+    "shrd   $0x34,%%r8,%%rbx\n"
+    "xor    %%r8,%%r8\n"
+    "adox   %%r11,%%r15\n"
+    "adox   %%r14,%%r12\n"
+    "adcx   %%r15,%%rbx\n"
+    "adc    $0x0,%%r12\n"
+    "mov    %%rbx,%%r14\n"
+    "and    %%r10,%%r14\n"
+    "shl    $0x4,%%rdx\n"
+    "mov    %%r13,%%r11\n"
+    "shr    $0x30,%%r11\n"
+    "shrd   $0x34,%%r12,%%rbx\n"
+    "lea    (%%rdx,%%r11,1),%%rdx\n"
+    "movabs $0x1000003d1,%%r15\n"
+    "mulx   %%r15,%%r12,%%r11\n"
+    "imul   $0x2,0x18(%%rsi),%%rdx\n"
+    "mulx   0x20(%%rsi),%%r8,%%r15\n"
+    "mov    (%%rsi),%%rdx\n"
+    "mulx   %%rdx,%%r10,%%rcx\n"
+    "xor    %%rdx,%%rdx\n"
+    "adox   %%r10,%%r12\n"
+    "adox   %%r11,%%rcx\n"
+    "adcx   %%r8,%%rbx\n"
+    "adc    $0x0,%%r15\n"
+    "mov    %%r12,%%r11\n"
+    "shrd   $0x34,%%rcx,%%r11\n"
+    "mov    0x8(%%rsi),%%rdx\n"
+    "mulx   %%r9,%%r8,%%r10\n"
+    "add    %%r8,%%r11\n"
+    "adc    $0x0,%%r10\n"
+    "mov    0x8(%%rsi),%%rdx\n"
+    "mulx   %%rdx,%%rcx,%%r8\n"
+    "mov    %%r9,%%rdx\n"
+    "mov    %%rdi,%q0\n"
+    "mulx   0x10(%%rsi),%%r9,%%rdi\n"
+    "movabs $0x1000003d10,%%rdx\n"
+    "mov    %%rbx,%q1\n"
+    "mov    %%r8,%q2\n"
+    "mulx   %%r14,%%rbx,%%r8\n"
+    "xor    %%r14,%%r14\n"
+    "adox   %%r11,%%rbx\n"
+    "adox   %%r8,%%r10\n"
+    "adcx   %%r9,%%rcx\n"
+    "adcx   %q2,%%rdi\n"
+    "mov    %%rbx,%%r11\n"
+    "shrd   $0x34,%%r10,%%r11\n"
+    "add    %%rcx,%%r11\n"
+    "adc    $0x0,%%rdi\n"
+    "mulx   %q1,%%r9,%%r8\n"
+    "xor    %%r10,%%r10\n"
+    "adox   %%r11,%%r9\n"
+    "adox   %%r8,%%rdi\n"
+    "movabs $0xfffffffffffff,%%r14\n"
+    "and    %%r14,%%rbx\n"
+    "mov    %%r9,%%rcx\n"
+    "shrd   $0x34,%%rdi,%%rcx\n"
+    "mov    %q0,%%r11\n"
+    "mov    %%rbx,0x8(%%r11)\n"
+    "movabs $0x1000003d10000,%%r8\n"
+    "mov    %%r15,%%rdx\n"
+    "mulx   %%r8,%%r15,%%rdi\n"
+    "and    %%r14,%%r9\n"
+    "mov    %%r9,0x10(%%r11)\n"
+    "and    %%r14,%%rax\n"
+    "lea    (%%rax,%%rcx,1),%%rax\n"
+    "adox   %%rax,%%r15\n"
+    "adox   %%r10,%%rdi\n"
+    "mov    %%r15,%%rbx\n"
+    "and    %%r14,%%rbx\n"
+    "shrd   $0x34,%%rdi,%%r15\n"
+    "mov    %%rbx,0x18(%%r11)\n"
+    "movabs $0xffffffffffff,%%rcx\n"
+    "and    %%rcx,%%r13\n"
+    "lea    0x0(%%r13,%%r15,1),%%r13\n"
+    "mov    %%r13,0x20(%%r11)\n"
+    "and    %%r14,%%r12\n"
+    "mov    %%r12,(%%r11)\n"
 
-#endif /* SECP256K1_FIELD_INNER5X52_IMPL_H */
+    : "=&m"(tmp0), "=&m"(tmp1), "=&m"(tmp2), "+D"(r)
+    : "D"(r), "S"(a)
+    : "rax", "rbx", "rcx", "rdx", "r8", "r9", "r10", "r11", "r12",
+      "r13", "r14", "r15", "cc", "memory");
+}
+#endif