visaljs 0.0.1 → 0.0.3

package/runtime/third_party/boringssl/gen/bcm/p256-x86_64-asm-linux.S

@@ -0,0 +1,4600 @@
+// This file is generated from a similarly-named Perl script in the BoringSSL
+// source tree. Do not edit by hand.
+
+#include <openssl/asm_base.h>
+
+#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64) && defined(__ELF__)
+.text	
+
+// The polynomial
+.section	.rodata
+.align	64
+p256_constants:
+.Lpoly:
+.quad	0xffffffffffffffff, 0x00000000ffffffff, 0x0000000000000000, 0xffffffff00000001
+
+.LOne:
+.long	1,1,1,1,1,1,1,1
+.LTwo:
+.long	2,2,2,2,2,2,2,2
+.LThree:
+.long	3,3,3,3,3,3,3,3
+.LONE_mont:
+.quad	0x0000000000000001, 0xffffffff00000000, 0xffffffffffffffff, 0x00000000fffffffe
+
+// Constants for computations modulo ord(p256)
+.Lord:
+.quad	0xf3b9cac2fc632551, 0xbce6faada7179e84, 0xffffffffffffffff, 0xffffffff00000000
+.LordK:
+.quad	0xccd1c8aaee00bc4f
+.text	
+
+// ############################################################################## 
+// void ecp_nistz256_neg(uint64_t res[4], uint64_t a[4]);
+.globl	ecp_nistz256_neg
+.hidden ecp_nistz256_neg
+.type	ecp_nistz256_neg,@function
+.align	32
+ecp_nistz256_neg:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-16
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-24
+.Lneg_body:
+
+	xorq	%r8,%r8
+	xorq	%r9,%r9
+	xorq	%r10,%r10
+	xorq	%r11,%r11
+	xorq	%r13,%r13
+
+	subq	0(%rsi),%r8
+	sbbq	8(%rsi),%r9
+	sbbq	16(%rsi),%r10
+	movq	%r8,%rax
+	sbbq	24(%rsi),%r11
+	leaq	.Lpoly(%rip),%rsi
+	movq	%r9,%rdx
+	sbbq	$0,%r13
+
+	addq	0(%rsi),%r8
+	movq	%r10,%rcx
+	adcq	8(%rsi),%r9
+	adcq	16(%rsi),%r10
+	movq	%r11,%r12
+	adcq	24(%rsi),%r11
+	testq	%r13,%r13
+
+	cmovzq	%rax,%r8
+	cmovzq	%rdx,%r9
+	movq	%r8,0(%rdi)
+	cmovzq	%rcx,%r10
+	movq	%r9,8(%rdi)
+	cmovzq	%r12,%r11
+	movq	%r10,16(%rdi)
+	movq	%r11,24(%rdi)
+
+	movq	0(%rsp),%r13
+.cfi_restore	%r13
+	movq	8(%rsp),%r12
+.cfi_restore	%r12
+	leaq	16(%rsp),%rsp
+.cfi_adjust_cfa_offset	-16
+.Lneg_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_neg,.-ecp_nistz256_neg
+// ############################################################################## 
+// void ecp_nistz256_ord_mul_mont(
+// uint64_t res[4],
+// uint64_t a[4],
+// uint64_t b[4]);
+
+.globl	ecp_nistz256_ord_mul_mont_nohw
+.hidden ecp_nistz256_ord_mul_mont_nohw
+.type	ecp_nistz256_ord_mul_mont_nohw,@function
+.align	32
+ecp_nistz256_ord_mul_mont_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lord_mul_body:
+
+	movq	0(%rdx),%rax
+	movq	%rdx,%rbx
+	leaq	.Lord(%rip),%r14
+	movq	.LordK(%rip),%r15
+
+// ################################ * b[0]
+	movq	%rax,%rcx
+	mulq	0(%rsi)
+	movq	%rax,%r8
+	movq	%rcx,%rax
+	movq	%rdx,%r9
+
+	mulq	8(%rsi)
+	addq	%rax,%r9
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	16(%rsi)
+	addq	%rax,%r10
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+
+	movq	%r8,%r13
+	imulq	%r15,%r8
+
+	movq	%rdx,%r11
+	mulq	24(%rsi)
+	addq	%rax,%r11
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r12
+
+// ################################ First reduction step
+	mulq	0(%r14)
+	movq	%r8,%rbp
+	addq	%rax,%r13  // guaranteed to be zero
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	subq	%r8,%r10
+	sbbq	$0,%r8  // can't borrow
+
+	mulq	8(%r14)
+	addq	%rcx,%r9
+	adcq	$0,%rdx
+	addq	%rax,%r9
+	movq	%rbp,%rax
+	adcq	%rdx,%r10
+	movq	%rbp,%rdx
+	adcq	$0,%r8  // can't overflow
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r11
+	movq	8(%rbx),%rax
+	sbbq	%rdx,%rbp  // can't borrow
+
+	addq	%r8,%r11
+	adcq	%rbp,%r12
+	adcq	$0,%r13
+
+// ################################ * b[1]
+	movq	%rax,%rcx
+	mulq	0(%rsi)
+	addq	%rax,%r9
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	8(%rsi)
+	addq	%rbp,%r10
+	adcq	$0,%rdx
+	addq	%rax,%r10
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	16(%rsi)
+	addq	%rbp,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+
+	movq	%r9,%rcx
+	imulq	%r15,%r9
+
+	movq	%rdx,%rbp
+	mulq	24(%rsi)
+	addq	%rbp,%r12
+	adcq	$0,%rdx
+	xorq	%r8,%r8
+	addq	%rax,%r12
+	movq	%r9,%rax
+	adcq	%rdx,%r13
+	adcq	$0,%r8
+
+// ################################ Second reduction step
+	mulq	0(%r14)
+	movq	%r9,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	%r9,%rax
+	adcq	%rdx,%rcx
+
+	subq	%r9,%r11
+	sbbq	$0,%r9  // can't borrow
+
+	mulq	8(%r14)
+	addq	%rcx,%r10
+	adcq	$0,%rdx
+	addq	%rax,%r10
+	movq	%rbp,%rax
+	adcq	%rdx,%r11
+	movq	%rbp,%rdx
+	adcq	$0,%r9  // can't overflow
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r12
+	movq	16(%rbx),%rax
+	sbbq	%rdx,%rbp  // can't borrow
+
+	addq	%r9,%r12
+	adcq	%rbp,%r13
+	adcq	$0,%r8
+
+// ################################# * b[2]
+	movq	%rax,%rcx
+	mulq	0(%rsi)
+	addq	%rax,%r10
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	8(%rsi)
+	addq	%rbp,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	16(%rsi)
+	addq	%rbp,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+
+	movq	%r10,%rcx
+	imulq	%r15,%r10
+
+	movq	%rdx,%rbp
+	mulq	24(%rsi)
+	addq	%rbp,%r13
+	adcq	$0,%rdx
+	xorq	%r9,%r9
+	addq	%rax,%r13
+	movq	%r10,%rax
+	adcq	%rdx,%r8
+	adcq	$0,%r9
+
+// ################################ Third reduction step
+	mulq	0(%r14)
+	movq	%r10,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	%r10,%rax
+	adcq	%rdx,%rcx
+
+	subq	%r10,%r12
+	sbbq	$0,%r10  // can't borrow
+
+	mulq	8(%r14)
+	addq	%rcx,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%rbp,%rax
+	adcq	%rdx,%r12
+	movq	%rbp,%rdx
+	adcq	$0,%r10  // can't overflow
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r13
+	movq	24(%rbx),%rax
+	sbbq	%rdx,%rbp  // can't borrow
+
+	addq	%r10,%r13
+	adcq	%rbp,%r8
+	adcq	$0,%r9
+
+// ################################ * b[3]
+	movq	%rax,%rcx
+	mulq	0(%rsi)
+	addq	%rax,%r11
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	8(%rsi)
+	addq	%rbp,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	16(%rsi)
+	addq	%rbp,%r13
+	adcq	$0,%rdx
+	addq	%rax,%r13
+	movq	%rcx,%rax
+	adcq	$0,%rdx
+
+	movq	%r11,%rcx
+	imulq	%r15,%r11
+
+	movq	%rdx,%rbp
+	mulq	24(%rsi)
+	addq	%rbp,%r8
+	adcq	$0,%rdx
+	xorq	%r10,%r10
+	addq	%rax,%r8
+	movq	%r11,%rax
+	adcq	%rdx,%r9
+	adcq	$0,%r10
+
+// ################################ Last reduction step
+	mulq	0(%r14)
+	movq	%r11,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	%r11,%rax
+	adcq	%rdx,%rcx
+
+	subq	%r11,%r13
+	sbbq	$0,%r11  // can't borrow
+
+	mulq	8(%r14)
+	addq	%rcx,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%rbp,%rax
+	adcq	%rdx,%r13
+	movq	%rbp,%rdx
+	adcq	$0,%r11  // can't overflow
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r8
+	sbbq	%rdx,%rbp  // can't borrow
+
+	addq	%r11,%r8
+	adcq	%rbp,%r9
+	adcq	$0,%r10
+
+// ################################ Subtract ord
+	movq	%r12,%rsi
+	subq	0(%r14),%r12
+	movq	%r13,%r11
+	sbbq	8(%r14),%r13
+	movq	%r8,%rcx
+	sbbq	16(%r14),%r8
+	movq	%r9,%rbp
+	sbbq	24(%r14),%r9
+	sbbq	$0,%r10
+
+	cmovcq	%rsi,%r12
+	cmovcq	%r11,%r13
+	cmovcq	%rcx,%r8
+	cmovcq	%rbp,%r9
+
+	movq	%r12,0(%rdi)
+	movq	%r13,8(%rdi)
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lord_mul_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_ord_mul_mont_nohw,.-ecp_nistz256_ord_mul_mont_nohw
+
+// ############################################################################## 
+// void ecp_nistz256_ord_sqr_mont(
+// uint64_t res[4],
+// uint64_t a[4],
+// uint64_t rep);
+
+.globl	ecp_nistz256_ord_sqr_mont_nohw
+.hidden ecp_nistz256_ord_sqr_mont_nohw
+.type	ecp_nistz256_ord_sqr_mont_nohw,@function
+.align	32
+ecp_nistz256_ord_sqr_mont_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lord_sqr_body:
+
+	movq	0(%rsi),%r8
+	movq	8(%rsi),%rax
+	movq	16(%rsi),%r14
+	movq	24(%rsi),%r15
+	leaq	.Lord(%rip),%rsi  // pointer to modulus
+	movq	%rdx,%rbx
+	jmp	.Loop_ord_sqr
+
+.align	32
+.Loop_ord_sqr:
+// ################################ a[1:] * a[0]
+	movq	%rax,%rbp  // put aside a[1]
+	mulq	%r8  // a[1] * a[0]
+	movq	%rax,%r9
+	movq	%rbp,%xmm1  // offload a[1]
+	movq	%r14,%rax
+	movq	%rdx,%r10
+
+	mulq	%r8  // a[2] * a[0]
+	addq	%rax,%r10
+	movq	%r15,%rax
+	movq	%r14,%xmm2  // offload a[2]
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%r8  // a[3] * a[0]
+	addq	%rax,%r11
+	movq	%r15,%rax
+	movq	%r15,%xmm3  // offload a[3]
+	adcq	$0,%rdx
+	movq	%rdx,%r12
+
+// ################################ a[3] * a[2]
+	mulq	%r14  // a[3] * a[2]
+	movq	%rax,%r13
+	movq	%r14,%rax
+	movq	%rdx,%r14
+
+// ################################ a[2:] * a[1]
+	mulq	%rbp  // a[2] * a[1]
+	addq	%rax,%r11
+	movq	%r15,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r15
+
+	mulq	%rbp  // a[3] * a[1]
+	addq	%rax,%r12
+	adcq	$0,%rdx
+
+	addq	%r15,%r12
+	adcq	%rdx,%r13
+	adcq	$0,%r14  // can't overflow
+
+// ################################ *2
+	xorq	%r15,%r15
+	movq	%r8,%rax
+	addq	%r9,%r9
+	adcq	%r10,%r10
+	adcq	%r11,%r11
+	adcq	%r12,%r12
+	adcq	%r13,%r13
+	adcq	%r14,%r14
+	adcq	$0,%r15
+
+// ################################ Missing products
+	mulq	%rax  // a[0] * a[0]
+	movq	%rax,%r8
+	movq	%xmm1,%rax
+	movq	%rdx,%rbp
+
+	mulq	%rax  // a[1] * a[1]
+	addq	%rbp,%r9
+	adcq	%rax,%r10
+	movq	%xmm2,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	%rax  // a[2] * a[2]
+	addq	%rbp,%r11
+	adcq	%rax,%r12
+	movq	%xmm3,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	movq	%r8,%rcx
+	imulq	32(%rsi),%r8  // *= .LordK
+
+	mulq	%rax  // a[3] * a[3]
+	addq	%rbp,%r13
+	adcq	%rax,%r14
+	movq	0(%rsi),%rax  // modulus[0]
+	adcq	%rdx,%r15  // can't overflow
+
+// ################################ First reduction step
+	mulq	%r8
+	movq	%r8,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	8(%rsi),%rax  // modulus[1]
+	adcq	%rdx,%rcx
+
+	subq	%r8,%r10
+	sbbq	$0,%rbp  // can't borrow
+
+	mulq	%r8
+	addq	%rcx,%r9
+	adcq	$0,%rdx
+	addq	%rax,%r9
+	movq	%r8,%rax
+	adcq	%rdx,%r10
+	movq	%r8,%rdx
+	adcq	$0,%rbp  // can't overflow
+
+	movq	%r9,%rcx
+	imulq	32(%rsi),%r9  // *= .LordK
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r11
+	movq	0(%rsi),%rax
+	sbbq	%rdx,%r8  // can't borrow
+
+	addq	%rbp,%r11
+	adcq	$0,%r8  // can't overflow
+
+// ################################ Second reduction step
+	mulq	%r9
+	movq	%r9,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	8(%rsi),%rax
+	adcq	%rdx,%rcx
+
+	subq	%r9,%r11
+	sbbq	$0,%rbp  // can't borrow
+
+	mulq	%r9
+	addq	%rcx,%r10
+	adcq	$0,%rdx
+	addq	%rax,%r10
+	movq	%r9,%rax
+	adcq	%rdx,%r11
+	movq	%r9,%rdx
+	adcq	$0,%rbp  // can't overflow
+
+	movq	%r10,%rcx
+	imulq	32(%rsi),%r10  // *= .LordK
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r8
+	movq	0(%rsi),%rax
+	sbbq	%rdx,%r9  // can't borrow
+
+	addq	%rbp,%r8
+	adcq	$0,%r9  // can't overflow
+
+// ################################ Third reduction step
+	mulq	%r10
+	movq	%r10,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	8(%rsi),%rax
+	adcq	%rdx,%rcx
+
+	subq	%r10,%r8
+	sbbq	$0,%rbp  // can't borrow
+
+	mulq	%r10
+	addq	%rcx,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%r10,%rax
+	adcq	%rdx,%r8
+	movq	%r10,%rdx
+	adcq	$0,%rbp  // can't overflow
+
+	movq	%r11,%rcx
+	imulq	32(%rsi),%r11  // *= .LordK
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r9
+	movq	0(%rsi),%rax
+	sbbq	%rdx,%r10  // can't borrow
+
+	addq	%rbp,%r9
+	adcq	$0,%r10  // can't overflow
+
+// ################################ Last reduction step
+	mulq	%r11
+	movq	%r11,%rbp
+	addq	%rax,%rcx  // guaranteed to be zero
+	movq	8(%rsi),%rax
+	adcq	%rdx,%rcx
+
+	subq	%r11,%r9
+	sbbq	$0,%rbp  // can't borrow
+
+	mulq	%r11
+	addq	%rcx,%r8
+	adcq	$0,%rdx
+	addq	%rax,%r8
+	movq	%r11,%rax
+	adcq	%rdx,%r9
+	movq	%r11,%rdx
+	adcq	$0,%rbp  // can't overflow
+
+	shlq	$32,%rax
+	shrq	$32,%rdx
+	subq	%rax,%r10
+	sbbq	%rdx,%r11  // can't borrow
+
+	addq	%rbp,%r10
+	adcq	$0,%r11  // can't overflow
+
+// ################################ Add bits [511:256] of the sqr result
+	xorq	%rdx,%rdx
+	addq	%r12,%r8
+	adcq	%r13,%r9
+	movq	%r8,%r12
+	adcq	%r14,%r10
+	adcq	%r15,%r11
+	movq	%r9,%rax
+	adcq	$0,%rdx
+
+// ################################ Compare to modulus
+	subq	0(%rsi),%r8
+	movq	%r10,%r14
+	sbbq	8(%rsi),%r9
+	sbbq	16(%rsi),%r10
+	movq	%r11,%r15
+	sbbq	24(%rsi),%r11
+	sbbq	$0,%rdx
+
+	cmovcq	%r12,%r8
+	cmovncq	%r9,%rax
+	cmovncq	%r10,%r14
+	cmovncq	%r11,%r15
+
+	decq	%rbx
+	jnz	.Loop_ord_sqr
+
+	movq	%r8,0(%rdi)
+	movq	%rax,8(%rdi)
+	pxor	%xmm1,%xmm1
+	movq	%r14,16(%rdi)
+	pxor	%xmm2,%xmm2
+	movq	%r15,24(%rdi)
+	pxor	%xmm3,%xmm3
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lord_sqr_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_ord_sqr_mont_nohw,.-ecp_nistz256_ord_sqr_mont_nohw
+// ############################################################################## 
+.globl	ecp_nistz256_ord_mul_mont_adx
+.hidden ecp_nistz256_ord_mul_mont_adx
+.type	ecp_nistz256_ord_mul_mont_adx,@function
+.align	32
+ecp_nistz256_ord_mul_mont_adx:
+.cfi_startproc	
+.Lecp_nistz256_ord_mul_mont_adx:
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lord_mulx_body:
+
+	movq	%rdx,%rbx
+	movq	0(%rdx),%rdx
+	movq	0(%rsi),%r9
+	movq	8(%rsi),%r10
+	movq	16(%rsi),%r11
+	movq	24(%rsi),%r12
+	leaq	-128(%rsi),%rsi  // control u-op density
+	leaq	.Lord-128(%rip),%r14
+	movq	.LordK(%rip),%r15
+
+// ################################ Multiply by b[0]
+	mulxq	%r9,%r8,%r9
+	mulxq	%r10,%rcx,%r10
+	mulxq	%r11,%rbp,%r11
+	addq	%rcx,%r9
+	mulxq	%r12,%rcx,%r12
+	movq	%r8,%rdx
+	mulxq	%r15,%rdx,%rax
+	adcq	%rbp,%r10
+	adcq	%rcx,%r11
+	adcq	$0,%r12
+
+// ################################ reduction
+	xorq	%r13,%r13  // %r13=0, cf=0, of=0
+	mulxq	0+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r8  // guaranteed to be zero
+	adoxq	%rbp,%r9
+
+	mulxq	8+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r9
+	adoxq	%rbp,%r10
+
+	mulxq	16+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	24+128(%r14),%rcx,%rbp
+	movq	8(%rbx),%rdx
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+	adcxq	%r8,%r12
+	adoxq	%r8,%r13
+	adcq	$0,%r13  // cf=0, of=0
+
+// ################################ Multiply by b[1]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r9
+	adoxq	%rbp,%r10
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r9,%rdx
+	mulxq	%r15,%rdx,%rax
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	adcxq	%r8,%r13
+	adoxq	%r8,%r8
+	adcq	$0,%r8  // cf=0, of=0
+
+// ################################ reduction
+	mulxq	0+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r9  // guaranteed to be zero
+	adoxq	%rbp,%r10
+
+	mulxq	8+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	16+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	24+128(%r14),%rcx,%rbp
+	movq	16(%rbx),%rdx
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+	adcxq	%r9,%r13
+	adoxq	%r9,%r8
+	adcq	$0,%r8  // cf=0, of=0
+
+// ################################ Multiply by b[2]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r10,%rdx
+	mulxq	%r15,%rdx,%rax
+	adcxq	%rcx,%r13
+	adoxq	%rbp,%r8
+
+	adcxq	%r9,%r8
+	adoxq	%r9,%r9
+	adcq	$0,%r9  // cf=0, of=0
+
+// ################################ reduction
+	mulxq	0+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r10  // guaranteed to be zero
+	adoxq	%rbp,%r11
+
+	mulxq	8+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	16+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	24+128(%r14),%rcx,%rbp
+	movq	24(%rbx),%rdx
+	adcxq	%rcx,%r13
+	adoxq	%rbp,%r8
+	adcxq	%r10,%r8
+	adoxq	%r10,%r9
+	adcq	$0,%r9  // cf=0, of=0
+
+// ################################ Multiply by b[3]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r13
+	adoxq	%rbp,%r8
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r11,%rdx
+	mulxq	%r15,%rdx,%rax
+	adcxq	%rcx,%r8
+	adoxq	%rbp,%r9
+
+	adcxq	%r10,%r9
+	adoxq	%r10,%r10
+	adcq	$0,%r10  // cf=0, of=0
+
+// ################################ reduction
+	mulxq	0+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r11  // guaranteed to be zero
+	adoxq	%rbp,%r12
+
+	mulxq	8+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	16+128(%r14),%rcx,%rbp
+	adcxq	%rcx,%r13
+	adoxq	%rbp,%r8
+
+	mulxq	24+128(%r14),%rcx,%rbp
+	leaq	128(%r14),%r14
+	movq	%r12,%rbx
+	adcxq	%rcx,%r8
+	adoxq	%rbp,%r9
+	movq	%r13,%rdx
+	adcxq	%r11,%r9
+	adoxq	%r11,%r10
+	adcq	$0,%r10
+
+// ################################
+// Branch-less conditional subtraction of P
+	movq	%r8,%rcx
+	subq	0(%r14),%r12
+	sbbq	8(%r14),%r13
+	sbbq	16(%r14),%r8
+	movq	%r9,%rbp
+	sbbq	24(%r14),%r9
+	sbbq	$0,%r10
+
+	cmovcq	%rbx,%r12
+	cmovcq	%rdx,%r13
+	cmovcq	%rcx,%r8
+	cmovcq	%rbp,%r9
+
+	movq	%r12,0(%rdi)
+	movq	%r13,8(%rdi)
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lord_mulx_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_ord_mul_mont_adx,.-ecp_nistz256_ord_mul_mont_adx
+
+.globl	ecp_nistz256_ord_sqr_mont_adx
+.hidden ecp_nistz256_ord_sqr_mont_adx
+.type	ecp_nistz256_ord_sqr_mont_adx,@function
+.align	32
+ecp_nistz256_ord_sqr_mont_adx:
+.cfi_startproc	
+_CET_ENDBR
+.Lecp_nistz256_ord_sqr_mont_adx:
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lord_sqrx_body:
+
+	movq	%rdx,%rbx
+	movq	0(%rsi),%rdx
+	movq	8(%rsi),%r14
+	movq	16(%rsi),%r15
+	movq	24(%rsi),%r8
+	leaq	.Lord(%rip),%rsi
+	jmp	.Loop_ord_sqrx
+
+.align	32
+.Loop_ord_sqrx:
+	mulxq	%r14,%r9,%r10  // a[0]*a[1]
+	mulxq	%r15,%rcx,%r11  // a[0]*a[2]
+	movq	%rdx,%rax  // offload a[0]
+	movq	%r14,%xmm1  // offload a[1]
+	mulxq	%r8,%rbp,%r12  // a[0]*a[3]
+	movq	%r14,%rdx
+	addq	%rcx,%r10
+	movq	%r15,%xmm2  // offload a[2]
+	adcq	%rbp,%r11
+	adcq	$0,%r12
+	xorq	%r13,%r13  // %r13=0,cf=0,of=0
+// ################################
+	mulxq	%r15,%rcx,%rbp  // a[1]*a[2]
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	%r8,%rcx,%rbp  // a[1]*a[3]
+	movq	%r15,%rdx
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+	adcq	$0,%r13
+// ################################
+	mulxq	%r8,%rcx,%r14  // a[2]*a[3]
+	movq	%rax,%rdx
+	movq	%r8,%xmm3  // offload a[3]
+	xorq	%r15,%r15  // %r15=0,cf=0,of=0
+	adcxq	%r9,%r9  // acc1:6<<1
+	adoxq	%rcx,%r13
+	adcxq	%r10,%r10
+	adoxq	%r15,%r14  // of=0
+
+// ################################ a[i]*a[i]
+	mulxq	%rdx,%r8,%rbp
+	movq	%xmm1,%rdx
+	adcxq	%r11,%r11
+	adoxq	%rbp,%r9
+	adcxq	%r12,%r12
+	mulxq	%rdx,%rcx,%rax
+	movq	%xmm2,%rdx
+	adcxq	%r13,%r13
+	adoxq	%rcx,%r10
+	adcxq	%r14,%r14
+	mulxq	%rdx,%rcx,%rbp
+.byte	0x67
+	movq	%xmm3,%rdx
+	adoxq	%rax,%r11
+	adcxq	%r15,%r15
+	adoxq	%rcx,%r12
+	adoxq	%rbp,%r13
+	mulxq	%rdx,%rcx,%rax
+	adoxq	%rcx,%r14
+	adoxq	%rax,%r15
+
+// ################################ reduction
+	movq	%r8,%rdx
+	mulxq	32(%rsi),%rdx,%rcx
+
+	xorq	%rax,%rax  // cf=0, of=0
+	mulxq	0(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r8  // guaranteed to be zero
+	adoxq	%rbp,%r9
+	mulxq	8(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r9
+	adoxq	%rbp,%r10
+	mulxq	16(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+	mulxq	24(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r8  // of=0
+	adcxq	%rax,%r8  // cf=0
+
+// ################################
+	movq	%r9,%rdx
+	mulxq	32(%rsi),%rdx,%rcx
+
+	mulxq	0(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r9  // guaranteed to be zero
+	adcxq	%rbp,%r10
+	mulxq	8(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r10
+	adcxq	%rbp,%r11
+	mulxq	16(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r11
+	adcxq	%rbp,%r8
+	mulxq	24(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r8
+	adcxq	%rbp,%r9  // cf=0
+	adoxq	%rax,%r9  // of=0
+
+// ################################
+	movq	%r10,%rdx
+	mulxq	32(%rsi),%rdx,%rcx
+
+	mulxq	0(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10  // guaranteed to be zero
+	adoxq	%rbp,%r11
+	mulxq	8(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r8
+	mulxq	16(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r8
+	adoxq	%rbp,%r9
+	mulxq	24(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r9
+	adoxq	%rbp,%r10  // of=0
+	adcxq	%rax,%r10  // cf=0
+
+// ################################
+	movq	%r11,%rdx
+	mulxq	32(%rsi),%rdx,%rcx
+
+	mulxq	0(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r11  // guaranteed to be zero
+	adcxq	%rbp,%r8
+	mulxq	8(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r8
+	adcxq	%rbp,%r9
+	mulxq	16(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r9
+	adcxq	%rbp,%r10
+	mulxq	24(%rsi),%rcx,%rbp
+	adoxq	%rcx,%r10
+	adcxq	%rbp,%r11
+	adoxq	%rax,%r11
+
+// ################################ accumulate upper half
+	addq	%r8,%r12  // add	%r12, %r8
+	adcq	%r13,%r9
+	movq	%r12,%rdx
+	adcq	%r14,%r10
+	adcq	%r15,%r11
+	movq	%r9,%r14
+	adcq	$0,%rax
+
+// ################################ compare to modulus
+	subq	0(%rsi),%r12
+	movq	%r10,%r15
+	sbbq	8(%rsi),%r9
+	sbbq	16(%rsi),%r10
+	movq	%r11,%r8
+	sbbq	24(%rsi),%r11
+	sbbq	$0,%rax
+
+	cmovncq	%r12,%rdx
+	cmovncq	%r9,%r14
+	cmovncq	%r10,%r15
+	cmovncq	%r11,%r8
+
+	decq	%rbx
+	jnz	.Loop_ord_sqrx
+
+	movq	%rdx,0(%rdi)
+	movq	%r14,8(%rdi)
+	pxor	%xmm1,%xmm1
+	movq	%r15,16(%rdi)
+	pxor	%xmm2,%xmm2
+	movq	%r8,24(%rdi)
+	pxor	%xmm3,%xmm3
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lord_sqrx_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_ord_sqr_mont_adx,.-ecp_nistz256_ord_sqr_mont_adx
+// ############################################################################## 
+// void ecp_nistz256_mul_mont(
+// uint64_t res[4],
+// uint64_t a[4],
+// uint64_t b[4]);
+
+.globl	ecp_nistz256_mul_mont_nohw
+.hidden ecp_nistz256_mul_mont_nohw
+.type	ecp_nistz256_mul_mont_nohw,@function
+.align	32
+ecp_nistz256_mul_mont_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lmul_body:
+	movq	%rdx,%rbx
+	movq	0(%rdx),%rax
+	movq	0(%rsi),%r9
+	movq	8(%rsi),%r10
+	movq	16(%rsi),%r11
+	movq	24(%rsi),%r12
+
+	call	__ecp_nistz256_mul_montq
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lmul_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_mul_mont_nohw,.-ecp_nistz256_mul_mont_nohw
+
+.type	__ecp_nistz256_mul_montq,@function
+.align	32
+__ecp_nistz256_mul_montq:
+.cfi_startproc	
+// #######################################################################
+// Multiply a by b[0]
+	movq	%rax,%rbp
+	mulq	%r9
+	movq	.Lpoly+8(%rip),%r14
+	movq	%rax,%r8
+	movq	%rbp,%rax
+	movq	%rdx,%r9
+
+	mulq	%r10
+	movq	.Lpoly+24(%rip),%r15
+	addq	%rax,%r9
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%r11
+	addq	%rax,%r10
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%r12
+	addq	%rax,%r11
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	xorq	%r13,%r13
+	movq	%rdx,%r12
+
+// #######################################################################
+// First reduction step
+// Basically now we want to multiply acc[0] by p256,
+// and add the result to the acc.
+// Due to the special form of p256 we do some optimizations
+// 
+// acc[0] x p256[0..1] = acc[0] x 2^96 - acc[0]
+// then we add acc[0] and get acc[0] x 2^96
+
+	movq	%r8,%rbp
+	shlq	$32,%r8
+	mulq	%r15
+	shrq	$32,%rbp
+	addq	%r8,%r9  // +=acc[0]<<96
+	adcq	%rbp,%r10
+	adcq	%rax,%r11
+	movq	8(%rbx),%rax
+	adcq	%rdx,%r12
+	adcq	$0,%r13
+	xorq	%r8,%r8
+
+// #######################################################################
+// Multiply by b[1]
+	movq	%rax,%rbp
+	mulq	0(%rsi)
+	addq	%rax,%r9
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	8(%rsi)
+	addq	%rcx,%r10
+	adcq	$0,%rdx
+	addq	%rax,%r10
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	16(%rsi)
+	addq	%rcx,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	24(%rsi)
+	addq	%rcx,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%r9,%rax
+	adcq	%rdx,%r13
+	adcq	$0,%r8
+
+// #######################################################################
+// Second reduction step
+	movq	%r9,%rbp
+	shlq	$32,%r9
+	mulq	%r15
+	shrq	$32,%rbp
+	addq	%r9,%r10
+	adcq	%rbp,%r11
+	adcq	%rax,%r12
+	movq	16(%rbx),%rax
+	adcq	%rdx,%r13
+	adcq	$0,%r8
+	xorq	%r9,%r9
+
+// #######################################################################
+// Multiply by b[2]
+	movq	%rax,%rbp
+	mulq	0(%rsi)
+	addq	%rax,%r10
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	8(%rsi)
+	addq	%rcx,%r11
+	adcq	$0,%rdx
+	addq	%rax,%r11
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	16(%rsi)
+	addq	%rcx,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	24(%rsi)
+	addq	%rcx,%r13
+	adcq	$0,%rdx
+	addq	%rax,%r13
+	movq	%r10,%rax
+	adcq	%rdx,%r8
+	adcq	$0,%r9
+
+// #######################################################################
+// Third reduction step
+	movq	%r10,%rbp
+	shlq	$32,%r10
+	mulq	%r15
+	shrq	$32,%rbp
+	addq	%r10,%r11
+	adcq	%rbp,%r12
+	adcq	%rax,%r13
+	movq	24(%rbx),%rax
+	adcq	%rdx,%r8
+	adcq	$0,%r9
+	xorq	%r10,%r10
+
+// #######################################################################
+// Multiply by b[3]
+	movq	%rax,%rbp
+	mulq	0(%rsi)
+	addq	%rax,%r11
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	8(%rsi)
+	addq	%rcx,%r12
+	adcq	$0,%rdx
+	addq	%rax,%r12
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	16(%rsi)
+	addq	%rcx,%r13
+	adcq	$0,%rdx
+	addq	%rax,%r13
+	movq	%rbp,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	24(%rsi)
+	addq	%rcx,%r8
+	adcq	$0,%rdx
+	addq	%rax,%r8
+	movq	%r11,%rax
+	adcq	%rdx,%r9
+	adcq	$0,%r10
+
+// #######################################################################
+// Final reduction step
+	movq	%r11,%rbp
+	shlq	$32,%r11
+	mulq	%r15
+	shrq	$32,%rbp
+	addq	%r11,%r12
+	adcq	%rbp,%r13
+	movq	%r12,%rcx
+	adcq	%rax,%r8
+	adcq	%rdx,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r10
+
+// #######################################################################
+// Branch-less conditional subtraction of P
+	subq	$-1,%r12  // .Lpoly[0]
+	movq	%r8,%rbx
+	sbbq	%r14,%r13  // .Lpoly[1]
+	sbbq	$0,%r8  // .Lpoly[2]
+	movq	%r9,%rdx
+	sbbq	%r15,%r9  // .Lpoly[3]
+	sbbq	$0,%r10
+
+	cmovcq	%rcx,%r12
+	cmovcq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rbx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%rdx,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_mul_montq,.-__ecp_nistz256_mul_montq
+
+// ############################################################################## 
+// void ecp_nistz256_sqr_mont(
+// uint64_t res[4],
+// uint64_t a[4]);
+
+// we optimize the square according to S.Gueron and V.Krasnov,
+// "Speeding up Big-Number Squaring"
+.globl	ecp_nistz256_sqr_mont_nohw
+.hidden ecp_nistz256_sqr_mont_nohw
+.type	ecp_nistz256_sqr_mont_nohw,@function
+.align	32
+ecp_nistz256_sqr_mont_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lsqr_body:
+	movq	0(%rsi),%rax
+	movq	8(%rsi),%r14
+	movq	16(%rsi),%r15
+	movq	24(%rsi),%r8
+
+	call	__ecp_nistz256_sqr_montq
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lsqr_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_sqr_mont_nohw,.-ecp_nistz256_sqr_mont_nohw
+
+.type	__ecp_nistz256_sqr_montq,@function
+.align	32
+__ecp_nistz256_sqr_montq:
+.cfi_startproc	
+	movq	%rax,%r13
+	mulq	%r14  // a[1]*a[0]
+	movq	%rax,%r9
+	movq	%r15,%rax
+	movq	%rdx,%r10
+
+	mulq	%r13  // a[0]*a[2]
+	addq	%rax,%r10
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%r13  // a[0]*a[3]
+	addq	%rax,%r11
+	movq	%r15,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r12
+
+// ################################
+	mulq	%r14  // a[1]*a[2]
+	addq	%rax,%r11
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rbp
+
+	mulq	%r14  // a[1]*a[3]
+	addq	%rax,%r12
+	movq	%r8,%rax
+	adcq	$0,%rdx
+	addq	%rbp,%r12
+	movq	%rdx,%r13
+	adcq	$0,%r13
+
+// ################################
+	mulq	%r15  // a[2]*a[3]
+	xorq	%r15,%r15
+	addq	%rax,%r13
+	movq	0(%rsi),%rax
+	movq	%rdx,%r14
+	adcq	$0,%r14
+
+	addq	%r9,%r9  // acc1:6<<1
+	adcq	%r10,%r10
+	adcq	%r11,%r11
+	adcq	%r12,%r12
+	adcq	%r13,%r13
+	adcq	%r14,%r14
+	adcq	$0,%r15
+
+	mulq	%rax
+	movq	%rax,%r8
+	movq	8(%rsi),%rax
+	movq	%rdx,%rcx
+
+	mulq	%rax
+	addq	%rcx,%r9
+	adcq	%rax,%r10
+	movq	16(%rsi),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	%rax
+	addq	%rcx,%r11
+	adcq	%rax,%r12
+	movq	24(%rsi),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rcx
+
+	mulq	%rax
+	addq	%rcx,%r13
+	adcq	%rax,%r14
+	movq	%r8,%rax
+	adcq	%rdx,%r15
+
+	movq	.Lpoly+8(%rip),%rsi
+	movq	.Lpoly+24(%rip),%rbp
+
+// #########################################
+// Now the reduction
+// First iteration
+	movq	%r8,%rcx
+	shlq	$32,%r8
+	mulq	%rbp
+	shrq	$32,%rcx
+	addq	%r8,%r9  // +=acc[0]<<96
+	adcq	%rcx,%r10
+	adcq	%rax,%r11
+	movq	%r9,%rax
+	adcq	$0,%rdx
+
+// #########################################
+// Second iteration
+	movq	%r9,%rcx
+	shlq	$32,%r9
+	movq	%rdx,%r8
+	mulq	%rbp
+	shrq	$32,%rcx
+	addq	%r9,%r10
+	adcq	%rcx,%r11
+	adcq	%rax,%r8
+	movq	%r10,%rax
+	adcq	$0,%rdx
+
+// #########################################
+// Third iteration
+	movq	%r10,%rcx
+	shlq	$32,%r10
+	movq	%rdx,%r9
+	mulq	%rbp
+	shrq	$32,%rcx
+	addq	%r10,%r11
+	adcq	%rcx,%r8
+	adcq	%rax,%r9
+	movq	%r11,%rax
+	adcq	$0,%rdx
+
+// ##########################################
+// Last iteration
+	movq	%r11,%rcx
+	shlq	$32,%r11
+	movq	%rdx,%r10
+	mulq	%rbp
+	shrq	$32,%rcx
+	addq	%r11,%r8
+	adcq	%rcx,%r9
+	adcq	%rax,%r10
+	adcq	$0,%rdx
+	xorq	%r11,%r11
+
+// ###########################################
+// Add the rest of the acc
+	addq	%r8,%r12
+	adcq	%r9,%r13
+	movq	%r12,%r8
+	adcq	%r10,%r14
+	adcq	%rdx,%r15
+	movq	%r13,%r9
+	adcq	$0,%r11
+
+	subq	$-1,%r12  // .Lpoly[0]
+	movq	%r14,%r10
+	sbbq	%rsi,%r13  // .Lpoly[1]
+	sbbq	$0,%r14  // .Lpoly[2]
+	movq	%r15,%rcx
+	sbbq	%rbp,%r15  // .Lpoly[3]
+	sbbq	$0,%r11
+
+	cmovcq	%r8,%r12
+	cmovcq	%r9,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%r10,%r14
+	movq	%r13,8(%rdi)
+	cmovcq	%rcx,%r15
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_sqr_montq,.-__ecp_nistz256_sqr_montq
+.globl	ecp_nistz256_mul_mont_adx
+.hidden ecp_nistz256_mul_mont_adx
+.type	ecp_nistz256_mul_mont_adx,@function
+.align	32
+ecp_nistz256_mul_mont_adx:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lmulx_body:
+	movq	%rdx,%rbx
+	movq	0(%rdx),%rdx
+	movq	0(%rsi),%r9
+	movq	8(%rsi),%r10
+	movq	16(%rsi),%r11
+	movq	24(%rsi),%r12
+	leaq	-128(%rsi),%rsi  // control u-op density
+
+	call	__ecp_nistz256_mul_montx
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lmulx_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_mul_mont_adx,.-ecp_nistz256_mul_mont_adx
+
+.type	__ecp_nistz256_mul_montx,@function
+.align	32
+__ecp_nistz256_mul_montx:
+.cfi_startproc	
+// #######################################################################
+// Multiply by b[0]
+	mulxq	%r9,%r8,%r9
+	mulxq	%r10,%rcx,%r10
+	movq	$32,%r14
+	xorq	%r13,%r13  // cf=0
+	mulxq	%r11,%rbp,%r11
+	movq	.Lpoly+24(%rip),%r15
+	adcq	%rcx,%r9
+	mulxq	%r12,%rcx,%r12
+	movq	%r8,%rdx
+	adcq	%rbp,%r10
+	shlxq	%r14,%r8,%rbp
+	adcq	%rcx,%r11
+	shrxq	%r14,%r8,%rcx
+	adcq	$0,%r12
+
+// #######################################################################
+// First reduction step
+	addq	%rbp,%r9
+	adcq	%rcx,%r10
+
+	mulxq	%r15,%rcx,%rbp
+	movq	8(%rbx),%rdx
+	adcq	%rcx,%r11
+	adcq	%rbp,%r12
+	adcq	$0,%r13
+	xorq	%r8,%r8  // %r8=0,cf=0,of=0
+
+// #######################################################################
+// Multiply by b[1]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r9
+	adoxq	%rbp,%r10
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r9,%rdx
+	adcxq	%rcx,%r12
+	shlxq	%r14,%r9,%rcx
+	adoxq	%rbp,%r13
+	shrxq	%r14,%r9,%rbp
+
+	adcxq	%r8,%r13
+	adoxq	%r8,%r8
+	adcq	$0,%r8
+
+// #######################################################################
+// Second reduction step
+	addq	%rcx,%r10
+	adcq	%rbp,%r11
+
+	mulxq	%r15,%rcx,%rbp
+	movq	16(%rbx),%rdx
+	adcq	%rcx,%r12
+	adcq	%rbp,%r13
+	adcq	$0,%r8
+	xorq	%r9,%r9  // %r9=0,cf=0,of=0
+
+// #######################################################################
+// Multiply by b[2]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r10
+	adoxq	%rbp,%r11
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r10,%rdx
+	adcxq	%rcx,%r13
+	shlxq	%r14,%r10,%rcx
+	adoxq	%rbp,%r8
+	shrxq	%r14,%r10,%rbp
+
+	adcxq	%r9,%r8
+	adoxq	%r9,%r9
+	adcq	$0,%r9
+
+// #######################################################################
+// Third reduction step
+	addq	%rcx,%r11
+	adcq	%rbp,%r12
+
+	mulxq	%r15,%rcx,%rbp
+	movq	24(%rbx),%rdx
+	adcq	%rcx,%r13
+	adcq	%rbp,%r8
+	adcq	$0,%r9
+	xorq	%r10,%r10  // %r10=0,cf=0,of=0
+
+// #######################################################################
+// Multiply by b[3]
+	mulxq	0+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	8+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+
+	mulxq	16+128(%rsi),%rcx,%rbp
+	adcxq	%rcx,%r13
+	adoxq	%rbp,%r8
+
+	mulxq	24+128(%rsi),%rcx,%rbp
+	movq	%r11,%rdx
+	adcxq	%rcx,%r8
+	shlxq	%r14,%r11,%rcx
+	adoxq	%rbp,%r9
+	shrxq	%r14,%r11,%rbp
+
+	adcxq	%r10,%r9
+	adoxq	%r10,%r10
+	adcq	$0,%r10
+
+// #######################################################################
+// Fourth reduction step
+	addq	%rcx,%r12
+	adcq	%rbp,%r13
+
+	mulxq	%r15,%rcx,%rbp
+	movq	%r12,%rbx
+	movq	.Lpoly+8(%rip),%r14
+	adcq	%rcx,%r8
+	movq	%r13,%rdx
+	adcq	%rbp,%r9
+	adcq	$0,%r10
+
+// #######################################################################
+// Branch-less conditional subtraction of P
+	xorl	%eax,%eax
+	movq	%r8,%rcx
+	sbbq	$-1,%r12  // .Lpoly[0]
+	sbbq	%r14,%r13  // .Lpoly[1]
+	sbbq	$0,%r8  // .Lpoly[2]
+	movq	%r9,%rbp
+	sbbq	%r15,%r9  // .Lpoly[3]
+	sbbq	$0,%r10
+
+	cmovcq	%rbx,%r12
+	cmovcq	%rdx,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%rbp,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_mul_montx,.-__ecp_nistz256_mul_montx
+
+.globl	ecp_nistz256_sqr_mont_adx
+.hidden ecp_nistz256_sqr_mont_adx
+.type	ecp_nistz256_sqr_mont_adx,@function
+.align	32
+ecp_nistz256_sqr_mont_adx:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+.Lsqrx_body:
+	movq	0(%rsi),%rdx
+	movq	8(%rsi),%r14
+	movq	16(%rsi),%r15
+	movq	24(%rsi),%r8
+	leaq	-128(%rsi),%rsi  // control u-op density
+
+	call	__ecp_nistz256_sqr_montx
+
+	movq	0(%rsp),%r15
+.cfi_restore	%r15
+	movq	8(%rsp),%r14
+.cfi_restore	%r14
+	movq	16(%rsp),%r13
+.cfi_restore	%r13
+	movq	24(%rsp),%r12
+.cfi_restore	%r12
+	movq	32(%rsp),%rbx
+.cfi_restore	%rbx
+	movq	40(%rsp),%rbp
+.cfi_restore	%rbp
+	leaq	48(%rsp),%rsp
+.cfi_adjust_cfa_offset	-48
+.Lsqrx_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_sqr_mont_adx,.-ecp_nistz256_sqr_mont_adx
+
+.type	__ecp_nistz256_sqr_montx,@function
+.align	32
+__ecp_nistz256_sqr_montx:
+.cfi_startproc	
+	mulxq	%r14,%r9,%r10  // a[0]*a[1]
+	mulxq	%r15,%rcx,%r11  // a[0]*a[2]
+	xorl	%eax,%eax
+	adcq	%rcx,%r10
+	mulxq	%r8,%rbp,%r12  // a[0]*a[3]
+	movq	%r14,%rdx
+	adcq	%rbp,%r11
+	adcq	$0,%r12
+	xorq	%r13,%r13  // %r13=0,cf=0,of=0
+
+// ################################
+	mulxq	%r15,%rcx,%rbp  // a[1]*a[2]
+	adcxq	%rcx,%r11
+	adoxq	%rbp,%r12
+
+	mulxq	%r8,%rcx,%rbp  // a[1]*a[3]
+	movq	%r15,%rdx
+	adcxq	%rcx,%r12
+	adoxq	%rbp,%r13
+	adcq	$0,%r13
+
+// ################################
+	mulxq	%r8,%rcx,%r14  // a[2]*a[3]
+	movq	0+128(%rsi),%rdx
+	xorq	%r15,%r15  // %r15=0,cf=0,of=0
+	adcxq	%r9,%r9  // acc1:6<<1
+	adoxq	%rcx,%r13
+	adcxq	%r10,%r10
+	adoxq	%r15,%r14  // of=0
+
+	mulxq	%rdx,%r8,%rbp
+	movq	8+128(%rsi),%rdx
+	adcxq	%r11,%r11
+	adoxq	%rbp,%r9
+	adcxq	%r12,%r12
+	mulxq	%rdx,%rcx,%rax
+	movq	16+128(%rsi),%rdx
+	adcxq	%r13,%r13
+	adoxq	%rcx,%r10
+	adcxq	%r14,%r14
+.byte	0x67
+	mulxq	%rdx,%rcx,%rbp
+	movq	24+128(%rsi),%rdx
+	adoxq	%rax,%r11
+	adcxq	%r15,%r15
+	adoxq	%rcx,%r12
+	movq	$32,%rsi
+	adoxq	%rbp,%r13
+.byte	0x67,0x67
+	mulxq	%rdx,%rcx,%rax
+	movq	.Lpoly+24(%rip),%rdx
+	adoxq	%rcx,%r14
+	shlxq	%rsi,%r8,%rcx
+	adoxq	%rax,%r15
+	shrxq	%rsi,%r8,%rax
+	movq	%rdx,%rbp
+
+// reduction step 1
+	addq	%rcx,%r9
+	adcq	%rax,%r10
+
+	mulxq	%r8,%rcx,%r8
+	adcq	%rcx,%r11
+	shlxq	%rsi,%r9,%rcx
+	adcq	$0,%r8
+	shrxq	%rsi,%r9,%rax
+
+// reduction step 2
+	addq	%rcx,%r10
+	adcq	%rax,%r11
+
+	mulxq	%r9,%rcx,%r9
+	adcq	%rcx,%r8
+	shlxq	%rsi,%r10,%rcx
+	adcq	$0,%r9
+	shrxq	%rsi,%r10,%rax
+
+// reduction step 3
+	addq	%rcx,%r11
+	adcq	%rax,%r8
+
+	mulxq	%r10,%rcx,%r10
+	adcq	%rcx,%r9
+	shlxq	%rsi,%r11,%rcx
+	adcq	$0,%r10
+	shrxq	%rsi,%r11,%rax
+
+// reduction step 4
+	addq	%rcx,%r8
+	adcq	%rax,%r9
+
+	mulxq	%r11,%rcx,%r11
+	adcq	%rcx,%r10
+	adcq	$0,%r11
+
+	xorq	%rdx,%rdx
+	addq	%r8,%r12  // accumulate upper half
+	movq	.Lpoly+8(%rip),%rsi
+	adcq	%r9,%r13
+	movq	%r12,%r8
+	adcq	%r10,%r14
+	adcq	%r11,%r15
+	movq	%r13,%r9
+	adcq	$0,%rdx
+
+	subq	$-1,%r12  // .Lpoly[0]
+	movq	%r14,%r10
+	sbbq	%rsi,%r13  // .Lpoly[1]
+	sbbq	$0,%r14  // .Lpoly[2]
+	movq	%r15,%r11
+	sbbq	%rbp,%r15  // .Lpoly[3]
+	sbbq	$0,%rdx
+
+	cmovcq	%r8,%r12
+	cmovcq	%r9,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%r10,%r14
+	movq	%r13,8(%rdi)
+	cmovcq	%r11,%r15
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_sqr_montx,.-__ecp_nistz256_sqr_montx
+// ############################################################################## 
+// void ecp_nistz256_select_w5_nohw(uint64_t *val, uint64_t *in_t, int index);
+.globl	ecp_nistz256_select_w5_nohw
+.hidden ecp_nistz256_select_w5_nohw
+.type	ecp_nistz256_select_w5_nohw,@function
+.align	32
+ecp_nistz256_select_w5_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	movdqa	.LOne(%rip),%xmm0
+	movd	%edx,%xmm1
+
+	pxor	%xmm2,%xmm2
+	pxor	%xmm3,%xmm3
+	pxor	%xmm4,%xmm4
+	pxor	%xmm5,%xmm5
+	pxor	%xmm6,%xmm6
+	pxor	%xmm7,%xmm7
+
+	movdqa	%xmm0,%xmm8
+	pshufd	$0,%xmm1,%xmm1
+
+	movq	$16,%rax
+.Lselect_loop_sse_w5:
+
+	movdqa	%xmm8,%xmm15
+	paddd	%xmm0,%xmm8
+	pcmpeqd	%xmm1,%xmm15
+
+	movdqa	0(%rsi),%xmm9
+	movdqa	16(%rsi),%xmm10
+	movdqa	32(%rsi),%xmm11
+	movdqa	48(%rsi),%xmm12
+	movdqa	64(%rsi),%xmm13
+	movdqa	80(%rsi),%xmm14
+	leaq	96(%rsi),%rsi
+
+	pand	%xmm15,%xmm9
+	pand	%xmm15,%xmm10
+	por	%xmm9,%xmm2
+	pand	%xmm15,%xmm11
+	por	%xmm10,%xmm3
+	pand	%xmm15,%xmm12
+	por	%xmm11,%xmm4
+	pand	%xmm15,%xmm13
+	por	%xmm12,%xmm5
+	pand	%xmm15,%xmm14
+	por	%xmm13,%xmm6
+	por	%xmm14,%xmm7
+
+	decq	%rax
+	jnz	.Lselect_loop_sse_w5
+
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+	movdqu	%xmm4,32(%rdi)
+	movdqu	%xmm5,48(%rdi)
+	movdqu	%xmm6,64(%rdi)
+	movdqu	%xmm7,80(%rdi)
+	ret
+.cfi_endproc	
+.LSEH_end_ecp_nistz256_select_w5_nohw:
+.size	ecp_nistz256_select_w5_nohw,.-ecp_nistz256_select_w5_nohw
+
+// ############################################################################## 
+// void ecp_nistz256_select_w7_nohw(uint64_t *val, uint64_t *in_t, int index);
+.globl	ecp_nistz256_select_w7_nohw
+.hidden ecp_nistz256_select_w7_nohw
+.type	ecp_nistz256_select_w7_nohw,@function
+.align	32
+ecp_nistz256_select_w7_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	movdqa	.LOne(%rip),%xmm8
+	movd	%edx,%xmm1
+
+	pxor	%xmm2,%xmm2
+	pxor	%xmm3,%xmm3
+	pxor	%xmm4,%xmm4
+	pxor	%xmm5,%xmm5
+
+	movdqa	%xmm8,%xmm0
+	pshufd	$0,%xmm1,%xmm1
+	movq	$64,%rax
+
+.Lselect_loop_sse_w7:
+	movdqa	%xmm8,%xmm15
+	paddd	%xmm0,%xmm8
+	movdqa	0(%rsi),%xmm9
+	movdqa	16(%rsi),%xmm10
+	pcmpeqd	%xmm1,%xmm15
+	movdqa	32(%rsi),%xmm11
+	movdqa	48(%rsi),%xmm12
+	leaq	64(%rsi),%rsi
+
+	pand	%xmm15,%xmm9
+	pand	%xmm15,%xmm10
+	por	%xmm9,%xmm2
+	pand	%xmm15,%xmm11
+	por	%xmm10,%xmm3
+	pand	%xmm15,%xmm12
+	por	%xmm11,%xmm4
+	prefetcht0	255(%rsi)
+	por	%xmm12,%xmm5
+
+	decq	%rax
+	jnz	.Lselect_loop_sse_w7
+
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+	movdqu	%xmm4,32(%rdi)
+	movdqu	%xmm5,48(%rdi)
+	ret
+.cfi_endproc	
+.LSEH_end_ecp_nistz256_select_w7_nohw:
+.size	ecp_nistz256_select_w7_nohw,.-ecp_nistz256_select_w7_nohw
+// ############################################################################## 
+// void ecp_nistz256_select_w5_avx2(uint64_t *val, uint64_t *in_t, int index);
+.globl	ecp_nistz256_select_w5_avx2
+.hidden ecp_nistz256_select_w5_avx2
+.type	ecp_nistz256_select_w5_avx2,@function
+.align	32
+ecp_nistz256_select_w5_avx2:
+.cfi_startproc	
+_CET_ENDBR
+	vzeroupper
+	vmovdqa	.LTwo(%rip),%ymm0
+
+	vpxor	%ymm2,%ymm2,%ymm2
+	vpxor	%ymm3,%ymm3,%ymm3
+	vpxor	%ymm4,%ymm4,%ymm4
+
+	vmovdqa	.LOne(%rip),%ymm5
+	vmovdqa	.LTwo(%rip),%ymm10
+
+	vmovd	%edx,%xmm1
+	vpermd	%ymm1,%ymm2,%ymm1
+
+	movq	$8,%rax
+.Lselect_loop_avx2_w5:
+
+	vmovdqa	0(%rsi),%ymm6
+	vmovdqa	32(%rsi),%ymm7
+	vmovdqa	64(%rsi),%ymm8
+
+	vmovdqa	96(%rsi),%ymm11
+	vmovdqa	128(%rsi),%ymm12
+	vmovdqa	160(%rsi),%ymm13
+
+	vpcmpeqd	%ymm1,%ymm5,%ymm9
+	vpcmpeqd	%ymm1,%ymm10,%ymm14
+
+	vpaddd	%ymm0,%ymm5,%ymm5
+	vpaddd	%ymm0,%ymm10,%ymm10
+	leaq	192(%rsi),%rsi
+
+	vpand	%ymm9,%ymm6,%ymm6
+	vpand	%ymm9,%ymm7,%ymm7
+	vpand	%ymm9,%ymm8,%ymm8
+	vpand	%ymm14,%ymm11,%ymm11
+	vpand	%ymm14,%ymm12,%ymm12
+	vpand	%ymm14,%ymm13,%ymm13
+
+	vpxor	%ymm6,%ymm2,%ymm2
+	vpxor	%ymm7,%ymm3,%ymm3
+	vpxor	%ymm8,%ymm4,%ymm4
+	vpxor	%ymm11,%ymm2,%ymm2
+	vpxor	%ymm12,%ymm3,%ymm3
+	vpxor	%ymm13,%ymm4,%ymm4
+
+	decq	%rax
+	jnz	.Lselect_loop_avx2_w5
+
+	vmovdqu	%ymm2,0(%rdi)
+	vmovdqu	%ymm3,32(%rdi)
+	vmovdqu	%ymm4,64(%rdi)
+	vzeroupper
+	ret
+.cfi_endproc	
+.LSEH_end_ecp_nistz256_select_w5_avx2:
+.size	ecp_nistz256_select_w5_avx2,.-ecp_nistz256_select_w5_avx2
+
+// ############################################################################## 
+// void ecp_nistz256_select_w7_avx2(uint64_t *val, uint64_t *in_t, int index);
+.globl	ecp_nistz256_select_w7_avx2
+.hidden ecp_nistz256_select_w7_avx2
+.type	ecp_nistz256_select_w7_avx2,@function
+.align	32
+ecp_nistz256_select_w7_avx2:
+.cfi_startproc	
+_CET_ENDBR
+	vzeroupper
+	vmovdqa	.LThree(%rip),%ymm0
+
+	vpxor	%ymm2,%ymm2,%ymm2
+	vpxor	%ymm3,%ymm3,%ymm3
+
+	vmovdqa	.LOne(%rip),%ymm4
+	vmovdqa	.LTwo(%rip),%ymm8
+	vmovdqa	.LThree(%rip),%ymm12
+
+	vmovd	%edx,%xmm1
+	vpermd	%ymm1,%ymm2,%ymm1
+// Skip index = 0, because it is implicitly the point at infinity
+
+	movq	$21,%rax
+.Lselect_loop_avx2_w7:
+
+	vmovdqa	0(%rsi),%ymm5
+	vmovdqa	32(%rsi),%ymm6
+
+	vmovdqa	64(%rsi),%ymm9
+	vmovdqa	96(%rsi),%ymm10
+
+	vmovdqa	128(%rsi),%ymm13
+	vmovdqa	160(%rsi),%ymm14
+
+	vpcmpeqd	%ymm1,%ymm4,%ymm7
+	vpcmpeqd	%ymm1,%ymm8,%ymm11
+	vpcmpeqd	%ymm1,%ymm12,%ymm15
+
+	vpaddd	%ymm0,%ymm4,%ymm4
+	vpaddd	%ymm0,%ymm8,%ymm8
+	vpaddd	%ymm0,%ymm12,%ymm12
+	leaq	192(%rsi),%rsi
+
+	vpand	%ymm7,%ymm5,%ymm5
+	vpand	%ymm7,%ymm6,%ymm6
+	vpand	%ymm11,%ymm9,%ymm9
+	vpand	%ymm11,%ymm10,%ymm10
+	vpand	%ymm15,%ymm13,%ymm13
+	vpand	%ymm15,%ymm14,%ymm14
+
+	vpxor	%ymm5,%ymm2,%ymm2
+	vpxor	%ymm6,%ymm3,%ymm3
+	vpxor	%ymm9,%ymm2,%ymm2
+	vpxor	%ymm10,%ymm3,%ymm3
+	vpxor	%ymm13,%ymm2,%ymm2
+	vpxor	%ymm14,%ymm3,%ymm3
+
+	decq	%rax
+	jnz	.Lselect_loop_avx2_w7
+
+
+	vmovdqa	0(%rsi),%ymm5
+	vmovdqa	32(%rsi),%ymm6
+
+	vpcmpeqd	%ymm1,%ymm4,%ymm7
+
+	vpand	%ymm7,%ymm5,%ymm5
+	vpand	%ymm7,%ymm6,%ymm6
+
+	vpxor	%ymm5,%ymm2,%ymm2
+	vpxor	%ymm6,%ymm3,%ymm3
+
+	vmovdqu	%ymm2,0(%rdi)
+	vmovdqu	%ymm3,32(%rdi)
+	vzeroupper
+	ret
+.cfi_endproc	
+.LSEH_end_ecp_nistz256_select_w7_avx2:
+.size	ecp_nistz256_select_w7_avx2,.-ecp_nistz256_select_w7_avx2
+.type	__ecp_nistz256_add_toq,@function
+.align	32
+__ecp_nistz256_add_toq:
+.cfi_startproc	
+	xorq	%r11,%r11
+	addq	0(%rbx),%r12
+	adcq	8(%rbx),%r13
+	movq	%r12,%rax
+	adcq	16(%rbx),%r8
+	adcq	24(%rbx),%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	cmovcq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_add_toq,.-__ecp_nistz256_add_toq
+
+.type	__ecp_nistz256_sub_fromq,@function
+.align	32
+__ecp_nistz256_sub_fromq:
+.cfi_startproc	
+	subq	0(%rbx),%r12
+	sbbq	8(%rbx),%r13
+	movq	%r12,%rax
+	sbbq	16(%rbx),%r8
+	sbbq	24(%rbx),%r9
+	movq	%r13,%rbp
+	sbbq	%r11,%r11
+
+	addq	$-1,%r12
+	movq	%r8,%rcx
+	adcq	%r14,%r13
+	adcq	$0,%r8
+	movq	%r9,%r10
+	adcq	%r15,%r9
+	testq	%r11,%r11
+
+	cmovzq	%rax,%r12
+	cmovzq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovzq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovzq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_sub_fromq,.-__ecp_nistz256_sub_fromq
+
+.type	__ecp_nistz256_subq,@function
+.align	32
+__ecp_nistz256_subq:
+.cfi_startproc	
+	subq	%r12,%rax
+	sbbq	%r13,%rbp
+	movq	%rax,%r12
+	sbbq	%r8,%rcx
+	sbbq	%r9,%r10
+	movq	%rbp,%r13
+	sbbq	%r11,%r11
+
+	addq	$-1,%rax
+	movq	%rcx,%r8
+	adcq	%r14,%rbp
+	adcq	$0,%rcx
+	movq	%r10,%r9
+	adcq	%r15,%r10
+	testq	%r11,%r11
+
+	cmovnzq	%rax,%r12
+	cmovnzq	%rbp,%r13
+	cmovnzq	%rcx,%r8
+	cmovnzq	%r10,%r9
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_subq,.-__ecp_nistz256_subq
+
+.type	__ecp_nistz256_mul_by_2q,@function
+.align	32
+__ecp_nistz256_mul_by_2q:
+.cfi_startproc	
+	xorq	%r11,%r11
+	addq	%r12,%r12  // a0:a3+a0:a3
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	cmovcq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_mul_by_2q,.-__ecp_nistz256_mul_by_2q
+.globl	ecp_nistz256_point_double_nohw
+.hidden ecp_nistz256_point_double_nohw
+.type	ecp_nistz256_point_double_nohw,@function
+.align	32
+ecp_nistz256_point_double_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$160+8,%rsp
+.cfi_adjust_cfa_offset	32*5+8
+.Lpoint_doubleq_body:
+
+.Lpoint_double_shortcutq:
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi.x
+	movq	%rsi,%rbx  // backup copy
+	movdqu	16(%rsi),%xmm1
+	movq	32+0(%rsi),%r12  // load in_y in "5-4-0-1" order
+	movq	32+8(%rsi),%r13
+	movq	32+16(%rsi),%r8
+	movq	32+24(%rsi),%r9
+	movq	.Lpoly+8(%rip),%r14
+	movq	.Lpoly+24(%rip),%r15
+	movdqa	%xmm0,96(%rsp)
+	movdqa	%xmm1,96+16(%rsp)
+	leaq	32(%rdi),%r10
+	leaq	64(%rdi),%r11
+	movq	%rdi,%xmm0
+	movq	%r10,%xmm1
+	movq	%r11,%xmm2
+
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2q  // p256_mul_by_2(S, in_y);
+
+	movq	64+0(%rsi),%rax
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	leaq	64-0(%rsi),%rsi
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Zsqr, in_z);
+
+	movq	0+0(%rsp),%rax
+	movq	8+0(%rsp),%r14
+	leaq	0+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(S, S);
+
+	movq	32(%rbx),%rax  // %rbx is still valid
+	movq	64+0(%rbx),%r9
+	movq	64+8(%rbx),%r10
+	movq	64+16(%rbx),%r11
+	movq	64+24(%rbx),%r12
+	leaq	64-0(%rbx),%rsi
+	leaq	32(%rbx),%rbx
+	movq	%xmm2,%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(res_z, in_z, in_y);
+	call	__ecp_nistz256_mul_by_2q  // p256_mul_by_2(res_z, res_z);
+
+	movq	96+0(%rsp),%r12  // "5-4-0-1" order
+	movq	96+8(%rsp),%r13
+	leaq	64(%rsp),%rbx
+	movq	96+16(%rsp),%r8
+	movq	96+24(%rsp),%r9
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_add_toq  // p256_add(M, in_x, Zsqr);
+
+	movq	96+0(%rsp),%r12  // "5-4-0-1" order
+	movq	96+8(%rsp),%r13
+	leaq	64(%rsp),%rbx
+	movq	96+16(%rsp),%r8
+	movq	96+24(%rsp),%r9
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(Zsqr, in_x, Zsqr);
+
+	movq	0+0(%rsp),%rax
+	movq	8+0(%rsp),%r14
+	leaq	0+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	movq	%xmm1,%rdi
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(res_y, S);
+	xorq	%r9,%r9
+	movq	%r12,%rax
+	addq	$-1,%r12
+	movq	%r13,%r10
+	adcq	%rsi,%r13
+	movq	%r14,%rcx
+	adcq	$0,%r14
+	movq	%r15,%r8
+	adcq	%rbp,%r15
+	adcq	$0,%r9
+	xorq	%rsi,%rsi  // borrow %rsi
+	testq	$1,%rax
+
+	cmovzq	%rax,%r12
+	cmovzq	%r10,%r13
+	cmovzq	%rcx,%r14
+	cmovzq	%r8,%r15
+	cmovzq	%rsi,%r9
+
+	movq	%r13,%rax  // a0:a3>>1
+	shrq	$1,%r12
+	shlq	$63,%rax
+	movq	%r14,%r10
+	shrq	$1,%r13
+	orq	%rax,%r12
+	shlq	$63,%r10
+	movq	%r15,%rcx
+	shrq	$1,%r14
+	orq	%r10,%r13
+	shlq	$63,%rcx
+	movq	%r12,0(%rdi)
+	shrq	$1,%r15
+	movq	%r13,8(%rdi)
+	shlq	$63,%r9
+	orq	%rcx,%r14
+	orq	%r9,%r15
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+	movq	64(%rsp),%rax
+	leaq	64(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(M, M, Zsqr);
+
+	leaq	128(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2q
+
+	leaq	32(%rsp),%rbx
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_add_toq  // p256_mul_by_3(M, M);
+
+	movq	96(%rsp),%rax
+	leaq	96(%rsp),%rbx
+	movq	0+0(%rsp),%r9
+	movq	8+0(%rsp),%r10
+	leaq	0+0(%rsp),%rsi
+	movq	16+0(%rsp),%r11
+	movq	24+0(%rsp),%r12
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S, S, in_x);
+
+	leaq	128(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2q  // p256_mul_by_2(tmp0, S);
+
+	movq	0+32(%rsp),%rax
+	movq	8+32(%rsp),%r14
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r15
+	movq	24+32(%rsp),%r8
+	movq	%xmm0,%rdi
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(res_x, M);
+
+	leaq	128(%rsp),%rbx
+	movq	%r14,%r8  // harmonize sqr output and sub input
+	movq	%r15,%r9
+	movq	%rsi,%r14
+	movq	%rbp,%r15
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_x, res_x, tmp0);
+
+	movq	0+0(%rsp),%rax
+	movq	0+8(%rsp),%rbp
+	movq	0+16(%rsp),%rcx
+	movq	0+24(%rsp),%r10  // "4-5-0-1" order
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_subq  // p256_sub(S, S, res_x);
+
+	movq	32(%rsp),%rax
+	leaq	32(%rsp),%rbx
+	movq	%r12,%r14  // harmonize sub output and mul input
+	xorl	%ecx,%ecx
+	movq	%r12,0+0(%rsp)  // have to save:-(
+	movq	%r13,%r10
+	movq	%r13,0+8(%rsp)
+	cmovzq	%r8,%r11
+	movq	%r8,0+16(%rsp)
+	leaq	0-0(%rsp),%rsi
+	cmovzq	%r9,%r12
+	movq	%r9,0+24(%rsp)
+	movq	%r14,%r9
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S, S, M);
+
+	movq	%xmm1,%rbx
+	movq	%xmm1,%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_y, S, res_y);
+
+	leaq	160+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lpoint_doubleq_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_double_nohw,.-ecp_nistz256_point_double_nohw
+.globl	ecp_nistz256_point_add_nohw
+.hidden ecp_nistz256_point_add_nohw
+.type	ecp_nistz256_point_add_nohw,@function
+.align	32
+ecp_nistz256_point_add_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$576+8,%rsp
+.cfi_adjust_cfa_offset	32*18+8
+.Lpoint_addq_body:
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	movdqu	48(%rsi),%xmm3
+	movdqu	64(%rsi),%xmm4
+	movdqu	80(%rsi),%xmm5
+	movq	%rsi,%rbx  // reassign
+	movq	%rdx,%rsi  // reassign
+	movdqa	%xmm0,384(%rsp)
+	movdqa	%xmm1,384+16(%rsp)
+	movdqa	%xmm2,416(%rsp)
+	movdqa	%xmm3,416+16(%rsp)
+	movdqa	%xmm4,448(%rsp)
+	movdqa	%xmm5,448+16(%rsp)
+	por	%xmm4,%xmm5
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rbx
+	pshufd	$0xb1,%xmm5,%xmm3
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	por	%xmm3,%xmm5
+	movdqu	48(%rsi),%xmm3
+	movq	64+0(%rsi),%rax  // load original in2_z
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	movdqa	%xmm0,480(%rsp)
+	pshufd	$0x1e,%xmm5,%xmm4
+	movdqa	%xmm1,480+16(%rsp)
+	movdqu	64(%rsi),%xmm0  // in2_z again
+	movdqu	80(%rsi),%xmm1
+	movdqa	%xmm2,512(%rsp)
+	movdqa	%xmm3,512+16(%rsp)
+	por	%xmm4,%xmm5
+	pxor	%xmm4,%xmm4
+	por	%xmm0,%xmm1
+	movq	%rdi,%xmm0  // save %rdi
+
+	leaq	64-0(%rsi),%rsi  // %rsi is still valid
+	movq	%rax,544+0(%rsp)  // make in2_z copy
+	movq	%r14,544+8(%rsp)
+	movq	%r15,544+16(%rsp)
+	movq	%r8,544+24(%rsp)
+	leaq	96(%rsp),%rdi  // Z2^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Z2sqr, in2_z);
+
+	pcmpeqd	%xmm4,%xmm5
+	pshufd	$0xb1,%xmm1,%xmm4
+	por	%xmm1,%xmm4
+	pshufd	$0,%xmm5,%xmm5  // in1infty
+	pshufd	$0x1e,%xmm4,%xmm3
+	por	%xmm3,%xmm4
+	pxor	%xmm3,%xmm3
+	pcmpeqd	%xmm3,%xmm4
+	pshufd	$0,%xmm4,%xmm4  // in2infty
+	movq	64+0(%rbx),%rax  // load original in1_z
+	movq	64+8(%rbx),%r14
+	movq	64+16(%rbx),%r15
+	movq	64+24(%rbx),%r8
+	movq	%rbx,%xmm1
+
+	leaq	64-0(%rbx),%rsi
+	leaq	32(%rsp),%rdi  // Z1^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Z1sqr, in1_z);
+
+	movq	544(%rsp),%rax
+	leaq	544(%rsp),%rbx
+	movq	0+96(%rsp),%r9
+	movq	8+96(%rsp),%r10
+	leaq	0+96(%rsp),%rsi
+	movq	16+96(%rsp),%r11
+	movq	24+96(%rsp),%r12
+	leaq	224(%rsp),%rdi  // S1 = Z2^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S1, Z2sqr, in2_z);
+
+	movq	448(%rsp),%rax
+	leaq	448(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	256(%rsp),%rdi  // S2 = Z1^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, Z1sqr, in1_z);
+
+	movq	416(%rsp),%rax
+	leaq	416(%rsp),%rbx
+	movq	0+224(%rsp),%r9
+	movq	8+224(%rsp),%r10
+	leaq	0+224(%rsp),%rsi
+	movq	16+224(%rsp),%r11
+	movq	24+224(%rsp),%r12
+	leaq	224(%rsp),%rdi  // S1 = Y1*Z2^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S1, S1, in1_y);
+
+	movq	512(%rsp),%rax
+	leaq	512(%rsp),%rbx
+	movq	0+256(%rsp),%r9
+	movq	8+256(%rsp),%r10
+	leaq	0+256(%rsp),%rsi
+	movq	16+256(%rsp),%r11
+	movq	24+256(%rsp),%r12
+	leaq	256(%rsp),%rdi  // S2 = Y2*Z1^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, S2, in2_y);
+
+	leaq	224(%rsp),%rbx
+	leaq	64(%rsp),%rdi  // R = S2 - S1
+	call	__ecp_nistz256_sub_fromq  // p256_sub(R, S2, S1);
+
+	orq	%r13,%r12  // see if result is zero
+	movdqa	%xmm4,%xmm2
+	orq	%r8,%r12
+	orq	%r9,%r12
+	por	%xmm5,%xmm2  // in1infty || in2infty
+	movq	%r12,%xmm3
+
+	movq	384(%rsp),%rax
+	leaq	384(%rsp),%rbx
+	movq	0+96(%rsp),%r9
+	movq	8+96(%rsp),%r10
+	leaq	0+96(%rsp),%rsi
+	movq	16+96(%rsp),%r11
+	movq	24+96(%rsp),%r12
+	leaq	160(%rsp),%rdi  // U1 = X1*Z2^2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(U1, in1_x, Z2sqr);
+
+	movq	480(%rsp),%rax
+	leaq	480(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	192(%rsp),%rdi  // U2 = X2*Z1^2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(U2, in2_x, Z1sqr);
+
+	leaq	160(%rsp),%rbx
+	leaq	0(%rsp),%rdi  // H = U2 - U1
+	call	__ecp_nistz256_sub_fromq  // p256_sub(H, U2, U1);
+
+	orq	%r13,%r12  // see if result is zero
+	orq	%r8,%r12
+	orq	%r9,%r12  // 
+
+	movq	%xmm2,%r8
+	movq	%xmm3,%r9
+	orq	%r8,%r12
+.byte	0x3e  // predict taken
+	jnz	.Ladd_proceedq  // 
+
+// We now know A = B or A = -B and neither is infinity. Compare the
+// y-coordinates via S1 and S2.
+	testq	%r9,%r9
+	jz	.Ladd_doubleq  // is_equal(S1, S2)
+
+// A = -B, so the result is infinity.
+// 
+// TODO(davidben): Does .Ladd_proceed handle this case? It seems to, in
+// which case we should eliminate this special-case and simplify the
+// timing analysis.
+	movq	%xmm0,%rdi  // restore %rdi
+	pxor	%xmm0,%xmm0
+	movdqu	%xmm0,0(%rdi)
+	movdqu	%xmm0,16(%rdi)
+	movdqu	%xmm0,32(%rdi)
+	movdqu	%xmm0,48(%rdi)
+	movdqu	%xmm0,64(%rdi)
+	movdqu	%xmm0,80(%rdi)
+	jmp	.Ladd_doneq
+
+.align	32
+.Ladd_doubleq:
+	movq	%xmm1,%rsi  // restore %rsi
+	movq	%xmm0,%rdi  // restore %rdi
+	addq	$416,%rsp  // difference in frame sizes
+.cfi_adjust_cfa_offset	-416
+	jmp	.Lpoint_double_shortcutq
+.cfi_adjust_cfa_offset	416
+
+.align	32
+.Ladd_proceedq:
+	movq	0+64(%rsp),%rax
+	movq	8+64(%rsp),%r14
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r15
+	movq	24+64(%rsp),%r8
+	leaq	96(%rsp),%rdi  // R^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Rsqr, R);
+
+	movq	448(%rsp),%rax
+	leaq	448(%rsp),%rbx
+	movq	0+0(%rsp),%r9
+	movq	8+0(%rsp),%r10
+	leaq	0+0(%rsp),%rsi
+	movq	16+0(%rsp),%r11
+	movq	24+0(%rsp),%r12
+	leaq	352(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(res_z, H, in1_z);
+
+	movq	0+0(%rsp),%rax
+	movq	8+0(%rsp),%r14
+	leaq	0+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	leaq	32(%rsp),%rdi  // H^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Hsqr, H);
+
+	movq	544(%rsp),%rax
+	leaq	544(%rsp),%rbx
+	movq	0+352(%rsp),%r9
+	movq	8+352(%rsp),%r10
+	leaq	0+352(%rsp),%rsi
+	movq	16+352(%rsp),%r11
+	movq	24+352(%rsp),%r12
+	leaq	352(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(res_z, res_z, in2_z);
+
+	movq	0(%rsp),%rax
+	leaq	0(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	128(%rsp),%rdi  // H^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(Hcub, Hsqr, H);
+
+	movq	160(%rsp),%rax
+	leaq	160(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	192(%rsp),%rdi  // U1*H^2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(U2, U1, Hsqr);
+// lea	192(%rsp), %rsi
+// lea	32(%rsp), %rdi	# 2*U1*H^2
+// call	__ecp_nistz256_mul_by_2	# ecp_nistz256_mul_by_2(Hsqr, U2);
+
+	xorq	%r11,%r11
+	addq	%r12,%r12  // a0:a3+a0:a3
+	leaq	96(%rsp),%rsi
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	movq	0(%rsi),%rax
+	cmovcq	%rbp,%r13
+	movq	8(%rsi),%rbp
+	cmovcq	%rcx,%r8
+	movq	16(%rsi),%rcx
+	cmovcq	%r10,%r9
+	movq	24(%rsi),%r10
+
+	call	__ecp_nistz256_subq  // p256_sub(res_x, Rsqr, Hsqr);
+
+	leaq	128(%rsp),%rbx
+	leaq	288(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_x, res_x, Hcub);
+
+	movq	192+0(%rsp),%rax
+	movq	192+8(%rsp),%rbp
+	movq	192+16(%rsp),%rcx
+	movq	192+24(%rsp),%r10
+	leaq	320(%rsp),%rdi
+
+	call	__ecp_nistz256_subq  // p256_sub(res_y, U2, res_x);
+
+	movq	%r12,0(%rdi)  // save the result, as
+	movq	%r13,8(%rdi)  // __ecp_nistz256_sub doesn't
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+	movq	128(%rsp),%rax
+	leaq	128(%rsp),%rbx
+	movq	0+224(%rsp),%r9
+	movq	8+224(%rsp),%r10
+	leaq	0+224(%rsp),%rsi
+	movq	16+224(%rsp),%r11
+	movq	24+224(%rsp),%r12
+	leaq	256(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, S1, Hcub);
+
+	movq	320(%rsp),%rax
+	leaq	320(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	320(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(res_y, R, res_y);
+
+	leaq	256(%rsp),%rbx
+	leaq	320(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_y, res_y, S2);
+
+	movq	%xmm0,%rdi  // restore %rdi
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_z, in2_z, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	352(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	352+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	544(%rsp),%xmm2
+	pand	544+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_z, in1_z, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	448(%rsp),%xmm2
+	pand	448+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,64(%rdi)
+	movdqu	%xmm3,80(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_x, in2_x, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	288(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	288+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	480(%rsp),%xmm2
+	pand	480+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_x, in1_x, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	384(%rsp),%xmm2
+	pand	384+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_y, in2_y, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	320(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	320+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	512(%rsp),%xmm2
+	pand	512+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_y, in1_y, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	416(%rsp),%xmm2
+	pand	416+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,32(%rdi)
+	movdqu	%xmm3,48(%rdi)
+
+.Ladd_doneq:
+	leaq	576+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lpoint_addq_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_add_nohw,.-ecp_nistz256_point_add_nohw
+.globl	ecp_nistz256_point_add_affine_nohw
+.hidden ecp_nistz256_point_add_affine_nohw
+.type	ecp_nistz256_point_add_affine_nohw,@function
+.align	32
+ecp_nistz256_point_add_affine_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$480+8,%rsp
+.cfi_adjust_cfa_offset	32*15+8
+.Ladd_affineq_body:
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi
+	movq	%rdx,%rbx  // reassign
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	movdqu	48(%rsi),%xmm3
+	movdqu	64(%rsi),%xmm4
+	movdqu	80(%rsi),%xmm5
+	movq	64+0(%rsi),%rax  // load original in1_z
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	movdqa	%xmm0,320(%rsp)
+	movdqa	%xmm1,320+16(%rsp)
+	movdqa	%xmm2,352(%rsp)
+	movdqa	%xmm3,352+16(%rsp)
+	movdqa	%xmm4,384(%rsp)
+	movdqa	%xmm5,384+16(%rsp)
+	por	%xmm4,%xmm5
+
+	movdqu	0(%rbx),%xmm0  // copy	*(P256_POINT_AFFINE *)%rbx
+	pshufd	$0xb1,%xmm5,%xmm3
+	movdqu	16(%rbx),%xmm1
+	movdqu	32(%rbx),%xmm2
+	por	%xmm3,%xmm5
+	movdqu	48(%rbx),%xmm3
+	movdqa	%xmm0,416(%rsp)
+	pshufd	$0x1e,%xmm5,%xmm4
+	movdqa	%xmm1,416+16(%rsp)
+	por	%xmm0,%xmm1
+	movq	%rdi,%xmm0  // save %rdi
+	movdqa	%xmm2,448(%rsp)
+	movdqa	%xmm3,448+16(%rsp)
+	por	%xmm2,%xmm3
+	por	%xmm4,%xmm5
+	pxor	%xmm4,%xmm4
+	por	%xmm1,%xmm3
+
+	leaq	64-0(%rsi),%rsi  // %rsi is still valid
+	leaq	32(%rsp),%rdi  // Z1^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Z1sqr, in1_z);
+
+	pcmpeqd	%xmm4,%xmm5
+	pshufd	$0xb1,%xmm3,%xmm4
+	movq	0(%rbx),%rax  // %rbx is still valid
+// lea	0x00(%rbx), %rbx
+	movq	%r12,%r9  // harmonize sqr output and mul input
+	por	%xmm3,%xmm4
+	pshufd	$0,%xmm5,%xmm5  // in1infty
+	pshufd	$0x1e,%xmm4,%xmm3
+	movq	%r13,%r10
+	por	%xmm3,%xmm4
+	pxor	%xmm3,%xmm3
+	movq	%r14,%r11
+	pcmpeqd	%xmm3,%xmm4
+	pshufd	$0,%xmm4,%xmm4  // in2infty
+
+	leaq	32-0(%rsp),%rsi
+	movq	%r15,%r12
+	leaq	0(%rsp),%rdi  // U2 = X2*Z1^2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(U2, Z1sqr, in2_x);
+
+	leaq	320(%rsp),%rbx
+	leaq	64(%rsp),%rdi  // H = U2 - U1
+	call	__ecp_nistz256_sub_fromq  // p256_sub(H, U2, in1_x);
+
+	movq	384(%rsp),%rax
+	leaq	384(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi  // S2 = Z1^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, Z1sqr, in1_z);
+
+	movq	384(%rsp),%rax
+	leaq	384(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	288(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(res_z, H, in1_z);
+
+	movq	448(%rsp),%rax
+	leaq	448(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	0+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi  // S2 = Y2*Z1^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, S2, in2_y);
+
+	leaq	352(%rsp),%rbx
+	leaq	96(%rsp),%rdi  // R = S2 - S1
+	call	__ecp_nistz256_sub_fromq  // p256_sub(R, S2, in1_y);
+
+	movq	0+64(%rsp),%rax
+	movq	8+64(%rsp),%r14
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r15
+	movq	24+64(%rsp),%r8
+	leaq	128(%rsp),%rdi  // H^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Hsqr, H);
+
+	movq	0+96(%rsp),%rax
+	movq	8+96(%rsp),%r14
+	leaq	0+96(%rsp),%rsi
+	movq	16+96(%rsp),%r15
+	movq	24+96(%rsp),%r8
+	leaq	192(%rsp),%rdi  // R^2
+	call	__ecp_nistz256_sqr_montq  // p256_sqr_mont(Rsqr, R);
+
+	movq	128(%rsp),%rax
+	leaq	128(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	160(%rsp),%rdi  // H^3
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(Hcub, Hsqr, H);
+
+	movq	320(%rsp),%rax
+	leaq	320(%rsp),%rbx
+	movq	0+128(%rsp),%r9
+	movq	8+128(%rsp),%r10
+	leaq	0+128(%rsp),%rsi
+	movq	16+128(%rsp),%r11
+	movq	24+128(%rsp),%r12
+	leaq	0(%rsp),%rdi  // U1*H^2
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(U2, in1_x, Hsqr);
+// lea	0(%rsp), %rsi
+// lea	128(%rsp), %rdi	# 2*U1*H^2
+// call	__ecp_nistz256_mul_by_2	# ecp_nistz256_mul_by_2(Hsqr, U2);
+
+	xorq	%r11,%r11
+	addq	%r12,%r12  // a0:a3+a0:a3
+	leaq	192(%rsp),%rsi
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	movq	0(%rsi),%rax
+	cmovcq	%rbp,%r13
+	movq	8(%rsi),%rbp
+	cmovcq	%rcx,%r8
+	movq	16(%rsi),%rcx
+	cmovcq	%r10,%r9
+	movq	24(%rsi),%r10
+
+	call	__ecp_nistz256_subq  // p256_sub(res_x, Rsqr, Hsqr);
+
+	leaq	160(%rsp),%rbx
+	leaq	224(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_x, res_x, Hcub);
+
+	movq	0+0(%rsp),%rax
+	movq	0+8(%rsp),%rbp
+	movq	0+16(%rsp),%rcx
+	movq	0+24(%rsp),%r10
+	leaq	64(%rsp),%rdi
+
+	call	__ecp_nistz256_subq  // p256_sub(H, U2, res_x);
+
+	movq	%r12,0(%rdi)  // save the result, as
+	movq	%r13,8(%rdi)  // __ecp_nistz256_sub doesn't
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+	movq	352(%rsp),%rax
+	leaq	352(%rsp),%rbx
+	movq	0+160(%rsp),%r9
+	movq	8+160(%rsp),%r10
+	leaq	0+160(%rsp),%rsi
+	movq	16+160(%rsp),%r11
+	movq	24+160(%rsp),%r12
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(S2, Hcub, in1_y);
+
+	movq	96(%rsp),%rax
+	leaq	96(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	0+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_mul_montq  // p256_mul_mont(H, H, R);
+
+	leaq	32(%rsp),%rbx
+	leaq	256(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromq  // p256_sub(res_y, H, S2);
+
+	movq	%xmm0,%rdi  // restore %rdi
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_z, ONE, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	288(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	288+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	.LONE_mont(%rip),%xmm2
+	pand	.LONE_mont+16(%rip),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_z, in1_z, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	384(%rsp),%xmm2
+	pand	384+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,64(%rdi)
+	movdqu	%xmm3,80(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_x, in2_x, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	224(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	224+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	416(%rsp),%xmm2
+	pand	416+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_x, in1_x, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	320(%rsp),%xmm2
+	pand	320+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_y, in2_y, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	256(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	256+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	448(%rsp),%xmm2
+	pand	448+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_y, in1_y, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	352(%rsp),%xmm2
+	pand	352+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,32(%rdi)
+	movdqu	%xmm3,48(%rdi)
+
+	leaq	480+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Ladd_affineq_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_add_affine_nohw,.-ecp_nistz256_point_add_affine_nohw
+.type	__ecp_nistz256_add_tox,@function
+.align	32
+__ecp_nistz256_add_tox:
+.cfi_startproc	
+	xorq	%r11,%r11
+	adcq	0(%rbx),%r12
+	adcq	8(%rbx),%r13
+	movq	%r12,%rax
+	adcq	16(%rbx),%r8
+	adcq	24(%rbx),%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	xorq	%r10,%r10
+	sbbq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	cmovcq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_add_tox,.-__ecp_nistz256_add_tox
+
+.type	__ecp_nistz256_sub_fromx,@function
+.align	32
+__ecp_nistz256_sub_fromx:
+.cfi_startproc	
+	xorq	%r11,%r11
+	sbbq	0(%rbx),%r12
+	sbbq	8(%rbx),%r13
+	movq	%r12,%rax
+	sbbq	16(%rbx),%r8
+	sbbq	24(%rbx),%r9
+	movq	%r13,%rbp
+	sbbq	$0,%r11
+
+	xorq	%r10,%r10
+	adcq	$-1,%r12
+	movq	%r8,%rcx
+	adcq	%r14,%r13
+	adcq	$0,%r8
+	movq	%r9,%r10
+	adcq	%r15,%r9
+
+	btq	$0,%r11
+	cmovncq	%rax,%r12
+	cmovncq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovncq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovncq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_sub_fromx,.-__ecp_nistz256_sub_fromx
+
+.type	__ecp_nistz256_subx,@function
+.align	32
+__ecp_nistz256_subx:
+.cfi_startproc	
+	xorq	%r11,%r11
+	sbbq	%r12,%rax
+	sbbq	%r13,%rbp
+	movq	%rax,%r12
+	sbbq	%r8,%rcx
+	sbbq	%r9,%r10
+	movq	%rbp,%r13
+	sbbq	$0,%r11
+
+	xorq	%r9,%r9
+	adcq	$-1,%rax
+	movq	%rcx,%r8
+	adcq	%r14,%rbp
+	adcq	$0,%rcx
+	movq	%r10,%r9
+	adcq	%r15,%r10
+
+	btq	$0,%r11
+	cmovcq	%rax,%r12
+	cmovcq	%rbp,%r13
+	cmovcq	%rcx,%r8
+	cmovcq	%r10,%r9
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_subx,.-__ecp_nistz256_subx
+
+.type	__ecp_nistz256_mul_by_2x,@function
+.align	32
+__ecp_nistz256_mul_by_2x:
+.cfi_startproc	
+	xorq	%r11,%r11
+	adcq	%r12,%r12  // a0:a3+a0:a3
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	xorq	%r10,%r10
+	sbbq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	cmovcq	%rbp,%r13
+	movq	%r12,0(%rdi)
+	cmovcq	%rcx,%r8
+	movq	%r13,8(%rdi)
+	cmovcq	%r10,%r9
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+
+	ret
+.cfi_endproc	
+.size	__ecp_nistz256_mul_by_2x,.-__ecp_nistz256_mul_by_2x
+.globl	ecp_nistz256_point_double_adx
+.hidden ecp_nistz256_point_double_adx
+.type	ecp_nistz256_point_double_adx,@function
+.align	32
+ecp_nistz256_point_double_adx:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$160+8,%rsp
+.cfi_adjust_cfa_offset	32*5+8
+.Lpoint_doublex_body:
+
+.Lpoint_double_shortcutx:
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi.x
+	movq	%rsi,%rbx  // backup copy
+	movdqu	16(%rsi),%xmm1
+	movq	32+0(%rsi),%r12  // load in_y in "5-4-0-1" order
+	movq	32+8(%rsi),%r13
+	movq	32+16(%rsi),%r8
+	movq	32+24(%rsi),%r9
+	movq	.Lpoly+8(%rip),%r14
+	movq	.Lpoly+24(%rip),%r15
+	movdqa	%xmm0,96(%rsp)
+	movdqa	%xmm1,96+16(%rsp)
+	leaq	32(%rdi),%r10
+	leaq	64(%rdi),%r11
+	movq	%rdi,%xmm0
+	movq	%r10,%xmm1
+	movq	%r11,%xmm2
+
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2x  // p256_mul_by_2(S, in_y);
+
+	movq	64+0(%rsi),%rdx
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	leaq	64-128(%rsi),%rsi
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Zsqr, in_z);
+
+	movq	0+0(%rsp),%rdx
+	movq	8+0(%rsp),%r14
+	leaq	-128+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(S, S);
+
+	movq	32(%rbx),%rdx  // %rbx is still valid
+	movq	64+0(%rbx),%r9
+	movq	64+8(%rbx),%r10
+	movq	64+16(%rbx),%r11
+	movq	64+24(%rbx),%r12
+	leaq	64-128(%rbx),%rsi
+	leaq	32(%rbx),%rbx
+	movq	%xmm2,%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(res_z, in_z, in_y);
+	call	__ecp_nistz256_mul_by_2x  // p256_mul_by_2(res_z, res_z);
+
+	movq	96+0(%rsp),%r12  // "5-4-0-1" order
+	movq	96+8(%rsp),%r13
+	leaq	64(%rsp),%rbx
+	movq	96+16(%rsp),%r8
+	movq	96+24(%rsp),%r9
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_add_tox  // p256_add(M, in_x, Zsqr);
+
+	movq	96+0(%rsp),%r12  // "5-4-0-1" order
+	movq	96+8(%rsp),%r13
+	leaq	64(%rsp),%rbx
+	movq	96+16(%rsp),%r8
+	movq	96+24(%rsp),%r9
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(Zsqr, in_x, Zsqr);
+
+	movq	0+0(%rsp),%rdx
+	movq	8+0(%rsp),%r14
+	leaq	-128+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	movq	%xmm1,%rdi
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(res_y, S);
+	xorq	%r9,%r9
+	movq	%r12,%rax
+	addq	$-1,%r12
+	movq	%r13,%r10
+	adcq	%rsi,%r13
+	movq	%r14,%rcx
+	adcq	$0,%r14
+	movq	%r15,%r8
+	adcq	%rbp,%r15
+	adcq	$0,%r9
+	xorq	%rsi,%rsi  // borrow %rsi
+	testq	$1,%rax
+
+	cmovzq	%rax,%r12
+	cmovzq	%r10,%r13
+	cmovzq	%rcx,%r14
+	cmovzq	%r8,%r15
+	cmovzq	%rsi,%r9
+
+	movq	%r13,%rax  // a0:a3>>1
+	shrq	$1,%r12
+	shlq	$63,%rax
+	movq	%r14,%r10
+	shrq	$1,%r13
+	orq	%rax,%r12
+	shlq	$63,%r10
+	movq	%r15,%rcx
+	shrq	$1,%r14
+	orq	%r10,%r13
+	shlq	$63,%rcx
+	movq	%r12,0(%rdi)
+	shrq	$1,%r15
+	movq	%r13,8(%rdi)
+	shlq	$63,%r9
+	orq	%rcx,%r14
+	orq	%r9,%r15
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+	movq	64(%rsp),%rdx
+	leaq	64(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(M, M, Zsqr);
+
+	leaq	128(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2x
+
+	leaq	32(%rsp),%rbx
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_add_tox  // p256_mul_by_3(M, M);
+
+	movq	96(%rsp),%rdx
+	leaq	96(%rsp),%rbx
+	movq	0+0(%rsp),%r9
+	movq	8+0(%rsp),%r10
+	leaq	-128+0(%rsp),%rsi
+	movq	16+0(%rsp),%r11
+	movq	24+0(%rsp),%r12
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S, S, in_x);
+
+	leaq	128(%rsp),%rdi
+	call	__ecp_nistz256_mul_by_2x  // p256_mul_by_2(tmp0, S);
+
+	movq	0+32(%rsp),%rdx
+	movq	8+32(%rsp),%r14
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r15
+	movq	24+32(%rsp),%r8
+	movq	%xmm0,%rdi
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(res_x, M);
+
+	leaq	128(%rsp),%rbx
+	movq	%r14,%r8  // harmonize sqr output and sub input
+	movq	%r15,%r9
+	movq	%rsi,%r14
+	movq	%rbp,%r15
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_x, res_x, tmp0);
+
+	movq	0+0(%rsp),%rax
+	movq	0+8(%rsp),%rbp
+	movq	0+16(%rsp),%rcx
+	movq	0+24(%rsp),%r10  // "4-5-0-1" order
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_subx  // p256_sub(S, S, res_x);
+
+	movq	32(%rsp),%rdx
+	leaq	32(%rsp),%rbx
+	movq	%r12,%r14  // harmonize sub output and mul input
+	xorl	%ecx,%ecx
+	movq	%r12,0+0(%rsp)  // have to save:-(
+	movq	%r13,%r10
+	movq	%r13,0+8(%rsp)
+	cmovzq	%r8,%r11
+	movq	%r8,0+16(%rsp)
+	leaq	0-128(%rsp),%rsi
+	cmovzq	%r9,%r12
+	movq	%r9,0+24(%rsp)
+	movq	%r14,%r9
+	leaq	0(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S, S, M);
+
+	movq	%xmm1,%rbx
+	movq	%xmm1,%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_y, S, res_y);
+
+	leaq	160+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lpoint_doublex_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_double_adx,.-ecp_nistz256_point_double_adx
+.globl	ecp_nistz256_point_add_adx
+.hidden ecp_nistz256_point_add_adx
+.type	ecp_nistz256_point_add_adx,@function
+.align	32
+ecp_nistz256_point_add_adx:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$576+8,%rsp
+.cfi_adjust_cfa_offset	32*18+8
+.Lpoint_addx_body:
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	movdqu	48(%rsi),%xmm3
+	movdqu	64(%rsi),%xmm4
+	movdqu	80(%rsi),%xmm5
+	movq	%rsi,%rbx  // reassign
+	movq	%rdx,%rsi  // reassign
+	movdqa	%xmm0,384(%rsp)
+	movdqa	%xmm1,384+16(%rsp)
+	movdqa	%xmm2,416(%rsp)
+	movdqa	%xmm3,416+16(%rsp)
+	movdqa	%xmm4,448(%rsp)
+	movdqa	%xmm5,448+16(%rsp)
+	por	%xmm4,%xmm5
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rbx
+	pshufd	$0xb1,%xmm5,%xmm3
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	por	%xmm3,%xmm5
+	movdqu	48(%rsi),%xmm3
+	movq	64+0(%rsi),%rdx  // load original in2_z
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	movdqa	%xmm0,480(%rsp)
+	pshufd	$0x1e,%xmm5,%xmm4
+	movdqa	%xmm1,480+16(%rsp)
+	movdqu	64(%rsi),%xmm0  // in2_z again
+	movdqu	80(%rsi),%xmm1
+	movdqa	%xmm2,512(%rsp)
+	movdqa	%xmm3,512+16(%rsp)
+	por	%xmm4,%xmm5
+	pxor	%xmm4,%xmm4
+	por	%xmm0,%xmm1
+	movq	%rdi,%xmm0  // save %rdi
+
+	leaq	64-128(%rsi),%rsi  // %rsi is still valid
+	movq	%rdx,544+0(%rsp)  // make in2_z copy
+	movq	%r14,544+8(%rsp)
+	movq	%r15,544+16(%rsp)
+	movq	%r8,544+24(%rsp)
+	leaq	96(%rsp),%rdi  // Z2^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Z2sqr, in2_z);
+
+	pcmpeqd	%xmm4,%xmm5
+	pshufd	$0xb1,%xmm1,%xmm4
+	por	%xmm1,%xmm4
+	pshufd	$0,%xmm5,%xmm5  // in1infty
+	pshufd	$0x1e,%xmm4,%xmm3
+	por	%xmm3,%xmm4
+	pxor	%xmm3,%xmm3
+	pcmpeqd	%xmm3,%xmm4
+	pshufd	$0,%xmm4,%xmm4  // in2infty
+	movq	64+0(%rbx),%rdx  // load original in1_z
+	movq	64+8(%rbx),%r14
+	movq	64+16(%rbx),%r15
+	movq	64+24(%rbx),%r8
+	movq	%rbx,%xmm1
+
+	leaq	64-128(%rbx),%rsi
+	leaq	32(%rsp),%rdi  // Z1^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Z1sqr, in1_z);
+
+	movq	544(%rsp),%rdx
+	leaq	544(%rsp),%rbx
+	movq	0+96(%rsp),%r9
+	movq	8+96(%rsp),%r10
+	leaq	-128+96(%rsp),%rsi
+	movq	16+96(%rsp),%r11
+	movq	24+96(%rsp),%r12
+	leaq	224(%rsp),%rdi  // S1 = Z2^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S1, Z2sqr, in2_z);
+
+	movq	448(%rsp),%rdx
+	leaq	448(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	256(%rsp),%rdi  // S2 = Z1^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, Z1sqr, in1_z);
+
+	movq	416(%rsp),%rdx
+	leaq	416(%rsp),%rbx
+	movq	0+224(%rsp),%r9
+	movq	8+224(%rsp),%r10
+	leaq	-128+224(%rsp),%rsi
+	movq	16+224(%rsp),%r11
+	movq	24+224(%rsp),%r12
+	leaq	224(%rsp),%rdi  // S1 = Y1*Z2^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S1, S1, in1_y);
+
+	movq	512(%rsp),%rdx
+	leaq	512(%rsp),%rbx
+	movq	0+256(%rsp),%r9
+	movq	8+256(%rsp),%r10
+	leaq	-128+256(%rsp),%rsi
+	movq	16+256(%rsp),%r11
+	movq	24+256(%rsp),%r12
+	leaq	256(%rsp),%rdi  // S2 = Y2*Z1^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, S2, in2_y);
+
+	leaq	224(%rsp),%rbx
+	leaq	64(%rsp),%rdi  // R = S2 - S1
+	call	__ecp_nistz256_sub_fromx  // p256_sub(R, S2, S1);
+
+	orq	%r13,%r12  // see if result is zero
+	movdqa	%xmm4,%xmm2
+	orq	%r8,%r12
+	orq	%r9,%r12
+	por	%xmm5,%xmm2  // in1infty || in2infty
+	movq	%r12,%xmm3
+
+	movq	384(%rsp),%rdx
+	leaq	384(%rsp),%rbx
+	movq	0+96(%rsp),%r9
+	movq	8+96(%rsp),%r10
+	leaq	-128+96(%rsp),%rsi
+	movq	16+96(%rsp),%r11
+	movq	24+96(%rsp),%r12
+	leaq	160(%rsp),%rdi  // U1 = X1*Z2^2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(U1, in1_x, Z2sqr);
+
+	movq	480(%rsp),%rdx
+	leaq	480(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	192(%rsp),%rdi  // U2 = X2*Z1^2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(U2, in2_x, Z1sqr);
+
+	leaq	160(%rsp),%rbx
+	leaq	0(%rsp),%rdi  // H = U2 - U1
+	call	__ecp_nistz256_sub_fromx  // p256_sub(H, U2, U1);
+
+	orq	%r13,%r12  // see if result is zero
+	orq	%r8,%r12
+	orq	%r9,%r12  // 
+
+	movq	%xmm2,%r8
+	movq	%xmm3,%r9
+	orq	%r8,%r12
+.byte	0x3e  // predict taken
+	jnz	.Ladd_proceedx  // 
+
+// We now know A = B or A = -B and neither is infinity. Compare the
+// y-coordinates via S1 and S2.
+	testq	%r9,%r9
+	jz	.Ladd_doublex  // is_equal(S1, S2)
+
+// A = -B, so the result is infinity.
+// 
+// TODO(davidben): Does .Ladd_proceed handle this case? It seems to, in
+// which case we should eliminate this special-case and simplify the
+// timing analysis.
+	movq	%xmm0,%rdi  // restore %rdi
+	pxor	%xmm0,%xmm0
+	movdqu	%xmm0,0(%rdi)
+	movdqu	%xmm0,16(%rdi)
+	movdqu	%xmm0,32(%rdi)
+	movdqu	%xmm0,48(%rdi)
+	movdqu	%xmm0,64(%rdi)
+	movdqu	%xmm0,80(%rdi)
+	jmp	.Ladd_donex
+
+.align	32
+.Ladd_doublex:
+	movq	%xmm1,%rsi  // restore %rsi
+	movq	%xmm0,%rdi  // restore %rdi
+	addq	$416,%rsp  // difference in frame sizes
+.cfi_adjust_cfa_offset	-416
+	jmp	.Lpoint_double_shortcutx
+.cfi_adjust_cfa_offset	416
+
+.align	32
+.Ladd_proceedx:
+	movq	0+64(%rsp),%rdx
+	movq	8+64(%rsp),%r14
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r15
+	movq	24+64(%rsp),%r8
+	leaq	96(%rsp),%rdi  // R^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Rsqr, R);
+
+	movq	448(%rsp),%rdx
+	leaq	448(%rsp),%rbx
+	movq	0+0(%rsp),%r9
+	movq	8+0(%rsp),%r10
+	leaq	-128+0(%rsp),%rsi
+	movq	16+0(%rsp),%r11
+	movq	24+0(%rsp),%r12
+	leaq	352(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(res_z, H, in1_z);
+
+	movq	0+0(%rsp),%rdx
+	movq	8+0(%rsp),%r14
+	leaq	-128+0(%rsp),%rsi
+	movq	16+0(%rsp),%r15
+	movq	24+0(%rsp),%r8
+	leaq	32(%rsp),%rdi  // H^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Hsqr, H);
+
+	movq	544(%rsp),%rdx
+	leaq	544(%rsp),%rbx
+	movq	0+352(%rsp),%r9
+	movq	8+352(%rsp),%r10
+	leaq	-128+352(%rsp),%rsi
+	movq	16+352(%rsp),%r11
+	movq	24+352(%rsp),%r12
+	leaq	352(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(res_z, res_z, in2_z);
+
+	movq	0(%rsp),%rdx
+	leaq	0(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	128(%rsp),%rdi  // H^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(Hcub, Hsqr, H);
+
+	movq	160(%rsp),%rdx
+	leaq	160(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	192(%rsp),%rdi  // U1*H^2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(U2, U1, Hsqr);
+// lea	192(%rsp), %rsi
+// lea	32(%rsp), %rdi	# 2*U1*H^2
+// call	__ecp_nistz256_mul_by_2	# ecp_nistz256_mul_by_2(Hsqr, U2);
+
+	xorq	%r11,%r11
+	addq	%r12,%r12  // a0:a3+a0:a3
+	leaq	96(%rsp),%rsi
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	movq	0(%rsi),%rax
+	cmovcq	%rbp,%r13
+	movq	8(%rsi),%rbp
+	cmovcq	%rcx,%r8
+	movq	16(%rsi),%rcx
+	cmovcq	%r10,%r9
+	movq	24(%rsi),%r10
+
+	call	__ecp_nistz256_subx  // p256_sub(res_x, Rsqr, Hsqr);
+
+	leaq	128(%rsp),%rbx
+	leaq	288(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_x, res_x, Hcub);
+
+	movq	192+0(%rsp),%rax
+	movq	192+8(%rsp),%rbp
+	movq	192+16(%rsp),%rcx
+	movq	192+24(%rsp),%r10
+	leaq	320(%rsp),%rdi
+
+	call	__ecp_nistz256_subx  // p256_sub(res_y, U2, res_x);
+
+	movq	%r12,0(%rdi)  // save the result, as
+	movq	%r13,8(%rdi)  // __ecp_nistz256_sub doesn't
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+	movq	128(%rsp),%rdx
+	leaq	128(%rsp),%rbx
+	movq	0+224(%rsp),%r9
+	movq	8+224(%rsp),%r10
+	leaq	-128+224(%rsp),%rsi
+	movq	16+224(%rsp),%r11
+	movq	24+224(%rsp),%r12
+	leaq	256(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, S1, Hcub);
+
+	movq	320(%rsp),%rdx
+	leaq	320(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	320(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(res_y, R, res_y);
+
+	leaq	256(%rsp),%rbx
+	leaq	320(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_y, res_y, S2);
+
+	movq	%xmm0,%rdi  // restore %rdi
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_z, in2_z, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	352(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	352+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	544(%rsp),%xmm2
+	pand	544+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_z, in1_z, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	448(%rsp),%xmm2
+	pand	448+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,64(%rdi)
+	movdqu	%xmm3,80(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_x, in2_x, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	288(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	288+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	480(%rsp),%xmm2
+	pand	480+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_x, in1_x, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	384(%rsp),%xmm2
+	pand	384+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_y, in2_y, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	320(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	320+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	512(%rsp),%xmm2
+	pand	512+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_y, in1_y, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	416(%rsp),%xmm2
+	pand	416+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,32(%rdi)
+	movdqu	%xmm3,48(%rdi)
+
+.Ladd_donex:
+	leaq	576+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lpoint_addx_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_add_adx,.-ecp_nistz256_point_add_adx
+.globl	ecp_nistz256_point_add_affine_adx
+.hidden ecp_nistz256_point_add_affine_adx
+.type	ecp_nistz256_point_add_affine_adx,@function
+.align	32
+ecp_nistz256_point_add_affine_adx:
+.cfi_startproc	
+_CET_ENDBR
+	pushq	%rbp
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbp,-16
+	pushq	%rbx
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%rbx,-24
+	pushq	%r12
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_adjust_cfa_offset	8
+.cfi_offset	%r15,-56
+	subq	$480+8,%rsp
+.cfi_adjust_cfa_offset	32*15+8
+.Ladd_affinex_body:
+
+	movdqu	0(%rsi),%xmm0  // copy	*(P256_POINT *)%rsi
+	movq	%rdx,%rbx  // reassign
+	movdqu	16(%rsi),%xmm1
+	movdqu	32(%rsi),%xmm2
+	movdqu	48(%rsi),%xmm3
+	movdqu	64(%rsi),%xmm4
+	movdqu	80(%rsi),%xmm5
+	movq	64+0(%rsi),%rdx  // load original in1_z
+	movq	64+8(%rsi),%r14
+	movq	64+16(%rsi),%r15
+	movq	64+24(%rsi),%r8
+	movdqa	%xmm0,320(%rsp)
+	movdqa	%xmm1,320+16(%rsp)
+	movdqa	%xmm2,352(%rsp)
+	movdqa	%xmm3,352+16(%rsp)
+	movdqa	%xmm4,384(%rsp)
+	movdqa	%xmm5,384+16(%rsp)
+	por	%xmm4,%xmm5
+
+	movdqu	0(%rbx),%xmm0  // copy	*(P256_POINT_AFFINE *)%rbx
+	pshufd	$0xb1,%xmm5,%xmm3
+	movdqu	16(%rbx),%xmm1
+	movdqu	32(%rbx),%xmm2
+	por	%xmm3,%xmm5
+	movdqu	48(%rbx),%xmm3
+	movdqa	%xmm0,416(%rsp)
+	pshufd	$0x1e,%xmm5,%xmm4
+	movdqa	%xmm1,416+16(%rsp)
+	por	%xmm0,%xmm1
+	movq	%rdi,%xmm0  // save %rdi
+	movdqa	%xmm2,448(%rsp)
+	movdqa	%xmm3,448+16(%rsp)
+	por	%xmm2,%xmm3
+	por	%xmm4,%xmm5
+	pxor	%xmm4,%xmm4
+	por	%xmm1,%xmm3
+
+	leaq	64-128(%rsi),%rsi  // %rsi is still valid
+	leaq	32(%rsp),%rdi  // Z1^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Z1sqr, in1_z);
+
+	pcmpeqd	%xmm4,%xmm5
+	pshufd	$0xb1,%xmm3,%xmm4
+	movq	0(%rbx),%rdx  // %rbx is still valid
+// lea	0x00(%rbx), %rbx
+	movq	%r12,%r9  // harmonize sqr output and mul input
+	por	%xmm3,%xmm4
+	pshufd	$0,%xmm5,%xmm5  // in1infty
+	pshufd	$0x1e,%xmm4,%xmm3
+	movq	%r13,%r10
+	por	%xmm3,%xmm4
+	pxor	%xmm3,%xmm3
+	movq	%r14,%r11
+	pcmpeqd	%xmm3,%xmm4
+	pshufd	$0,%xmm4,%xmm4  // in2infty
+
+	leaq	32-128(%rsp),%rsi
+	movq	%r15,%r12
+	leaq	0(%rsp),%rdi  // U2 = X2*Z1^2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(U2, Z1sqr, in2_x);
+
+	leaq	320(%rsp),%rbx
+	leaq	64(%rsp),%rdi  // H = U2 - U1
+	call	__ecp_nistz256_sub_fromx  // p256_sub(H, U2, in1_x);
+
+	movq	384(%rsp),%rdx
+	leaq	384(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi  // S2 = Z1^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, Z1sqr, in1_z);
+
+	movq	384(%rsp),%rdx
+	leaq	384(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	288(%rsp),%rdi  // Z3 = H*Z1*Z2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(res_z, H, in1_z);
+
+	movq	448(%rsp),%rdx
+	leaq	448(%rsp),%rbx
+	movq	0+32(%rsp),%r9
+	movq	8+32(%rsp),%r10
+	leaq	-128+32(%rsp),%rsi
+	movq	16+32(%rsp),%r11
+	movq	24+32(%rsp),%r12
+	leaq	32(%rsp),%rdi  // S2 = Y2*Z1^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, S2, in2_y);
+
+	leaq	352(%rsp),%rbx
+	leaq	96(%rsp),%rdi  // R = S2 - S1
+	call	__ecp_nistz256_sub_fromx  // p256_sub(R, S2, in1_y);
+
+	movq	0+64(%rsp),%rdx
+	movq	8+64(%rsp),%r14
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r15
+	movq	24+64(%rsp),%r8
+	leaq	128(%rsp),%rdi  // H^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Hsqr, H);
+
+	movq	0+96(%rsp),%rdx
+	movq	8+96(%rsp),%r14
+	leaq	-128+96(%rsp),%rsi
+	movq	16+96(%rsp),%r15
+	movq	24+96(%rsp),%r8
+	leaq	192(%rsp),%rdi  // R^2
+	call	__ecp_nistz256_sqr_montx  // p256_sqr_mont(Rsqr, R);
+
+	movq	128(%rsp),%rdx
+	leaq	128(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	160(%rsp),%rdi  // H^3
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(Hcub, Hsqr, H);
+
+	movq	320(%rsp),%rdx
+	leaq	320(%rsp),%rbx
+	movq	0+128(%rsp),%r9
+	movq	8+128(%rsp),%r10
+	leaq	-128+128(%rsp),%rsi
+	movq	16+128(%rsp),%r11
+	movq	24+128(%rsp),%r12
+	leaq	0(%rsp),%rdi  // U1*H^2
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(U2, in1_x, Hsqr);
+// lea	0(%rsp), %rsi
+// lea	128(%rsp), %rdi	# 2*U1*H^2
+// call	__ecp_nistz256_mul_by_2	# ecp_nistz256_mul_by_2(Hsqr, U2);
+
+	xorq	%r11,%r11
+	addq	%r12,%r12  // a0:a3+a0:a3
+	leaq	192(%rsp),%rsi
+	adcq	%r13,%r13
+	movq	%r12,%rax
+	adcq	%r8,%r8
+	adcq	%r9,%r9
+	movq	%r13,%rbp
+	adcq	$0,%r11
+
+	subq	$-1,%r12
+	movq	%r8,%rcx
+	sbbq	%r14,%r13
+	sbbq	$0,%r8
+	movq	%r9,%r10
+	sbbq	%r15,%r9
+	sbbq	$0,%r11
+
+	cmovcq	%rax,%r12
+	movq	0(%rsi),%rax
+	cmovcq	%rbp,%r13
+	movq	8(%rsi),%rbp
+	cmovcq	%rcx,%r8
+	movq	16(%rsi),%rcx
+	cmovcq	%r10,%r9
+	movq	24(%rsi),%r10
+
+	call	__ecp_nistz256_subx  // p256_sub(res_x, Rsqr, Hsqr);
+
+	leaq	160(%rsp),%rbx
+	leaq	224(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_x, res_x, Hcub);
+
+	movq	0+0(%rsp),%rax
+	movq	0+8(%rsp),%rbp
+	movq	0+16(%rsp),%rcx
+	movq	0+24(%rsp),%r10
+	leaq	64(%rsp),%rdi
+
+	call	__ecp_nistz256_subx  // p256_sub(H, U2, res_x);
+
+	movq	%r12,0(%rdi)  // save the result, as
+	movq	%r13,8(%rdi)  // __ecp_nistz256_sub doesn't
+	movq	%r8,16(%rdi)
+	movq	%r9,24(%rdi)
+	movq	352(%rsp),%rdx
+	leaq	352(%rsp),%rbx
+	movq	0+160(%rsp),%r9
+	movq	8+160(%rsp),%r10
+	leaq	-128+160(%rsp),%rsi
+	movq	16+160(%rsp),%r11
+	movq	24+160(%rsp),%r12
+	leaq	32(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(S2, Hcub, in1_y);
+
+	movq	96(%rsp),%rdx
+	leaq	96(%rsp),%rbx
+	movq	0+64(%rsp),%r9
+	movq	8+64(%rsp),%r10
+	leaq	-128+64(%rsp),%rsi
+	movq	16+64(%rsp),%r11
+	movq	24+64(%rsp),%r12
+	leaq	64(%rsp),%rdi
+	call	__ecp_nistz256_mul_montx  // p256_mul_mont(H, H, R);
+
+	leaq	32(%rsp),%rbx
+	leaq	256(%rsp),%rdi
+	call	__ecp_nistz256_sub_fromx  // p256_sub(res_y, H, S2);
+
+	movq	%xmm0,%rdi  // restore %rdi
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_z, ONE, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	288(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	288+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	.LONE_mont(%rip),%xmm2
+	pand	.LONE_mont+16(%rip),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_z, in1_z, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	384(%rsp),%xmm2
+	pand	384+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,64(%rdi)
+	movdqu	%xmm3,80(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_x, in2_x, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	224(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	224+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	416(%rsp),%xmm2
+	pand	416+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_x, in1_x, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	320(%rsp),%xmm2
+	pand	320+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,0(%rdi)
+	movdqu	%xmm3,16(%rdi)
+
+	movdqa	%xmm5,%xmm0  // copy_conditional(res_y, in2_y, in1infty);
+	movdqa	%xmm5,%xmm1
+	pandn	256(%rsp),%xmm0
+	movdqa	%xmm5,%xmm2
+	pandn	256+16(%rsp),%xmm1
+	movdqa	%xmm5,%xmm3
+	pand	448(%rsp),%xmm2
+	pand	448+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+
+	movdqa	%xmm4,%xmm0  // copy_conditional(res_y, in1_y, in2infty);
+	movdqa	%xmm4,%xmm1
+	pandn	%xmm2,%xmm0
+	movdqa	%xmm4,%xmm2
+	pandn	%xmm3,%xmm1
+	movdqa	%xmm4,%xmm3
+	pand	352(%rsp),%xmm2
+	pand	352+16(%rsp),%xmm3
+	por	%xmm0,%xmm2
+	por	%xmm1,%xmm3
+	movdqu	%xmm2,32(%rdi)
+	movdqu	%xmm3,48(%rdi)
+
+	leaq	480+56(%rsp),%rsi
+.cfi_def_cfa	%rsi,8
+	movq	-48(%rsi),%r15
+.cfi_restore	%r15
+	movq	-40(%rsi),%r14
+.cfi_restore	%r14
+	movq	-32(%rsi),%r13
+.cfi_restore	%r13
+	movq	-24(%rsi),%r12
+.cfi_restore	%r12
+	movq	-16(%rsi),%rbx
+.cfi_restore	%rbx
+	movq	-8(%rsi),%rbp
+.cfi_restore	%rbp
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Ladd_affinex_epilogue:
+	ret
+.cfi_endproc	
+.size	ecp_nistz256_point_add_affine_adx,.-ecp_nistz256_point_add_affine_adx
+#endif