visaljs 0.0.2 → 0.0.4

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

@@ -0,0 +1,1237 @@
+// 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	
+
+.globl	bn_mul_mont_nohw
+.hidden bn_mul_mont_nohw
+.type	bn_mul_mont_nohw,@function
+.align	16
+bn_mul_mont_nohw:
+.cfi_startproc	
+_CET_ENDBR
+	movl	%r9d,%r9d
+	movq	%rsp,%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+
+	negq	%r9
+	movq	%rsp,%r11
+	leaq	-16(%rsp,%r9,8),%r10  // future alloca(8*(num+2))
+	negq	%r9  // restore %r9
+	andq	$-1024,%r10  // minimize TLB usage
+
+// An OS-agnostic version of __chkstk.
+// 
+// Some OSes (Windows) insist on stack being "wired" to
+// physical memory in strictly sequential manner, i.e. if stack
+// allocation spans two pages, then reference to farmost one can
+// be punishable by SEGV. But page walking can do good even on
+// other OSes, because it guarantees that villain thread hits
+// the guard page before it can make damage to innocent one...
+	subq	%r10,%r11
+	andq	$-4096,%r11
+	leaq	(%r10,%r11,1),%rsp
+	movq	(%rsp),%r11
+	cmpq	%r10,%rsp
+	ja	.Lmul_page_walk
+	jmp	.Lmul_page_walk_done
+
+.align	16
+.Lmul_page_walk:
+	leaq	-4096(%rsp),%rsp
+	movq	(%rsp),%r11
+	cmpq	%r10,%rsp
+	ja	.Lmul_page_walk
+.Lmul_page_walk_done:
+
+	movq	%rax,8(%rsp,%r9,8)  // tp[num+1]=%rsp
+.cfi_escape	0x0f,0x0a,0x77,0x08,0x79,0x00,0x38,0x1e,0x22,0x06,0x23,0x08
+.Lmul_body:
+	movq	%rdx,%r12  // reassign %rdx
+	movq	(%r8),%r8  // pull n0[0] value
+	movq	(%r12),%rbx  // m0=bp[0]
+	movq	(%rsi),%rax
+
+	xorq	%r14,%r14  // i=0
+	xorq	%r15,%r15  // j=0
+
+	movq	%r8,%rbp
+	mulq	%rbx  // ap[0]*bp[0]
+	movq	%rax,%r10
+	movq	(%rcx),%rax
+
+	imulq	%r10,%rbp  // "tp[0]"*n0
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[0]*m1
+	addq	%rax,%r10  // discarded
+	movq	8(%rsi),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r13
+
+	leaq	1(%r15),%r15  // j++
+	jmp	.L1st_enter
+
+.align	16
+.L1st:
+	addq	%rax,%r13
+	movq	(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%r13  // np[j]*m1+ap[j]*bp[0]
+	movq	%r10,%r11
+	adcq	$0,%rdx
+	movq	%r13,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+.L1st_enter:
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r11
+	movq	(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	leaq	1(%r15),%r15  // j++
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	cmpq	%r9,%r15
+	jne	.L1st
+
+	addq	%rax,%r13
+	movq	(%rsi),%rax  // ap[0]
+	adcq	$0,%rdx
+	addq	%r11,%r13  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%r13,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+	movq	%r10,%r11
+
+	xorq	%rdx,%rdx
+	addq	%r11,%r13
+	adcq	$0,%rdx
+	movq	%r13,-8(%rsp,%r9,8)
+	movq	%rdx,(%rsp,%r9,8)  // store upmost overflow bit
+
+	leaq	1(%r14),%r14  // i++
+	jmp	.Louter
+.align	16
+.Louter:
+	movq	(%r12,%r14,8),%rbx  // m0=bp[i]
+	xorq	%r15,%r15  // j=0
+	movq	%r8,%rbp
+	movq	(%rsp),%r10
+	mulq	%rbx  // ap[0]*bp[i]
+	addq	%rax,%r10  // ap[0]*bp[i]+tp[0]
+	movq	(%rcx),%rax
+	adcq	$0,%rdx
+
+	imulq	%r10,%rbp  // tp[0]*n0
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[0]*m1
+	addq	%rax,%r10  // discarded
+	movq	8(%rsi),%rax
+	adcq	$0,%rdx
+	movq	8(%rsp),%r10  // tp[1]
+	movq	%rdx,%r13
+
+	leaq	1(%r15),%r15  // j++
+	jmp	.Linner_enter
+
+.align	16
+.Linner:
+	addq	%rax,%r13
+	movq	(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13  // np[j]*m1+ap[j]*bp[i]+tp[j]
+	movq	(%rsp,%r15,8),%r10
+	adcq	$0,%rdx
+	movq	%r13,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+.Linner_enter:
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r11
+	movq	(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%r10  // ap[j]*bp[i]+tp[j]
+	movq	%rdx,%r11
+	adcq	$0,%r11
+	leaq	1(%r15),%r15  // j++
+
+	mulq	%rbp  // np[j]*m1
+	cmpq	%r9,%r15
+	jne	.Linner
+
+	addq	%rax,%r13
+	movq	(%rsi),%rax  // ap[0]
+	adcq	$0,%rdx
+	addq	%r10,%r13  // np[j]*m1+ap[j]*bp[i]+tp[j]
+	movq	(%rsp,%r15,8),%r10
+	adcq	$0,%rdx
+	movq	%r13,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+	xorq	%rdx,%rdx
+	addq	%r11,%r13
+	adcq	$0,%rdx
+	addq	%r10,%r13  // pull upmost overflow bit
+	adcq	$0,%rdx
+	movq	%r13,-8(%rsp,%r9,8)
+	movq	%rdx,(%rsp,%r9,8)  // store upmost overflow bit
+
+	leaq	1(%r14),%r14  // i++
+	cmpq	%r9,%r14
+	jb	.Louter
+
+	xorq	%r14,%r14  // i=0 and clear CF
+	movq	(%rsp),%rax  // tp[0]
+	movq	%r9,%r15  // j=num
+
+.align	16
+.Lsub:	sbbq	(%rcx,%r14,8),%rax
+	movq	%rax,(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	movq	8(%rsp,%r14,8),%rax  // tp[i+1]
+	leaq	1(%r14),%r14  // i++
+	decq	%r15  // doesn't affect CF
+	jnz	.Lsub
+
+	sbbq	$0,%rax  // handle upmost overflow bit
+	movq	$-1,%rbx
+	xorq	%rax,%rbx  // not %rax
+	xorq	%r14,%r14
+	movq	%r9,%r15  // j=num
+
+.Lcopy:  // conditional copy
+	movq	(%rdi,%r14,8),%rcx
+	movq	(%rsp,%r14,8),%rdx
+	andq	%rbx,%rcx
+	andq	%rax,%rdx
+	movq	%r9,(%rsp,%r14,8)  // zap temporary vector
+	orq	%rcx,%rdx
+	movq	%rdx,(%rdi,%r14,8)  // rp[i]=tp[i]
+	leaq	1(%r14),%r14
+	subq	$1,%r15
+	jnz	.Lcopy
+
+	movq	8(%rsp,%r9,8),%rsi  // restore %rsp
+.cfi_def_cfa	%rsi,8
+// No return value
+	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),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lmul_epilogue:
+	ret
+.cfi_endproc	
+.size	bn_mul_mont_nohw,.-bn_mul_mont_nohw
+.globl	bn_mul4x_mont
+.hidden bn_mul4x_mont
+.type	bn_mul4x_mont,@function
+.align	16
+bn_mul4x_mont:
+.cfi_startproc	
+_CET_ENDBR
+	movl	%r9d,%r9d
+	movq	%rsp,%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+
+	negq	%r9
+	movq	%rsp,%r11
+	leaq	-32(%rsp,%r9,8),%r10  // future alloca(8*(num+4))
+	negq	%r9  // restore
+	andq	$-1024,%r10  // minimize TLB usage
+
+	subq	%r10,%r11
+	andq	$-4096,%r11
+	leaq	(%r10,%r11,1),%rsp
+	movq	(%rsp),%r11
+	cmpq	%r10,%rsp
+	ja	.Lmul4x_page_walk
+	jmp	.Lmul4x_page_walk_done
+
+.Lmul4x_page_walk:
+	leaq	-4096(%rsp),%rsp
+	movq	(%rsp),%r11
+	cmpq	%r10,%rsp
+	ja	.Lmul4x_page_walk
+.Lmul4x_page_walk_done:
+
+	movq	%rax,8(%rsp,%r9,8)  // tp[num+1]=%rsp
+.cfi_escape	0x0f,0x0a,0x77,0x08,0x79,0x00,0x38,0x1e,0x22,0x06,0x23,0x08
+.Lmul4x_body:
+	movq	%rdi,16(%rsp,%r9,8)  // tp[num+2]=%rdi
+	movq	%rdx,%r12  // reassign %r12
+	movq	(%r8),%r8  // pull n0[0] value
+	movq	(%r12),%rbx  // m0=bp[0]
+	movq	(%rsi),%rax
+
+	xorq	%r14,%r14  // i=0
+	xorq	%r15,%r15  // j=0
+
+	movq	%r8,%rbp
+	mulq	%rbx  // ap[0]*bp[0]
+	movq	%rax,%r10
+	movq	(%rcx),%rax
+
+	imulq	%r10,%rbp  // "tp[0]"*n0
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[0]*m1
+	addq	%rax,%r10  // discarded
+	movq	8(%rsi),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rdi
+
+	mulq	%rbx
+	addq	%rax,%r11
+	movq	8(%rcx),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%rbp
+	addq	%rax,%rdi
+	movq	16(%rsi),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi
+	leaq	4(%r15),%r15  // j++
+	adcq	$0,%rdx
+	movq	%rdi,(%rsp)
+	movq	%rdx,%r13
+	jmp	.L1st4x
+.align	16
+.L1st4x:
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r10
+	movq	-16(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	-8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%r13,-24(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r11
+	movq	-8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%rdi,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r10
+	movq	(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%r13,-8(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r11
+	movq	8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	leaq	4(%r15),%r15  // j++
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	-16(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%rdi,-32(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+	cmpq	%r9,%r15
+	jb	.L1st4x
+
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r10
+	movq	-16(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	-8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%r13,-24(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[0]
+	addq	%rax,%r11
+	movq	-8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	(%rsi),%rax  // ap[0]
+	adcq	$0,%rdx
+	addq	%r11,%rdi  // np[j]*m1+ap[j]*bp[0]
+	adcq	$0,%rdx
+	movq	%rdi,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+	xorq	%rdi,%rdi
+	addq	%r10,%r13
+	adcq	$0,%rdi
+	movq	%r13,-8(%rsp,%r15,8)
+	movq	%rdi,(%rsp,%r15,8)  // store upmost overflow bit
+
+	leaq	1(%r14),%r14  // i++
+.align	4
+.Louter4x:
+	movq	(%r12,%r14,8),%rbx  // m0=bp[i]
+	xorq	%r15,%r15  // j=0
+	movq	(%rsp),%r10
+	movq	%r8,%rbp
+	mulq	%rbx  // ap[0]*bp[i]
+	addq	%rax,%r10  // ap[0]*bp[i]+tp[0]
+	movq	(%rcx),%rax
+	adcq	$0,%rdx
+
+	imulq	%r10,%rbp  // tp[0]*n0
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[0]*m1
+	addq	%rax,%r10  // "%r13", discarded
+	movq	8(%rsi),%rax
+	adcq	$0,%rdx
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r11
+	movq	8(%rcx),%rax
+	adcq	$0,%rdx
+	addq	8(%rsp),%r11  // +tp[1]
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	16(%rsi),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi  // np[j]*m1+ap[j]*bp[i]+tp[j]
+	leaq	4(%r15),%r15  // j+=2
+	adcq	$0,%rdx
+	movq	%rdi,(%rsp)  // tp[j-1]
+	movq	%rdx,%r13
+	jmp	.Linner4x
+.align	16
+.Linner4x:
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r10
+	movq	-16(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	-16(%rsp,%r15,8),%r10  // ap[j]*bp[i]+tp[j]
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	-8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13
+	adcq	$0,%rdx
+	movq	%r13,-24(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r11
+	movq	-8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	-8(%rsp,%r15,8),%r11
+	adcq	$0,%rdx
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi
+	adcq	$0,%rdx
+	movq	%rdi,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r10
+	movq	(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	(%rsp,%r15,8),%r10  // ap[j]*bp[i]+tp[j]
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13
+	adcq	$0,%rdx
+	movq	%r13,-8(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r11
+	movq	8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	8(%rsp,%r15,8),%r11
+	adcq	$0,%rdx
+	leaq	4(%r15),%r15  // j++
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	-16(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r11,%rdi
+	adcq	$0,%rdx
+	movq	%rdi,-32(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+	cmpq	%r9,%r15
+	jb	.Linner4x
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r10
+	movq	-16(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	-16(%rsp,%r15,8),%r10  // ap[j]*bp[i]+tp[j]
+	adcq	$0,%rdx
+	movq	%rdx,%r11
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%r13
+	movq	-8(%rsi,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	%r10,%r13
+	adcq	$0,%rdx
+	movq	%r13,-24(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%rdi
+
+	mulq	%rbx  // ap[j]*bp[i]
+	addq	%rax,%r11
+	movq	-8(%rcx,%r15,8),%rax
+	adcq	$0,%rdx
+	addq	-8(%rsp,%r15,8),%r11
+	adcq	$0,%rdx
+	leaq	1(%r14),%r14  // i++
+	movq	%rdx,%r10
+
+	mulq	%rbp  // np[j]*m1
+	addq	%rax,%rdi
+	movq	(%rsi),%rax  // ap[0]
+	adcq	$0,%rdx
+	addq	%r11,%rdi
+	adcq	$0,%rdx
+	movq	%rdi,-16(%rsp,%r15,8)  // tp[j-1]
+	movq	%rdx,%r13
+
+	xorq	%rdi,%rdi
+	addq	%r10,%r13
+	adcq	$0,%rdi
+	addq	(%rsp,%r9,8),%r13  // pull upmost overflow bit
+	adcq	$0,%rdi
+	movq	%r13,-8(%rsp,%r15,8)
+	movq	%rdi,(%rsp,%r15,8)  // store upmost overflow bit
+
+	cmpq	%r9,%r14
+	jb	.Louter4x
+	movq	16(%rsp,%r9,8),%rdi  // restore %rdi
+	leaq	-4(%r9),%r15
+	movq	0(%rsp),%rax  // tp[0]
+	movq	8(%rsp),%rdx  // tp[1]
+	shrq	$2,%r15  // j=num/4-1
+	leaq	(%rsp),%rsi  // borrow ap for tp
+	xorq	%r14,%r14  // i=0 and clear CF
+
+	subq	0(%rcx),%rax
+	movq	16(%rsi),%rbx  // tp[2]
+	movq	24(%rsi),%rbp  // tp[3]
+	sbbq	8(%rcx),%rdx
+
+.Lsub4x:
+	movq	%rax,0(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	movq	%rdx,8(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	sbbq	16(%rcx,%r14,8),%rbx
+	movq	32(%rsi,%r14,8),%rax  // tp[i+1]
+	movq	40(%rsi,%r14,8),%rdx
+	sbbq	24(%rcx,%r14,8),%rbp
+	movq	%rbx,16(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	movq	%rbp,24(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	sbbq	32(%rcx,%r14,8),%rax
+	movq	48(%rsi,%r14,8),%rbx
+	movq	56(%rsi,%r14,8),%rbp
+	sbbq	40(%rcx,%r14,8),%rdx
+	leaq	4(%r14),%r14  // i++
+	decq	%r15  // doesn't affect CF
+	jnz	.Lsub4x
+
+	movq	%rax,0(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	movq	32(%rsi,%r14,8),%rax  // load overflow bit
+	sbbq	16(%rcx,%r14,8),%rbx
+	movq	%rdx,8(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	sbbq	24(%rcx,%r14,8),%rbp
+	movq	%rbx,16(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+
+	sbbq	$0,%rax  // handle upmost overflow bit
+	movq	%rbp,24(%rdi,%r14,8)  // rp[i]=tp[i]-np[i]
+	pxor	%xmm0,%xmm0
+	movq	%rax,%xmm4
+	pcmpeqd	%xmm5,%xmm5
+	pshufd	$0,%xmm4,%xmm4
+	movq	%r9,%r15
+	pxor	%xmm4,%xmm5
+	shrq	$2,%r15  // j=num/4
+	xorl	%eax,%eax  // i=0
+
+	jmp	.Lcopy4x
+.align	16
+.Lcopy4x:  // conditional copy
+	movdqa	(%rsp,%rax,1),%xmm1
+	movdqu	(%rdi,%rax,1),%xmm2
+	pand	%xmm4,%xmm1
+	pand	%xmm5,%xmm2
+	movdqa	16(%rsp,%rax,1),%xmm3
+	movdqa	%xmm0,(%rsp,%rax,1)
+	por	%xmm2,%xmm1
+	movdqu	16(%rdi,%rax,1),%xmm2
+	movdqu	%xmm1,(%rdi,%rax,1)
+	pand	%xmm4,%xmm3
+	pand	%xmm5,%xmm2
+	movdqa	%xmm0,16(%rsp,%rax,1)
+	por	%xmm2,%xmm3
+	movdqu	%xmm3,16(%rdi,%rax,1)
+	leaq	32(%rax),%rax
+	decq	%r15
+	jnz	.Lcopy4x
+	movq	8(%rsp,%r9,8),%rsi  // restore %rsp
+.cfi_def_cfa	%rsi, 8
+// No return value
+	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),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lmul4x_epilogue:
+	ret
+.cfi_endproc	
+.size	bn_mul4x_mont,.-bn_mul4x_mont
+.extern	bn_sqrx8x_internal
+.hidden bn_sqrx8x_internal  // see x86_64-mont5 module
+.extern	bn_sqr8x_internal
+.hidden bn_sqr8x_internal  // see x86_64-mont5 module
+
+.globl	bn_sqr8x_mont
+.hidden bn_sqr8x_mont
+.type	bn_sqr8x_mont,@function
+.align	32
+bn_sqr8x_mont:
+.cfi_startproc	
+_CET_ENDBR
+	movl	%r9d,%r9d
+	movq	%rsp,%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+.Lsqr8x_prologue:
+
+	movl	%r9d,%r10d
+	shll	$3,%r9d  // convert %r9 to bytes
+	shlq	$3+2,%r10  // 4*%r9
+	negq	%r9
+
+// #############################################################
+// ensure that stack frame doesn't alias with %rsi modulo
+// 4096. this is done to allow memory disambiguation logic
+// do its job.
+// 
+	leaq	-64(%rsp,%r9,2),%r11
+	movq	%rsp,%rbp
+	movq	(%r8),%r8  // *n0
+	subq	%rsi,%r11
+	andq	$4095,%r11
+	cmpq	%r11,%r10
+	jb	.Lsqr8x_sp_alt
+	subq	%r11,%rbp  // align with %rsi
+	leaq	-64(%rbp,%r9,2),%rbp  // future alloca(frame+2*%r9)
+	jmp	.Lsqr8x_sp_done
+
+.align	32
+.Lsqr8x_sp_alt:
+	leaq	4096-64(,%r9,2),%r10  // 4096-frame-2*%r9
+	leaq	-64(%rbp,%r9,2),%rbp  // future alloca(frame+2*%r9)
+	subq	%r10,%r11
+	movq	$0,%r10
+	cmovcq	%r10,%r11
+	subq	%r11,%rbp
+.Lsqr8x_sp_done:
+	andq	$-64,%rbp
+	movq	%rsp,%r11
+	subq	%rbp,%r11
+	andq	$-4096,%r11
+	leaq	(%r11,%rbp,1),%rsp
+	movq	(%rsp),%r10
+	cmpq	%rbp,%rsp
+	ja	.Lsqr8x_page_walk
+	jmp	.Lsqr8x_page_walk_done
+
+.align	16
+.Lsqr8x_page_walk:
+	leaq	-4096(%rsp),%rsp
+	movq	(%rsp),%r10
+	cmpq	%rbp,%rsp
+	ja	.Lsqr8x_page_walk
+.Lsqr8x_page_walk_done:
+
+	movq	%r9,%r10
+	negq	%r9
+
+	movq	%r8,32(%rsp)
+	movq	%rax,40(%rsp)  // save original %rsp
+.cfi_escape	0x0f,0x05,0x77,0x28,0x06,0x23,0x08
+.Lsqr8x_body:
+
+	movq	%rcx,%xmm2  // save pointer to modulus
+	pxor	%xmm0,%xmm0
+	movq	%rdi,%xmm1  // save %rdi
+	movq	%r10,%xmm3  // -%r9
+	testq	%rdx,%rdx
+	jz	.Lsqr8x_nox
+
+	call	bn_sqrx8x_internal  // see x86_64-mont5 module
+// %rax	top-most carry
+// %rbp	nptr
+// %rcx	-8*num
+// %r8	end of tp[2*num]
+	leaq	(%r8,%rcx,1),%rbx
+	movq	%rcx,%r9
+	movq	%rcx,%rdx
+	movq	%xmm1,%rdi
+	sarq	$3+2,%rcx  // %cf=0
+	jmp	.Lsqr8x_sub
+
+.align	32
+.Lsqr8x_nox:
+	call	bn_sqr8x_internal  // see x86_64-mont5 module
+// %rax	top-most carry
+// %rbp	nptr
+// %r8	-8*num
+// %rdi	end of tp[2*num]
+	leaq	(%rdi,%r9,1),%rbx
+	movq	%r9,%rcx
+	movq	%r9,%rdx
+	movq	%xmm1,%rdi
+	sarq	$3+2,%rcx  // %cf=0
+	jmp	.Lsqr8x_sub
+
+.align	32
+.Lsqr8x_sub:
+	movq	0(%rbx),%r12
+	movq	8(%rbx),%r13
+	movq	16(%rbx),%r14
+	movq	24(%rbx),%r15
+	leaq	32(%rbx),%rbx
+	sbbq	0(%rbp),%r12
+	sbbq	8(%rbp),%r13
+	sbbq	16(%rbp),%r14
+	sbbq	24(%rbp),%r15
+	leaq	32(%rbp),%rbp
+	movq	%r12,0(%rdi)
+	movq	%r13,8(%rdi)
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+	leaq	32(%rdi),%rdi
+	incq	%rcx  // preserves %cf
+	jnz	.Lsqr8x_sub
+
+	sbbq	$0,%rax  // top-most carry
+	leaq	(%rbx,%r9,1),%rbx  // rewind
+	leaq	(%rdi,%r9,1),%rdi  // rewind
+
+	movq	%rax,%xmm1
+	pxor	%xmm0,%xmm0
+	pshufd	$0,%xmm1,%xmm1
+	movq	40(%rsp),%rsi  // restore %rsp
+.cfi_def_cfa	%rsi,8
+	jmp	.Lsqr8x_cond_copy
+
+.align	32
+.Lsqr8x_cond_copy:
+	movdqa	0(%rbx),%xmm2
+	movdqa	16(%rbx),%xmm3
+	leaq	32(%rbx),%rbx
+	movdqu	0(%rdi),%xmm4
+	movdqu	16(%rdi),%xmm5
+	leaq	32(%rdi),%rdi
+	movdqa	%xmm0,-32(%rbx)  // zero tp
+	movdqa	%xmm0,-16(%rbx)
+	movdqa	%xmm0,-32(%rbx,%rdx,1)
+	movdqa	%xmm0,-16(%rbx,%rdx,1)
+	pcmpeqd	%xmm1,%xmm0
+	pand	%xmm1,%xmm2
+	pand	%xmm1,%xmm3
+	pand	%xmm0,%xmm4
+	pand	%xmm0,%xmm5
+	pxor	%xmm0,%xmm0
+	por	%xmm2,%xmm4
+	por	%xmm3,%xmm5
+	movdqu	%xmm4,-32(%rdi)
+	movdqu	%xmm5,-16(%rdi)
+	addq	$32,%r9
+	jnz	.Lsqr8x_cond_copy
+
+// No return value
+	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),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lsqr8x_epilogue:
+	ret
+.cfi_endproc	
+.size	bn_sqr8x_mont,.-bn_sqr8x_mont
+.globl	bn_mulx4x_mont
+.hidden bn_mulx4x_mont
+.type	bn_mulx4x_mont,@function
+.align	32
+bn_mulx4x_mont:
+.cfi_startproc	
+_CET_ENDBR
+	movq	%rsp,%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+.Lmulx4x_prologue:
+
+	shll	$3,%r9d  // convert %r9 to bytes
+	xorq	%r10,%r10
+	subq	%r9,%r10  // -%r9
+	movq	(%r8),%r8  // *n0
+	leaq	-72(%rsp,%r10,1),%rbp  // future alloca(frame+%r9+8)
+	andq	$-128,%rbp
+	movq	%rsp,%r11
+	subq	%rbp,%r11
+	andq	$-4096,%r11
+	leaq	(%r11,%rbp,1),%rsp
+	movq	(%rsp),%r10
+	cmpq	%rbp,%rsp
+	ja	.Lmulx4x_page_walk
+	jmp	.Lmulx4x_page_walk_done
+
+.align	16
+.Lmulx4x_page_walk:
+	leaq	-4096(%rsp),%rsp
+	movq	(%rsp),%r10
+	cmpq	%rbp,%rsp
+	ja	.Lmulx4x_page_walk
+.Lmulx4x_page_walk_done:
+
+	leaq	(%rdx,%r9,1),%r10
+// #############################################################
+// Stack layout
+// +0	num
+// +8	off-loaded &b[i]
+// +16	end of b[num]
+// +24	saved n0
+// +32	saved rp
+// +40	saved %rsp
+// +48	inner counter
+// +56
+// +64	tmp[num+1]
+// 
+	movq	%r9,0(%rsp)  // save %r9
+	shrq	$5,%r9
+	movq	%r10,16(%rsp)  // end of b[num]
+	subq	$1,%r9
+	movq	%r8,24(%rsp)  // save *n0
+	movq	%rdi,32(%rsp)  // save %rdi
+	movq	%rax,40(%rsp)  // save original %rsp
+.cfi_escape	0x0f,0x05,0x77,0x28,0x06,0x23,0x08
+	movq	%r9,48(%rsp)  // inner counter
+	jmp	.Lmulx4x_body
+
+.align	32
+.Lmulx4x_body:
+	leaq	8(%rdx),%rdi
+	movq	(%rdx),%rdx  // b[0], %rdx==%rdx actually
+	leaq	64+32(%rsp),%rbx
+	movq	%rdx,%r9
+
+	mulxq	0(%rsi),%r8,%rax  // a[0]*b[0]
+	mulxq	8(%rsi),%r11,%r14  // a[1]*b[0]
+	addq	%rax,%r11
+	movq	%rdi,8(%rsp)  // off-load &b[i]
+	mulxq	16(%rsi),%r12,%r13  // ...
+	adcq	%r14,%r12
+	adcq	$0,%r13
+
+	movq	%r8,%rdi  // borrow %rdi
+	imulq	24(%rsp),%r8  // "t[0]"*n0
+	xorq	%rbp,%rbp  // cf=0, of=0
+
+	mulxq	24(%rsi),%rax,%r14
+	movq	%r8,%rdx
+	leaq	32(%rsi),%rsi
+	adcxq	%rax,%r13
+	adcxq	%rbp,%r14  // cf=0
+
+	mulxq	0(%rcx),%rax,%r10
+	adcxq	%rax,%rdi  // discarded
+	adoxq	%r11,%r10
+	mulxq	8(%rcx),%rax,%r11
+	adcxq	%rax,%r10
+	adoxq	%r12,%r11
+.byte	0xc4,0x62,0xfb,0xf6,0xa1,0x10,0x00,0x00,0x00  // mulx	2*8(%rcx),%rax,%r12
+	movq	48(%rsp),%rdi  // counter value
+	movq	%r10,-32(%rbx)
+	adcxq	%rax,%r11
+	adoxq	%r13,%r12
+	mulxq	24(%rcx),%rax,%r15
+	movq	%r9,%rdx
+	movq	%r11,-24(%rbx)
+	adcxq	%rax,%r12
+	adoxq	%rbp,%r15  // of=0
+	leaq	32(%rcx),%rcx
+	movq	%r12,-16(%rbx)
+
+	jmp	.Lmulx4x_1st
+
+.align	32
+.Lmulx4x_1st:
+	adcxq	%rbp,%r15  // cf=0, modulo-scheduled
+	mulxq	0(%rsi),%r10,%rax  // a[4]*b[0]
+	adcxq	%r14,%r10
+	mulxq	8(%rsi),%r11,%r14  // a[5]*b[0]
+	adcxq	%rax,%r11
+	mulxq	16(%rsi),%r12,%rax  // ...
+	adcxq	%r14,%r12
+	mulxq	24(%rsi),%r13,%r14
+.byte	0x67,0x67
+	movq	%r8,%rdx
+	adcxq	%rax,%r13
+	adcxq	%rbp,%r14  // cf=0
+	leaq	32(%rsi),%rsi
+	leaq	32(%rbx),%rbx
+
+	adoxq	%r15,%r10
+	mulxq	0(%rcx),%rax,%r15
+	adcxq	%rax,%r10
+	adoxq	%r15,%r11
+	mulxq	8(%rcx),%rax,%r15
+	adcxq	%rax,%r11
+	adoxq	%r15,%r12
+	mulxq	16(%rcx),%rax,%r15
+	movq	%r10,-40(%rbx)
+	adcxq	%rax,%r12
+	movq	%r11,-32(%rbx)
+	adoxq	%r15,%r13
+	mulxq	24(%rcx),%rax,%r15
+	movq	%r9,%rdx
+	movq	%r12,-24(%rbx)
+	adcxq	%rax,%r13
+	adoxq	%rbp,%r15
+	leaq	32(%rcx),%rcx
+	movq	%r13,-16(%rbx)
+
+	decq	%rdi  // of=0, pass cf
+	jnz	.Lmulx4x_1st
+
+	movq	0(%rsp),%rax  // load num
+	movq	8(%rsp),%rdi  // re-load &b[i]
+	adcq	%rbp,%r15  // modulo-scheduled
+	addq	%r15,%r14
+	sbbq	%r15,%r15  // top-most carry
+	movq	%r14,-8(%rbx)
+	jmp	.Lmulx4x_outer
+
+.align	32
+.Lmulx4x_outer:
+	movq	(%rdi),%rdx  // b[i]
+	leaq	8(%rdi),%rdi  // b++
+	subq	%rax,%rsi  // rewind %rsi
+	movq	%r15,(%rbx)  // save top-most carry
+	leaq	64+32(%rsp),%rbx
+	subq	%rax,%rcx  // rewind %rcx
+
+	mulxq	0(%rsi),%r8,%r11  // a[0]*b[i]
+	xorl	%ebp,%ebp  // xor	%rbp,%rbp	# cf=0, of=0
+	movq	%rdx,%r9
+	mulxq	8(%rsi),%r14,%r12  // a[1]*b[i]
+	adoxq	-32(%rbx),%r8
+	adcxq	%r14,%r11
+	mulxq	16(%rsi),%r15,%r13  // ...
+	adoxq	-24(%rbx),%r11
+	adcxq	%r15,%r12
+	adoxq	-16(%rbx),%r12
+	adcxq	%rbp,%r13
+	adoxq	%rbp,%r13
+
+	movq	%rdi,8(%rsp)  // off-load &b[i]
+	movq	%r8,%r15
+	imulq	24(%rsp),%r8  // "t[0]"*n0
+	xorl	%ebp,%ebp  // xor	%rbp,%rbp	# cf=0, of=0
+
+	mulxq	24(%rsi),%rax,%r14
+	movq	%r8,%rdx
+	adcxq	%rax,%r13
+	adoxq	-8(%rbx),%r13
+	adcxq	%rbp,%r14
+	leaq	32(%rsi),%rsi
+	adoxq	%rbp,%r14
+
+	mulxq	0(%rcx),%rax,%r10
+	adcxq	%rax,%r15  // discarded
+	adoxq	%r11,%r10
+	mulxq	8(%rcx),%rax,%r11
+	adcxq	%rax,%r10
+	adoxq	%r12,%r11
+	mulxq	16(%rcx),%rax,%r12
+	movq	%r10,-32(%rbx)
+	adcxq	%rax,%r11
+	adoxq	%r13,%r12
+	mulxq	24(%rcx),%rax,%r15
+	movq	%r9,%rdx
+	movq	%r11,-24(%rbx)
+	leaq	32(%rcx),%rcx
+	adcxq	%rax,%r12
+	adoxq	%rbp,%r15  // of=0
+	movq	48(%rsp),%rdi  // counter value
+	movq	%r12,-16(%rbx)
+
+	jmp	.Lmulx4x_inner
+
+.align	32
+.Lmulx4x_inner:
+	mulxq	0(%rsi),%r10,%rax  // a[4]*b[i]
+	adcxq	%rbp,%r15  // cf=0, modulo-scheduled
+	adoxq	%r14,%r10
+	mulxq	8(%rsi),%r11,%r14  // a[5]*b[i]
+	adcxq	0(%rbx),%r10
+	adoxq	%rax,%r11
+	mulxq	16(%rsi),%r12,%rax  // ...
+	adcxq	8(%rbx),%r11
+	adoxq	%r14,%r12
+	mulxq	24(%rsi),%r13,%r14
+	movq	%r8,%rdx
+	adcxq	16(%rbx),%r12
+	adoxq	%rax,%r13
+	adcxq	24(%rbx),%r13
+	adoxq	%rbp,%r14  // of=0
+	leaq	32(%rsi),%rsi
+	leaq	32(%rbx),%rbx
+	adcxq	%rbp,%r14  // cf=0
+
+	adoxq	%r15,%r10
+	mulxq	0(%rcx),%rax,%r15
+	adcxq	%rax,%r10
+	adoxq	%r15,%r11
+	mulxq	8(%rcx),%rax,%r15
+	adcxq	%rax,%r11
+	adoxq	%r15,%r12
+	mulxq	16(%rcx),%rax,%r15
+	movq	%r10,-40(%rbx)
+	adcxq	%rax,%r12
+	adoxq	%r15,%r13
+	mulxq	24(%rcx),%rax,%r15
+	movq	%r9,%rdx
+	movq	%r11,-32(%rbx)
+	movq	%r12,-24(%rbx)
+	adcxq	%rax,%r13
+	adoxq	%rbp,%r15
+	leaq	32(%rcx),%rcx
+	movq	%r13,-16(%rbx)
+
+	decq	%rdi  // of=0, pass cf
+	jnz	.Lmulx4x_inner
+
+	movq	0(%rsp),%rax  // load num
+	movq	8(%rsp),%rdi  // re-load &b[i]
+	adcq	%rbp,%r15  // modulo-scheduled
+	subq	0(%rbx),%rbp  // pull top-most carry
+	adcq	%r15,%r14
+	sbbq	%r15,%r15  // top-most carry
+	movq	%r14,-8(%rbx)
+
+	cmpq	16(%rsp),%rdi
+	jne	.Lmulx4x_outer
+
+	leaq	64(%rsp),%rbx
+	subq	%rax,%rcx  // rewind %rcx
+	negq	%r15
+	movq	%rax,%rdx
+	shrq	$3+2,%rax  // %cf=0
+	movq	32(%rsp),%rdi  // restore rp
+	jmp	.Lmulx4x_sub
+
+.align	32
+.Lmulx4x_sub:
+	movq	0(%rbx),%r11
+	movq	8(%rbx),%r12
+	movq	16(%rbx),%r13
+	movq	24(%rbx),%r14
+	leaq	32(%rbx),%rbx
+	sbbq	0(%rcx),%r11
+	sbbq	8(%rcx),%r12
+	sbbq	16(%rcx),%r13
+	sbbq	24(%rcx),%r14
+	leaq	32(%rcx),%rcx
+	movq	%r11,0(%rdi)
+	movq	%r12,8(%rdi)
+	movq	%r13,16(%rdi)
+	movq	%r14,24(%rdi)
+	leaq	32(%rdi),%rdi
+	decq	%rax  // preserves %cf
+	jnz	.Lmulx4x_sub
+
+	sbbq	$0,%r15  // top-most carry
+	leaq	64(%rsp),%rbx
+	subq	%rdx,%rdi  // rewind
+
+	movq	%r15,%xmm1
+	pxor	%xmm0,%xmm0
+	pshufd	$0,%xmm1,%xmm1
+	movq	40(%rsp),%rsi  // restore %rsp
+.cfi_def_cfa	%rsi,8
+	jmp	.Lmulx4x_cond_copy
+
+.align	32
+.Lmulx4x_cond_copy:
+	movdqa	0(%rbx),%xmm2
+	movdqa	16(%rbx),%xmm3
+	leaq	32(%rbx),%rbx
+	movdqu	0(%rdi),%xmm4
+	movdqu	16(%rdi),%xmm5
+	leaq	32(%rdi),%rdi
+	movdqa	%xmm0,-32(%rbx)  // zero tp
+	movdqa	%xmm0,-16(%rbx)
+	pcmpeqd	%xmm1,%xmm0
+	pand	%xmm1,%xmm2
+	pand	%xmm1,%xmm3
+	pand	%xmm0,%xmm4
+	pand	%xmm0,%xmm5
+	pxor	%xmm0,%xmm0
+	por	%xmm2,%xmm4
+	por	%xmm3,%xmm5
+	movdqu	%xmm4,-32(%rdi)
+	movdqu	%xmm5,-16(%rdi)
+	subq	$32,%rdx
+	jnz	.Lmulx4x_cond_copy
+
+	movq	%rdx,(%rbx)
+
+// No return value
+	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),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	leaq	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
+.Lmulx4x_epilogue:
+	ret
+.cfi_endproc	
+.size	bn_mulx4x_mont,.-bn_mulx4x_mont
+.byte	77,111,110,116,103,111,109,101,114,121,32,77,117,108,116,105,112,108,105,99,97,116,105,111,110,32,102,111,114,32,120,56,54,95,54,52,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0
+.align	16
+#endif