STPW.P	(R6, R7), 8(R10)
		VMOV	R7, V2.S[0]
		WORD	$0x4E030042 //TBL	V3.B16, [V2.B16], V2.B16
		AESE	V0.B16, V2.B16
		EORW	R13, R0
		LSLW	$1, R13
		SUBS	$1, R8
		VMOV	V2.S[0], R9
		EORW	R9, R0
		EORW	R0, R1
		EORW	R1, R2
		EORW	R2, R3
		VMOV	R3, V2.S[0]
		WORD	$0x4E040042 //TBL	V3.B16, [V2.B16], V2.B16
		AESE	V0.B16, V2.B16
		VMOV	V2.S[0], R9
		EORW	R9, R4
		EORW	R4, R5
		EORW	R5, R6

	ROUND2(Rd, Re, Ra, Rb, Rc)
	ROUND2(Rc, Rd, Re, Ra, Rb)
	ROUND2(Rb, Rc, Rd, Re, Ra)
	SUB.S	$1, Rctr
	BNE	loop2

	MOVW	$0x8F1BBCDC, Rconst
	MOVW	$4, Rctr
loop3:	ROUND3(Ra, Rb, Rc, Rd, Re)
	ROUND3(Re, Ra, Rb, Rc, Rd)
	ROUND3(Rd, Re, Ra, Rb, Rc)
	ROUND3(Rc, Rd, Re, Ra, Rb)
	ROUND3(Rb, Rc, Rd, Re, Ra)
	SUB.S	$1, Rctr
	BNE	loop3

	MOVW	$0xCA62C1D6, Rconst
	MOVW	$4, Rctr

// Code generated by command: go run fe_amd64_asm.go -out ../fe_amd64.s -stubs ../fe_amd64.go -pkg field. DO NOT EDIT.

//go:build !purego

#include "textflag.h"

// func feMul(out *Element, a *Element, b *Element)
TEXT ·feMul(SB), NOSPLIT, $0-24
	MOVQ a+8(FP), CX
	MOVQ b+16(FP), BX

	// r0 = a0×b0
	MOVQ (CX), AX
	MULQ (BX)
	MOVQ AX, DI
	MOVQ DX, SI

	// r0 += 19×a1×b4
	MOVQ   8(CX), AX
	IMUL3Q $0x13, AX, AX

	MOVL	$0, gobuf_sp(BX)	// clear to help garbage collector
	MOVL	$0, gobuf_ret(BX)
	MOVL	$0, gobuf_ctxt(BX)
	MOVL	gobuf_pc(BX), BX
	JMP	BX

// func mcall(fn func(*g))
// Switch to m->g0's stack, call fn(g).
// Fn must never return. It should gogo(&g->sched)
// to keep running g.
TEXT runtime·mcall(SB), NOSPLIT, $0-4
	MOVL	fn+0(FP), DI

	get_tls(DX)
	MOVL	g(DX), AX	// save state in g->sched

	MOVD	$0, gobuf_lr(R5)
	MOVD	$0, gobuf_ctxt(R5)
	CMP	R0, R0 // set condition codes for == test, needed by stack split
	MOVD	gobuf_pc(R5), R6
	BR	(R6)

// void mcall(fn func(*g))
// Switch to m->g0's stack, call fn(g).
// Fn must never return.  It should gogo(&g->sched)
// to keep running g.
TEXT runtime·mcall(SB), NOSPLIT, $-8-8
	// Save caller state in g->sched
	MOVD	R15, (g_sched+gobuf_sp)(g)

//go:build !purego

// This is an optimized implementation of AES-GCM using AES-NI and CLMUL-NI
// The implementation uses some optimization as described in:
// [1] Gueron, S., Kounavis, M.E.: Intel® Carry-Less Multiplication
//     Instruction and its Usage for Computing the GCM Mode rev. 2.02
// [2] Gueron, S., Krasnov, V.: Speeding up Counter Mode in Software and
//     Hardware

#include "textflag.h"

#define B0 X0
#define B1 X1
#define B2 X2

	MOVW	p+4(FP), Rdata	// pointer to the data
	MOVW	p_len+8(FP), Rt0	// number of bytes
	ADD	Rdata, Rt0
	MOVW	Rt0, p_end	// pointer to end of data

loop:
	MOVW	Rdata, p_data	// Save Rdata
	AND.S	$3, Rdata, Rt0	// TST $3, Rdata not working see issue 5921
	BEQ	aligned			// aligned detected - skip copy

	// Copy the unaligned source data into the aligned temporary buffer

	SRD     R9, R11, R11    // x[len(z)-2] >> s
	SLD     $3, R8, R12
	MOVD    (R6)(R12), R12
	SLD     R5, R12, R12    // x[len(z)-1]<<ŝ
	OR      R12, R11, R11   // x[len(z)-2]>>s | x[len(z)-1]<<ŝ
	MOVD    R11, (R3)(R10)  // z[len(z)-2]=x[len(z)-2]>>s | x[len(z)-1]<<ŝ
loopexit:
	ADD     $-1, R4
	SLD     $3, R4
	MOVD    (R6)(R4), R5
	SRD     R9, R5, R5      // x[len(z)-1]>>s
	MOVD    R5, (R3)(R4)    // z[len(z)-1]=x[len(z)-1]>>s

TEXT runtime·usleep(SB),NOSPLIT,$16-4
	MOVW	usec+0(FP), R3

	// Use magic constant 0x8637bd06 and shift right 51
	// to perform usec/1000000.
	MOVD	$0x8637bd06, R4
	MULLD	R3, R4, R4	// Convert usec to S.
	SRD	$51, R4, R4
	MOVD	R4, 8(R1)	// Store to tv_sec

	MOVD	$1000000, R5
	MULLW	R4, R5, R5	// Convert tv_sec back into uS
	SUB	R5, R3, R5	// Compute remainder uS.
	MULLD	$1000, R5, R5	// Convert to nsec

#include "funcdata.h"

// bool armcas(int32 *val, int32 old, int32 new)
// Atomically:
//	if(*val == old){
//		*val = new;
//		return 1;
//	}else
//		return 0;
//
// To implement ·cas in sys_$GOOS_arm.s
// using the native instructions, use:
//
//	TEXT ·cas(SB),NOSPLIT,$0
//		B	·armcas(SB)
//
TEXT ·armcas(SB),NOSPLIT,$0-13
	MOVW	ptr+0(FP), R1
	MOVW	old+4(FP), R2
	MOVW	new+8(FP), R3
casl:
	LDREX	(R1), R0