sparcv8plus.s

mediastreamer2是开源的网络传输媒体流的库

	add	c_3,t_2,c_3
	srlx	t_1,32,c_12	!=
	stuw	t_1,rp(1)	!r[1]=c2;
	or	c_12,c_3,c_12

	mulx	a_2,b_0,t_1	!mul_add_c(a[2],b[0],c3,c1,c2);
	addcc	c_12,t_1,c_12	!=
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	bp(2),b_2	!=
	mulx	a_1,b_1,t_1	!mul_add_c(a[1],b[1],c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3	!=
	lduw	bp(3),b_3
	mulx	a_0,b_2,t_1	!mul_add_c(a[0],b[2],c3,c1,c2);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8	!=
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(2)	!r[2]=c3;
	or	c_12,c_3,c_12	!=

	mulx	a_0,b_3,t_1	!mul_add_c(a[0],b[3],c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8	!=
	add	c_3,t_2,c_3
	mulx	a_1,b_2,t_1	!mul_add_c(a[1],b[2],c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8	!=
	add	c_3,t_2,c_3
	lduw	ap(3),a_3
	mulx	a_2,b_1,t_1	!mul_add_c(a[2],b[1],c1,c2,c3);
	addcc	c_12,t_1,c_12	!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_3,b_0,t_1	!mul_add_c(a[3],b[0],c1,c2,c3);!=
	addcc	c_12,t_1,t_1	!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(3)	!=!r[3]=c1;
	or	c_12,c_3,c_12

	mulx	a_3,b_1,t_1	!mul_add_c(a[3],b[1],c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3		!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_2,b_2,t_1	!mul_add_c(a[2],b[2],c2,c3,c1);
	addcc	c_12,t_1,c_12	!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_1,b_3,t_1	!mul_add_c(a[1],b[3],c2,c3,c1);
	addcc	c_12,t_1,t_1	!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(4)	!=!r[4]=c2;
	or	c_12,c_3,c_12

	mulx	a_2,b_3,t_1	!mul_add_c(a[2],b[3],c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3		!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_3,b_2,t_1	!mul_add_c(a[3],b[2],c3,c1,c2);
	addcc	c_12,t_1,t_1	!=
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(5)	!=!r[5]=c3;
	or	c_12,c_3,c_12

	mulx	a_3,b_3,t_1	!mul_add_c(a[3],b[3],c1,c2,c3);
	addcc	c_12,t_1,t_1
	srlx	t_1,32,c_12	!=
	stuw	t_1,rp(6)	!r[6]=c1;
	stuw	c_12,rp(7)	!r[7]=c2;
	
	ret
	restore	%g0,%g0,%o0

.type	bn_mul_comba4,#function
.size	bn_mul_comba4,(.-bn_mul_comba4)

.align	32

.global bn_sqr_comba8
bn_sqr_comba8:
	save	%sp,FRAME_SIZE,%sp
	mov	1,t_2
	lduw	ap(0),a_0
	sllx	t_2,32,t_2
	lduw	ap(1),a_1
	mulx	a_0,a_0,t_1	!sqr_add_c(a,0,c1,c2,c3);
	srlx	t_1,32,c_12
	stuw	t_1,rp(0)	!r[0]=c1;

	lduw	ap(2),a_2
	mulx	a_0,a_1,t_1	!=!sqr_add_c2(a,1,0,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(1)	!r[1]=c2;
	or	c_12,c_3,c_12

	mulx	a_2,a_0,t_1	!sqr_add_c2(a,2,0,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(3),a_3
	mulx	a_1,a_1,t_1	!sqr_add_c(a,1,c3,c1,c2);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(2)	!r[2]=c3;
	or	c_12,c_3,c_12

	mulx	a_0,a_3,t_1	!sqr_add_c2(a,3,0,c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(4),a_4
	mulx	a_1,a_2,t_1	!sqr_add_c2(a,2,1,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	st	t_1,rp(3)	!r[3]=c1;
	or	c_12,c_3,c_12

	mulx	a_4,a_0,t_1	!sqr_add_c2(a,4,0,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_3,a_1,t_1	!sqr_add_c2(a,3,1,c2,c3,c1);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(5),a_5
	mulx	a_2,a_2,t_1	!sqr_add_c(a,2,c2,c3,c1);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(4)	!r[4]=c2;
	or	c_12,c_3,c_12

	mulx	a_0,a_5,t_1	!sqr_add_c2(a,5,0,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_1,a_4,t_1	!sqr_add_c2(a,4,1,c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(6),a_6
	mulx	a_2,a_3,t_1	!sqr_add_c2(a,3,2,c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(5)	!r[5]=c3;
	or	c_12,c_3,c_12

	mulx	a_6,a_0,t_1	!sqr_add_c2(a,6,0,c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_5,a_1,t_1	!sqr_add_c2(a,5,1,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_4,a_2,t_1	!sqr_add_c2(a,4,2,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(7),a_7
	mulx	a_3,a_3,t_1	!=!sqr_add_c(a,3,c1,c2,c3);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(6)	!r[6]=c1;
	or	c_12,c_3,c_12

	mulx	a_0,a_7,t_1	!sqr_add_c2(a,7,0,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_1,a_6,t_1	!sqr_add_c2(a,6,1,c2,c3,c1);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_2,a_5,t_1	!sqr_add_c2(a,5,2,c2,c3,c1);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_3,a_4,t_1	!sqr_add_c2(a,4,3,c2,c3,c1);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(7)	!r[7]=c2;
	or	c_12,c_3,c_12

	mulx	a_7,a_1,t_1	!sqr_add_c2(a,7,1,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_6,a_2,t_1	!sqr_add_c2(a,6,2,c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_5,a_3,t_1	!sqr_add_c2(a,5,3,c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_4,a_4,t_1	!sqr_add_c(a,4,c3,c1,c2);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(8)	!r[8]=c3;
	or	c_12,c_3,c_12

	mulx	a_2,a_7,t_1	!sqr_add_c2(a,7,2,c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_3,a_6,t_1	!sqr_add_c2(a,6,3,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_4,a_5,t_1	!sqr_add_c2(a,5,4,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(9)	!r[9]=c1;
	or	c_12,c_3,c_12

	mulx	a_7,a_3,t_1	!sqr_add_c2(a,7,3,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_6,a_4,t_1	!sqr_add_c2(a,6,4,c2,c3,c1);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_5,a_5,t_1	!sqr_add_c(a,5,c2,c3,c1);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(10)	!r[10]=c2;
	or	c_12,c_3,c_12

	mulx	a_4,a_7,t_1	!sqr_add_c2(a,7,4,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_5,a_6,t_1	!sqr_add_c2(a,6,5,c3,c1,c2);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(11)	!r[11]=c3;
	or	c_12,c_3,c_12

	mulx	a_7,a_5,t_1	!sqr_add_c2(a,7,5,c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_6,a_6,t_1	!sqr_add_c(a,6,c1,c2,c3);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(12)	!r[12]=c1;
	or	c_12,c_3,c_12

	mulx	a_6,a_7,t_1	!sqr_add_c2(a,7,6,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(13)	!r[13]=c2;
	or	c_12,c_3,c_12

	mulx	a_7,a_7,t_1	!sqr_add_c(a,7,c3,c1,c2);
	addcc	c_12,t_1,t_1
	srlx	t_1,32,c_12
	stuw	t_1,rp(14)	!r[14]=c3;
	stuw	c_12,rp(15)	!r[15]=c1;

	ret
	restore	%g0,%g0,%o0

.type	bn_sqr_comba8,#function
.size	bn_sqr_comba8,(.-bn_sqr_comba8)

.align	32

.global bn_sqr_comba4
/*
 * void bn_sqr_comba4(r,a)
 * BN_ULONG *r,*a;
 */
bn_sqr_comba4:
	save	%sp,FRAME_SIZE,%sp
	mov	1,t_2
	lduw	ap(0),a_0
	sllx	t_2,32,t_2
	lduw	ap(1),a_1
	mulx	a_0,a_0,t_1	!sqr_add_c(a,0,c1,c2,c3);
	srlx	t_1,32,c_12
	stuw	t_1,rp(0)	!r[0]=c1;

	lduw	ap(2),a_2
	mulx	a_0,a_1,t_1	!sqr_add_c2(a,1,0,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(1)	!r[1]=c2;
	or	c_12,c_3,c_12

	mulx	a_2,a_0,t_1	!sqr_add_c2(a,2,0,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	lduw	ap(3),a_3
	mulx	a_1,a_1,t_1	!sqr_add_c(a,1,c3,c1,c2);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(2)	!r[2]=c3;
	or	c_12,c_3,c_12

	mulx	a_0,a_3,t_1	!sqr_add_c2(a,3,0,c1,c2,c3);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_1,a_2,t_1	!sqr_add_c2(a,2,1,c1,c2,c3);
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(3)	!r[3]=c1;
	or	c_12,c_3,c_12

	mulx	a_3,a_1,t_1	!sqr_add_c2(a,3,1,c2,c3,c1);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,c_12
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	mulx	a_2,a_2,t_1	!sqr_add_c(a,2,c2,c3,c1);
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(4)	!r[4]=c2;
	or	c_12,c_3,c_12

	mulx	a_2,a_3,t_1	!sqr_add_c2(a,3,2,c3,c1,c2);
	addcc	c_12,t_1,c_12
	clr	c_3
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	addcc	c_12,t_1,t_1
	bcs,a	%xcc,.+8
	add	c_3,t_2,c_3
	srlx	t_1,32,c_12
	stuw	t_1,rp(5)	!r[5]=c3;
	or	c_12,c_3,c_12

	mulx	a_3,a_3,t_1	!sqr_add_c(a,3,c1,c2,c3);
	addcc	c_12,t_1,t_1
	srlx	t_1,32,c_12
	stuw	t_1,rp(6)	!r[6]=c1;
	stuw	c_12,rp(7)	!r[7]=c2;
	
	ret
	restore	%g0,%g0,%o0

.type	bn_sqr_comba4,#function
.size	bn_sqr_comba4,(.-bn_sqr_comba4)

.align	32