michael@0: /* This Source Code Form is subject to the terms of the Mozilla Public
michael@0:  * License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0:  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
michael@0: #include <regdef.h>
michael@0:         .set    noreorder
michael@0:         .set    noat
michael@0: 
michael@0:         .section        .text, 1, 0x00000006, 4, 4
michael@0: .text:
michael@0:         .section        .text
michael@0: 
michael@0:         .ent    s_mpv_mul_d_add
michael@0:         .globl  s_mpv_mul_d_add
michael@0: 
michael@0: s_mpv_mul_d_add: 
michael@0:  #/* c += a * b */
michael@0:  #void s_mpv_mul_d_add(const mp_digit *a, mp_size a_len, mp_digit b, 
michael@0:  #			      mp_digit *c)
michael@0:  #{
michael@0:  #  mp_digit   a0, a1;	regs a4, a5
michael@0:  #  mp_digit   c0, c1;  regs a6, a7
michael@0:  #  mp_digit   cy = 0;  reg t2
michael@0:  #  mp_word    w0, w1;  regs t0, t1
michael@0:  #
michael@0:  #  if (a_len) {
michael@0: 	beq	a1,zero,.L.1
michael@0: 	move	t2,zero		# cy = 0
michael@0: 	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
michael@0: 	dsrl32	a2,a2,0		# This clears the upper 32 bits.
michael@0:  #    a0 = a[0];
michael@0: 	lwu	a4,0(a0)
michael@0:  #    w0 = ((mp_word)b * a0);
michael@0: 	dmultu	a2,a4
michael@0:  #    if (--a_len) {
michael@0: 	addiu	a1,a1,-1
michael@0: 	beq	a1,zero,.L.2
michael@0:  #      while (a_len >= 2) {
michael@0: 	sltiu	t3,a1,2
michael@0: 	bne	t3,zero,.L.3
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0: .L.4:
michael@0:  #	  a_len -= 2;
michael@0:         addiu	a1,a1,-2
michael@0:  #	  c0     = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #	  w0    += cy;
michael@0: 	mflo	t0
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5			#
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  a0     = a[2];
michael@0: 	lwu	a4,8(a0)
michael@0:  #	  a     += 2;
michael@0: 	addiu	a0,a0,8
michael@0:  #	  c1     = c[1];
michael@0: 	lwu	a7,4(a3)
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  w1    += c1;
michael@0: 	daddu	t1,t1,a7
michael@0:  #	  w0     = (mp_word)b * a0;
michael@0: 	dmultu	a2,a4			#
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  c     += 2;
michael@0: 	addiu	a3,a3,8
michael@0: 	sltiu	t3,a1,2
michael@0: 	beq	t3,zero,.L.4
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0:  #      }
michael@0: .L.3:
michael@0:  #      c0       = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #      w0      += cy;
michael@0:  #      if (a_len) {
michael@0: 	mflo	t0
michael@0: 	beq	a1,zero,.L.5
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6		#
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  c1     = c[1];
michael@0: 	lwu	a7,4(a3)
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  w1    += c1;
michael@0: 	daddu	t1,t1,a7
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  c     += 1;
michael@0: 	b	.L.6
michael@0: 	addiu	a3,a3,4
michael@0:  #      } else {
michael@0: .L.5:
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	b	.L.6
michael@0: 	dsrl32	t2,t0,0
michael@0:  #      }
michael@0:  #    } else {
michael@0: .L.2:
michael@0:  #      c0     = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #      w0    += c0;
michael@0: 	mflo	t0
michael@0: 	daddu	t0,t0,a6
michael@0:  #      c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #      cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #    }
michael@0: .L.6:
michael@0:  #    c[1] = cy;
michael@0: 	jr	ra
michael@0: 	sw	t2,4(a3)
michael@0:  #  }
michael@0: .L.1:
michael@0: 	jr	ra
michael@0: 	nop
michael@0:  #}
michael@0:  #
michael@0:         .end    s_mpv_mul_d_add
michael@0: 
michael@0:         .ent    s_mpv_mul_d_add_prop
michael@0:         .globl  s_mpv_mul_d_add_prop
michael@0: 
michael@0: s_mpv_mul_d_add_prop: 
michael@0:  #/* c += a * b */
michael@0:  #void s_mpv_mul_d_add_prop(const mp_digit *a, mp_size a_len, mp_digit b, 
michael@0:  #			      mp_digit *c)
michael@0:  #{
michael@0:  #  mp_digit   a0, a1;	regs a4, a5
michael@0:  #  mp_digit   c0, c1;  regs a6, a7
michael@0:  #  mp_digit   cy = 0;  reg t2
michael@0:  #  mp_word    w0, w1;  regs t0, t1
michael@0:  #
michael@0:  #  if (a_len) {
michael@0: 	beq	a1,zero,.M.1
michael@0: 	move	t2,zero		# cy = 0
michael@0: 	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
michael@0: 	dsrl32	a2,a2,0		# This clears the upper 32 bits.
michael@0:  #    a0 = a[0];
michael@0: 	lwu	a4,0(a0)
michael@0:  #    w0 = ((mp_word)b * a0);
michael@0: 	dmultu	a2,a4
michael@0:  #    if (--a_len) {
michael@0: 	addiu	a1,a1,-1
michael@0: 	beq	a1,zero,.M.2
michael@0:  #      while (a_len >= 2) {
michael@0: 	sltiu	t3,a1,2
michael@0: 	bne	t3,zero,.M.3
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0: .M.4:
michael@0:  #	  a_len -= 2;
michael@0:         addiu	a1,a1,-2
michael@0:  #	  c0     = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #	  w0    += cy;
michael@0: 	mflo	t0
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5			#
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  a0     = a[2];
michael@0: 	lwu	a4,8(a0)
michael@0:  #	  a     += 2;
michael@0: 	addiu	a0,a0,8
michael@0:  #	  c1     = c[1];
michael@0: 	lwu	a7,4(a3)
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  w1    += c1;
michael@0: 	daddu	t1,t1,a7
michael@0:  #	  w0     = (mp_word)b * a0;
michael@0: 	dmultu	a2,a4			#
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  c     += 2;
michael@0: 	addiu	a3,a3,8
michael@0: 	sltiu	t3,a1,2
michael@0: 	beq	t3,zero,.M.4
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0:  #      }
michael@0: .M.3:
michael@0:  #      c0       = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #      w0      += cy;
michael@0:  #      if (a_len) {
michael@0: 	mflo	t0
michael@0: 	beq	a1,zero,.M.5
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6		#
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  c1     = c[1];
michael@0: 	lwu	a7,4(a3)
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  w1    += c1;
michael@0: 	daddu	t1,t1,a7
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  c     += 1;
michael@0: 	b	.M.6
michael@0: 	addiu	a3,a3,8
michael@0:  #      } else {
michael@0: .M.5:
michael@0:  #	  w0    += c0;
michael@0: 	daddu	t0,t0,a6
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0: 	b	.M.6
michael@0: 	addiu	a3,a3,4
michael@0:  #      }
michael@0:  #    } else {
michael@0: .M.2:
michael@0:  #      c0     = c[0];
michael@0: 	lwu	a6,0(a3)
michael@0:  #      w0    += c0;
michael@0: 	mflo	t0
michael@0: 	daddu	t0,t0,a6
michael@0:  #      c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #      cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0: 	addiu	a3,a3,4
michael@0:  #    }
michael@0: .M.6:
michael@0: 
michael@0:  #    while (cy) {
michael@0: 	beq	t2,zero,.M.1
michael@0: 	nop
michael@0: .M.7:
michael@0:  #      mp_word w = (mp_word)*c + cy;
michael@0: 	lwu	a6,0(a3)
michael@0: 	daddu	t2,t2,a6
michael@0:  #      *c++ = ACCUM(w);
michael@0: 	sw	t2,0(a3)
michael@0:  #      cy = CARRYOUT(w);
michael@0: 	dsrl32	t2,t2,0
michael@0: 	bne	t2,zero,.M.7
michael@0: 	addiu	a3,a3,4
michael@0: 
michael@0:  #  }
michael@0: .M.1:
michael@0: 	jr	ra
michael@0: 	nop
michael@0:  #}
michael@0:  #
michael@0:         .end    s_mpv_mul_d_add_prop
michael@0: 
michael@0:         .ent    s_mpv_mul_d
michael@0:         .globl  s_mpv_mul_d
michael@0: 
michael@0: s_mpv_mul_d: 
michael@0:  #/* c = a * b */
michael@0:  #void s_mpv_mul_d(const mp_digit *a, mp_size a_len, mp_digit b, 
michael@0:  #			      mp_digit *c)
michael@0:  #{
michael@0:  #  mp_digit   a0, a1;	regs a4, a5
michael@0:  #  mp_digit   cy = 0;  reg t2
michael@0:  #  mp_word    w0, w1;  regs t0, t1
michael@0:  #
michael@0:  #  if (a_len) {
michael@0: 	beq	a1,zero,.N.1
michael@0: 	move	t2,zero		# cy = 0
michael@0: 	dsll32	a2,a2,0		# "b" is sometimes negative (?!?!)
michael@0: 	dsrl32	a2,a2,0		# This clears the upper 32 bits.
michael@0:  #    a0 = a[0];
michael@0: 	lwu	a4,0(a0)
michael@0:  #    w0 = ((mp_word)b * a0);
michael@0: 	dmultu	a2,a4
michael@0:  #    if (--a_len) {
michael@0: 	addiu	a1,a1,-1
michael@0: 	beq	a1,zero,.N.2
michael@0:  #      while (a_len >= 2) {
michael@0: 	sltiu	t3,a1,2
michael@0: 	bne	t3,zero,.N.3
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0: .N.4:
michael@0:  #	  a_len -= 2;
michael@0:         addiu	a1,a1,-2
michael@0:  #	  w0    += cy;
michael@0: 	mflo	t0
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5	
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  a0     = a[2];
michael@0: 	lwu	a4,8(a0)
michael@0:  #	  a     += 2;
michael@0: 	addiu	a0,a0,8
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  w0     = (mp_word)b * a0;
michael@0: 	dmultu	a2,a4	
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  c     += 2;
michael@0: 	addiu	a3,a3,8
michael@0: 	sltiu	t3,a1,2
michael@0: 	beq	t3,zero,.N.4
michael@0:  #	  a1     = a[1];
michael@0: 	lwu	a5,4(a0)
michael@0:  #      }
michael@0: .N.3:
michael@0:  #      w0      += cy;
michael@0:  #      if (a_len) {
michael@0: 	mflo	t0
michael@0: 	beq	a1,zero,.N.5
michael@0: 	daddu	t0,t0,t2
michael@0:  #	  w1     = (mp_word)b * a1; 
michael@0: 	dmultu	a2,a5			#
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  w1    += cy;
michael@0: 	mflo	t1
michael@0: 	daddu	t1,t1,t2
michael@0:  #	  c[1]   = ACCUM(w1);
michael@0: 	sw	t1,4(a3)
michael@0:  #	  cy     = CARRYOUT(w1);
michael@0: 	dsrl32	t2,t1,0
michael@0:  #	  c     += 1;
michael@0: 	b	.N.6
michael@0: 	addiu	a3,a3,4
michael@0:  #      } else {
michael@0: .N.5:
michael@0:  #	  c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #	  cy     = CARRYOUT(w0);
michael@0: 	b	.N.6
michael@0: 	dsrl32	t2,t0,0
michael@0:  #      }
michael@0:  #    } else {
michael@0: .N.2:
michael@0: 	mflo	t0
michael@0:  #      c[0]   = ACCUM(w0);
michael@0: 	sw	t0,0(a3)
michael@0:  #      cy     = CARRYOUT(w0);
michael@0: 	dsrl32	t2,t0,0
michael@0:  #    }
michael@0: .N.6:
michael@0:  #    c[1] = cy;
michael@0: 	jr	ra
michael@0: 	sw	t2,4(a3)
michael@0:  #  }
michael@0: .N.1:
michael@0: 	jr	ra
michael@0: 	nop
michael@0:  #}
michael@0:  #
michael@0:         .end    s_mpv_mul_d
michael@0: 
michael@0: 
michael@0:         .ent    s_mpv_sqr_add_prop
michael@0:         .globl  s_mpv_sqr_add_prop
michael@0:  #void   s_mpv_sqr_add_prop(const mp_digit *a, mp_size a_len, mp_digit *sqrs);
michael@0:  #	registers
michael@0:  #	a0		*a
michael@0:  #	a1		a_len
michael@0:  #	a2		*sqr
michael@0:  #	a3		digit from *a, a_i
michael@0:  #	a4		square of digit from a
michael@0:  #	a5,a6		next 2 digits in sqr
michael@0:  #	a7,t0		carry 
michael@0: s_mpv_sqr_add_prop:
michael@0: 	move	a7,zero
michael@0: 	move	t0,zero
michael@0: 	lwu	a3,0(a0)
michael@0: 	addiu	a1,a1,-1	# --a_len
michael@0: 	dmultu	a3,a3
michael@0: 	beq	a1,zero,.P.3	# jump if we've already done the only sqr
michael@0: 	addiu	a0,a0,4		# ++a
michael@0: .P.2:
michael@0:         lwu	a5,0(a2)
michael@0:         lwu	a6,4(a2)
michael@0: 	addiu	a2,a2,8		# sqrs += 2;
michael@0: 	dsll32	a6,a6,0
michael@0: 	daddu	a5,a5,a6
michael@0: 	lwu	a3,0(a0)
michael@0: 	addiu	a0,a0,4		# ++a
michael@0: 	mflo	a4
michael@0: 	daddu	a6,a5,a4
michael@0: 	sltu	a7,a6,a5	# a7 = a6 < a5	detect overflow
michael@0: 	dmultu	a3,a3
michael@0: 	daddu	a4,a6,t0
michael@0: 	sltu	t0,a4,a6
michael@0: 	add	t0,t0,a7
michael@0: 	sw	a4,-8(a2)
michael@0: 	addiu	a1,a1,-1	# --a_len
michael@0: 	dsrl32	a4,a4,0
michael@0: 	bne	a1,zero,.P.2	# loop if a_len > 0
michael@0: 	sw	a4,-4(a2)
michael@0: .P.3:
michael@0:         lwu	a5,0(a2)
michael@0:         lwu	a6,4(a2)
michael@0: 	addiu	a2,a2,8		# sqrs += 2;
michael@0: 	dsll32	a6,a6,0
michael@0: 	daddu	a5,a5,a6
michael@0: 	mflo	a4
michael@0: 	daddu	a6,a5,a4
michael@0: 	sltu	a7,a6,a5	# a7 = a6 < a5	detect overflow
michael@0: 	daddu	a4,a6,t0
michael@0: 	sltu	t0,a4,a6
michael@0: 	add	t0,t0,a7
michael@0: 	sw	a4,-8(a2)
michael@0: 	beq	t0,zero,.P.9	# jump if no carry
michael@0: 	dsrl32	a4,a4,0
michael@0: .P.8:
michael@0: 	sw	a4,-4(a2)
michael@0: 	/* propagate final carry */
michael@0: 	lwu	a5,0(a2)
michael@0: 	daddu	a6,a5,t0
michael@0: 	sltu	t0,a6,a5
michael@0: 	bne	t0,zero,.P.8	# loop if carry persists
michael@0: 	addiu	a2,a2,4		# sqrs++
michael@0: .P.9:
michael@0: 	jr	ra
michael@0: 	sw	a4,-4(a2)
michael@0: 
michael@0:         .end    s_mpv_sqr_add_prop