-.align 16
-.Lsqr4x_shift_n_add:
- lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[1] # | t[2*i]>>63
- mov -16($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch
- mov $A0[1],$shift # shift=t[2*i+1]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- mov -8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch
- adc %rax,$S[0]
- mov -8($aptr,$i),%rax # a[i+1] # prefetch
- mov $S[0],-32($tptr,$i,2)
- adc %rdx,$S[1]
-
- lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift
- mov $S[1],-24($tptr,$i,2)
- sbb $carry,$carry # mov cf,$carry
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[3] # | t[2*i]>>63
- mov 0($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch
- mov $A0[1],$shift # shift=t[2*i+1]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- mov 8($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch
- adc %rax,$S[2]
- mov 0($aptr,$i),%rax # a[i+1] # prefetch
- mov $S[2],-16($tptr,$i,2)
- adc %rdx,$S[3]
-
- lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
- mov $S[3],-8($tptr,$i,2)
- sbb $carry,$carry # mov cf,$carry
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[1] # | t[2*i]>>63
- mov 16($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch
- mov $A0[1],$shift # shift=t[2*i+1]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- mov 24($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch
- adc %rax,$S[0]
- mov 8($aptr,$i),%rax # a[i+1] # prefetch
- mov $S[0],0($tptr,$i,2)
- adc %rdx,$S[1]
-
- lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift
- mov $S[1],8($tptr,$i,2)
- sbb $carry,$carry # mov cf,$carry
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[3] # | t[2*i]>>63
- mov 32($tptr,$i,2),$A0[0] # t[2*i+2] # prefetch
- mov $A0[1],$shift # shift=t[2*i+1]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- mov 40($tptr,$i,2),$A0[1] # t[2*i+2+1] # prefetch
- adc %rax,$S[2]
- mov 16($aptr,$i),%rax # a[i+1] # prefetch
- mov $S[2],16($tptr,$i,2)
- adc %rdx,$S[3]
- mov $S[3],24($tptr,$i,2)
- sbb $carry,$carry # mov cf,$carry
- add \$32,$i
- jnz .Lsqr4x_shift_n_add
-
- lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[1] # | t[2*i]>>63
- mov -16($tptr),$A0[0] # t[2*i+2] # prefetch
- mov $A0[1],$shift # shift=t[2*i+1]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- mov -8($tptr),$A0[1] # t[2*i+2+1] # prefetch
- adc %rax,$S[0]
- mov -8($aptr),%rax # a[i+1] # prefetch
- mov $S[0],-32($tptr)
- adc %rdx,$S[1]
-
- lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1|shift
- mov $S[1],-24($tptr)
- sbb $carry,$carry # mov cf,$carry
- shr \$63,$A0[0]
- lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
- shr \$63,$A0[1]
- or $A0[0],$S[3] # | t[2*i]>>63
- mul %rax # a[i]*a[i]
- neg $carry # mov $carry,cf
- adc %rax,$S[2]
- adc %rdx,$S[3]
- mov $S[2],-16($tptr)
- mov $S[3],-8($tptr)
-___
-}\f
-##############################################################
-# Montgomery reduction part, "word-by-word" algorithm.
-#
-{
-my ($topbit,$nptr)=("%rbp",$aptr);
-my ($m0,$m1)=($a0,$a1);
-my @Ni=("%rbx","%r9");
-$code.=<<___;
- mov 40(%rsp),$nptr # restore $nptr
- mov 48(%rsp),$n0 # restore *n0
- xor $j,$j
- mov $num,0(%rsp) # save $num
- sub $num,$j # $j=-$num
- mov 64(%rsp),$A0[0] # t[0] # modsched #
- mov $n0,$m0 # # modsched #
- lea 64(%rsp,$num,2),%rax # end of t[] buffer
- lea 64(%rsp,$num),$tptr # end of t[] window
- mov %rax,8(%rsp) # save end of t[] buffer
- lea ($nptr,$num),$nptr # end of n[] buffer
- xor $topbit,$topbit # $topbit=0
-
- mov 0($nptr,$j),%rax # n[0] # modsched #
- mov 8($nptr,$j),$Ni[1] # n[1] # modsched #
- imulq $A0[0],$m0 # m0=t[0]*n0 # modsched #
- mov %rax,$Ni[0] # # modsched #
- jmp .Lsqr4x_mont_outer