bn/asm/x86_64-mont.pl

701 my ($i,$j,$tptr)=("%rbp","%rcx",$rptr);
755 	lea	64(%rsp,$num,2),$tptr	# end of tp[] buffer, &tp[2*$num]
757 	lea	-32($tptr,$i),$tptr	# end of tp[] window, &tp[2*$num-"$i"]
765 	mov	$A0[0],-24($tptr,$i)	# t[1]
772 	mov	$A0[1],-16($tptr,$i)	# t[2]
791 	mov	$A0[0],-8($tptr,$j)	# t[3]
810 	mov	$A0[1],($tptr,$j)	# t[4]
827 	mov	$A0[0],8($tptr,$j)	# t[5]
843 	mov	$A0[1],16($tptr,$j)	# t[6]
861 	mov	$A0[0],-8($tptr,$j)	# t[7]
873 	mov	$A1[1],($tptr)		# t[8]
875 	mov	$A1[0],8($tptr)		# t[9]
881 	lea	64(%rsp,$num,2),$tptr	# end of tp[] buffer, &tp[2*$num]
883 	lea	-32($tptr,$i),$tptr	# end of tp[] window, &tp[2*$num-"$i"]
887 	mov	-24($tptr,$i),$A0[0]	# t[1]
893 	mov	$A0[0],-24($tptr,$i)	# t[1]
896 	add	-16($tptr,$i),$A0[1]	# a[2]*a[0]+t[2]
902 	mov	$A0[1],-16($tptr,$i)	# t[2]
910 	add	8($tptr,$j),$A1[0]
924 	mov	$A0[0],8($tptr,$j)	# t[3]
933 	add	($tptr,$j),$A1[1]
947 	mov	$A0[1],($tptr,$j)	# t[4]
951 	add	8($tptr,$j),$A1[0]
966 	mov	$A0[0],-8($tptr,$j)	# t[5], "preloaded t[1]" below
978 	mov	$A1[1],($tptr)		# t[6], "preloaded t[2]" below
979 	mov	$A1[0],8($tptr)		# t[7], "preloaded t[3]" below
986 	lea	64(%rsp,$num,2),$tptr	# end of tp[] buffer, &tp[2*$num]
988 	lea	-32($tptr,$i),$tptr	# end of tp[] window, &tp[2*$num-"$i"]
997 	mov	$A0[0],-24($tptr)	# t[1]
1006 	mov	$A0[1],-16($tptr)	# t[2]
1022 	mov	$A0[0],-8($tptr)	# t[3]
1032 	mov	$A1[1],($tptr)		# t[4]
1033 	mov	$A1[0],8($tptr)		# t[5]
1048 	mov	%rax,8($tptr)		# t[5]
1049 	mov	%rdx,16($tptr)		# t[6]
1050 	mov	$carry,24($tptr)	# t[7]
1053 	lea	64(%rsp,$num,2),$tptr
1055 	 mov	-24($tptr,$i,2),$A0[1]	# t[1]
1062 	 mov	-16($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1066 	 mov	-8($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1069 	mov	$S[0],-32($tptr,$i,2)
1073 	 mov	$S[1],-24($tptr,$i,2)
1079 	 mov	0($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1083 	 mov	8($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1086 	mov	$S[2],-16($tptr,$i,2)
1089 	mov	$S[3],-40($tptr,$i,2)
1100 	 mov	-16($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1104 	 mov	-8($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1107 	mov	$S[0],-32($tptr,$i,2)
1111 	 mov	$S[1],-24($tptr,$i,2)
1117 	 mov	0($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1121 	 mov	8($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1124 	mov	$S[2],-16($tptr,$i,2)
1128 	 mov	$S[3],-8($tptr,$i,2)
1134 	 mov	16($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1138 	 mov	24($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1141 	mov	$S[0],0($tptr,$i,2)
1145 	 mov	$S[1],8($tptr,$i,2)
1151 	 mov	32($tptr,$i,2),$A0[0]	# t[2*i+2]	# prefetch
1155 	 mov	40($tptr,$i,2),$A0[1]	# t[2*i+2+1]	# prefetch
1158 	mov	$S[2],16($tptr,$i,2)
1160 	mov	$S[3],24($tptr,$i,2)
1170 	 mov	-16($tptr),$A0[0]	# t[2*i+2]	# prefetch
1174 	 mov	-8($tptr),$A0[1]	# t[2*i+2+1]	# prefetch
1177 	mov	$S[0],-32($tptr)
1181 	 mov	$S[1],-24($tptr)
1191 	mov	$S[2],-16($tptr)
1192 	mov	$S[3],-8($tptr)
1212 	lea	64(%rsp,$num),$tptr	# end of t[] window
1233 	add	8($tptr,$j),$A0[1]
1250 	mov	$A1[0],8($tptr,$j)	# "t[1]"
1253 	add	16($tptr,$j),$A0[0]
1268 	mov	$A1[1],16($tptr,$j)	# "t[2]"
1271 	add	24($tptr,$j),$A0[1]
1290 	mov	$A1[0],-8($tptr,$j)	# "t[3]"
1293 	add	($tptr,$j),$A0[0]
1308 	mov	$A1[1],($tptr,$j)	# "t[4]"
1311 	add	8($tptr,$j),$A0[1]
1327 	mov	$A1[0],8($tptr,$j)	# "t[5]"
1330 	add	16($tptr,$j),$A0[0]
1345 	mov	$A1[1],16($tptr,$j)	# "t[6]"
1348 	add	24($tptr,$j),$A0[1]
1368 	mov	$A1[0],-8($tptr)	# "t[7]"
1371 	add	($tptr),$A0[0]		# +t[8]
1377 	 imulq	16($tptr,$j),$m0	# m0=t[0]*n0	# modsched #
1381 	 mov	16($tptr,$j),$A0[0]	# t[0]		# modsched #
1387 	mov	$A1[1],($tptr)		# "t[8]"
1390 	add	8($tptr),$A1[0]		# +t[9]
1393 	lea	16($tptr),$tptr		# "t[$num]>>128"
1395 	mov	$A1[0],-8($tptr)	# "t[9]"
1396 	cmp	8(%rsp),$tptr		# are we done?
1400 	mov	$topbit,($tptr)		# save $topbit
1408 my ($tptr,$nptr)=("%rbx",$aptr);
1412 	lea	64(%rsp,$num),$tptr	# upper half of t[2*$num] holds result
1415 	mov	8($tptr),@ri[1]		# t[1]
1420 	mov	16($tptr),@ri[2]	# t[2]
1421 	mov	24($tptr),@ri[3]	# t[3]
1430 	mov	32($tptr,$i,8),@ri[0]	# tp[i+1]
1431 	mov	40($tptr,$i,8),@ri[1]
1436 	mov	48($tptr,$i,8),@ri[2]
1437 	mov	56($tptr,$i,8),@ri[3]
1444 	mov	32($tptr,$i,8),@ri[0]	# load overflow bit
1453 	and	@ri[0],$tptr
1458 	or	$nptr,$tptr		# tp=borrow?tp:rp
1462 	movdqu	($tptr),%xmm1
1470 	movdqu	16($tptr,$i),%xmm2
1471 	movdqu	32($tptr,$i),%xmm1
1482 	movdqu	16($tptr,$i),%xmm2