changes

来自「Arithmetic for integers of almost unlimi」· 代码 · 共 86 行

Bugfix in Intel586NT\idigitveccsubto.c submitted
by Grant Atoyan re. missing DWORD PTR.   cludwig (October 4, 2002)


Changes	Integer-2.0 --> Integer-2.1

T.P: Modified Makefiles for libLiDIA.a

new division !
	DigitVecMultSub				ok
	normdiv					ok
	DigitVecDiv

remove old code					ok

square - code for karatsuba ? ---> Isquare ?
	( 0 multiplication test ? )
	( parameters for various CPUs )

DigitDiv a la Juffa

DigitDiv without rem ?				no

Macros for Add/Sub ?	(idigit.h)		no

xgcd:	only one coefficient computed

bgcd:	remove "else" clause in main loop	ok

386 Functions from ALI ?

PLAN:
=====

subdirs:

Sparc7
Sparc8
Mips
Power
M68k-sun3
M68k-NeXT

T800-Parc
I86-PC

with: idigit* - functions (complete set)

Makefile:	DOBJ=Sparc7/*.o
			etc.


load delay penalty improvement in	(15.2.93)
DigitVecCsubto
DigitVecSri


squaring:	(idea: 15.2.93)

IasIsq
IsqasI
DigitVecKaratsubaQ

better: (10.5.93)	automatic detection in mul - frontend.


xgcd:

only one coefficient!
Lehmer Trick! ?

Gcc version 2.4.5 (or higher) on Sparc:
Use of gcc's unsigned long long after casts from unsigned long
--> the right machine multiplication is generated! (with -mv8 at least)
(16.7.93)

unsigned long long version of functions using multiplication!
(e.g. for gcc, Sun acc)


Include modular arithmetic for odd modules, Montgomerys method
for multiplication.	imod.h imod.c				RD, 19.7.93

Include bit operations for large integers (ibit.c)              TP, 5.10.94