mtlb_int32

Description

Matlab and Scilab int32 behave differently :

• Beyong the main interval bounds, Scilab's int32() wraps, while Matlab's mtlb_int32() saturates.
• For non integer values, Scilab's int32() truncates the fractional part, while Matlab's mtlb_int32() rounds to the nearest integer.

Let imin = -(2^16) = -2147483648 and imax = 2^16 - 1 = 2147483647.

x	int32(x)	mtlb_int32(x)
%nan	0	0
-%inf	imin	imin
imin-2	imax-1	imin
imin-1	imax	imin
imin	imin	imin
imin-1	imin-1	imin-1
-80.7	-80	-81
-80.5	-80	-81
-80.4	-80	-80
10.2	10	10
10.5	10	11
10.51	10	11
imax	imax	imax
imax+1	imin	imax
imax+2	imin+1	imax
imax+3	imin+2	imax
%inf	imax	imax

mtlb_int32 has not to be used for hand coded functions. The function mtlb_int32(A) is used by mfile2sci to replace int32(A) when it is not possible to know what is the input while porting Matlab code to Scilab. This function then determines the correct semantic at run time.