The IBM i standard library provides the _DecimalT class for working with decimal numbers. However, compared to working on this type from C or RPG programs, working through the _DecimalT class is significantly inferior in performance (for a division operation, for example, about 10 times slower). The specified class uses function calls from standard service programs to perform operations on the decimal type. I think that slowdown happens just because of calls of these functions. Does anyone have any idea how to work with the decimal type from C++ programs more efficiently? Is it possible to use (directly or indirectly) the MI DIV instruction?
1 Answers
I do not think the MI DIV instruction is useable from ILE code. There is no bound program access provided for that instruction.
Look at the CPYNV instruction. Note it has a built in number used for access from ILE code:
https://www.ibm.com/docs/en/i/7.2?topic=instructions-copy-numeric-value-cpynv
But not all MI instructions have an ILE built in number. The documentation on DIV does not include mention of a built in number.
Here is RPG code that runs the MI CPYNV instruction to copy from one RPG packed decimal variable to another. But without the built in number, I do not think you can do something comparable from ILE RPG or ILE C for the DIV instruction.
** --------------------- conMiScalar -----------------------
** value of the Type field in refScalarAttr struct.
dconMiScalar_None...
d c x'ff'
dconMiScalar_Bin c x'00'
dconMiScalar_Zoned...
d c x'02'
dconMiScalar_Packed...
d c x'03'
dconMiScalar_Char...
d c x'04'
dconMiScalar_UBin...
d c x'0a'
** ----------------------- refScalar ---------------------------
** MI scalar operand.
drefScalar ds qualified based(@)
d type 1a
d lgth 5i 0
d prec 3i 0 overlay(lgth:1)
d digits 3i 0 overlay(lgth:2)
d rsv1 10i 0
d csno s 10p 0
d pvCsno s 10p 0
d scalar_csno ds likeds(refScalar)
d scalar_pvCsno ds likeds(refScalar)
d msg s 50a varying
/free
csno = 25 ;
clear scalar_csno ;
scalar_csno.type = conMiScalar_packed ;
scalar_csno.prec = 0 ;
scalar_csno.digits = 10 ;
clear scalar_pvCsno ;
scalar_pvCsno.type = conMiScalar_packed ;
scalar_pvCsno.prec = 0 ;
scalar_pvCsno.digits = 10 ;
pr_LbCpyNv( %addr(pvCsno): %addr(scalar_pvCsno):
%addr(csno): %addr(scalar_csno)) ;
msg = 'pvCsno:' + %char(pvCsno) ;
dsply msg ;
Related