VSI BASIC for OpenVMS I64 Systems is a native compiler that runs
on and generates code to run on OpenVMS I64 systems. Version
V1.8-004 is a port of the Alpha BASIC compiler to the Itanium
architecture.
1 – General Notes
o Throughout these notes, the term VSI BASIC or the name BASIC
used alone, applies to VSI BASIC for OpenVMS I64 Systems, HP
BASIC for OpenVMS Alpha Systems, and HP BASIC for OpenVMS VAX
Systems. The following terms refer to information that applies
specifically to the corresponding product.
I64 BASIC - VSI BASIC for OpenVMS I64 Systems
Alpha BASIC - HP BASIC for OpenVMS Alpha Systems
VAX BASIC - HP BASIC for OpenVMS VAX Systems
o I64 BASIC V1.8-004 minimally requires OpenVMS I64 Version
8.4-1H1 to run.
o The default data type for I64 BASIC is IEEE S_floating point
whereas for Alpha BASIC and VAX BASIC the default is VAX F_
floating point. To continue use of the VAX floating point data
types you must use the /REAL_SIZE={SINGLE | DOUBLE | GFLOAT}
qualifier, or the OPTION SIZE=REAL {SINGLE | DOUBLE | GFLOAT}
statement.
Since the VAX floating point data types are not available
in the Itanium hardware, they are supported by converting
the values to an appropriate IEEE data type, performing
the calculation, and then converting the values back to VAX
floating point data type. This process can cause rounding
errors and the results to differ slightly from those obtained
using VAX floating point directly.
Note that constants in programs are represented in the
default floating point data type unless explicitly specified
otherwise, and they may be subject to conversion to IEEE data
types.
See the white paper, "OpenVMS floating-point arithmetic on the
Intel[R] Itanium[R] architecture" at
http://www.hp.com/products1/evolution/alpha_retaintrust/download/i64-floating-pt-wp.pdf
for more information.
o By default, programs are compiled at optimization level 4
(/OPT=LEVEL=4), the highest level of optimization. If you
encounter problems with the generated code, try compiling the
program with a lower level of optimization. This may correct
the problem. Please report any problems to HP, even if the
program works correctly at a lower optimization level.
2 – Feature enhancements
Some minor feature enhancements have been made for version 1.8-
004. These include:
o Routine parameters that are passed by reference whose address
is not aligned no longer generate alignment faults.
o Various diagnostic messages have been improved.
3 – Problems fixed
A number of problems have been fixed in I64 BASIC V1.8-004. The
known problems that have been fixed include:
o A dump was getting generated when the program was trying to
perform copy in an optimized mode. This issue has been fixed.
o Extremely long command lines no longer cause a compiler
internal error.
o The display of S-Float constants in listing files no longer
causes an error.
4 – Known problems and Restrictions
4.1 – General Problems
The following are known general problems with I64 BASIC V1.8-004.
o Floating point numbers displayed in E-format may differ.
In some cases floating point numbers displayed in E-format may
contain an additional zero in the exponent field. For example,
6.0E+002 instead of 6.0E+02
o Floating point exceptions not accurately reported.
In some cases a floating point exception may not be accurately
reported. For example: a divide-by-zero exception may be
reported as an invalid-number.
o Inability to open UNDEFINED type file for output.
It is currently not possible to open a file of type UNDEFINED
for output.
Workaround: Create an empty file with the desired attributes
by some other means, for example with a program in a different
language or with the DCL command CREATE/FDL. Then in BASIC
open the file for input with the APPEND attribute. The
following code fragment illustrates the technique.
%INCLUDE "$FDLDEF" %FROM %LIBRARY "SYS$LIBRARY:BASIC$STARLET"
EXTERNAL INTEGER FUNCTION FDL$CREATE
MAP (outbuff) STRING myrec = 512
DECLARE INTEGER i, fdlstat
DECLARE STRING resultname
fdlstat = FDL$CREATE ("RECORD; FORMAT UNDEFINED;", &
"myfilename", ".DAT", resultname,, &
FDL$M_FDL_STRING + &
FDL$M_DEFAULT_STRING + &
FDL$M_FULL_OUTPUT)
OPEN resultname FOR INPUT AS FILE #3, ORGANIZATION UNDEFINED, &
ACCESS APPEND, &
MAP outbuff, &
RECORDTYPE ANY
myrec = ...
PUT #3
o DET incorrect with default XFLOAT type.
When the default floating point type is set to XFLOAT, the
result of the determinant statement (DET) is incorrect.
Workaround: Use explicitly declared variables of XFLOAT type,
and avoid using XFLOAT type as the default floating point type
when the DET statement is used.
o The most-negative integer constants cannot be represented.
The compiler incorrectly gives an integer overflow message
when the most negative integer constants:
BYTE -128%
WORD -32768%
LONG -2147483648%
are used.
Workaround: Use the appropriate expression:.
BYTE -127% - 1%
WORD -32767% - 1%
LONG -2147483647% - 1%
o Passed by value quadword arguments in STARLET are not yet
supported.
Since routines that declare pass by value quadword parameters
are not yet supported by BASIC, declarations of routines that
do this are not included in BASIC$STARLET.TLB.
Workaround: Use of the these routines should be avoided until
full support of 64-bit quantities is provided.
o The CMA facility found in BASIC$STARLET.TLB is incompatible
with BASIC.
BASIC is not currently thread safe and will not work with the
CMA$DEF module in BASIC$STARLET.TLB. Compiling it with BASIC
results in a number of Record type undefined errors.
Workaround: BASIC does not support the CMA facility and
compiling the CMA$DEF module should be avoided.
4.2 – Debugger problems
With this release of the compiler, debugging support for I64
BASIC programs is not complete. The "OpenVMS I64 V8.4-1H1 Release
Notes" document contains general release notes on the debugger.
This section describes additional problems and workarounds
specific to I64 BASIC programs.
o Examining a symbol declared with the DIMENSION keyword
sometimes causes a debugger access violation.
Workaround: Type SHOW SYMBOL/ADDRESS then examine the memory
location directly.
o Examining a field in a record sometimes fails with the error:
%DEBUG-W-NOFIELD, 'xxx' is not a field in this record
Workaround: Examine the entire record.
o In some cases the Debugger can not determine the end of user
code. In these cases, the following message may be produced if
you STEP at the last line of user application code:
DBG> Step
%DEBUG-W-SCRNOSRCLIN, no source line for address xxxxxxxx for display in SRC
You can ignore this message.
o Examining some types of variables is not possible, including:
A RECORD variable, or an array of RECORDs which is declared
in a MAP DYNAMIC and also passed as a parameter to another
routine.
o Examining EXTERNAL constants with a plain EVALUATE command
results in a warning message. Instead the EVALUATE/ADDRESS
command should be used. Note that it is not possible to
display EXTERNAL constants in floating point format.
o Problems with examining dynamic objects in the debugger.
Certain types of dynamic objects referenced by descriptors
cannot be displayed in the OpenVMS debugger.
Workaround: The inability to display these kinds of objects is
a restriction.
5 – Documentation
5.1 – Release notes
I64 BASIC provides the following online information, which is
copied to the user's system during installation:
o The VSI BASIC for OpenVMS I64 Systems online Release Notes
o I64 BASIC DCL Help
o I64 BASIC Language Topics Help
5.1.1 – I64 BASIC Language Topics Help
HELP on BASIC language topics, available from the VAX BASIC
environment is accessible from DCL in I64 BASIC. For help on
language features, type the following command at the DCL prompt:
$ HELP/LIBRARY=BASICHELP
Alternately, a symbol can be defined to access help on BASIC
language features directly. For example, the DCL command:
$ BASHELP == "HELP /LIBR=BASICHELP"
will allow you to type:
$ BASHELP BASIC-topic
5.2 – Documentation corrections
The following corrections should be made to the Alpha BASIC
manuals.
5.2.1 – Reference Manual
Corrections to the online HTML documentation HP BASIC for OpenVMS
Systems Reference Manual.
o The Table of Result Data Types in Alpha BASIC Expressions in
Section 1.6.1.2 has two incorrect entries in the online HTML
documentation. The table in the paper and PDF documentation
however is correct. The entries for GFLOAT/SFLOAT and
SFLOAT/GFLOAT should be GFLOAT.
5.2.2 – User Guide
Corrections to the HP BASIC for OpenVMS Systems User Manual.
o The conditional at the end of Example 13-1, in the "File
Input and Output" chapter should have an equal sign instead
of "AND". It should read as Rms_status = Rms$_normal instead
of Rms_status AND Rms$_normal.
6 – Feedback
If you experence a problem, please file a PTR problem report,
including enough information to reproduce the problem.
Thank you for your continued support of VSI BASIC for OpenVMS I64
Systems.