Forte Developer 7 New Features

This chapter describes the new features of the Forte trademark Developer 7 compilers and command-line tools. The primary focus of this release is improved performance, additional OpenMP support, and the ability to debug code written in the Javaprogramming language (Java code) as well as in the C programming language, C++ programming language, and Fortran programming language.

This chapter has the following sections:

C Compiler

C++ Compiler

Fortran Compiler

dbx Command-Line Debugger

Sun Performance Library

Interval Arithmetic

Performance Analysis Tools

Documentation

In most sections, there is a table that lists the new features of that component. The table has either two columns or three columns:

Two-column table. The left-hand column provides a short description of the feature, and the right-hand column has a longer description.

Three-column table. The left-hand column provides a short description of the feature, the middle column lists the relevant command or option, and the right-hand column has a longer description of the new feature.

Note - To find the Forte Developer documentation described in this chapter, see the documentation index installed with the product software at /opt/SUNWspro/docs/index.html. If your software is not installed in the /opt directory, contact your system administrator for the equivalent path on your system or network.

C Compiler

TABLE 1 lists the new features for this release of the C compiler. For more information about each of the new features listed in TABLE 1, see the index of the C User's Guide.


Feature	Option	Description
Enhanced interprocedural optimizations (SPARC)	`-xipo=2`	Performs inlining across all source files or performs interprocedural aliasing analysis as well as optimizations of memory allocation and layout to improve cache performance.
New prefetch levels (SPARC)	`-xprefetch_level=`n	Allows you to control the aggressiveness of the automatic insertion of prefetch instructions as determined by `-xprefetch=auto`.
Stack-overflow checking (SPARC)	`-xcheck=stkovf`	Adds a runtime check for stack overflow of the main thread in a singly-threaded program as well as slave-thread stacks in a multithreaded program.
Labels as values (computed `goto`)		Enables runtime determination of branching destinations. The address of a label can be acquired by using the `&&` operator and assigned to a pointer of type `void *`.
Support for additional C99 features	`-xc99`	Enables support of various implemented C99 features.
Improved formatting control of error messages	`-errfmt` `-errshort`	The compiler now offers two new commands that you can use to change the formatting of error messages: `-errfmt` and `-errshort`.
Support for UltraSPARC IIICu processor		The compiler's `-xtarget`, `-xchip`, and `-xcache` options support the UltraSPARC® IIICu processor.
Compiling for the Native Connector Tool	`-xnativeconnect`	Includes interface information inside object files and subsequent shared libraries so that the shared library can interface with Java code. The Native Connector Tool (NCT) enables the automatic generation of Java code and the Java Native Interface so that C shared libraries can be called from Java code.

C++ Compiler

TABLE 2 lists the new features available with this release (version 5.4) of the C++ compiler. All of the features are described in the C++ User's Guide.


Feature	Option	Description
Support for OpenMP (SPARC)	`-xopenmp`	Implements the OpenMP interface for explicit parallelization, including a set of source code directives, run-time library routines, and environment variables.
Type-based alias analysis and optimizations (SPARC)	`-xalias_level`	Allows the compiler to perform type-based alias analysis and optimizations.
Compiling for the Native Connector Tool	`-xnativeconnect`	Includes interface information inside object files and subsequent shared libraries so that the shared library can interface with Java code. The Native Connector Tool enables the automatic generation of Java code and the Java Native Interface so that C++ shared libraries can be called from Java code.
Enhanced interprocedural optimizations (SPARC)	`-xipo=2`	Performs inlining across all source files or performs interprocedural aliasing analysis as well as optimizations of memory allocation and layout to improve cache performance.
New prefetch levels (SPARC)	`-xprefetch_level=`n	Allows you to control the aggressiveness of the automatic insertion of prefetch instructions as determined by `-xprefetch=auto`.
Stack-overflow checking (SPARC)	`-xcheck=stkovf`	Adds a runtime check for stack overflow of the main thread in a singly-threaded program as well as slave-thread stacks in a multithreaded program.
Support for STLport Standard Library implementation version 4.5.2	`-library=stlport4`	Supports STLport's Standard Library implementation version 4.5.2.
`+w` behavior change	`+w`	No longer reports that a function is too large to inline and no longer reports that parameters are unused in order to reduce output messages and simplify the use of the `+w` option in routine builds. These warnings are still available with the `+w2` option.
`+w2` behavior change	`+w2`	No longer reports the use of implementation-dependent constructs in the system header files in order to simplify the use of the `+w2` option in routine builds.
`#error` directive aborts compilation		Issues an error message and immediately halts compilation. The compiler quits and reports the failure.

Fortran Compiler

The Forte Developer 7 release provides a Fortran 95 compiler, f95, with compatibility support for legacy Fortran 77 programs. See the chapter "FORTRAN 77 Compatibility: Migrating to Fortran 95" in the Fortran User's Guide for details on porting legacy Fortran 77 programs to the Fortran 95 compiler.

TABLE 3 lists the new features of the Fortran 95 compiler. See the Fortran User's Guide, Fortran Programming Guide, and the Fortran Library Reference for details.


Feature	Option	Description
Fortran 77 compatibility flag	`-f77`	Accepts many Fortran 77 constructs that are normally incompatible with Fortran 95.
Forcing implicit USE statements	`-use`	Forces one or more implicit USE statements into each subprogram or module subprogram compiled with the `-use` flag.
Enhanced `MODULE` features	`-moddir=`directory-name	Controls where `f95` writes compiled `MODULE` subprograms (`.mod` files). There is a new `MODDIR` environment variable. The `-M` option now accepts an archive (`.a`) file or module (`.mod`) file as an argument, as well as a directory name. The way the compiler searches for modules has also changed, and the way the compiler records module and include file dependences in makefiles with the `.KEEP_STATE` target has also improved.
`-Xlist` enhancements	`-XlistMP`	Provides verification of OpenMP directives. Global program analysis with the `-Xlist` flag has a number of new checks.
Identify known libraries	`-xknown_lib=`keywords	Directs the compiler to treat references to certain known libraries as intrinsics, ignoring any user-supplied versions. This enables the compiler to perform optimizations over calls to library routines based on special knowledge of the library.
Ignoring dummy argument type in interfaces	`!$PRAGMA` `IGNORE_TKR {`name `{,` name`} ...}`	Causes the compiler to ignore the type, kind, and rank of the specified dummy argument names appearing in a generic procedure interface when resolving a specific call.
`-xalias` for permitting nonstandard aliasing	`-xalias`	Advises the compiler about how far the program deviates from the aliasing restrictions required by the Fortran 95 standard. Useful when porting legacy Fortran 77 code that utilizes nonstandard programming techniques.
Enhancements to `-C` runtime checking	`-C`	Now also performs runtime array conformance checking.
The `BYTE` data type		Accepts the `BYTE` data type and treats `BYTE` as `INTEGER` `*1`.
Fortran 2000 features		New formatted I/O specifiers implemented in `f95`: `DECIMAL=`, `ROUND=`, and `IOMSG=`.
Rounding in formatted I/O	`-iorounding`	Sets the default rounding mode for formatted I/O to `processor-defined` or `compatible`. These modes correspond to the `ROUND=` specifier described in the Fortran 2000 features in this table.
Stack-overflow checking (SPARC)	`-xcheck=stkovf`	Adds a runtime check for stack overflow of the main thread in a single-threaded program as well as slave-thread stacks in a multithreaded program.
New default thread stack size		Increased to 4 megabytes on SPARC V8 platforms and 8 megabytes on SPARC V9 platforms.
Enhanced interprocedural optimizations (SPARC)	`-xipo=`n	Performs inlining across all source files or performs interprocedural aliasing analysis as well as optimizations of memory allocation and layout to improve cache performance.
New prefetch levels (SPARC)	`-xprefetch_level=`n	Allows you to control the automatic insertion of prefetch instructions.

`dbx` Command-Line Debugger

TABLE 4 lists the new features in this release of the dbx command-line debugger. For more information about these features, see the Debugging a Program With dbx manual.


Feature	Description
Debug programs with mixed-language code	The `dbx` command-line debugger can now debug applications running in the Solaris operating environment that are a mixture of Java code and C code or C++ code using the Java Native Interface (JNI). The JNI allows Java code and applications and libraries written in native languages to operate together.
Changes to `dbx` `collector` command	You can set the approximate limit to the amount of data collected. Address space data is no longer supported. There are subcommands to record a sample and to control whether `dbx` records a sample when it stops the target process. For more information about new features of the performance analysis tools, see TABLE 7.
`dbx` environment variables removed	The following environment variables have been removed: `allow_critical_exclusion`, `aout_cache_size`, `locache_enable`, and `delay_xs`.
Changes to `loadobject` command	There is a new `loadobject` command to control and list the symbols loaded by `dbx`. The new command replaces both the previous `loadobject` and `loadobjects` commands.
Change to default `dbx` prompt	The default `dbx` prompt is now the last path component of `argv[0]` instead of the entire string in `argv[0]`.

Sun Performance Library

Sun Performance Library trademark is a set of optimized, high-speed mathematical subroutines for solving linear algebra problems and other numerically intensive problems. Sun Performance Library is based on a collection of public domain applications available from Netlib (at http://www.netlib.org). These routines have been enhanced and bundled as the Sun Performance Library.

TABLE 5 lists the new features in this release of the Sun Performance Library. See the Sun Performance Library User's Guide and the section 3p man pages for more information.


Feature	Description
New FFT interfaces supersede a subset of `FFTPACK` and `VFFTPACK` Routines	See the `fft`(3p) man page and the Sun Performance Library User's Guide for descriptions of the new FFT routines. `FFTPACK` and `VFFTPACK` routines are included with this version of Sun Performance Library, but they are no longer supported.
Interval `BLAS` routines added	Interval `BLAS` routines operate on interval scalars, interval vectors, and interval matrices (dense, banded, and triangular).
Sort and permute routines added	The following routines have been added: `blas_dsort`, `blas_isort`, `blas_ssort`, `blas_dsortv`, `blas_isortv`, `blas_ssortv`, `blas_dpermute`, `blas_ipermute`, and `blas_spermute`.
Select sparse BLAS routines have been parallelized	All of the sparse matrix-matrix multiply (`mm`) routines and the sparse triangular solve (`sm`) routines have been parallelized. New sparse matrix-matrix multiply and sparse triangular solve routines have also been added to this release.
Sparse solver support for real, complex, and double complex data types	Sparse solver routines provide support for real, complex, and double complex data types.
`LINPACK` removed from Sun Performance Library	`LINPACK` is no longer included in the library. `LAPACK` version 3.0 supersedes `LINPACK` and all previous versions of `LAPACK`. If code that calls `LINPACK` routines cannot be modified to use `LAPACK` routines, the public domain version of `LINPACK` can still be obtained from Netlib (`http://www.netlib.org/linpack`).
Legacy `so.2` and `so.3` libraries not included with Sun Performance Library	To reduce the size of the Sun Performance Library for Forte Developer 7, only the `so.4` libraries have been included, and the legacy `so.2` and `so.3` libraries have been removed from this release.

Interval Arithmetic

TABLE 6 lists the Forte Developer 7 Interval Arithmetic new features.


Feature	Description
Interval BLAS routines added to Sun Performance Library	Interval `BLAS` routines operate on interval scalars, interval vectors, and interval matrices (dense, banded, and triangular). For more information on the Interval BLAS routines, see the Sun Performance Library User's Guide.
Two new template classes added to C++ interval arithmetic	A C++ interface to the C++ interval arithmetic library is extended with the `nvector` and `nmatrix` template classes that operate on interval vectors and interval matrices. See the `nvector`(3C++) and `nmatrix`(3C++) man pages for more information.
Interval-specific GPC functions added	Several interval-specific checks are implemented in GPC.

Performance Analysis Tools

TABLE 7 lists the new data collection and presentation features in the Forte Developer 7 release of the performance analysis tools. For more information, see the following man pages:

collect(1)

collector(1)

er_print(1)

libcollector(3)


Feature	Command and Option	Description
Simultaneous clock-based profiling and hardware-counter profiling		Collects clock-based profiling data and hardware-counter profiling data simultaneously.
Low-resolution profiling options added		Reduces the amount of data collected on executables that run for a long time.
Periodic sampling available with the `collect` command	`collect` `-S`	Collects periodic samples at the interval specified.
New sampling controls in `dbx`	`dbx` `collector` `dbxsample` `dbx` `collector` `sample` `record`	Records a sample and controls whether `dbx` records a sample when it stops the target process.
Approximate limit to the amount of data collected	`collect` `-L` `dbx` `collector` `limit`	Limits the amount of profiling data and tracing data recorded to a specified number of megabytes.
Record profiling data for Java programs through the `collect` command		Records profiling data on Java methods compiled by the Java HotSpot virtual machine.
Experiments on descendant processes recorded	`collect` `-F`	Controls whether or not descendant processes should have their data recorded. Performance data can be collected on descendant processes created by calls to `fork`(2), `fork1`(2), `vfork`(2), `fork`(3F), and `exec`(2) and its variants. Only available from the `collect` command.
Fortran API to collector library		Has Fortran API subroutines for recording samples, pause and resume, and closing an experiment.
Support for dynamically compiled functions		C and C++ API functions provide data to the collector on dynamically compiled functions and dynamically loaded modules.
Tracing of memory allocations and deallocations (heap tracing) recorded	`collect` `-H` `dbx` `collector` `heaptrace`	Collects heap tracing data. Data is converted into allocation and leak metrics. Display of metrics is supported by the `leaks` and `allocs` commands in `er_print`.
Tracing of MPI calls separated from synchronization delay tracing	`collect` `-m` `dbx` `collector` `mpitrace`	Collects more data on the MPI calls that are traced and displays them separately from synchronization delay data.
New meaning for `collect -n`		Prints details of the experiment that would have been run. It no longer runs the target program.
Address space data collection unavailable	`collect` `-a` `dbx` `collector` `address_space`	The collection and display of address space data is no longer supported.
Experiments no longer kept in a hidden directory		The experiment name is now the name of an ordinary UNIX® software directory. Older versions of the tools can not read newer experiments.
Improved robustness of data collection		When an application installs a signal handler, the collector re-installs its signal handler as the primary handler and passes signals to the application's handler so that profiling signals are not lost. The collector also prevents an application from using the hardware counters if hardware-counter overflow profiling is enabled and ensures that a application cannot interfere with clock-based profiling.
Performance tools readme available from the command line	`collect` `-R`	Displays the text version of the performance tools readme file.
Changed handling of load objects	`er_print` `object_list` `er_print` `object_select`	Aggregates and presents data for the load object as a whole instead of removing data for a load object.
New commands to set thresholds for highlighting high-metric code lines	`er_print` `sthresh` `er_print` `dthresh`	Sets the thresholds for flagging important lines in annotated source code and annotated disassembly code.
Function and callers-callees data printed	`er_print` `fsingle` `er_print` `csingle`	Prints function summary and callers-callees data for a single function.

Documentation

This section describes Forte Developer 7 documentation new features.

The OpenMP API User's Guide is a new manual for this release. This manual summarizes the OpenMP directives and pragmas accepted by the compilers, describes implementation details, and provides guidelines for converting legacy Sun parallelization directives to OpenMP.

The OpenMP Application Program Interface (API) supports multiplatform shared-memory parallel programming in C, C++, and Fortran. OpenMP is a portable, scalable model that gives programmers an explicit and flexible interface for developing parallel applications on shared-memory mulitprocessor platforms. In this release, the C compiler, C++ compiler, and Fortran 95 compiler feature full native support on SPARC® platforms for the OpenMP C/C++ API Version 1.0 and Fortran 95 API Version 2.0 specifications. For PDF versions of the specifications, go to the http://www.openmp.org web site.

The Fortran User's Guide has a new chapter entitled "FORTRAN 77 Compatibility: Migrating to Fortran 95" that provides details on porting legacy Fortran 77 programs to the Fortran 95 compiler.

The Debugging a Program With dbx manual has a new chapter entitled "Debugging a Java Application With dbx." The chapter describes how you can use the dbx command-line debugger to debug applications running in the Solaris operating environment that are a mixture of Java code and C code or C++ code using the Java Native Interface (JNI). The JNI allows Java code and applications and libraries written in native languages to operate together.

The Analyzing Program Performance manual is now entitled Program Performance Analysis Tools.

The following manuals have been discontinued and were not updated for this release:

Sun WorkShop 6 update 2 Installation and Licensing Reference (see the Forte Developer 7: Installation Guide for installation and licensing information)

About Sun WorkShop 6 update 2 Documentation

Introduction to Sun WorkShop

Previous releases of the FORTRAN 77 Language Reference are available through the docs.sun.com web site. This manual has not been updated for this release.

Forte Developer 7 product documentation is provided in formats that are readable by assistive technologies for users with disabilities. For more information, see Product Documentation in Accessible Formats.