forked from Imagelibrary/rtems
4e4e6911e5
FP issues on this target: The default variants of libc, libm and libgcc assume that a 68881 coprocessor is present. Without the FPSP, any floating point operation, including printf() with a "%f" format specifier, is likely to cause an unimplemented instruction exception. The FPSP works with the default variants of libc, libm and libgcc. It does not work in conjunction with the msoft-float variants. The paranoia test goes into an infinite loop at milestone 40. I am guessing that floor() is returning an incorrect value. The msoft-float variants of libc, libm and libgcc appear to do floating point I/O properly. They only failed in paranoia. Offhand, I can't think of why they would conflict with the FPSP, so I think that there is something wrong with the msoft-float code. It might be my installation. Given my experiences, I decided to install the FPSP in bsp_start(), and to link against the default variants of libc, libm and libgcc. This causes the executables to increase in size by about 60 KB. The README file and the mvme167.cfg specify how to remove the FPSP, and how to link against the msoft-float variants of the libraries. This is not what Eric Norum had done: on my host, his gen68360_040 port links RTEMS code with the msoft-float variants of libc and libm, and the default variant of libgcc. In this configuration, the output of printf() with "%f" is garbage on my target.
#
# $Id$
#
Building RTEMS
==============
See the file README.configure.
Directory Overview
==================
This is the top level of the RTEMS directory structure. The following
is a description of the files and directories in this directory:
INSTALL
Rudimentary installation instructions. For more detailed
information please see the Release Notes. The Postscript
version of this manual can be found in the file
c_or_ada/doc/relnotes.tgz.
LICENSE
Required legalese.
README
This file.
c
This directory contains the source code for the C
implementation of RTEMS as well as the test suites, sample
applications, Board Support Packages, Device Drivers, and
support libraries.
doc
This directory contains the PDL for the RTEMS executive.
Ada versus C
============
There are two implementations of RTEMS in this source tree --
in Ada and in C. These two implementations are functionally
and structurally equivalent. The C implementation follows
the packaging conventions and hiearchical nature of the Ada
implementation. In addition, a style has been followed which
allows one to easily find the corresponding Ada and C
implementations.
File names in C and code placement was carefully designed to insure
a close mapping to the Ada implementation. The following file name
extensions are used:
.adb - Ada body
.ads - Ada specification
.adp - Ada body requiring preprocessing
.inc - include file for .adp files
.c - C body (non-inlined routines)
.inl - C body (inlined routines)
.h - C specification
In the executive source, XYZ.c and XYZ.inl correspond directly to a
single XYZ.adb or XYZ.adp file. A .h file corresponds directly to
the .ads file. There are only a handful of .inc files in the
Ada source and these are used to insure that the desired simple
inline textual expansion is performed. This avoids scoping and
calling convention side-effects in carefully constructed tests
which usually test context switch behavior.
In addition, in Ada code and data name references are always fully
qualified as PACKAGE.NAME. In C, this convention is followed
by having the package name as part of the name itself and using a
capital letter to indicate the presence of a "." level. So we have
PACKAGE.NAME in Ada and _Package_Name in C. The leading "_" in C
is used to avoid naming conflicts between RTEMS and user variables.
By using these conventions, one can easily compare the C and Ada
implementations.
The most noticeable difference between the C and Ada83 code is
the inability to easily obtain a "typed pointer" in Ada83.
Using the "&" operator in C yields a pointer with a specific type.
The 'Address attribute is the closest feature in Ada83. This
returns a System.Address and this must be coerced via Unchecked_Conversion
into an access type of the desired type. It is easy to view
System.Address as similar to a "void *" in C, but this is not the case.
A "void *" can be assigned to any other pointer type without an
explicit conversion.
The solution adopted to this problem was to provide two routines for
each access type in the Ada implementation -- one to convert from
System.Address to the access type and another to go the opposite
direction. This results in code which accomplishes the same thing
as the corresponding C but it is easier to get lost in the clutter
of the apparent subprogram invocations than the "less bulky"
C equivalent.
A related difference is the types which are only in Ada which are used
for pointers to arrays. These types do not exist and are not needed
in the C implementation.