Skip to content

Latest commit

 

History

History
99 lines (59 loc) · 3.24 KB

CODING.rst

File metadata and controls

99 lines (59 loc) · 3.24 KB
Raw

RRPGE Minimal library & Host prototype, Coding style and conventions

Author

Sandor Zsuga (Jubatian)

Copyright

2013 - 2014, GNU GPLv3 (version 3 of the GNU General Public License) extended as RRPGEvt (temporary version of the RRPGE License): see LICENSE.GPLv3 and LICENSE.RRPGEvt in the project root.

Introduction

This file describes some cruical coding style elements and conventions utilized for the development of this library, primarily those which may not be apparent from the code or it's comments alone.

Freestanding

The code in "librrpge" consists the freestanding RRPGE emulator library. This is supposed to be compile-able with any compiler capable to process code following the C89 standard.

Standard C library functions must also be avoided, so making the library capable to support environments where the standard C library does not exists. Only the type definitions from <stdint.h> may be used to define own types, so in a C89 environment they can be substituted.

RRPGE library interface

The headers in "librrpge" with an "rrpge" prefix are the interface of the RRPGE library. This component is still in development. The RRPGE library is meant to be accessed only through these, later, this part should be standardised to provide a common interface for probable different implementations.

Types

Compiler "int" width

The compiler's "int" width is assumed to be at least 32 bits. This most importantly assumes the followings to be true:

  • Integer promotion promotes towards an at least 32 bit wide type.
  • Constants are at least 32 bit wide.

There is no point in supporting narrower "int" widths as it is unlikely feasible emulation is possible on such architectures.

The "auint" type

The "auint" type must be declared to be at least as wide as the compiler's native "int" width, ensuring no promotions (into signed integer) happen on a variable of type "auint". The type may be wider than the compiler's native "int" width.

Fixed width types

Fixed width types (8, 16, 32 bits) should only be used for working with arrays in memory.

The types must be assumed to hold at least the indicated amount of bits. They may hold more.

When moving values into a fixed with type, if necessary, an appropriate mask should ensure the value is not out of range for the type. Normally compilers supporting the given type proper may discard this mask, so it has no effect on performance or compiled binary size.

When reading a fixed with type, if necessary, the range should be ensured by an appropriate mask. Normally compilers should be capable to discard this mask if the type has the appropriate fixed width.

These rules ensure that even when a suitable fixed width type is not available, the emulator may still operate properly using wider types.