Path: ...!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Thiago Adams <thiago.adams@gmail.com>
Newsgroups: comp.lang.c
Subject: transpiling to low level C
Date: Sun, 15 Dec 2024 00:05:13 -0300
Organization: A noiseless patient Spider
Lines: 78
Message-ID: <vjlh19$8j4k$1@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Injection-Date: Sun, 15 Dec 2024 04:05:14 +0100 (CET)
Injection-Info: dont-email.me; posting-host="75cbf010cfc8fccc260a6c4cc0d7f911";
	logging-data="281748"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1/STxH8W3SRyICgNER6Jg8/RDM/MmmGZXY="
User-Agent: Mozilla Thunderbird
Cancel-Lock: sha1:RgBLG47t8N0w8odfO+MN/aQsXc4=
Content-Language: en-GB
Bytes: 2819


I am working on a C backend that generates simple C code.

You can test it here:
http://thradams.com/cake/playground.html

Objective
- Shift all the complexity of the C compiler to the frontend.
   Why? This approach simplifies having one frontend and multiple 
backends.
- Use C as an intermediate language to feed a backend (any C compiler 
can act as the backend).
   The backend can then focus solely on code generation.
- Code is not portable

Removed C89 Features
- Preprocessor
- sizeof
- typedef
- enum
- Constant expressions (the final result is precomputed during earlier 
stages)
- const
- I may also remove switch.
- struct are declared only at external scope

Most features from C99, C11, and C23 are not included.
However, features directly related to code generation may be supported, 
in the future such as:
- restrict
- inline
- atomic
- _NoReturn

Current Status
This project is very new and still contains bugs.


   When generating a constant like '(unsigned char) 1', the cast is 
removed.  The rationale is that type-specific representations aren't 
necessary, as values are promoted automatically.
   Similarly, 'nullptr' is replaced by '0' instead of '(void*)0'.
   Constructs dependent on type (like 'sizeof', '_Generic', etc.) have 
already been removed in earlier stages.

Example:

   unsigned char c = u8'~';

Generated Code:

   unsigned char c = 126;

Instead of:

   unsigned char c = ((unsigned char)126);

For larger types like 'unsigned long long', a suffix is still included.

Initialization Strategy

   Currently, I use 'memset' to initialize structures to zero. For 
instance:

   struct X x = {};
   memset(&x, 0, 8 /*bytes*/);


Boolean Conversion
   The behavior of converting integers to 'bool' is still incomplete.
   For example:

   bool b = 123;

Here, 'b' should have the value '1' but today it is 123.