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: Re: transpiling to low level C
Date: Mon, 16 Dec 2024 08:17:45 -0300
Organization: A noiseless patient Spider
Lines: 64
Message-ID: <vjp28q$13k4m$1@dont-email.me>
References: <vjlh19$8j4k$1@dont-email.me> <vjlm2b$dfpo$1@dont-email.me>
 <vjmbuh$hhg1$1@dont-email.me> <vjnkqh$op3c$2@dont-email.me>
 <vjnocf$pme5$1@dont-email.me> <vjojaj$11811$1@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Mon, 16 Dec 2024 12:17:46 +0100 (CET)
Injection-Info: dont-email.me; posting-host="9df6c4f7516d4e1a3aa2c9bdac164fe4";
	logging-data="1167510"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1+RxuJTL3itKldKO5Gh3ERkHZ0CCpx6OHY="
User-Agent: Mozilla Thunderbird
Cancel-Lock: sha1:LRt120zwnixQ/3nFdc0bCJcjHSc=
Content-Language: en-US
In-Reply-To: <vjojaj$11811$1@dont-email.me>
Bytes: 3413

On 16/12/2024 04:02, BGB wrote:
> On 12/15/2024 5:22 PM, Thiago Adams wrote:
>> Em 12/15/2024 7:22 PM, Lawrence D'Oliveiro escreveu:
>>> On Sun, 15 Dec 2024 07:44:34 -0300, Thiago Adams wrote:
>>>
>>>> I also think LLVM is too big.
>>>
>>> How about QBE, then <https://c9x.me/compile/>.
>>
>> QBE is just for linux and it does not generate code directly. C is 
>> everywhere.
>> But I am reading QBE docs to see what is necessary in a IL.
>>
> 
> I agree that both LLVM and GCC are too big.
> 
> I am not sure the minimum size of a usable C compiler, my current 
> estimate is probably somewhere in the area of 50-100 kLOC.
> 

My hope is 10-30K lines. But remember this is a simpler version of C. 
This is the objective. It also may work in one pass, without building a 
lot of data structures for the AST.


> My current compiler is a bit bigger than this, and my own past attempt 
> to implement a C compiler in under 30k lines had failed.
> 
> My current compiler use to be smaller, but multiple backends and similar 
> did make it bigger:
>    SH-2, SH-4, BJX1 (older)
>    BJX2, RISC-V (current)
> 
> 
> 
> If I were to estimate a feature set for a 3AC IL:
>    Basic UNARY and BINARY operators;
>      Also COMPARE and similar;
>    Load and Store Index operators;
>      With support for an immediate index;
>    GOTO
>    IF-GOTO
>    CALL (with a list of argument variables and a target variable).
> 
> One doesn't need any higher-level control flow constructs.
> 
> Any built-in support for "switch()" can be optional (the main reason to 
> have so is that it can make switch a little faster, but isn't required).
> 
> Earlier on, my compiler didn't have any backend support for this, and 
> would instead decompose the switch into "if/else/goto" logic. You can 
> then use recursive binary subdivision to implement the list of case 
> labels (or linear probes if N is small).
> 
> BGBCC still uses binary subdivide if the case-labels don't satisfy the 
> constraints for a branch-table (Say, needs to be 16..256 case labels 
> with a density over 75%).
> 


Yes, switch is something I can remove.
C2Y is adding range in case. 
(https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3370.htm)