Extremely CISC instructions- C compilers.

[ben]

On 2021-08-25 10:27 a.m., Todd Goodman via cctalk wrote:
> 
> On 8/25/2021 11:49 AM, ben via cctalk wrote:
>> [..SNIP..]
>> C uses cheap tricks for speed. 8 bit bytes, 32 bit integers, taken 
>> from B. I have 21 bit CPU, with 3 7 bit bytes/word. Algol would have a 
>> PACK/UPACK function, and be fairly portable. C on the other hand a mess.
>> Ok. I don't have 21 bit cpu, but I have this spare FPGA card ...
> 
> Nope, the standard doesn't specify those bit sizes.

> Back in the 80s I was using the BBN C Machine with 10-bit bytes and 
> happily building from source I picked up on the newsgroups with little 
> issue
>
What is the BBN C Machine?

>>
>> > world chose to standardize on C.
>> More like the same 32 bit/ 8 bit bytes vanilla cpu, with push and pop
>> on the stack. Can't have near/far pointers with some intel products
>> so we need new standard for the non PDP/11 or VAX computers, and again
>> and again...

> No, the same standard works with different architectures.  Have you 
> looked into GCC in the past 20 years or so?

I looked at it 20 years ago, I did not like it then. Why .
Read all the source into memory. A memory HOG! ONE PASS COMPILE!

About the mid 70's every new compiler C/PASCAL/ALGOL/ASM  wished be a 1 
pass compiler for speed. Slow compile times are from a floppy or a 
overloaded time sharing system or too small a memory. Hardware
not software issues.
Recursive decent compilers, are good at parsing but evaluate ()'s
wrong in code generation. The PDP 11 compiler is a 3 pass compiler,
macro, recursive decent,tree evaluation. Clean and simple and small.

For all the gripes I have about C, it is best of all the languages
I can think of from the 1970's.

Ben.
PS: How about C to FORTRAN IV compiler. :)