Re: Computer architects leaving Intel...

On 9/4/2024 1:13 PM, MitchAlsup1 wrote:

On Wed, 4 Sep 2024 17:53:13 +0000, David Brown wrote:
 

On 04/09/2024 18:07, Tim Rentsch wrote:

Terje Mathisen <terje.mathisen@tmsw.no> writes:
>

Michael S wrote:
>

On Tue, 3 Sep 2024 17:41:40 +0200
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
>

Michael S wrote:
>

3 years ago Terje Mathisen wrote that many years ago he read that
behaviour of memcpy() with overlappped src/dst was defined.
https://groups.google.com/g/comp.arch/c/rSk8c7Urd_Y/m/ZWEG5V1KAQAJ
Mitch Alsup answered "That was true in 1983".
So, two people of different age living in different parts of the
world are telling the same story.  May be, there exist old popular
book that said that it was defined?
    >>

>
It probably wasn't written in the official C standard, which I
couldn't have afforded to buy/read, but in a compiler runtime doc?
>
Specifying that it would always copy from beginning to end of the
source buffer, in increasing address order meant that it was
guaranteed safe when used to compact buffers.

>
What is "compact buffers" ?

>
Assume a buffer consisting of records of some type, some of them
marked as deleted.  Iterating over them while removing the gaps means
that you are always copying to a destination lower in memory, right?

>
If all the records are in one large array, there is a simple
test to see if memcpy() must work or whether some alternative
should be used instead.

>
Such tests are usually built into implementations of memmove(), which
will chose to run forwards or backwards as needed.  So you might as well
just call memmove() any time you are not sure memcpy() is safe and
appropriate.

 Memmove() is always appropriate unless you are doing something
nefarious.
So:
# define memcpy memomve
and move forward with life--for the 2 extra cycles memmove costs it
saves everyone long term grief.
 

Pretty reasonable.
If you want to move intact data around within overlapping buffers or arrays, then yes, memmove makes sense.

When you need the nefarious activities of memcpy write it as a
for loop by yourself and comment the nafariousness of the use.

I had added _memlzcpy() to my C library implementation:
It specifically mimics the behavior of a loop copying bytes one-at-a-time, as typically needed for LZ77 implementations, but aiming to be moderately efficient about it (with the same general interface as memcpy and memmove).
You don't want an actual "byte at a time" copy loop, because this is painfully slow, so it turns into a bunch of hair to mimic the behavior while copying typically 16 or 32 byte blocks.
Then, also, _memlzcpyf(), which is similar, but may access and stomp 16-32 bytes past the end of the source and destination (in these sorts of memory copies, dealing accurately with the last "mod N" bytes can be fairly expensive).
It can also be used for multi-byte memset.
   _memlzcpy(dst+4, dst, 16380);
Will fill the destination with 16K worth of copies of the first 4 bytes.
But, the downside is that these have a higher constant-time cost (vs memcpy and memmove), and are generally also slower for bulk copy (vs memmove or memcpy). Some of this is partly because getting an accurate self-overlapping copy set up is more involved than a plain memory copy.
Generally, memmove only really needs to compare the addresses:
   dst<src:
     Copy in forward direction;
   dst>src:
     Copy in backwards direction;
   dst==src:
     Do nothing.