Re: 80286 protected mode

Liste des GroupesRevenir à c arch 
Sujet : Re: 80286 protected mode
De : david.brown (at) *nospam* hesbynett.no (David Brown)
Groupes : comp.arch
Date : 13. Oct 2024, 13:10:20
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <vegdbc$ltgv$2@dont-email.me>
References : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
User-Agent : Mozilla Thunderbird
On 13/10/2024 11:00, Michael S wrote:
On Fri, 11 Oct 2024 16:54:13 +0200
David Brown <david.brown@hesbynett.no> wrote:
 
On 11/10/2024 14:13, Michael S wrote:
On Fri, 11 Oct 2024 13:37:03 +0200
David Brown <david.brown@hesbynett.no> wrote:
  
On 10/10/2024 23:19, Brian G. Lucas wrote:
>
Not applicable.
     
>
I don't understand what you mean by that.  /What/ is not applicable
to /what/ ?
 
>
Brian probably meant to say that that it is not applicable to his
my66k LLVM back end.
>
But I am pretty sure that what you suggest is applicable, but bad
idea for memcpy/memmove routine that targets Arm+SVE.
Dynamic dispatch based on concrete core features/identification,
i.e. exactly the same mechanism that is done on "non-scalable"
architectures, would provide better performance. And memcpy/memmove
is certainly sufficiently important to justify an additional
development effort.
  
>
That explanation helps a little, but only a little.  I wasn't
suggesting anything - or if I was, it was several posts ago and the
context has long since been snipped.
 You suggested that "scalable" vector extension are preferable for
memcpy/memmove implementation over "non-scalable" SIMD.
I certainly suggested that they have some advantages, yes.  I don't know nearly enough details about implementations and practical usage to know if scalable vector instructions are /always/ better than non-scalable SIMD with fixed-size registers, either from the viewpoint of their efficiency at runtime or their implementation in hardware.
It seems to me that if the compiler knows the size of a memcpy/memmove, then the best results would probably be achieved by the compiler inlining the copy using fixed size registers of a suitable size.  If it does not know the size, then I would expect (but I don't know for sure) that a hardware scalable vector instruction should be more efficient than using fixed-size registers.  If that were not the case, then I wonder why scalable vector hardware has become popular recently in ISAs.
If you - or someone else - knows enough to say more about this, then I'd be glad to learn about it.

 
Can you be more explicit about
what you think I was suggesting, and why it might not be a good idea
for targeting a "my66k" ISA?  (That is not a processor I have heard
of, so you'll have to give a brief summary of any particular features
that are relevant here.)
>
 The proper spelling appears to be My 66000.
For starter, My 66000 has no SIMD. It does not even have dedicated FP
register file. Both FP and Int share common 32x64bit register space.
 
OK.

More importantly, it has dedicate instruction with exactly the same
semantics as memmove(). Pretty much the same as ARM64. In both cases
instruction is defined, but not yet implemented in production silicon.
The difference is that in case of ARM64 we can be reasonably sure that
eventually it will be implemented in production silicon. Which means
that in at least several out of multitude of implementations it will
suck.
 
So if I understand you correctly, your argument is that scalable vector instructions - at least for copying memory - is slow in hardware implementations, and thus it would be better to simply copy memory in a loop using larger fixed-size registers?  I would find that surprising, but as I said, I don't know the details of implementations.
(I do know that in the 68k family, the hardware division instruction was dropped for later devices after it was realised that a software division routine was faster than the hardware instruction.  So such strange things have happened.)

Date Sujet#  Auteur
16 Apr 24 * Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)237Lawrence D'Oliveiro
16 Apr 24 `* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)236David Brown
16 Apr 24  +- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1MitchAlsup1
26 May 24  +- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1MitchAlsup1
1 Oct 24  `* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)233MitchAlsup1
1 Oct 24   `* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)232Thomas Koenig
1 Oct 24    +* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)225MitchAlsup1
2 Oct 24    i+* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)223Brett
3 Oct 24    ii`* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)222Lawrence D'Oliveiro
3 Oct 24    ii +- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1Brett
3 Oct 24    ii +- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1Anton Ertl
3 Oct 24    ii `* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)219David Brown
3 Oct 24    ii  `* Byte ordering (was: Whether something is RISC or not)218Anton Ertl
3 Oct 24    ii   +- Re: Byte ordering (was: Whether something is RISC or not)1David Brown
4 Oct 24    ii   +* Re: Byte ordering (was: Whether something is RISC or not)215Lawrence D'Oliveiro
4 Oct 24    ii   i+- Re: Byte ordering1Lynn Wheeler
4 Oct 24    ii   i+* Re: Byte ordering (was: Whether something is RISC or not)211David Brown
4 Oct 24    ii   ii`* Re: Byte ordering (was: Whether something is RISC or not)210Anton Ertl
4 Oct 24    ii   ii +* Re: Byte ordering5BGB
5 Oct 24    ii   ii i`* Re: Byte ordering4MitchAlsup1
5 Oct 24    ii   ii i +* Re: Byte ordering2BGB
5 Oct 24    ii   ii i i`- Re: Byte ordering1Lawrence D'Oliveiro
5 Oct 24    ii   ii i `- Re: Byte ordering1Lawrence D'Oliveiro
5 Oct 24    ii   ii +* Re: Byte ordering (was: Whether something is RISC or not)13Lawrence D'Oliveiro
5 Oct 24    ii   ii i`* Re: Byte ordering (was: Whether something is RISC or not)12Brett
5 Oct 24    ii   ii i `* Re: Byte ordering (was: Whether something is RISC or not)11Anton Ertl
5 Oct 24    ii   ii i  `* Re: Byte ordering (was: Whether something is RISC or not)10Michael S
6 Oct 24    ii   ii i   +- Re: Byte ordering1Terje Mathisen
6 Oct 24    ii   ii i   `* Re: Byte ordering (was: Whether something is RISC or not)8Brett
7 Oct 24    ii   ii i    `* Re: Byte ordering (was: Whether something is RISC or not)7Lawrence D'Oliveiro
7 Oct 24    ii   ii i     `* Re: Byte ordering (was: Whether something is RISC or not)6Brett
7 Oct 24    ii   ii i      `* Re: Byte ordering (was: Whether something is RISC or not)5Michael S
7 Oct 24    ii   ii i       +* Re: Byte ordering2Stefan Monnier
7 Oct 24    ii   ii i       i`- Re: Byte ordering1Michael S
7 Oct 24    ii   ii i       `* Re: Byte ordering (was: Whether something is RISC or not)2Lawrence D'Oliveiro
8 Oct 24    ii   ii i        `- Re: Byte ordering1Terje Mathisen
6 Oct 24    ii   ii `* Re: Byte ordering191David Brown
6 Oct 24    ii   ii  `* Re: Byte ordering190Anton Ertl
6 Oct 24    ii   ii   `* Re: Byte ordering189John Dallman
7 Oct 24    ii   ii    +* Re: Byte ordering20Lawrence D'Oliveiro
8 Oct 24    ii   ii    i`* Re: Byte ordering19John Dallman
9 Oct 24    ii   ii    i +- VMS/NT memory management (was: Byte ordering)1Stefan Monnier
15 Oct 24    ii   ii    i +* Re: Byte ordering2Lawrence D'Oliveiro
15 Oct 24    ii   ii    i i`- Re: Byte ordering1MitchAlsup1
15 Oct 24    ii   ii    i `* Re: Byte ordering15Lawrence D'Oliveiro
15 Oct 24    ii   ii    i  +* Re: Byte ordering3Michael S
15 Oct 24    ii   ii    i  i+- Re: Byte ordering1John Dallman
18 Oct 24    ii   ii    i  i`- Re: Byte ordering1Lawrence D'Oliveiro
15 Oct 24    ii   ii    i  +* Re: Byte ordering9John Dallman
16 Oct 24    ii   ii    i  i+* Re: Byte ordering7George Neuner
16 Oct 24    ii   ii    i  ii`* Re: Byte ordering6Terje Mathisen
16 Oct 24    ii   ii    i  ii `* Re: Byte ordering5David Brown
17 Oct 24    ii   ii    i  ii  +* Re: Byte ordering2George Neuner
17 Oct 24    ii   ii    i  ii  i`- Re: Byte ordering1David Brown
17 Oct 24    ii   ii    i  ii  `* Re: clouds, not Byte ordering2John Levine
17 Oct 24    ii   ii    i  ii   `- Re: clouds, not Byte ordering1David Brown
18 Oct 24    ii   ii    i  i`- Re: Byte ordering1Lawrence D'Oliveiro
16 Oct 24    ii   ii    i  `* Re: Byte ordering2Paul A. Clayton
18 Oct 24    ii   ii    i   `- Re: Microkernels & Capabilities (was Re: Byte ordering)1Lawrence D'Oliveiro
7 Oct 24    ii   ii    `* 80286 protected mode168Anton Ertl
7 Oct 24    ii   ii     +* Re: 80286 protected mode5Lars Poulsen
7 Oct 24    ii   ii     i`* Re: 80286 protected mode4Terje Mathisen
7 Oct 24    ii   ii     i +- Re: 80286 protected mode1Michael S
7 Oct 24    ii   ii     i `* Re: 80286 protected mode2Lawrence D'Oliveiro
8 Oct 24    ii   ii     i  `- Re: 80286 protected mode1Terje Mathisen
7 Oct 24    ii   ii     +* Re: 80286 protected mode3Brett
7 Oct 24    ii   ii     i`* Re: 80286 protected mode2Michael S
7 Oct 24    ii   ii     i `- Re: 80286 protected mode1Brett
7 Oct 24    ii   ii     +- Re: 80286 protected mode1Lawrence D'Oliveiro
8 Oct 24    ii   ii     +* Re: 80286 protected mode152MitchAlsup1
8 Oct 24    ii   ii     i+* Re: 80286 protected mode4Lawrence D'Oliveiro
8 Oct 24    ii   ii     ii`* Re: 80286 protected mode3MitchAlsup1
9 Oct 24    ii   ii     ii +- Re: 80286 protected mode1David Brown
15 Oct 24    ii   ii     ii `- Re: 80286 protected mode1Lawrence D'Oliveiro
8 Oct 24    ii   ii     i`* Re: 80286 protected mode147Anton Ertl
8 Oct 24    ii   ii     i +- Re: 80286 protected mode1Robert Finch
9 Oct 24    ii   ii     i `* Re: 80286 protected mode145David Brown
9 Oct 24    ii   ii     i  +* Re: 80286 protected mode79MitchAlsup1
9 Oct 24    ii   ii     i  i`* Re: 80286 protected mode78David Brown
9 Oct 24    ii   ii     i  i `* Re: 80286 protected mode77Stephen Fuld
10 Oct 24    ii   ii     i  i  +* Re: 80286 protected mode2MitchAlsup1
10 Oct 24    ii   ii     i  i  i`- Re: 80286 protected mode1David Brown
10 Oct 24    ii   ii     i  i  +- Re: 80286 protected mode1David Brown
11 Oct 24    ii   ii     i  i  `* Re: 80286 protected mode73Tim Rentsch
15 Oct 24    ii   ii     i  i   `* Re: 80286 protected mode72Stefan Monnier
15 Oct 24    ii   ii     i  i    +* Re: 80286 protected mode30MitchAlsup1
16 Oct 24    ii   ii     i  i    i+* Re: 80286 protected mode25MitchAlsup1
16 Oct 24    ii   ii     i  i    ii+* Re: C and turtles, 80286 protected mode13John Levine
16 Oct 24    ii   ii     i  i    iii+* Re: C and turtles, 80286 protected mode7MitchAlsup1
16 Oct 24    ii   ii     i  i    iiii`* Re: C and turtles, 80286 protected mode6John Levine
17 Oct 24    ii   ii     i  i    iiii `* Re: C and turtles, 80286 protected mode5Thomas Koenig
20 Oct 24    ii   ii     i  i    iiii  `* Re: C and turtles, 80286 protected mode4Lawrence D'Oliveiro
20 Oct 24    ii   ii     i  i    iiii   `* Re: C and turtles, 80286 protected mode3George Neuner
22 Oct 24    ii   ii     i  i    iiii    `* Re: C and turtles, 80286 protected mode2Tim Rentsch
22 Oct 24    ii   ii     i  i    iiii     `- Re: C and turtles, 80286 protected mode1George Neuner
16 Oct 24    ii   ii     i  i    iii+- Re: C and turtles, 80286 protected mode1David Brown
16 Oct 24    ii   ii     i  i    iii`* Re: C and turtles, 80286 protected mode4Paul A. Clayton
17 Oct 24    ii   ii     i  i    iii +- Re: C and turtles, 80286 protected mode1David Brown
20 Oct 24    ii   ii     i  i    iii `* Re: C and turtles, 80286 protected mode2Lawrence D'Oliveiro
20 Oct 24    ii   ii     i  i    iii  `- Re: C and turtles, 80286 protected mode1Paul A. Clayton
16 Oct 24    ii   ii     i  i    ii+* Re: 80286 protected mode7Thomas Koenig
17 Oct 24    ii   ii     i  i    ii+* Re: 80286 protected mode3George Neuner
17 Oct 24    ii   ii     i  i    ii`- Re: 80286 protected mode1Tim Rentsch
16 Oct 24    ii   ii     i  i    i+* Re: 80286 protected mode3David Brown
17 Oct 24    ii   ii     i  i    i`- Re: 80286 protected mode1Tim Rentsch
16 Oct 24    ii   ii     i  i    `* Re: 80286 protected mode41David Brown
9 Oct 24    ii   ii     i  +* Re: 80286 protected mode51Thomas Koenig
13 Oct 24    ii   ii     i  `* Re: 80286 protected mode14Anton Ertl
8 Oct 24    ii   ii     `* Re: 80286 protected mode6John Levine
6 Oct 24    ii   i`* Re: Byte ordering (was: Whether something is RISC or not)2Michael S
4 Oct 24    ii   `- Re: Byte ordering (was: Whether something is RISC or not)1John Dallman
2 Oct 24    i`- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1Thomas Koenig
2 Oct 24    +* Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)5David Schultz
3 Oct 24    `- Re: Whether something is RISC or not (Re: PDP-8 theology, not Concertina II Progress)1Lawrence D'Oliveiro

Haut de la page

Les messages affichés proviennent d'usenet.

NewsPortal