C++ Logo

STD-PROPOSALS

Advanced search

Subject: Re: [std-proposals] Parralized instructions in language in CPU
From: Wesley Oliver (wesley.olis_at_[hidden])
Date: 2021-01-18 11:27:29


The other question depending on application spesific, can one have both a
sentinate and length or size for optimal performance. Where for certain
applications could determine that max value integer could represent could
be nil sentinal vavule loose 1value of resolution, which for most practical
applications loss of 1valvue never be a problem if garrantee usefully
values are lower than the max, it never be used, if must be used then,
better use a larger data type with greater resolution as the algorithm that
can design with sentinal will be much faster.

As well as how code re-ordering can be redone, to combine range checking
and Conditiinal checking. Always with compiler optermizations. Squeeze more
performance out. Could go 32bit to 40bit but have segfsult. And alignment
issues, but mabye in the future things will Improve that be less relavent
as cpu be more flexible.

Is the c++ language allowed to do experimental features or plugins so can
experiment more with things like this.

Kind Regards,

Wesley Oliver

On Mon, 18 Jan 2021, 18:21 Thiago Macieira via Std-Proposals, <
std-proposals_at_[hidden]> wrote:

> On Monday, 18 January 2021 00:12:18 PST Wesley Oliver via Std-Proposals
> wrote:
> > Hi,
> >
> > I would like to look at how to achieve the same performance that c++
> would
> > be capable of achieving with characters that are '\0' terminate
> physically
> > by there last data position. Because data that is null terminate, doesn't
> > require range checking.
>
> That's an incorrect statement. Just because the bounds are implicit does
> not
> mean range checking is optional. Anywhere where range checking should have
> been done, it still needs to be done. The only difference is that you must
> calculate the range before doing the checking.
>
> As a consequence of having to iterate to find the boundary, some
> operations
> become different.
>
> But it's not true that implicit termination or termination by sentinel
> makes
> things faster. In fact, from experience, it's quite the opposite. So
> you'll
> need to prove your hypothesis with data if it is the motivation factor for
> this.
>
> > So my question is how could we improve things, such that the typically
> > conditional bounds checking statements for int array or similar could be
> > reduce or written in slightly different form,
> > such that we can achieve the same performance as null terminated data.
>
> Please consider that "achieve the same performance as null-terminated
> data"
> currently means "reducing performance". It's not what you want.
>
> > For this to happen, it would require improved compiler and also
> > hypothetically invisigaing new cpu wiring or logic, that in 2025 years
> time
> > would give us massive performance boost, as we have figure out how to
> write
> > could that has many performance knocks in better way, by reduction or
> what
> > every that technic is, to reduce the number of instructions required.
>
> Investigating a new CPU is out of scope. Moreover, I don't think you've
> investigated CPUs sufficiently, since they do run superscalar and
> pipelined,
> meaning they will run a few instructions ahead. For example, you said:
>
> > So the ideas I have from above, would be conditional statements that
> could
> > be parralized, with out changing the logic of the program. so think that
> > both
> > numArray[i] == numArray2[j]
> > i < maxlen
>
> They ALREADY are parallelised today. Like I said, I don't think you've
> done
> enough investigation of current CPUs before making this proposal. If you
> want
> to talk about how to improve CPUs, aside from locating the right people to
> talk to, you need to come with hard data showing where they can do better
> and
> where time is spent in overhead. Similalry, if you want to get compilers
> to
> improve, show some hard numbers on how they are not using the existing CPU
> capabilities to the full extent.
>
> From experience talking to CPU architects (and I have!), they have enough
> areas to currently address that can improve code by more than 1%. So your
> barrier to get their attention is to show that they could get at least
> that
> much.
>
> > function match(char* str) {
> >
> > char* matchme = "matchme\0";
> >
> > let countMatch = 0;
> >
> > for(char* ch = str, char chd = ch*;chd != '\0';ch++, chd = *ch)
> > {
> > // could make this an inline function.
> >
> > char* chs = str, char chds = chs*;
> > char* mech = matchme, char mechd = *mech ;
> > while(true) { // sure that by now compiler optermized, could just say
> > loop
> > if (chds != mechd) {
> > if(mechd == '\0') {
> > break;
> > continue; // kicks out of the loop and skip the rest of the
> > parent look code, for a case that fail, in the case of successfully
> match,
> > then countMatch++ will execute.
> > }
> > }
> > mech++; mechd = * mech; chs++; chds = *chs ;
> > }
> > countMatch++;
> > }
> > }
>
> One other thing: if you're going to post code for which performance can be
> imiproved, you have to start with the state of the art. Your code above
> for
> substring matching is not the most optimal today.
>
> --
> Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
> Software Architect - Intel DPG Cloud Engineering
>
>
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>



STD-PROPOSALS list run by std-proposals-owner@lists.isocpp.org

Standard Proposals Archives on Google Groups