Date: Fri, 3 Apr 2026 11:35:54 +0500
I am really really sorry in the formating mistakes of my 16 points and for
posting the benchmark twice. Like I should've though thrice before
posting(proof read my post) so a million sorries.
Regards, Muneem.
On Fri, 3 Apr 2026, 9:39 am Muneem, <itfllow123_at_[hidden]> wrote:
> no one asked for this reply but I just wanted to keep the records straight.
>
> Here is why anyone who has reflections in mind (as a base for this
> solution) might be wrong:
> 1.reflectors aren't for subscripting(are a completely different thing(they
> are like unions and std::variants)), but rather a better version of
> std::visit, so subscripting using reflectors is using the wrong tool for
> the wrong solution and dosent send proper intent to the compiler .
> 2. Even if reflectors were optimized enough to be fast even when I use
> proxy objects in it(to act as indexes into a heterogeneous list), the
> compiler can't optimize freely without it backfiring(code bloating). Why?
> Because my tool has at least two contexts available to it (more if you guys
> propose), the context of defining the object, the context of using that
> object, and (if the subscript can be user defined objects that can return a
> integer through some semantical promise (using concepts to enforce)) the
> context of the subscript object and it's type. All this would give enough
> context for the compiler to optimize without backfiring, that is to
> optimize for the whole project, rather than one expression.
> 3. Its less verbose and would incur less compile time overhead for more
> potential rewards.
>
> Anyone who thought of this is a genius, but I just wanted to clarify on
> why he might not be right.
>
>
>
>
> Regards, Muneem.
>
> On Fri, 3 Apr 2026, 7:43 am Muneem, <itfllow123_at_[hidden]> wrote:
>
>> Hi!
>> 1. Even if it wasn't(and I was wrong), there are many more
>> examples(spaceship operator, constexpr if statements, consteval, the list
>> is too big to list all in the open). The point isn't that I am perfect but
>> rather that my proposal is right.
>> 2. I said what I said because Bjarne Strourstrup in his book (C++
>> programming 4th edition, page 675, Section 23.3.2 ) " That point is called
>> the first point of instantiation (ยง26.3.3). The implementation is allowed
>> to postpone essentially all checking until the program is linked, and for
>> some errors link time is also the earliest point when complete checking is
>> possible. Independently of when checking is done, the same set of rules is
>> checked. ", but maybe I was wrong.
>> 3. I am really really really sorry for sending too many emails.
>> 4. In the real world(I know you know more (intel goated) than me but let
>> me clarify), you dont have constant, in real world, the index can be real
>> time, infact, for anything constant, tuples work even for
>> heterogeneous sets. I dont know how you get a constant but if your input is
>> coming a real time source then in my opinion it is impossible to make it be
>> constant.
>> 5. I used the cpu profiling of visual studio 2022
>> Time taken for looping 100000000000 times over a line that adds into a
>> volatile x from a index in obj read by std::cin:
>>
>> 17-20s switch 2 minutes 20s for visit, 11 for ternary statements, 20 for
>> subscript.
>>
>> CPU(termed as processer usage in visual studio) usage (in percetager):
>> [4-12] percent for switch.
>> [8- 20] for std::visit
>> [4-9] percent for ternary.
>> [4-12] for subscripting.
>>
>> 6. I updated my proposal so I highly suggest that you ask AI to summarize
>> the thing or read it if you will.
>>
>> 7. I used the cpu profiling of visual studio 2022
>> Time taken for looping 100000000000 times over a line that adds into a
>> volatile x from a index in obj read by std::cin:
>>
>> 17-20s switch 2 minutes 20s for visit, 11 for ternary statements, 20 for
>> subscript.
>>
>> CPU(termed as processer usage in visual studio) usage (in percetager):
>> [4-12] percent for switch.
>> [8- 20] for std::visit
>> [4-9] percent for ternary.
>> [4-12] for subscripting.
>>
>> 8. The point on optimization dosent just apply on making subscripting
>> into heterogeneous lists like existing indexing, but to rather make a
>> syntactic construct that makes certain semantic promises(that I hope you
>> guys propose) which would help the compiler optimize it for the specific
>> cases where we want a balance between code bloat and runtime performance
>> (ie, we dont know which one to use swtich case or std::visit). Adding
>> semantical "promises" to constructs make constructs powerful than raw
>> assembly. Recall spaceship operators and the fact that they have semantic
>> promises(through the return type) that make the compiler know what it can
>> expect better, and leads to less verbose code than having to define every
>> comparison opreator's default function.
>> 9. I do read assembly, but in most real world examples, the assembly can
>> vary widely so you cant depend on the assembly code to know if your aproach
>> works or not. I showed you the benchmarks, you said that they can lie, so
>> can assembly then, like compilers have potential layers of translation
>> inside each phase, which coupled with unspecified behaviour can mean that
>> assembly can lie.
>> 10. My main goal is to get someone to vouch for my proposal and add more
>> semantical requirements, not to defend whether templates can be
>> instantiated or not.
>> 11. The reason why I made std::visit static (in my benchmarks yesterday
>> and in the benchmarks on the latest proposal on my github:
>> https://github.com/HjaldrKhilji/Future-potential-ISO-porposals/blob/main/Extension%20to%20runtime%20polymorphism.txt)
>> was because the variables that the switch and ternary returned were also
>> global variables, so I was trying to level the field by making the
>> variables global, variant static (but return literals).
>> 12.The code that you might find fair, but I find too much in favor of
>> std::visit, but std::visit still loses:https://godbolt.org/z/aP488n1cj,
>> benchmarks dont change accoridng to visual studio profiler.
>> 13. I used std::variant to show that just in case someone asks me to use
>> proxies or a type erasure system like using references in std:variant or
>> using virtual functions, the performance would be bad and the code would
>> still be verbose, leaving only to the verbose switch/if/ternary statements.
>> 14. The microbench is exactly what I am trying to prove that a new
>> syntactic construct won't just help with a few nanoseconds, but help with
>> many cases of the same thing to work better in conjunction with the rest of
>> the project.
>> 15. The reason that I didn't make the fact that (according to you) the
>> switch statement had the outer loop optimized to 3 loops does dont branch
>> is because that argument is pointless, my point isn't that why switch
>> statements are better, but rather on why my solution is better, and this
>> benchmark was to show that std::visit and polymorphism does not address all
>> problems that manual branching has (code bloat (potentially) causing worse
>> performance than redirection (huge respect and credit to adrian for making
>> me remember this point)).
>> 16. Again, if I paste a whole large example, I wouldn't convey my point
>> which is that we need something to convey intent better when subscripting
>> heterogeneous sets.
>> std::visit fails, and switch case statements backfire because of
>> potential code bloat. The benchmark was to show that the compiler can
>> optimize performance for smaller projects where It knows that it can be
>> faster in one specific case, a new construct would help the compiler know
>> what the goal of using that construct is, so that it can make it faster (if
>> it can) or even make it slower if it helps the rest of the project. Again,
>> the construct wont make a small example fast, but rather a whole ton of.
>> The reason that I didn't show a real world example for switch statements
>> is that I am not arguing for switch statements to be used everywhere, but
>> rather the opposite.
>>
>>
>> On Fri, Apr 3, 2026 at 6:31 AM Thiago Macieira via Std-Proposals <
>> std-proposals_at_[hidden]> wrote:
>>
>>> On Thursday, 2 April 2026 12:32:34 Pacific Daylight Time Muneem via Std-
>>> Proposals wrote:
>>> > It's different for templates because the linker has to plug in some
>>> type,
>>>
>>> No. There is no linker in template expansion.
>>>
>>> --
>>> Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
>>> Principal Engineer - Intel Data Center - Platform & Sys. Eng.
>>> --
>>> Std-Proposals mailing list
>>> Std-Proposals_at_[hidden]
>>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>>
>>
posting the benchmark twice. Like I should've though thrice before
posting(proof read my post) so a million sorries.
Regards, Muneem.
On Fri, 3 Apr 2026, 9:39 am Muneem, <itfllow123_at_[hidden]> wrote:
> no one asked for this reply but I just wanted to keep the records straight.
>
> Here is why anyone who has reflections in mind (as a base for this
> solution) might be wrong:
> 1.reflectors aren't for subscripting(are a completely different thing(they
> are like unions and std::variants)), but rather a better version of
> std::visit, so subscripting using reflectors is using the wrong tool for
> the wrong solution and dosent send proper intent to the compiler .
> 2. Even if reflectors were optimized enough to be fast even when I use
> proxy objects in it(to act as indexes into a heterogeneous list), the
> compiler can't optimize freely without it backfiring(code bloating). Why?
> Because my tool has at least two contexts available to it (more if you guys
> propose), the context of defining the object, the context of using that
> object, and (if the subscript can be user defined objects that can return a
> integer through some semantical promise (using concepts to enforce)) the
> context of the subscript object and it's type. All this would give enough
> context for the compiler to optimize without backfiring, that is to
> optimize for the whole project, rather than one expression.
> 3. Its less verbose and would incur less compile time overhead for more
> potential rewards.
>
> Anyone who thought of this is a genius, but I just wanted to clarify on
> why he might not be right.
>
>
>
>
> Regards, Muneem.
>
> On Fri, 3 Apr 2026, 7:43 am Muneem, <itfllow123_at_[hidden]> wrote:
>
>> Hi!
>> 1. Even if it wasn't(and I was wrong), there are many more
>> examples(spaceship operator, constexpr if statements, consteval, the list
>> is too big to list all in the open). The point isn't that I am perfect but
>> rather that my proposal is right.
>> 2. I said what I said because Bjarne Strourstrup in his book (C++
>> programming 4th edition, page 675, Section 23.3.2 ) " That point is called
>> the first point of instantiation (ยง26.3.3). The implementation is allowed
>> to postpone essentially all checking until the program is linked, and for
>> some errors link time is also the earliest point when complete checking is
>> possible. Independently of when checking is done, the same set of rules is
>> checked. ", but maybe I was wrong.
>> 3. I am really really really sorry for sending too many emails.
>> 4. In the real world(I know you know more (intel goated) than me but let
>> me clarify), you dont have constant, in real world, the index can be real
>> time, infact, for anything constant, tuples work even for
>> heterogeneous sets. I dont know how you get a constant but if your input is
>> coming a real time source then in my opinion it is impossible to make it be
>> constant.
>> 5. I used the cpu profiling of visual studio 2022
>> Time taken for looping 100000000000 times over a line that adds into a
>> volatile x from a index in obj read by std::cin:
>>
>> 17-20s switch 2 minutes 20s for visit, 11 for ternary statements, 20 for
>> subscript.
>>
>> CPU(termed as processer usage in visual studio) usage (in percetager):
>> [4-12] percent for switch.
>> [8- 20] for std::visit
>> [4-9] percent for ternary.
>> [4-12] for subscripting.
>>
>> 6. I updated my proposal so I highly suggest that you ask AI to summarize
>> the thing or read it if you will.
>>
>> 7. I used the cpu profiling of visual studio 2022
>> Time taken for looping 100000000000 times over a line that adds into a
>> volatile x from a index in obj read by std::cin:
>>
>> 17-20s switch 2 minutes 20s for visit, 11 for ternary statements, 20 for
>> subscript.
>>
>> CPU(termed as processer usage in visual studio) usage (in percetager):
>> [4-12] percent for switch.
>> [8- 20] for std::visit
>> [4-9] percent for ternary.
>> [4-12] for subscripting.
>>
>> 8. The point on optimization dosent just apply on making subscripting
>> into heterogeneous lists like existing indexing, but to rather make a
>> syntactic construct that makes certain semantic promises(that I hope you
>> guys propose) which would help the compiler optimize it for the specific
>> cases where we want a balance between code bloat and runtime performance
>> (ie, we dont know which one to use swtich case or std::visit). Adding
>> semantical "promises" to constructs make constructs powerful than raw
>> assembly. Recall spaceship operators and the fact that they have semantic
>> promises(through the return type) that make the compiler know what it can
>> expect better, and leads to less verbose code than having to define every
>> comparison opreator's default function.
>> 9. I do read assembly, but in most real world examples, the assembly can
>> vary widely so you cant depend on the assembly code to know if your aproach
>> works or not. I showed you the benchmarks, you said that they can lie, so
>> can assembly then, like compilers have potential layers of translation
>> inside each phase, which coupled with unspecified behaviour can mean that
>> assembly can lie.
>> 10. My main goal is to get someone to vouch for my proposal and add more
>> semantical requirements, not to defend whether templates can be
>> instantiated or not.
>> 11. The reason why I made std::visit static (in my benchmarks yesterday
>> and in the benchmarks on the latest proposal on my github:
>> https://github.com/HjaldrKhilji/Future-potential-ISO-porposals/blob/main/Extension%20to%20runtime%20polymorphism.txt)
>> was because the variables that the switch and ternary returned were also
>> global variables, so I was trying to level the field by making the
>> variables global, variant static (but return literals).
>> 12.The code that you might find fair, but I find too much in favor of
>> std::visit, but std::visit still loses:https://godbolt.org/z/aP488n1cj,
>> benchmarks dont change accoridng to visual studio profiler.
>> 13. I used std::variant to show that just in case someone asks me to use
>> proxies or a type erasure system like using references in std:variant or
>> using virtual functions, the performance would be bad and the code would
>> still be verbose, leaving only to the verbose switch/if/ternary statements.
>> 14. The microbench is exactly what I am trying to prove that a new
>> syntactic construct won't just help with a few nanoseconds, but help with
>> many cases of the same thing to work better in conjunction with the rest of
>> the project.
>> 15. The reason that I didn't make the fact that (according to you) the
>> switch statement had the outer loop optimized to 3 loops does dont branch
>> is because that argument is pointless, my point isn't that why switch
>> statements are better, but rather on why my solution is better, and this
>> benchmark was to show that std::visit and polymorphism does not address all
>> problems that manual branching has (code bloat (potentially) causing worse
>> performance than redirection (huge respect and credit to adrian for making
>> me remember this point)).
>> 16. Again, if I paste a whole large example, I wouldn't convey my point
>> which is that we need something to convey intent better when subscripting
>> heterogeneous sets.
>> std::visit fails, and switch case statements backfire because of
>> potential code bloat. The benchmark was to show that the compiler can
>> optimize performance for smaller projects where It knows that it can be
>> faster in one specific case, a new construct would help the compiler know
>> what the goal of using that construct is, so that it can make it faster (if
>> it can) or even make it slower if it helps the rest of the project. Again,
>> the construct wont make a small example fast, but rather a whole ton of.
>> The reason that I didn't show a real world example for switch statements
>> is that I am not arguing for switch statements to be used everywhere, but
>> rather the opposite.
>>
>>
>> On Fri, Apr 3, 2026 at 6:31 AM Thiago Macieira via Std-Proposals <
>> std-proposals_at_[hidden]> wrote:
>>
>>> On Thursday, 2 April 2026 12:32:34 Pacific Daylight Time Muneem via Std-
>>> Proposals wrote:
>>> > It's different for templates because the linker has to plug in some
>>> type,
>>>
>>> No. There is no linker in template expansion.
>>>
>>> --
>>> Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
>>> Principal Engineer - Intel Data Center - Platform & Sys. Eng.
>>> --
>>> Std-Proposals mailing list
>>> Std-Proposals_at_[hidden]
>>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>>
>>
Received on 2026-04-03 06:36:10