On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
> Are there any examples of current platforms/architectures where
> sizeof(void*) > sizeof(int*) ?
>
> - Alf

Are there any old platforms where this was the case also ?

Lynn

On 8/30/2021 7:59 PM, Lynn McGuire wrote:
> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>> Are there any examples of current platforms/architectures where
>> sizeof(void*) > sizeof(int*) ?
>>
>> - Alf
>
> Are there any old platforms where this was the case also ?
>
> Lynn
>

Quoted from SO

https://stackoverflow.com/questions/24867814/printfp-and-casting-to-void/24867850#24867850

===
An example of an implementation with different pointer types
representation is Cray PVP where the representation of pointer types is
64-bit for `void *` and `char *` but 32-bit for the other pointer types.

See "Cray C/C++ Reference Manual", Table 3. in "9.1.2.2"
http://docs.cray.com/books/004-2179-003/004-2179-003-manual.pdf
===

--
Best regards,
Andrey Tarasevich

On 8/30/2021 9:53 PM, Andrey Tarasevich wrote:
> On 8/30/2021 7:59 PM, Lynn McGuire wrote:
>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>> Are there any examples of current platforms/architectures where
>>> sizeof(void*) > sizeof(int*) ?
>>>
>>> - Alf
>>
>> Are there any old platforms where this was the case also ?
>>
>> Lynn
>>
>
> Quoted from SO
>
> https://stackoverflow.com/questions/24867814/printfp-and-casting-to-void/24867850#24867850
>
>
> ===
> An example of an implementation with different pointer types
> representation is Cray PVP where the representation of pointer types is
> 64-bit for `void *` and `char *` but 32-bit for the other pointer types.
>
> See "Cray C/C++ Reference Manual", Table 3. in "9.1.2.2"
> http://docs.cray.com/books/004-2179-003/004-2179-003-manual.pdf
> ===
>

The link to the Reference Manual is broken, but the same manual is
currently available from here

https://manualzz.com/doc/13740819/cray-c-c---reference-manual-004–2179–003

Also, 32 bits were apparently used in value representation of those
"other pointer types", while the object representation was still the
same 64 bits, implying 32 padding bits.

--
Best regards,
Andrey Tarasevich

On 2021-08-31 at 04:59, Lynn McGuire wrote:
> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>> Are there any examples of current platforms/architectures where
>> sizeof(void*) > sizeof(int*) ?
>>
>> - Alf
>
> Are there any old platforms where this was the case also ?
>
> Lynn
>

It would have been word-addressed machines, where char* (and thus void*)
needed a part-word indicator in addition to the word address.

Haven't seen any of those lately.

On 8/31/2021 2:01 AM, Bo Persson wrote:
> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>> Are there any examples of current platforms/architectures where
>>> sizeof(void*) > sizeof(int*) ?
>>>
>>> - Alf
>>
>> Are there any old platforms where this was the case also ?
>>
>> Lynn
>>
>
> It would have been word-addressed machines, where char* (and thus void*)
> needed a part-word indicator in addition to the word address.
>
> Haven't seen any of those lately.

Yeah, can't think of a reason why to build anything other than a byte
addressable machine. I don't even see expansion of the byte from 8 bits
to 16 bits.

Lynn

On 01/09/2021 06:46, Lynn McGuire wrote:
> On 8/31/2021 2:01 AM, Bo Persson wrote:
>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>> Are there any examples of current platforms/architectures where
>>>> sizeof(void*) > sizeof(int*) ?
>>>>
>>>> - Alf
>>>
>>> Are there any old platforms where this was the case also ?
>>>
>>> Lynn
>>>
>>
>> It would have been word-addressed machines, where char* (and thus
>> void*) needed a part-word indicator in addition to the word address.
>>
>> Haven't seen any of those lately.
>
> Yeah, can't think of a reason why to build anything other than a byte
> addressable machine.  I don't even see expansion of the byte from 8 bits
> to 16 bits.

By the definition of "byte", all C and C++ targets are byte addressable.

>
> Lynn
>

There are plenty of them - they are all around you, but you don't notice
them. DSP's are often word addressable, where "word" can be anything
from 12 to perhaps 36 bits, with a range of values in between. And
there are very good reasons for not making them 8-bit byte addressable.

But it is relatively rare that there are C++ compilers for these
systems, and they have larger CHAR_BIT rather than having different
sized pointers for char* and int*.

On Tuesday, 31 August 2021 at 04:00:06 UTC+1, Lynn McGuire wrote:
> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
> > Are there any examples of current platforms/architectures where
> > sizeof(void*) > sizeof(int*) ?
> >
> > - Alf
>
> Are there any old platforms where this was the case also ?

Historically Prime (aka Pr1me) 50-series machines had this.

The underlying architecture was based on 16-bit "half-words", with the memory organized into segments.

As I recall, the standard 32-pointer held a 2-bit protection/privilege ring number, 14-bits of segment number and 16-bits of offset to a half-word in the segment.
Accessing 8-bit bytes was via a 48-bit pointer with the final 16-bits dedicated to holding 0 or 1 as the offset of the byte in the half-word (honestly).

In addition, 8-bit ASCII had the sign bit on rather than off, and native I/O was half-word oriented.

Porting C to it was, I gather, a bit of a pain. Oh - and when you incremented a pointer past the end of a segment, it would wrap back to the beginning of that segment, not move on to the next, if I recall correctly.

For a later C release, an alternative form of 32-bit addressing was added - I can't remember the details, but think the ring number bits were repurposed.

On Wednesday, 1 September 2021 at 11:30:12 UTC+1, Mark Bluemel wrote:
> On Tuesday, 31 August 2021 at 04:00:06 UTC+1, Lynn McGuire wrote:
> > On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
> > > Are there any examples of current platforms/architectures where
> > > sizeof(void*) > sizeof(int*) ?
> > >
> > > - Alf
> >
> > Are there any old platforms where this was the case also ?
> Historically Prime (aka Pr1me) 50-series machines had this.
>
> The underlying architecture was based on 16-bit "half-words", with the memory organized into segments.
>
> As I recall, the standard 32-pointer held a 2-bit protection/privilege ring number, 14-bits of segment number and 16-bits of offset to a half-word in the segment.
> Accessing 8-bit bytes was via a 48-bit pointer with the final 16-bits dedicated to holding 0 or 1 as the offset of the byte in the half-word (honestly).
>
> In addition, 8-bit ASCII had the sign bit on rather than off, and native I/O was half-word oriented.
>
> Porting C to it was, I gather, a bit of a pain. Oh - and when you incremented a pointer past the end of a segment, it would wrap back to the beginning of that segment, not move on to the next, if I recall correctly.
>
> For a later C release, an alternative form of 32-bit addressing was added - I can't remember the details, but think the ring number bits were repurposed.

See also <http://www.c-faq.com/null/machexamp.html>

Lynn McGuire <lynnmcguire5@gmail.com> writes:
>On 8/31/2021 2:01 AM, Bo Persson wrote:
>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>> Are there any examples of current platforms/architectures where
>>>> sizeof(void*) > sizeof(int*) ?
>>>>
>>>> - Alf
>>>
>>> Are there any old platforms where this was the case also ?
>>>
>>> Lynn
>>>
>>
>> It would have been word-addressed machines, where char* (and thus void*)
>> needed a part-word indicator in addition to the word address.
>>
>> Haven't seen any of those lately.
>
>Yeah, can't think of a reason why to build anything other than a byte
>addressable machine.

Strong words from a fortran programmer. How many CPUs have you
designed in the last forty years?

As David points out, many DSPs are not byte addressable, and there
are other machines _STILL IN DAILY PRODUCTION_ that don't have
byte addressability, yet support C compilers (e.g. both flavors
of Unisys Clearpath).

A modern DPU:

https://www.anandtech.com/show/16790/marvell-announces-octeon-10-dpu-family-first-to-5nm-with-n2-cpus

Am 01.09.2021 um 17:32 schrieb Scott Lurndal:
> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>> On 8/31/2021 2:01 AM, Bo Persson wrote:
>>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>>> Are there any examples of current platforms/architectures where
>>>>> sizeof(void*) > sizeof(int*) ?
>>>>>
>>>>> - Alf
>>>>
>>>> Are there any old platforms where this was the case also ?
>>>>
>>>> Lynn
>>>>
>>>
>>> It would have been word-addressed machines, where char* (and thus void*)
>>> needed a part-word indicator in addition to the word address.
>>>
>>> Haven't seen any of those lately.
>>
>> Yeah, can't think of a reason why to build anything other than a byte
>> addressable machine.
>
> Strong words from a fortran programmer. How many CPUs have you
> designed in the last forty years?
>
> As David points out, many DSPs are not byte addressable, and there
> are other machines _STILL IN DAILY PRODUCTION_ that don't have
> byte addressability, yet support C compilers (e.g. both flavors
> of Unisys Clearpath).

These are not "many" but according to the examples given yet
a homeopathic part.

On 01/09/2021 19:20, Bonita Montero wrote:
> Am 01.09.2021 um 17:32 schrieb Scott Lurndal:
>> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>>> On 8/31/2021 2:01 AM, Bo Persson wrote:
>>>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>>>> Are there any examples of current platforms/architectures where
>>>>>> sizeof(void*) > sizeof(int*) ?
>>>>>>
>>>>>> - Alf
>>>>>
>>>>> Are there any old platforms where this was the case also ?
>>>>>
>>>>> Lynn
>>>>>
>>>>
>>>> It would have been word-addressed machines, where char* (and thus
>>>> void*)
>>>> needed a part-word indicator in addition to the word address.
>>>>
>>>> Haven't seen any of those lately.
>>>
>>> Yeah, can't think of a reason why to build anything other than a byte
>>> addressable machine.
>>
>> Strong words from a fortran programmer.   How many CPUs have you
>> designed in the last forty years?
>>
>> As David points out, many DSPs are not byte addressable, and there
>> are other machines _STILL IN DAILY PRODUCTION_ that don't have
>> byte addressability, yet support C compilers (e.g. both flavors
>> of Unisys Clearpath).
>
> These are not "many" but according to the examples given yet
> a homeopathic part.
>

I believe there are more dedicated DSP's produced per year than 64-bit
cpus. There may even be more 4-bit microcontrollers still produced than
64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
(according to a quick google search). There are DSPs and other
specialist microcontrollers with minimum order quantities of millions.
I know of microcontrollers (programmable in C, and in this case with
8-bit char) that cost $0.03 each.

I don't know what you mean by "not many". In terms of numbers of units,
8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
devices by orders of magnitude - and 32-bit devices similarly outsell PC
processors. In terms of numbers of designs or chips available, it's the
same story. PC processors lead in terms of total cost, and lead hugely
in terms of the number of different people programming on them.

On 9/1/2021 5:30 AM, Mark Bluemel wrote:
> On Tuesday, 31 August 2021 at 04:00:06 UTC+1, Lynn McGuire wrote:
>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>> Are there any examples of current platforms/architectures where
>>> sizeof(void*) > sizeof(int*) ?
>>>
>>> - Alf
>>
>> Are there any old platforms where this was the case also ?
>
> Historically Prime (aka Pr1me) 50-series machines had this.
>
> The underlying architecture was based on 16-bit "half-words", with the memory organized into segments.
>
> As I recall, the standard 32-pointer held a 2-bit protection/privilege ring number, 14-bits of segment number and 16-bits of offset to a half-word in the segment.
> Accessing 8-bit bytes was via a 48-bit pointer with the final 16-bits dedicated to holding 0 or 1 as the offset of the byte in the half-word (honestly).
>
> In addition, 8-bit ASCII had the sign bit on rather than off, and native I/O was half-word oriented.
>
> Porting C to it was, I gather, a bit of a pain. Oh - and when you incremented a pointer past the end of a segment, it would wrap back to the beginning of that segment, not move on to the next, if I recall correctly.
>
> For a later C release, an alternative form of 32-bit addressing was added - I can't remember the details, but think the ring number bits were repurposed.

Good night, what a mess !

We had a Prime model 450 in 1977 and a model 750 in 1980 ???. And then
a model 2xxx ??? in 1985 ??? The only compilers they had that I knew of
were the Fortran, Cobol, and PL/I ???

Lynn

On 9/1/2021 12:44 PM, David Brown wrote:
> On 01/09/2021 19:20, Bonita Montero wrote:
>> Am 01.09.2021 um 17:32 schrieb Scott Lurndal:
>>> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>>>> On 8/31/2021 2:01 AM, Bo Persson wrote:
>>>>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>>>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>>>>> Are there any examples of current platforms/architectures where
>>>>>>> sizeof(void*) > sizeof(int*) ?
>>>>>>>
>>>>>>> - Alf
>>>>>>
>>>>>> Are there any old platforms where this was the case also ?
>>>>>>
>>>>>> Lynn
>>>>>>
>>>>>
>>>>> It would have been word-addressed machines, where char* (and thus
>>>>> void*)
>>>>> needed a part-word indicator in addition to the word address.
>>>>>
>>>>> Haven't seen any of those lately.
>>>>
>>>> Yeah, can't think of a reason why to build anything other than a byte
>>>> addressable machine.
>>>
>>> Strong words from a fortran programmer.   How many CPUs have you
>>> designed in the last forty years?
>>>
>>> As David points out, many DSPs are not byte addressable, and there
>>> are other machines _STILL IN DAILY PRODUCTION_ that don't have
>>> byte addressability, yet support C compilers (e.g. both flavors
>>> of Unisys Clearpath).
>>
>> These are not "many" but according to the examples given yet
>> a homeopathic part.
>>
>
> I believe there are more dedicated DSP's produced per year than 64-bit
> cpus. There may even be more 4-bit microcontrollers still produced than
> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> (according to a quick google search). There are DSPs and other
> specialist microcontrollers with minimum order quantities of millions.
> I know of microcontrollers (programmable in C, and in this case with
> 8-bit char) that cost $0.03 each.
>
> I don't know what you mean by "not many". In terms of numbers of units,
> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
> devices by orders of magnitude - and 32-bit devices similarly outsell PC
> processors. In terms of numbers of designs or chips available, it's the
> same story. PC processors lead in terms of total cost, and lead hugely
> in terms of the number of different people programming on them.

Your numbers for AMD and Intel are WAY off. There were 275 million PCs
sold in 2020 containing Intel or AMD cpus.
https://www.statista.com/statistics/273495/global-shipments-of-personal-computers-since-2006/

Lynn

On 01/09/2021 20:17, Lynn McGuire wrote:
> On 9/1/2021 12:44 PM, David Brown wrote:
>> On 01/09/2021 19:20, Bonita Montero wrote:
>>> Am 01.09.2021 um 17:32 schrieb Scott Lurndal:
>>>> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>>>>> On 8/31/2021 2:01 AM, Bo Persson wrote:
>>>>>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>>>>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>>>>>> Are there any examples of current platforms/architectures where
>>>>>>>> sizeof(void*) > sizeof(int*) ?
>>>>>>>>
>>>>>>>> - Alf
>>>>>>>
>>>>>>> Are there any old platforms where this was the case also ?
>>>>>>>
>>>>>>> Lynn
>>>>>>>
>>>>>>
>>>>>> It would have been word-addressed machines, where char* (and thus
>>>>>> void*)
>>>>>> needed a part-word indicator in addition to the word address.
>>>>>>
>>>>>> Haven't seen any of those lately.
>>>>>
>>>>> Yeah, can't think of a reason why to build anything other than a byte
>>>>> addressable machine.
>>>>
>>>> Strong words from a fortran programmer.   How many CPUs have you
>>>> designed in the last forty years?
>>>>
>>>> As David points out, many DSPs are not byte addressable, and there
>>>> are other machines _STILL IN DAILY PRODUCTION_ that don't have
>>>> byte addressability, yet support C compilers (e.g. both flavors
>>>> of Unisys Clearpath).
>>>
>>> These are not "many" but according to the examples given yet
>>> a homeopathic part.
>>>
>>
>> I believe there are more dedicated DSP's produced per year than 64-bit
>> cpus.  There may even be more 4-bit microcontrollers still produced than
>> 64-bit cpus.  AMD and Intel each sell perhaps 500,000 processors a year
>> (according to a quick google search).  There are DSPs and other
>> specialist microcontrollers with minimum order quantities of millions.
>> I know of microcontrollers (programmable in C, and in this case with
>> 8-bit char) that cost $0.03 each.
>>
>> I don't know what you mean by "not many".  In terms of numbers of units,
>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>> processors.  In terms of numbers of designs or chips available, it's the
>> same story.  PC processors lead in terms of total cost, and lead hugely
>> in terms of the number of different people programming on them.
>
> Your numbers for AMD and Intel are WAY off.  There were 275 million PCs
> sold in 2020 containing Intel or AMD cpus.
>
> https://www.statista.com/statistics/273495/global-shipments-of-personal-computers-since-2006/
>

They did look odd when I read them - perhaps the statistics page I found
was referring to a limited type of chip. Anyway, small devices still
massively outsell them in numbers.

Am 01.09.2021 um 19:44 schrieb David Brown:

> I believe there are more dedicated DSP's produced per year than 64-bit
> cpus. There may even be more 4-bit microcontrollers still produced than
> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> (according to a quick google search). There are DSPs and other
> specialist microcontrollers with minimum order quantities of millions.
> I know of microcontrollers (programmable in C, and in this case with
> 8-bit char) that cost $0.03 each.

The issue is not about whether they're can't address raw bytes but just
larger quantities. The issue is about machines with typed pointers, i.e.
when you cast a pointer they change their bit-representation, and these
exist in a homeopathic amount.

Lynn McGuire <lynnmcguire5@gmail.com> writes:
>On 9/1/2021 12:44 PM, David Brown wrote:

>> I believe there are more dedicated DSP's produced per year than 64-bit
>> cpus. There may even be more 4-bit microcontrollers still produced than
>> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
>> (according to a quick google search). There are DSPs and other
>> specialist microcontrollers with minimum order quantities of millions.
>> I know of microcontrollers (programmable in C, and in this case with
>> 8-bit char) that cost $0.03 each.
>>
>> I don't know what you mean by "not many". In terms of numbers of units,
>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>> processors. In terms of numbers of designs or chips available, it's the
>> same story. PC processors lead in terms of total cost, and lead hugely
>> in terms of the number of different people programming on them.
>
>Your numbers for AMD and Intel are WAY off. There were 275 million PCs
>sold in 2020 containing Intel or AMD cpus.

300 million X86 (intel, amd) processors per year, including servers.
6 billion arm processors per _quarter_.
1.5 billion ARC processors per year.
400 million MIPS processors per year.

On 01/09/2021 18:44, David Brown wrote:
> On 01/09/2021 19:20, Bonita Montero wrote:
>> Am 01.09.2021 um 17:32 schrieb Scott Lurndal:
>>> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>>>> On 8/31/2021 2:01 AM, Bo Persson wrote:
>>>>> On 2021-08-31 at 04:59, Lynn McGuire wrote:
>>>>>> On 8/29/2021 1:57 PM, Alf P. Steinbach wrote:
>>>>>>> Are there any examples of current platforms/architectures where
>>>>>>> sizeof(void*) > sizeof(int*) ?
>>>>>>>
>>>>>>> - Alf
>>>>>>
>>>>>> Are there any old platforms where this was the case also ?
>>>>>>
>>>>>> Lynn
>>>>>>
>>>>>
>>>>> It would have been word-addressed machines, where char* (and thus
>>>>> void*)
>>>>> needed a part-word indicator in addition to the word address.
>>>>>
>>>>> Haven't seen any of those lately.
>>>>
>>>> Yeah, can't think of a reason why to build anything other than a byte
>>>> addressable machine.
>>>
>>> Strong words from a fortran programmer.   How many CPUs have you
>>> designed in the last forty years?
>>>
>>> As David points out, many DSPs are not byte addressable, and there
>>> are other machines _STILL IN DAILY PRODUCTION_ that don't have
>>> byte addressability, yet support C compilers (e.g. both flavors
>>> of Unisys Clearpath).
>>
>> These are not "many" but according to the examples given yet
>> a homeopathic part.
>>
>
> I believe there are more dedicated DSP's produced per year than 64-bit
> cpus. There may even be more 4-bit microcontrollers still produced than
> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> (according to a quick google search). There are DSPs and other
> specialist microcontrollers with minimum order quantities of millions.
> I know of microcontrollers (programmable in C, and in this case with
> 8-bit char) that cost $0.03 each.
>
> I don't know what you mean by "not many". In terms of numbers of units,
> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
> devices by orders of magnitude - and 32-bit devices similarly outsell PC
> processors. In terms of numbers of designs or chips available, it's the
> same story. PC processors lead in terms of total cost, and lead hugely
> in terms of the number of different people programming on them.

So, you want a programming language still extensively used for
conventional 64-bit computer chips (eg. the x64s and ARMs) to have its
design dictated by (or even crippled by) the limitations of a 4-bit
microcontroller used in some toaster?

It's likely anyway that that 4-bit C compiler will be implementing a
rather specific dialect of C, without worrying too much about how well
it might work on 64-bit machines.

But we're not allowed to do the same when it's the other way around?

On Wednesday, September 1, 2021 at 2:37:13 PM UTC-4, Bonita Montero wrote:
> Am 01.09.2021 um 19:44 schrieb David Brown:
>
> > I believe there are more dedicated DSP's produced per year than 64-bit
> > cpus. There may even be more 4-bit microcontrollers still produced than
> > 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> > (according to a quick google search). There are DSPs and other
> > specialist microcontrollers with minimum order quantities of millions.
> > I know of microcontrollers (programmable in C, and in this case with
> > 8-bit char) that cost $0.03 each.
> The issue is not about whether they're can't address raw bytes but just
> larger quantities. The issue is about machines with typed pointers, i.e.
> when you cast a pointer they change their bit-representation, and these
> exist in a homeopathic amount.

Keep in mind that not being able to address bytes is the single most
common reason for having two or more different sizes of pointers.
Still, it would have been more on topic if had been able to directly confirm
that the platforms he was referring to had implementations that used two
or more pointer representations.
Describing these as "typed pointers" makes it sound as if the pointers
contain information about the type that they represent. That's no the case
being described. They typically have two, and only two, representations:
the large version consists of a word address and an explicit byte offset
from the start of the specified word, while the short form contains only
the word address, with an offset of 0 being implicit when such pointers
are used.

On Wednesday, September 1, 2021 at 3:35:17 PM UTC-4, Bart wrote:
> On 01/09/2021 18:44, David Brown wrote:
....
> > I believe there are more dedicated DSP's produced per year than 64-bit
> > cpus. There may even be more 4-bit microcontrollers still produced than
> > 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> > (according to a quick google search). There are DSPs and other
> > specialist microcontrollers with minimum order quantities of millions.
> > I know of microcontrollers (programmable in C, and in this case with
> > 8-bit char) that cost $0.03 each.
> >
> > I don't know what you mean by "not many". In terms of numbers of units,
> > 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
> > devices by orders of magnitude - and 32-bit devices similarly outsell PC
> > processors. In terms of numbers of designs or chips available, it's the
> > same story. PC processors lead in terms of total cost, and lead hugely
> > in terms of the number of different people programming on them.
> So, you want a programming language still extensively used for
> conventional 64-bit computer chips (eg. the x64s and ARMs) to have its
> design dictated by (or even crippled by) the limitations of a 4-bit
> microcontroller used in some toaster?

How exactly is it crippled?

Nothing we've been talking about involves dictating the design of the
language for 64-bit chips. It involves deliberately NOT dictating that
compilers targeting smaller chips make the same design choices when
implementating that language.

The behavior can be unspecified, and even undefined when you write
code that makes the unportable assumption that all pointers have the same
representation. However, for an implementation where all pointers do have the
same representation, such code has behavior which is defined, by the
implementation, even if not by the standard. There's nothing wrong with
writing unportable code, so long as it can be ported to every platform that it
needs to be ported to.

On 9/1/2021 2:00 PM, Scott Lurndal wrote:
> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>> On 9/1/2021 12:44 PM, David Brown wrote:
>
>>> I believe there are more dedicated DSP's produced per year than 64-bit
>>> cpus. There may even be more 4-bit microcontrollers still produced than
>>> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
>>> (according to a quick google search). There are DSPs and other
>>> specialist microcontrollers with minimum order quantities of millions.
>>> I know of microcontrollers (programmable in C, and in this case with
>>> 8-bit char) that cost $0.03 each.
>>>
>>> I don't know what you mean by "not many". In terms of numbers of units,
>>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>>> processors. In terms of numbers of designs or chips available, it's the
>>> same story. PC processors lead in terms of total cost, and lead hugely
>>> in terms of the number of different people programming on them.
>>
>> Your numbers for AMD and Intel are WAY off. There were 275 million PCs
>> sold in 2020 containing Intel or AMD cpus.
>
> 300 million X86 (intel, amd) processors per year, including servers.
> 6 billion arm processors per _quarter_.
> 1.5 billion ARC processors per year.
> 400 million MIPS processors per year.

Yup, not surprised. What is your source ?

Lynn

Lynn McGuire <lynnmcguire5@gmail.com> writes:
>On 9/1/2021 2:00 PM, Scott Lurndal wrote:
>> Lynn McGuire <lynnmcguire5@gmail.com> writes:
>>> On 9/1/2021 12:44 PM, David Brown wrote:
>>
>>>> I believe there are more dedicated DSP's produced per year than 64-bit
>>>> cpus. There may even be more 4-bit microcontrollers still produced than
>>>> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
>>>> (according to a quick google search). There are DSPs and other
>>>> specialist microcontrollers with minimum order quantities of millions.
>>>> I know of microcontrollers (programmable in C, and in this case with
>>>> 8-bit char) that cost $0.03 each.
>>>>
>>>> I don't know what you mean by "not many". In terms of numbers of units,
>>>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>>>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>>>> processors. In terms of numbers of designs or chips available, it's the
>>>> same story. PC processors lead in terms of total cost, and lead hugely
>>>> in terms of the number of different people programming on them.
>>>
>>> Your numbers for AMD and Intel are WAY off. There were 275 million PCs
>>> sold in 2020 containing Intel or AMD cpus.
>>
>> 300 million X86 (intel, amd) processors per year, including servers.
>> 6 billion arm processors per _quarter_.
>> 1.5 billion ARC processors per year.
>> 400 million MIPS processors per year.
>
>Yup, not surprised. What is your source ?

Industry analyst documents that I'm not free to share.

But if you look at Toms Hardware you may find similar numbers.

Bart <bc@freeuk.com> writes:
>On 01/09/2021 18:44, David Brown wrote:

>> I don't know what you mean by "not many". In terms of numbers of units,
>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>> processors. In terms of numbers of designs or chips available, it's the
>> same story. PC processors lead in terms of total cost, and lead hugely
>> in terms of the number of different people programming on them.
>
>So, you want a programming language still extensively used for
>conventional 64-bit computer chips (eg. the x64s and ARMs) to have its
>design dictated by (or even crippled by) the limitations of a 4-bit
>microcontroller used in some toaster?

The design of C, and the standards predate 64-bit computer chips
by some time.

>
>It's likely anyway that that 4-bit C compiler will be implementing a
>rather specific dialect of C, without worrying too much about how well
>it might work on 64-bit machines.

Actually, the typical C programmer wants his program to run on
any system that supports the standard regardless of the underlying
hardware. The vast majority of C programmers don't even care what
the underlying hardware is.

>
>But we're not allowed to do the same when it's the other way around?

You can complain all you want if it pleases you. It won't change
anything, so it seems somewhat pointless.

On Wednesday, September 1, 2021 at 11:15:02 PM UTC+3, Scott Lurndal wrote:
> Lynn McGuire <lynnmc...@gmail.com> writes:
> >On 9/1/2021 2:00 PM, Scott Lurndal wrote:
> >> Lynn McGuire <lynnmc...@gmail.com> writes:
> >>> On 9/1/2021 12:44 PM, David Brown wrote:
> >>
> >>>> I believe there are more dedicated DSP's produced per year than 64-bit
> >>>> cpus. There may even be more 4-bit microcontrollers still produced than
> >>>> 64-bit cpus. AMD and Intel each sell perhaps 500,000 processors a year
> >>>> (according to a quick google search). There are DSPs and other
> >>>> specialist microcontrollers with minimum order quantities of millions.
> >>>> I know of microcontrollers (programmable in C, and in this case with
> >>>> 8-bit char) that cost $0.03 each.
> >>>>
> >>>> I don't know what you mean by "not many". In terms of numbers of units,
> >>>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
> >>>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
> >>>> processors. In terms of numbers of designs or chips available, it's the
> >>>> same story. PC processors lead in terms of total cost, and lead hugely
> >>>> in terms of the number of different people programming on them.
> >>>
> >>> Your numbers for AMD and Intel are WAY off. There were 275 million PCs
> >>> sold in 2020 containing Intel or AMD cpus.
> >>
> >> 300 million X86 (intel, amd) processors per year, including servers.
> >> 6 billion arm processors per _quarter_.
> >> 1.5 billion ARC processors per year.
> >> 400 million MIPS processors per year.
> >
> >Yup, not surprised. What is your source ?
> Industry analyst documents that I'm not free to share.
>
> But if you look at Toms Hardware you may find similar numbers.

Somehow I don't think the MIPS figure is up to date.
Right now MIPS is in *very* bad financial shape and changing owners approximately 5 times per decade.

Also, it does not look like your Intel figures include network storage appliances and that your AMD figures include gaming consoles.
Each of them is likely close to 20M per year.

On Wednesday, September 1, 2021 at 11:18:51 PM UTC+3, Scott Lurndal wrote:
> Bart <b...@freeuk.com> writes:
> >On 01/09/2021 18:44, David Brown wrote:
>
> >> I don't know what you mean by "not many". In terms of numbers of units,
> >> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
> >> devices by orders of magnitude - and 32-bit devices similarly outsell PC
> >> processors. In terms of numbers of designs or chips available, it's the
> >> same story. PC processors lead in terms of total cost, and lead hugely
> >> in terms of the number of different people programming on them.
> >
> >So, you want a programming language still extensively used for
> >conventional 64-bit computer chips (eg. the x64s and ARMs) to have its
> >design dictated by (or even crippled by) the limitations of a 4-bit
> >microcontroller used in some toaster?
> The design of C, and the standards predate 64-bit computer chips
> by some time.
> >
> >It's likely anyway that that 4-bit C compiler will be implementing a
> >rather specific dialect of C, without worrying too much about how well
> >it might work on 64-bit machines.
> Actually, the typical C programmer wants his program to run on
> any system that supports the standard regardless of the underlying
> hardware. The vast majority of C programmers don't even care what
> the underlying hardware is.

I didn't yet encounter in real physical life a C programmer that cared in a slightest about portability or C Standard.
Only on Internet.

> >
> >But we're not allowed to do the same when it's the other way around?
> You can complain all you want if it pleases you. It won't change
> anything, so it seems somewhat pointless.

On 01/09/2021 22:07, Michael S wrote:
> On Wednesday, September 1, 2021 at 11:18:51 PM UTC+3, Scott Lurndal wrote:
>> Bart <b...@freeuk.com> writes:
>>> On 01/09/2021 18:44, David Brown wrote:
>>
>>>> I don't know what you mean by "not many". In terms of numbers of units,
>>>> 8-bit devices and "hidden" microcontrollers and DSPs outsell 32-bit
>>>> devices by orders of magnitude - and 32-bit devices similarly outsell PC
>>>> processors. In terms of numbers of designs or chips available, it's the
>>>> same story. PC processors lead in terms of total cost, and lead hugely
>>>> in terms of the number of different people programming on them.
>>>
>>> So, you want a programming language still extensively used for
>>> conventional 64-bit computer chips (eg. the x64s and ARMs) to have its
>>> design dictated by (or even crippled by) the limitations of a 4-bit
>>> microcontroller used in some toaster?
>> The design of C, and the standards predate 64-bit computer chips
>> by some time.
>>>
>>> It's likely anyway that that 4-bit C compiler will be implementing a
>>> rather specific dialect of C, without worrying too much about how well
>>> it might work on 64-bit machines.
>> Actually, the typical C programmer wants his program to run on
>> any system that supports the standard regardless of the underlying
>> hardware. The vast majority of C programmers don't even care what
>> the underlying hardware is.
>
> I didn't yet encounter in real physical life a C programmer that cared in a slightest about portability or C Standard.
> Only on Internet.

They might care when UB is used as an excuse for a compiler to do
unexpected things, when that UB is actually well-defined on their
platforms of interest.