A concern about mixing C and C++

[ziman137]

Hello all,

I have a question and am seeking for some advice.

I am currently working to implement an algorithmic library. Because
the performance is the most important factor in later applications, I
decide to write it in C instead of C++. However, I thought it might be
convenient to use some C++ code at some misc places. I'm aware that, I
could always use the C++ compiler to get it work.

My concerns are:

1) Would the way I mix C and C++ code have any potential drawbacks to
the performance?
2) Would the way I mix C and C++ code have any potential drawbacks for
the future users to use the library?

My intention for choosing C interface instead of C++ OOD is to gain the
maximum performance as possible, yet I still like to use some C++
coding features (e.g., "const", reference instead of pointers, ...).

Thanks,

Gary

[Ian Collins]

Malcolm wrote:

"Ian Collins" <ia******@hotma il.comwrote in message
news:4j******** *****@individua l.net...

>>Malcolm wrote:

>>>"ziman137" <ga********@yah oo.comwrote in message

1) Would the way I mix C and C++ code have any potential drawbacks to
the performance?
Potentiall y yes. It must be compiled with a C++ compiler, and it is not
impossible that the C++ compiler on a given system is inferior to the C
compiler. However it is not very likely, often they share substantially
the
same code.
If you use C++ constructs in your code you will find them harder to
optimise
than the C constructs, because C++ is designed to present a nice
interface
to the programmer rather than to expose basic processor operations.
However
you are probably aware of that already.

Eh? C++ can wrap basic processor operations in a higher level
abstraction , but it can still expose then in the same way C does.

Can, yes. But Bjarne Strousup believes that malloc(), for example, should
not be used in new C++ code.

Correct, but you can provide your own version of new/delete, which puts
the developer back in the driver's seat.

Modern C++ using the standard library will have set of abstracted interfaces
to basic structures. They work well and might even be more efficient than
hand-coded similar C structures. However it is not possible for the
programmer to exert fine control over them. So for instance if he wants to
interate over an array, he will use the vector class iterator. This may be a
bare pointer, but probably it carries some run time checks with it to
protect the pointer from going out of bounds. Generally a good thing, but
harder to optimise.

He probably would, but if push came to shove, he could take the address
of the first element and use a plain pointer. Iterators tend to be well
optimised. Some implementations use raw pointers and none are required
to do any bounds checking, although this may be available as an
extension for testing.

C++ gives the developer a choice between the C way of doing something
and the C++ way.

--
Ian Collins.

[Keith Thompson]

Richard Heathfield <in*****@invali d.invalidwrites :

jacob navia said:

>There may be situations
where C++ templates outperform assembly language,

Jacob's talking through his hat again. His claim is easily disproved. All we
have to do is tell the compiler to generate assembly language from the C++
template code. We now have assembly language that performs exactly as well
as the C++ template code. And, given an experienced assembly language
programmer (to match the experienced C++ programmer who produces such
astoundingly quick code), we can almost certainly find an optimisation,
however trivial, that will make the assembly language version at least a
little faster than the C++ template version from which it was generated.

On the other hand, for any language at a higher level than assembler,
the compiler may find opportunities for optimizations that someone
creating hand-written assembler wouldn't find, or would choose not to
use.

For example, if I'm writing low-level code that I expect to be
maintained, I'm not likely to perform a pervasive optimization that
works only if an array size is a power of 2.

A skilled assembly language programmer is likely to be much smarter
than an optimizing compiler, but an optimizer has a different set of
tradeoffs to consider. A programmer will perform optimizations over
time as he's working on a piece of code; an optimizing compiler will
redo all its work from scratch every time it's invoked.

Yes, you can use an optimizing compiler as a way to generate assembly
language -- but that argument doesn't apply to hand-written assembly
language.

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
San Diego Supercomputer Center <* <http://users.sdsc.edu/~kst>
We must do something. This is something. Therefore, we must do this.

[Richard Heathfield]

Keith Thompson said:

<snip>

>
On the other hand, for any language at a higher level than assembler,
the compiler may find opportunities for optimizations that someone
creating hand-written assembler wouldn't find, or would choose not to
use.

Sure, but perhaps you missed my point, which is that the /starting point/
for the assembly language programmer would be the super-efficient C++
template code, expressed in assembly language form. Given that starting
base, (a) it's bound to be as fast as the C++ template code, because it's
the same code that the C++ template code is compiled to - which is in
itself sufficient to destroy Mr Navia's point, and (b) it would be quite
surprising if the skilled assembly language programmer couldn't find /one/
hand-optimisation to make the asm form just a tiny smidgenette faster.

<snip>

Yes, you can use an optimizing compiler as a way to generate assembly
language -- but that argument doesn't apply to hand-written assembly
language.

It does if the starting point for the handwriting is the generated code.

--
Richard Heathfield
"Usenet is a strange place" - dmr 29/7/1999
http://www.cpax.org.uk
email: rjh at above domain (but drop the www, obviously)

[Ian Collins]

Richard Heathfield wrote:

Keith Thompson said:

<snip>

>>On the other hand, for any language at a higher level than assembler,
the compiler may find opportunities for optimizations that someone
creating hand-written assembler wouldn't find, or would choose not to
use.

Sure, but perhaps you missed my point, which is that the /starting point/
for the assembly language programmer would be the super-efficient C++
template code, expressed in assembly language form. Given that starting
base, (a) it's bound to be as fast as the C++ template code, because it's
the same code that the C++ template code is compiled to - which is in
itself sufficient to destroy Mr Navia's point, and (b) it would be quite
surprising if the skilled assembly language programmer couldn't find /one/
hand-optimisation to make the asm form just a tiny smidgenette faster.

I don't know if you do this, but I often find I can obtain more
productive optimisations by using the generated assembly language to
tune the high level language code rather than hand editing the assembly.

One area optimising compilers have an edge and will continue to improve
is global optimisations, there is only so much code a human can retain
in working memory. We can often do better with micro-optimisations, but
the compiler is better equipped to see the big picture.

--
Ian Collins.

[Keith Thompson]

Richard Heathfield <in*****@invali d.invalidwrites :

Keith Thompson said:

<snip>

>>
On the other hand, for any language at a higher level than assembler,
the compiler may find opportunities for optimizations that someone
creating hand-written assembler wouldn't find, or would choose not to
use.

Sure, but perhaps you missed my point, which is that the /starting point/
for the assembly language programmer would be the super-efficient C++
template code, expressed in assembly language form. Given that starting
base, (a) it's bound to be as fast as the C++ template code, because it's
the same code that the C++ template code is compiled to - which is in
itself sufficient to destroy Mr Navia's point, and (b) it would be quite
surprising if the skilled assembly language programmer couldn't find /one/
hand-optimisation to make the asm form just a tiny smidgenette faster.

<snip>

>Yes, you can use an optimizing compiler as a way to generate assembly
language -- but that argument doesn't apply to hand-written assembly
language.

It does if the starting point for the handwriting is the generated code.

If the optimizing compiler generates code that pervasively depends on
some semantically trivial aspect of the high-level source, the
generated assembly won't be a good basis for maintenance. For
example, a compiler might generate radically different code if an
array size is a power of two than if it isn't, even if there's only,
say, a 1% advantage. If you change the array declaration, the
compiler will happily re-write the entire application when you
recompile it. A sane assembly language programmer would probably
avoid such an optimization if it would hurt maintainability that much.

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
San Diego Supercomputer Center <* <http://users.sdsc.edu/~kst>
We must do something. This is something. Therefore, we must do this.

[Ark]

Richard Heathfield wrote:

Keith Thompson said:

<snip>

>On the other hand, for any language at a higher level than assembler,
the compiler may find opportunities for optimizations that someone
creating hand-written assembler wouldn't find, or would choose not to
use.

Sure, but perhaps you missed my point, which is that the /starting point/
for the assembly language programmer would be the super-efficient C++
template code, expressed in assembly language form. Given that starting
base, (a) it's bound to be as fast as the C++ template code, because it's
the same code that the C++ template code is compiled to - which is in
itself sufficient to destroy Mr Navia's point, and (b) it would be quite
surprising if the skilled assembly language programmer couldn't find /one/
hand-optimisation to make the asm form just a tiny smidgenette faster.

<snip>

>Yes, you can use an optimizing compiler as a way to generate assembly
language -- but that argument doesn't apply to hand-written assembly
language.

It does if the starting point for the handwriting is the generated code.

The real point is that the compiler-generated assembly code ain't
necessarily a good starting point, for it perhaps takes advantages of
accidental coincidences of data that a human programmer will not indulge
in. So (maintainable) assembly code _may_ be worse than
compiler-generated, the point Keith Thompson had made pretty clear.
Example:
The line
int x = NUMBER;
is compiled (for ARM) differently depending on
#define NUMBER 1
#define NUMBER 257
#define NUMBER 0x101010101
A human would stick to a single implementation.

[Ark]

Ian Collins wrote:

Malcolm wrote:

>"Ian Collins" <ia******@hotma il.comwrote in message
news:4j******* ******@individu al.net...

>>Malcolm wrote:

"ziman137" <ga********@yah oo.comwrote in message

1) Would the way I mix C and C++ code have any potential drawbacks to
the performance?
>
Potentiall y yes. It must be compiled with a C++ compiler, and it is not
impossible that the C++ compiler on a given system is inferior to the C
compiler. However it is not very likely, often they share substantially
the
same code.
If you use C++ constructs in your code you will find them harder to
optimise
than the C constructs, because C++ is designed to present a nice
interface
to the programmer rather than to expose basic processor operations.
However
you are probably aware of that already.
Eh? C++ can wrap basic processor operations in a higher level
abstraction , but it can still expose then in the same way C does.

Can, yes. But Bjarne Strousup believes that malloc(), for example, should
not be used in new C++ code.

Correct, but you can provide your own version of new/delete, which puts
the developer back in the driver's seat.

>Modern C++ using the standard library will have set of abstracted interfaces
to basic structures. They work well and might even be more efficient than
hand-coded similar C structures. However it is not possible for the
programmer to exert fine control over them. So for instance if he wants to
interate over an array, he will use the vector class iterator. This may be a
bare pointer, but probably it carries some run time checks with it to
protect the pointer from going out of bounds. Generally a good thing, but
harder to optimise.

He probably would, but if push came to shove, he could take the address
of the first element and use a plain pointer. Iterators tend to be well
optimised. Some implementations use raw pointers and none are required
to do any bounds checking, although this may be available as an
extension for testing.

C++ gives the developer a choice between the C way of doing something
and the C++ way.

Ehmmm... please help me here: I am not a C++ guy.
AFAIK, C++ does not allow all the C way (example: tentative declarations).
Then, if I write C-style in C++, at the very minimum the C++ compiler
doesn't know e.g. about any exceptions that may be thrown in a function
I call and which is in a different translation unit. So it just has to
bring in the heavy machinery in just in case. Am I missing something?

[Ian Collins]

Ark wrote:

Ian Collins wrote:

>C++ gives the developer a choice between the C way of doing something
and the C++ way.

Ehmmm... please help me here: I am not a C++ guy.
AFAIK, C++ does not allow all the C way (example: tentative declarations).
Then, if I write C-style in C++, at the very minimum the C++ compiler
doesn't know e.g. about any exceptions that may be thrown in a function
I call and which is in a different translation unit. So it just has to
bring in the heavy machinery in just in case. Am I missing something?

If my understanding of tentative declarations is correct, they are
synonymous with forward declarations in C++ due to the differing use of
'struct' in a declaration, that is:

struct X;

forward declares the type X.

The compiler doesn't have to bother with unexpected exceptions when
compiling idiomatic C code, any unhandled exception can be caught by
whatever calls main().

As a trivial example, on my platform the following generated essentially
the same code when compiled as C or C++:

extern int fn(void);
extern void fn1(int);

typedef struct { int n; } X;

int main(void)
{
X x;

x.n = fn();

fn1( x.n );
}

--
Ian Collins.

[Ark]

Ian Collins wrote:

Ark wrote:

>Ian Collins wrote:

>>C++ gives the developer a choice between the C way of doing something
and the C++ way.

Ehmmm... please help me here: I am not a C++ guy.
AFAIK, C++ does not allow all the C way (example: tentative declarations).
Then, if I write C-style in C++, at the very minimum the C++ compiler
doesn't know e.g. about any exceptions that may be thrown in a function
I call and which is in a different translation unit. So it just has to
bring in the heavy machinery in just in case. Am I missing something?

If my understanding of tentative declarations is correct, they are
synonymous with forward declarations in C++ due to the differing use of
'struct' in a declaration, that is:

struct X;

forward declares the type X.

The compiler doesn't have to bother with unexpected exceptions when
compiling idiomatic C code, any unhandled exception can be caught by
whatever calls main().

As a trivial example, on my platform the following generated essentially
the same code when compiled as C or C++:

extern int fn(void);
extern void fn1(int);

typedef struct { int n; } X;

int main(void)
{
X x;

x.n = fn();

fn1( x.n );
}

1. Tentative declarations of variables are (in C) indistinguishab le from
definitions until the end of the translation unit; they AFAIK expressly
prohibited in C++. E.g., a (const) circular data structure like
typedef struct X {int x, const struct X *next} X;
const X x;
const X y;
const X z = {5, &x};
const X y = {6, &z};
const X x = {7, &y};
is a valid C but not C++.
2. Interesting example; just curious what the differences are and
whether it matters that your function is called main, not, say, foo.

[Ian Collins]

Ark wrote:

Ian Collins wrote:

>Ark wrote:

>>Ian Collins wrote:

C++ gives the developer a choice between the C way of doing something
and the C++ way.

Ehmmm... please help me here: I am not a C++ guy.
AFAIK, C++ does not allow all the C way (example: tentative
declarations) .
Then, if I write C-style in C++, at the very minimum the C++ compiler
doesn't know e.g. about any exceptions that may be thrown in a function
I call and which is in a different translation unit. So it just has to
bring in the heavy machinery in just in case. Am I missing something?

If my understanding of tentative declarations is correct, they are
synonymous with forward declarations in C++ due to the differing use of
'struct' in a declaration, that is:

struct X;

forward declares the type X.

The compiler doesn't have to bother with unexpected exceptions when
compiling idiomatic C code, any unhandled exception can be caught by
whatever calls main().

As a trivial example, on my platform the following generated essentially
the same code when compiled as C or C++:

extern int fn(void);
extern void fn1(int);

typedef struct { int n; } X;

int main(void)
{
X x;

x.n = fn();

fn1( x.n );
}

1. Tentative declarations of variables are (in C) indistinguishab le from
definitions until the end of the translation unit; they AFAIK expressly
prohibited in C++. E.g., a (const) circular data structure like
typedef struct X {int x, const struct X *next} X;

typedef struct X {int x; const struct X *next;} X;

Is valid C but not C++.

typedef struct X {int x; const X *next;} X;

Is valid C++ but not C.

typedef struct X_t {int x; const struct X_t *next;} X;

Is valid C++ and C.

const X x;
const X y;
const X z = {5, &x};
const X y = {6, &z};
const X x = {7, &y};
is a valid C but not C++.

Correct , C++ prohibits uninitialised consts. Easily fixed for both thus:

extern const X x;
extern const X y;
const X z = {5, &x};
const X y = {6, &z};
const X x = {7, &y};

2. Interesting example; just curious what the differences are and
whether it matters that your function is called main, not, say, foo.

The C++ compiler pushed one extra register on the stack. The function
name does not make any difference.

--
Ian Collins.