Can I printf a half-precision floating-point value?

Question

I have a _Float16 half-precision variable named x in my C program, and would like to printf() it. Now, I can write: printf("%f", (double) x);, and this will work; but - can I printf x directly? Without the cast, but rather with an appropriate format specifier?

Looking for an answer regarding glibc mostly, but other C standard library implementations are also interesting.

Note: _Float16 is not automatically promoted to a double.

You can't even pass a single-precision float to printf. There's only the L modifier for long double. — nwellnhof
– nwellnhof, Commented Jul 15 at 12:14
Isn't the _Float16 automatically promoted to double (for a variadic function) without the cast? — Weather Vane
– Weather Vane, Commented Jul 15 at 12:14
Since it's not a standard C type, the best bet would be to read the compiler and library documentation. Or possibly even the source itself, if it's available (which it is for glibc). — Some programmer dude
– Some programmer dude, Commented Jul 15 at 12:32
@Someprogrammerdude "it's not a standard C type" --> _Float16 is in the current spec § H.11 7 — chux
– chux, Commented Jul 15 at 17:29
@WeatherVane "Isn't the _Float16 automatically promoted to double" --> No. "The integer promotions are performed on each trailing argument, and trailing arguments that have type float are promoted to double. These are called the default argument promotions. No other conversions are performed implicitly." § 6.5 6 — chux
– chux, Commented Jul 15 at 17:31

KamilCuk · Accepted Answer · 2025-07-15 18:40:15Z

5

can I printf x directly?

Not out of the box.

I do not think https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3184.pdf was merged.

looking for an answer regarding glibc mostly

Implement https://www.gnu.org/software/libc/manual/html_node/Customizing-Printf.html with strfromf16 .

Note that technically lowercase letters in printf format specifiers are reserved https://port70.net/~nsz/c/c11/n1570.html#7.31.11 . Consider uppercase letter, but no one cares. The C23 standard introduced upper case letters H D R K B, so I do not understand how it works.

Note: _Float16 is not promoted to double in variadic argument list.

edited Jul 15 at 18:40

answered Jul 15 at 12:14

KamilCuk

146k8 gold badges86 silver badges155 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

zwol Jul 15 at 12:58

Argument promotion, as described in the other answer, means you cannot get a _Float16 into printf in the first place.

KamilCuk Jul 15 at 14:04

Then pass struct myfloat { _Float16 f; }, for such problem I believe OP will found a workaround.

KamilCuk Jul 15 at 18:31

I was also under the impression that _Float16 is not promoted, which is why I did not include the text on the first version of the answer.

einpoklum Jul 16 at 8:29

Thanks for that; but - can you carry a little more text from the links into this answer?

nwellnhof · Accepted Answer · 2025-07-15 12:33:23Z

-2

This is impossible in C. The variadic arguments of the printf function are promoted ("default argument promotion"), so float and presumably _Float16 are always converted to double before the function is called. Just like char and short are promoted to int.

edited Jul 15 at 12:33

answered Jul 15 at 12:26

nwellnhof

34.1k7 gold badges97 silver badges121 bronze badges

9 Comments

interjay Jul 15 at 14:36

It doesn't seem to automatically promote _Float16: godbolt.org/z/8xTP3xhMe

zwol Jul 15 at 15:14

That's almost certainly a bug. The only way it would not be a bug is if the C standard specifically exempted _Float16 from the default argument promotions ... which I would like to say is out of the question, but I don't trust the post-C2011 C committee to be as conservative as they ought to be. I don't know where to find a publicly available copy of the first rev of the C standard that includes _Float16 so I can't check.

Eric Postpischil Jul 15 at 16:38

@zwol: It is not a bug, and the default argument promotions do not convert _Float16 to double. First, _Float16 is not a standard type, so the standard says nothing about it. Second, the default argument promotions arose out of earlier era when promoting arguments to a standard width for their domain made sense. So the default promotions only mention, for the floating-point, converting float to double, as there was no narrower type then. As new types are added, there is no need to add them to the promotions. There is now a standard narrow type, _Float16_t, but it is not in the

Weather Vane Jul 15 at 15:24

@zwol in my draft versions, C18 (n2310) does not mention _Float16 and C23 (n3096) does.

interjay Jul 15 at 15:30

@zwol It isn't a bug. The C23 standard says that only integers and float are promoted as variadic parameters: "The integer promotions are performed on each trailing argument, and trailing arguments that have type float are promoted to double. These are called the default argument promotions. No other conversions are performed implicitly.".

zwol Jul 15 at 15:40

@interjay Blech. I'm tempted to file a defect report.

Eric Postpischil Jul 15 at 16:43

This answer is wrong. Of the floating-point types, the C standard only says float is promoted in the default argument promotions. It is certainly not impossible in C for another floating-point type to be passed directly to printf. All that would be needed is a conversion specification (the printf formatting code) for it.

|

Collectives™ on Stack Overflow

Can I printf a half-precision floating-point value?

2 Answers 2

4 Comments

9 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

9 Comments

Your Answer

Sign up or log in

Post as a guest

Related