Add FMT_REDUCE_INT_INSTANTIATIONS flag

[kammce]

Reduce code bloat by removing multiple instantiation of int_writer based
on the parameter.

Rationale:

• The only functions that gains a speedup by int size would be
  int_writer::on_dec()'s call to count_digits which uses CLZ. Thus to
  still take advantage of this speedup, we store the size of the int
  so we can use a switch statement to call the correct count_digits.
• All other implementations of count_digits require some sort of looping
  that terminates when the value hits zero regardless of what sized int
  it is.

Caveats:

• There is a performance hit when dealing with and passing around
  64-bit/128-bit values compared to 32-bit values on 32-bit platforms,
  and with 64-bit values on 64-bit systems. But this should not reduce the
  performance that dramatically.
• There is also a performance hit for on_dec() due to the addition of a
  switch case. But, due to it size, this should reduce to a jump table.

Resolves #1778

I agree that my contributions are licensed under the {fmt} license, and agree to future changes to the licensing.

[kammce]

@vitaut is there a make target to run a performance test?

[vitaut]

There is a performance hit when dealing with and passing around
64-bit/128-bit values compared to 32-bit values on 32-bit platforms,
and with 64-bit values on 64-bit systems.

This is a non-starter.

A better option would be to keep int_writer parameterized on integer type and add a configuration option to make uint32_or_64_or_128_t only map to the largest type. This won't add overhead for the common case.

[kammce]

Hmmm, I like it. I can make that change.

[vitaut]

Looks good but I suggest renaming FMT_USE_SMALLEST_INT to something that more clearly describes the effect of this option. Maybe FMT_REDUCE_INT_INSTANTIATIONS?

[vitaut]

Thanks!

[vitaut]

@kammce, BTW for future reference, what platform are you targeting and what are the savings with -DFMT_REDUCE_INT_INSTANTIATIONS=1?

[kammce]

TL;DR: Saved 2024 bytes

I'm surprised I didn't mention that at the start.

Target CPU: Cortex M3 (although I test on M4F, M7 etc)
Target MCU: stm32f103 (64kB of flash)
Compiler: gcc-arm-none-eabi-9-2020-q2-update
Optimization Level: -Os

Following flags were used to get the binary as small as possible during testing:

#define FMT_STATIC_THOUSANDS_SEPARATOR ','
#define FMT_USE_FLOAT 0
#define FMT_USE_DOUBLE 0
#define FMT_USE_LONG_DOUBLE 0
#define FMT_REDUCE_INT_INSTANTIATIONS 1

Binary increased by 10,812 bytes when using fmt with the above flags turned on.
12,836 bytes is the increase with -D FMT_REDUCE_INT_INSTANTIATIONS 0.

Thus the binary amount saved is 2024 bytes.