Changed the clz implementations to use xor instead of subtraction so that when
count_digits "undoes" the BSR -> CLZ translation, the optimizer is more
willing to recognize the equivalence.
Changed the data array in bsr2log10 to static since otherwise MSVC generates
code to build the array every time the function is called.
* Disable fallthrough attributes for the Intel compilers
On MacOS and Linux the Intel compilers may be identified as the
host compilers (Clang or GNU) but do not support the corresponding
compiler specific fallthrough attributes.
* Rearrange ifdef logic for excluding pre-C++17 fallthrough attributes
This puts Intel and PGI compilers into a separate group
and thus makes the intent and logic more obvious.
* add support for fallback_formatter in detail::write
* add ToString test into OStreamTest
to check fmt::to_string() with class that has output stream operator
* add WithOstreamOperator test into CompileTest
to check fmt::format() with FMT_COMPILE() and class that has output stream operator
* use conditional_t inside detail::write instead of 2 overloads
* Revert "add WithOstreamOperator test into CompileTest"
* remove Context from template parameters in detail::write
The intel-19 compiler warns about hidden variables:
```
/s/dev/nightly/libraries/ioss/src/fmt/format.h(2689): warning #1599: declaration hides variable "begin" (declared at line 2668)
FMT_CONSTEXPR void operator()(const Char* begin, const Char* end) {
^
detected during:
instantiation of "Context::iterator fmt::v7::vformat_to<ArgFormatter,Char,Context>(ArgFormatter::iterator, fmt::v7::basic_string_view<Char>, fmt::v7::basic_format_args<Context>, fmt::v7::detail::locale_ref) [with ArgFormatter=fmt::v7::detail::arg_formatter<fmt::v7::detail::buffer_appender<char>, char>, Char=char, Context=fmt::v7::format_context]" at line 3492
instantiation of "fmt::v7::detail::buffer_appender<Char> fmt::v7::detail::vformat_to(fmt::v7::detail::buffer<Char> &, fmt::v7::basic_string_view<Char>, fmt::v7::basic_format_args<fmt::v7::basic_format_context<fmt::v7::detail::buffer_appender<fmt::v7::type_identity_t<Char>>, fmt::v7::type_identity_t<Char>>>) [with Char=char]" at line 1413 of "/s/dev/nightly/libraries/ioss/src/fmt/format-inl.h"
/s/dev/nightly/libraries/ioss/src/fmt/format.h(2689): warning #1599: declaration hides variable "end" (declared at line 2669)
FMT_CONSTEXPR void operator()(const Char* begin, const Char* end) {
^
detected during:
instantiation of "Context::iterator fmt::v7::vformat_to<ArgFormatter,Char,Context>(ArgFormatter::iterator, fmt::v7::basic_string_view<Char>, fmt::v7::basic_format_args<Context>, fmt::v7::detail::locale_ref) [with ArgFormatter=fmt::v7::detail::arg_formatter<fmt::v7::detail::buffer_appender<char>, char>, Char=char, Context=fmt::v7::format_context]" at line 3492
instantiation of "fmt::v7::detail::buffer_appender<Char> fmt::v7::detail::vformat_to(fmt::v7::detail::buffer<Char> &, fmt::v7::basic_string_view<Char>, fmt::v7::basic_format_args<fmt::v7::basic_format_context<fmt::v7::detail::buffer_appender<fmt::v7::type_identity_t<Char>>, fmt::v7::type_identity_t<Char>>>) [with Char=char]" at line 1413 of "/s/dev/nightly/libraries/ioss/src/fmt/format-inl.h"
```
Rename the second set of variables to `pbegin` and `pend` to eliminate warning.
Our nvcc compilers (10.1.243 and 9.2.X) do not define the correct value for `FMT_USE_UDL_TEMPLATE` and then end up with an error later on in the build. Explicitly search for `__NVCC__` symbol not being defined. Might want to instead use `FMT_NVCC` or some other check, but the raw `__EDG_VERSION__` check is not working correctly for nvcc.
* include/fmt/format.h: int_writer: removed unnecessary iterator type re-declaration (prevents shadow-waringing in clang)
* include/fmt/format.h: int_writer: correctly cast signed integer to unsigned to prevent 'implicit conversion changes signedness'-warnings in clang.
Co-authored-by: Martin Wührer <martin.wuehrer@artech.at>
* Remove <typename UInt> from int_writer
Reduce code bloat by removing multiple instantiation of int_writer based
on the <typename UInt> 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
* Add FMT_USE_SMALLEST_INT flag
When defined and set to zero, will use the largest available integer
container for writing ints. The has the benefit of reducing instances
the of int_writer class which will reduce the binary cost.
* Rename flag to FMT_REDUCE_INT_INSTANTIATIONS
Add comment above FMT_REDUCE_INT_INSTANTIATIONS definition describing
why a developer would use it.
* Move FMT_REDUCE_INT_INSTANTIATIONS to format.h
Co-authored-by: Khalil Estell <kammce@google.com>
