michaelh/zephyr
1
0
Fork 0
Code Releases Activity
zephyr/lib/os/Kconfig.cbprintf
Keith Packard 7a5fcb8c60 libc/picolibc: Support 'long long' and 'minimal' printf variants 
Picolibc's 'minimal' printf mode reduces functionality and size even more
than the 'integer' mode. Use this where memory is at a premium and where
the application knows that it does not require exact printf semantics.

1.8.5 adds two more printf variants, 'long long' and 'minimal'. The 'long
long' variant is the same as the 'integer' variant but with long long
support enabled. The 'minimal' variant reduces functionality and size even
more than the 'integer' mode. Applications can use this where memory is at
a premium and where the application does not require exact printf
semantics.

With these two added variants, the SDK has enough options so that all of
the cbprintf modes can be supported with the pre-compiled bits:

 1. CBPRINTF_NANO - picolibc's 'minimal' variant
 2. CBPRINTF_REDUCED_INTEGRAL - picolibc's 'integer' variant
 3. CBPRINTF_FULL_INTEGRAL - picolibc's 'long long' variant
 4. CBPRINTF_FB_SUPPORT - picolibc's 'double' variant

This patch makes the cbprintf Kconfig values drive the default picolibc
variant, disables picolibc variants not capable of supporting the required
cbprintf level, but allows applications to select more functionality in
picolibc than cbprintf requires.

Note that this depends on the SDK including picolibc 1.8.5. Without that,
selecting the 'minimal' or 'long long' variant in Zephyr will end up with
the default variant from picolibc, which is the full version with floating
point support. When using the module things will work as specified.

Signed-off-by: Keith Packard <keithp@keithp.com>
2023-11-20 06:07:58 -05:00

175 lines
6.1 KiB
Plaintext
Raw Permalink Blame History

# Copyright (c) 2020 Nordic Semiconductor ASA
# SPDX-License-Identifier: Apache-2.0
choice CBPRINTF_IMPLEMENTATION
prompt "Capabilities of cbprintf implementation"
default CBPRINTF_COMPLETE
config CBPRINTF_COMPLETE
bool "All selected features"
help
Select this for an implementation that supports all potential
conversions, with Kconfig options to control availability at build
time.
# 80: -53% / 982 B (80 / 00)
config CBPRINTF_NANO
bool "Space-optimized but feature-limited"
# nano needs to count characters if it's the formatter for libc
imply CBPRINTF_LIBC_SUBSTS if MINIMAL_LIBC
help
If selected a completely different implementation of the core
formatting capability is substituted. This has a much smaller code
footprint, but provides fewer capabilities.
endchoice # CBPRINTF_IMPLEMENTATION
choice CBPRINTF_INTEGRAL_CONV
prompt "Control range of convertible integer values"
default CBPRINTF_FULL_INTEGRAL
# 01: 0% / 0 B (01 / 00)
config CBPRINTF_FULL_INTEGRAL
bool "Convert the full range of integer values"
select PICOLIBC_IO_MINIMAL_LONG_LONG if PICOLIBC_IO_MINIMAL && PICOLIBC_USE_MODULE
help
Build cbprintf with buffers sized to support converting the full
range of all integral and pointer values.
Selecting this may increase code size on 32-bit systems as
GCC's __udivdi3 helper could end up being pulled into the
final binary. In any case, this will increase call stack size
by a few words.
# 00:
config CBPRINTF_REDUCED_INTEGRAL
bool "Convert only integer values that fit in 32 bits"
help
Build cbprintf with buffers sized to support converting integer
values with no more than 32 bits.
This will decrease stack space, but affects conversion of any type
with more than 32 bits. This includes not only intmax_t but any
type that can be converted to an integral representation including
size_t and pointers.
With CBPRINTF_COMPLETE conversions that may result in value-specific
truncation are not supported, and the generated text will be the
specification (e.g. %jd).
With CBPRINTF_NANO all conversions will be attempted but values that
cannot fit will be silently truncated.
endchoice
# 02: 82% / 1530 B (02 / 00)
config CBPRINTF_FP_SUPPORT
bool "Floating point formatting in cbprintf"
default y if FPU
depends on CBPRINTF_COMPLETE
help
Build the cbprintf utility function with support for floating
point format specifiers. Selecting this increases stack size
requirements slightly, but increases code size significantly.
# 04: 13% / 456 B (07 / 03)
config CBPRINTF_FP_A_SUPPORT
bool "Floating point %a conversions"
depends on CBPRINTF_FULL_INTEGRAL
select CBPRINTF_FP_SUPPORT
help
The %a hexadecimal format for floating point value conversion was
added in C99, but the output is not easily understood so it rarely
appears in application code.
Selecting this adds support for the conversion, but increases the
overall code size related to FP support.
# 40: -15% / -508 B (46 / 06)
config CBPRINTF_FP_ALWAYS_A
bool "Select %a format for all floating point specifications"
select CBPRINTF_FP_A_SUPPORT
depends on !PICOLIBC
help
The %a format for floats requires significantly less code than the
standard decimal representations (%f, %e, %g). Selecting this
option implicitly uses %a (or %A) for all decimal floating
conversions. The precision of the original specification is
ignored.
Selecting this decreases code size when FP_SUPPORT is enabled.
# 08: 3% / 60 B (08 / 00)
config CBPRINTF_N_SPECIFIER
bool "Support %n specifications"
depends on CBPRINTF_COMPLETE
depends on !PICOLIBC
default y
help
If selected %n can be used to determine the number of characters
emitted. If enabled there is a small increase in code size.
Picolibc does not support this feature for security reasons.
# 180: 18% / 138 B (180 / 80) [NANO]
config CBPRINTF_LIBC_SUBSTS
bool "Generate C-library compatible functions using cbprintf"
help
If selected wrappers are generated for various C library functions
using the cbprintf formatter underneath. The wrappers use the C
library function name with a cb suffix; e.g. printfcb() or
vsnprintfcb().
When used with CBPRINTF_NANO this increases the implementation code
size by a small amount.
When used with picolibc, this option generates cbprintf-compatible
functions using stdio, effectively inverting the sense above.
module = CBPRINTF_PACKAGE
module-str = cbprintf_package
source "subsys/logging/Kconfig.template.log_config"
config CBPRINTF_PACKAGE_LONGDOUBLE
bool "Support packaging of long doubles"
help
Option impact required alignment for buffers used for packaging
(CBPRINTF_PACKAGE_ALIGNMENT). On most platforms long doubles
requires buffer to be 16 bytes aligned. Long doubles are rarely used
so such alignment is an unnecessary waste. If option is disabled,
then compilation fails if long double is used.
config CBPRINTF_STATIC_PACKAGE_CHECK_ALIGNMENT
bool "Validate alignment of a static package buffer"
# To avoid self referential macro when printk is redirected to logging
depends on !LOG_PRINTK
help
When enabled, CBPRINTF_STATIC_PACKAGE asserts when buffer is not
properly aligned. If macro is widely used then assert may impact
memory footprint.
config CBPRINTF_PACKAGE_HEADER_STORE_CREATION_FLAGS
bool
help
Enable this option to store the flags used to create the package
into the package itself.
config CBPRINTF_PACKAGE_SUPPORT_TAGGED_ARGUMENTS
bool
depends on !PICOLIBC
select CBPRINTF_PACKAGE_HEADER_STORE_CREATION_FLAGS
help
Hidden option to support tagged arguments in cbvprint_package().
If enabled and cbvprint_package() is called with the corresponding
flag, the packaging function no longer looks at the format strings
to determine the types of arguments, but instead, each argument is
tagged with a type by preceding it with another argument as type
(integer).
config CBPRINTF_CONVERT_CHECK_PTR
bool
default y if !LOG_FMT_SECTION_STRIP
help
If enabled, cbprintf_package_convert() supports checking if string
candidate is used for %p format specifier. Check cannot be performed
if string is not present in the memory.
View git blame Copy permalink