xref: /linux/Documentation/core-api/floating-point.rst (revision c532de5a67a70f8533d495f8f2aaa9a0491c3ad0)
1.. SPDX-License-Identifier: GPL-2.0+
2
3Floating-point API
4==================
5
6Kernel code is normally prohibited from using floating-point (FP) registers or
7instructions, including the C float and double data types. This rule reduces
8system call overhead, because the kernel does not need to save and restore the
9userspace floating-point register state.
10
11However, occasionally drivers or library functions may need to include FP code.
12This is supported by isolating the functions containing FP code to a separate
13translation unit (a separate source file), and saving/restoring the FP register
14state around calls to those functions. This creates "critical sections" of
15floating-point usage.
16
17The reason for this isolation is to prevent the compiler from generating code
18touching the FP registers outside these critical sections. Compilers sometimes
19use FP registers to optimize inlined ``memcpy`` or variable assignment, as
20floating-point registers may be wider than general-purpose registers.
21
22Usability of floating-point code within the kernel is architecture-specific.
23Additionally, because a single kernel may be configured to support platforms
24both with and without a floating-point unit, FPU availability must be checked
25both at build time and at run time.
26
27Several architectures implement the generic kernel floating-point API from
28``linux/fpu.h``, as described below. Some other architectures implement their
29own unique APIs, which are documented separately.
30
31Build-time API
32--------------
33
34Floating-point code may be built if the option ``ARCH_HAS_KERNEL_FPU_SUPPORT``
35is enabled. For C code, such code must be placed in a separate file, and that
36file must have its compilation flags adjusted using the following pattern::
37
38    CFLAGS_foo.o += $(CC_FLAGS_FPU)
39    CFLAGS_REMOVE_foo.o += $(CC_FLAGS_NO_FPU)
40
41Architectures are expected to define one or both of these variables in their
42top-level Makefile as needed. For example::
43
44    CC_FLAGS_FPU := -mhard-float
45
46or::
47
48    CC_FLAGS_NO_FPU := -msoft-float
49
50Normal kernel code is assumed to use the equivalent of ``CC_FLAGS_NO_FPU``.
51
52Runtime API
53-----------
54
55The runtime API is provided in ``linux/fpu.h``. This header cannot be included
56from files implementing FP code (those with their compilation flags adjusted as
57above). Instead, it must be included when defining the FP critical sections.
58
59.. c:function:: bool kernel_fpu_available( void )
60
61        This function reports if floating-point code can be used on this CPU or
62        platform. The value returned by this function is not expected to change
63        at runtime, so it only needs to be called once, not before every
64        critical section.
65
66.. c:function:: void kernel_fpu_begin( void )
67                void kernel_fpu_end( void )
68
69        These functions create a floating-point critical section. It is only
70        valid to call ``kernel_fpu_begin()`` after a previous call to
71        ``kernel_fpu_available()`` returned ``true``. These functions are only
72        guaranteed to be callable from (preemptible or non-preemptible) process
73        context.
74
75        Preemption may be disabled inside critical sections, so their size
76        should be minimized. They are *not* required to be reentrant. If the
77        caller expects to nest critical sections, it must implement its own
78        reference counting.
79