| aarch32_cpudetect.c |
|
2605 |
| aarch64_cpudetect.c |
|
6122 |
| arm.h |
|
1304 |
| arm_cpudetect.h |
|
1481 |
| asmdefs_mmi.h |
HAVE_MIPS64 |
2704 |
| bitops.h |
|
2223 |
| compiler_attributes.h |
|
2630 |
| emmintrin_compat.h |
From emmintrin.h (gcc 4.5.3) |
1682 |
| emms_mmx.asm |
|
539 |
| emms_mmx.c |
|
523 |
| float_control_word.asm |
|
770 |
| loongarch.h |
|
760 |
| loongarch_cpudetect.c |
end __loongarch__ && __linux__ |
1296 |
| mem.h |
Shift down with rounding |
1703 |
| mem_ops.h |
\file
\brief Provides portable memory access primitives
This function provides portable primitives for getting and setting of
signed and unsigned integers in 16, 24, and 32 bit sizes. The operations
can be performed on unaligned data regardless of hardware support for
unaligned accesses.
The type used to pass the integral values may be changed by defining
MEM_VALUE_T with the appropriate type. The type given must be an integral
numeric type.
The actual functions instantiated have the MEM_VALUE_T type name pasted
on to the symbol name. This allows the developer to instantiate these
operations for multiple types within the same translation unit. This is
of somewhat questionable utility, but the capability exists nonetheless.
Users not making use of this functionality should call the functions
without the type name appended, and the preprocessor will take care of
it.
NOTE: This code is not supported on platforms where char > 1 octet ATM.
|
6785 |
| mem_ops_aligned.h |
\file
\brief Provides portable memory access primitives for operating on aligned
data
This file is split from mem_ops.h for easier maintenance. See mem_ops.h
for a more detailed description of these primitives.
|
7220 |
| mips.h |
|
659 |
| mips_cpudetect.c |
end __mips__ && __linux__ |
1867 |
| msvc.h |
|
879 |
| ppc.h |
|
682 |
| ppc_cpudetect.c |
|
1860 |
| static_assert.h |
|
1190 |
| system_state.h |
|
830 |
| vpx_once.h |
Implement a function wrapper to guarantee initialization
thread-safety for library singletons.
NOTE: These functions use static locks, and can only be
used with one common argument per compilation unit. So
file1.c:
vpx_once(foo);
...
vpx_once(foo);
file2.c:
vpx_once(bar);
will ensure foo() and bar() are each called only once, but in
file1.c:
vpx_once(foo);
vpx_once(bar):
bar() will never be called because the lock is used up
by the call to foo().
|
3453 |
| vpx_ports.mk |
|
1617 |
| vpx_timer.h |
Win32 specific includes
|
2675 |
| x86.h |
For __cpuidex, __rdtsc |
12452 |
| x86_abi_support.asm |
|
9860 |