2021-09-28 05:58:51 +00:00
|
|
|
/*-*- mode:c;indent-tabs-mode:nil;c-basic-offset:2;tab-width:8;coding:utf-8 -*-│
|
|
|
|
│ vi: set et ft=c ts=2 sts=2 sw=2 fenc=utf-8 :vi │
|
|
|
|
╞══════════════════════════════════════════════════════════════════════════════╡
|
|
|
|
│ Copyright 2021 Justine Alexandra Roberts Tunney │
|
|
|
|
│ │
|
|
|
|
│ Permission to use, copy, modify, and/or distribute this software for │
|
|
|
|
│ any purpose with or without fee is hereby granted, provided that the │
|
|
|
|
│ above copyright notice and this permission notice appear in all copies. │
|
|
|
|
│ │
|
|
|
|
│ THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL │
|
|
|
|
│ WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED │
|
|
|
|
│ WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE │
|
|
|
|
│ AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL │
|
|
|
|
│ DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR │
|
|
|
|
│ PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER │
|
|
|
|
│ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR │
|
|
|
|
│ PERFORMANCE OF THIS SOFTWARE. │
|
|
|
|
╚─────────────────────────────────────────────────────────────────────────────*/
|
|
|
|
#include "libc/assert.h"
|
|
|
|
#include "libc/dce.h"
|
|
|
|
#include "libc/nexgen32e/nexgen32e.h"
|
|
|
|
#include "libc/nexgen32e/x86feature.h"
|
|
|
|
#include "libc/str/str.h"
|
2023-05-15 08:51:29 +00:00
|
|
|
#ifndef __aarch64__
|
2021-09-28 05:58:51 +00:00
|
|
|
|
|
|
|
typedef long long xmm_t __attribute__((__vector_size__(16), __aligned__(1)));
|
|
|
|
typedef long long xmm_a __attribute__((__vector_size__(16), __aligned__(16)));
|
|
|
|
|
Release Cosmopolitan v3.3
This change upgrades to GCC 12.3 and GNU binutils 2.42. The GNU linker
appears to have changed things so that only a single de-duplicated str
table is present in the binary, and it gets placed wherever the linker
wants, regardless of what the linker script says. To cope with that we
need to stop using .ident to embed licenses. As such, this change does
significant work to revamp how third party licenses are defined in the
codebase, using `.section .notice,"aR",@progbits`.
This new GCC 12.3 toolchain has support for GNU indirect functions. It
lets us support __target_clones__ for the first time. This is used for
optimizing the performance of libc string functions such as strlen and
friends so far on x86, by ensuring AVX systems favor a second codepath
that uses VEX encoding. It shaves some latency off certain operations.
It's a useful feature to have for scientific computing for the reasons
explained by the test/libcxx/openmp_test.cc example which compiles for
fifteen different microarchitectures. Thanks to the upgrades, it's now
also possible to use newer instruction sets, such as AVX512FP16, VNNI.
Cosmo now uses the %gs register on x86 by default for TLS. Doing it is
helpful for any program that links `cosmo_dlopen()`. Such programs had
to recompile their binaries at startup to change the TLS instructions.
That's not great, since it means every page in the executable needs to
be faulted. The work of rewriting TLS-related x86 opcodes, is moved to
fixupobj.com instead. This is great news for MacOS x86 users, since we
previously needed to morph the binary every time for that platform but
now that's no longer necessary. The only platforms where we need fixup
of TLS x86 opcodes at runtime are now Windows, OpenBSD, and NetBSD. On
Windows we morph TLS to point deeper into the TIB, based on a TlsAlloc
assignment, and on OpenBSD/NetBSD we morph %gs back into %fs since the
kernels do not allow us to specify a value for the %gs register.
OpenBSD users are now required to use APE Loader to run Cosmo binaries
and assimilation is no longer possible. OpenBSD kernel needs to change
to allow programs to specify a value for the %gs register, or it needs
to stop marking executable pages loaded by the kernel as mimmutable().
This release fixes __constructor__, .ctor, .init_array, and lastly the
.preinit_array so they behave the exact same way as glibc.
We no longer use hex constants to define math.h symbols like M_PI.
2024-02-20 19:12:09 +00:00
|
|
|
static __vex void *__memmove(void *dst, const void *src, size_t n) {
|
2021-10-08 15:11:51 +00:00
|
|
|
char *d;
|
2021-09-28 05:58:51 +00:00
|
|
|
size_t i;
|
|
|
|
const char *s;
|
|
|
|
uint64_t a, b;
|
2023-09-02 03:49:13 +00:00
|
|
|
xmm_t v, w, x, y, V, W, X, Y;
|
2021-09-28 05:58:51 +00:00
|
|
|
d = dst;
|
|
|
|
s = src;
|
2023-05-02 02:43:59 +00:00
|
|
|
|
2023-07-30 18:05:05 +00:00
|
|
|
#if defined(__x86_64__) && !defined(__chibicc__)
|
2021-09-28 05:58:51 +00:00
|
|
|
if (IsTiny()) {
|
2021-10-04 10:23:31 +00:00
|
|
|
uint16_t w1, w2;
|
|
|
|
uint32_t l1, l2;
|
|
|
|
uint64_t q1, q2;
|
|
|
|
if (n <= 16) {
|
|
|
|
if (n >= 8) {
|
|
|
|
__builtin_memcpy(&q1, s, 8);
|
|
|
|
__builtin_memcpy(&q2, s + n - 8, 8);
|
|
|
|
__builtin_memcpy(d, &q1, 8);
|
|
|
|
__builtin_memcpy(d + n - 8, &q2, 8);
|
|
|
|
} else if (n >= 4) {
|
|
|
|
__builtin_memcpy(&l1, s, 4);
|
|
|
|
__builtin_memcpy(&l2, s + n - 4, 4);
|
|
|
|
__builtin_memcpy(d, &l1, 4);
|
|
|
|
__builtin_memcpy(d + n - 4, &l2, 4);
|
|
|
|
} else if (n >= 2) {
|
|
|
|
__builtin_memcpy(&w1, s, 2);
|
|
|
|
__builtin_memcpy(&w2, s + n - 2, 2);
|
|
|
|
__builtin_memcpy(d, &w1, 2);
|
|
|
|
__builtin_memcpy(d + n - 2, &w2, 2);
|
|
|
|
} else if (n) {
|
|
|
|
*d = *s;
|
|
|
|
}
|
2021-09-28 05:58:51 +00:00
|
|
|
} else {
|
2021-10-08 15:11:51 +00:00
|
|
|
if (d <= s) {
|
|
|
|
asm("rep movsb"
|
|
|
|
: "+D"(d), "+S"(s), "+c"(n), "=m"(*(char(*)[n])dst)
|
|
|
|
: "m"(*(char(*)[n])src));
|
|
|
|
} else {
|
|
|
|
d += n - 1;
|
|
|
|
s += n - 1;
|
|
|
|
asm("std\n\t"
|
|
|
|
"rep movsb\n\t"
|
|
|
|
"cld"
|
|
|
|
: "+D"(d), "+S"(s), "+c"(n), "=m"(*(char(*)[n])dst)
|
|
|
|
: "m"(*(char(*)[n])src));
|
|
|
|
}
|
2021-09-28 05:58:51 +00:00
|
|
|
}
|
|
|
|
return dst;
|
|
|
|
}
|
2023-05-02 02:43:59 +00:00
|
|
|
#endif
|
|
|
|
|
2021-09-28 05:58:51 +00:00
|
|
|
switch (n) {
|
|
|
|
case 0:
|
|
|
|
return d;
|
|
|
|
case 1:
|
|
|
|
*d = *s;
|
|
|
|
return d;
|
|
|
|
case 2:
|
|
|
|
__builtin_memcpy(&a, s, 2);
|
|
|
|
__builtin_memcpy(d, &a, 2);
|
|
|
|
return d;
|
|
|
|
case 3:
|
|
|
|
__builtin_memcpy(&a, s, 2);
|
|
|
|
__builtin_memcpy(&b, s + 1, 2);
|
|
|
|
__builtin_memcpy(d, &a, 2);
|
|
|
|
__builtin_memcpy(d + 1, &b, 2);
|
|
|
|
return d;
|
|
|
|
case 4:
|
|
|
|
__builtin_memcpy(&a, s, 4);
|
|
|
|
__builtin_memcpy(d, &a, 4);
|
|
|
|
return d;
|
|
|
|
case 5 ... 7:
|
|
|
|
__builtin_memcpy(&a, s, 4);
|
|
|
|
__builtin_memcpy(&b, s + n - 4, 4);
|
|
|
|
__builtin_memcpy(d, &a, 4);
|
|
|
|
__builtin_memcpy(d + n - 4, &b, 4);
|
|
|
|
return d;
|
|
|
|
case 8:
|
|
|
|
__builtin_memcpy(&a, s, 8);
|
|
|
|
__builtin_memcpy(d, &a, 8);
|
|
|
|
return d;
|
|
|
|
case 9 ... 15:
|
|
|
|
__builtin_memcpy(&a, s, 8);
|
|
|
|
__builtin_memcpy(&b, s + n - 8, 8);
|
|
|
|
__builtin_memcpy(d, &a, 8);
|
|
|
|
__builtin_memcpy(d + n - 8, &b, 8);
|
|
|
|
return d;
|
|
|
|
case 16:
|
|
|
|
*(xmm_t *)d = *(xmm_t *)s;
|
|
|
|
return d;
|
|
|
|
case 17 ... 32:
|
|
|
|
v = *(xmm_t *)s;
|
|
|
|
w = *(xmm_t *)(s + n - 16);
|
|
|
|
*(xmm_t *)d = v;
|
|
|
|
*(xmm_t *)(d + n - 16) = w;
|
|
|
|
return d;
|
|
|
|
case 33 ... 64:
|
|
|
|
v = *(xmm_t *)s;
|
|
|
|
w = *(xmm_t *)(s + 16);
|
|
|
|
x = *(xmm_t *)(s + n - 32);
|
|
|
|
y = *(xmm_t *)(s + n - 16);
|
|
|
|
*(xmm_t *)d = v;
|
|
|
|
*(xmm_t *)(d + 16) = w;
|
|
|
|
*(xmm_t *)(d + n - 32) = x;
|
|
|
|
*(xmm_t *)(d + n - 16) = y;
|
|
|
|
return d;
|
|
|
|
case 65 ... 127:
|
|
|
|
v = *(xmm_t *)s;
|
|
|
|
w = *(xmm_t *)(s + 16);
|
|
|
|
x = *(xmm_t *)(s + 32);
|
|
|
|
y = *(xmm_t *)(s + 48);
|
|
|
|
V = *(xmm_t *)(s + n - 64);
|
|
|
|
W = *(xmm_t *)(s + n - 48);
|
|
|
|
X = *(xmm_t *)(s + n - 32);
|
|
|
|
Y = *(xmm_t *)(s + n - 16);
|
|
|
|
*(xmm_t *)d = v;
|
|
|
|
*(xmm_t *)(d + 16) = w;
|
|
|
|
*(xmm_t *)(d + 32) = x;
|
|
|
|
*(xmm_t *)(d + 48) = y;
|
|
|
|
*(xmm_t *)(d + n - 64) = V;
|
|
|
|
*(xmm_t *)(d + n - 48) = W;
|
|
|
|
*(xmm_t *)(d + n - 32) = X;
|
|
|
|
*(xmm_t *)(d + n - 16) = Y;
|
|
|
|
return d;
|
|
|
|
default:
|
|
|
|
if (d == s)
|
|
|
|
return d;
|
2023-05-02 02:43:59 +00:00
|
|
|
|
2023-07-30 18:05:05 +00:00
|
|
|
#if defined(__x86_64__) && !defined(__chibicc__)
|
2021-09-28 05:58:51 +00:00
|
|
|
if (n < kHalfCache3 || !kHalfCache3) {
|
|
|
|
if (d > s) {
|
2024-06-22 12:45:49 +00:00
|
|
|
if (n < 900 || !X86_HAVE(ERMS)) {
|
2021-09-28 05:58:51 +00:00
|
|
|
do {
|
|
|
|
n -= 32;
|
|
|
|
v = *(const xmm_t *)(s + n);
|
|
|
|
w = *(const xmm_t *)(s + n + 16);
|
|
|
|
*(xmm_t *)(d + n) = v;
|
|
|
|
*(xmm_t *)(d + n + 16) = w;
|
|
|
|
} while (n >= 32);
|
|
|
|
} else {
|
|
|
|
asm("std\n\t"
|
|
|
|
"rep movsb\n\t"
|
|
|
|
"cld"
|
|
|
|
: "=D"(d), "=S"(s), "+c"(n), "=m"(*(char(*)[n])d)
|
|
|
|
: "0"(d + n - 1), "1"(s + n - 1), "m"(*(char(*)[n])s));
|
2021-10-08 15:11:51 +00:00
|
|
|
return dst;
|
2021-09-28 05:58:51 +00:00
|
|
|
}
|
|
|
|
} else {
|
2024-06-22 12:45:49 +00:00
|
|
|
if (n < 900 || !X86_HAVE(ERMS)) {
|
2021-09-28 05:58:51 +00:00
|
|
|
i = 0;
|
|
|
|
do {
|
|
|
|
v = *(const xmm_t *)(s + i);
|
|
|
|
w = *(const xmm_t *)(s + i + 16);
|
|
|
|
*(xmm_t *)(d + i) = v;
|
|
|
|
*(xmm_t *)(d + i + 16) = w;
|
|
|
|
} while ((i += 32) + 32 <= n);
|
|
|
|
d += i;
|
|
|
|
s += i;
|
|
|
|
n -= i;
|
|
|
|
} else {
|
|
|
|
asm("rep movsb"
|
|
|
|
: "+D"(d), "+S"(s), "+c"(n), "=m"(*(char(*)[n])d)
|
|
|
|
: "m"(*(char(*)[n])s));
|
2021-10-08 15:11:51 +00:00
|
|
|
return dst;
|
2021-09-28 05:58:51 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (d > s) {
|
|
|
|
while ((uintptr_t)(d + n) & 15) {
|
|
|
|
--n;
|
|
|
|
d[n] = s[n];
|
|
|
|
}
|
|
|
|
do {
|
|
|
|
n -= 32;
|
|
|
|
v = *(const xmm_t *)(s + n);
|
|
|
|
w = *(const xmm_t *)(s + n + 16);
|
|
|
|
__builtin_ia32_movntdq((xmm_a *)(d + n), v);
|
|
|
|
__builtin_ia32_movntdq((xmm_a *)(d + n + 16), w);
|
|
|
|
} while (n >= 32);
|
|
|
|
} else {
|
|
|
|
i = 0;
|
|
|
|
while ((uintptr_t)(d + i) & 15) {
|
|
|
|
d[i] = s[i];
|
|
|
|
++i;
|
|
|
|
}
|
|
|
|
do {
|
|
|
|
v = *(const xmm_t *)(s + i);
|
|
|
|
w = *(const xmm_t *)(s + i + 16);
|
|
|
|
__builtin_ia32_movntdq((xmm_a *)(d + i), v);
|
|
|
|
__builtin_ia32_movntdq((xmm_a *)(d + i + 16), w);
|
|
|
|
} while ((i += 32) + 32 <= n);
|
|
|
|
d += i;
|
|
|
|
s += i;
|
|
|
|
n -= i;
|
|
|
|
}
|
|
|
|
asm("sfence");
|
|
|
|
}
|
2023-05-02 02:43:59 +00:00
|
|
|
#else
|
|
|
|
|
|
|
|
if (d > s) {
|
|
|
|
do {
|
|
|
|
n -= 32;
|
|
|
|
v = *(const xmm_t *)(s + n);
|
|
|
|
w = *(const xmm_t *)(s + n + 16);
|
|
|
|
*(xmm_t *)(d + n) = v;
|
|
|
|
*(xmm_t *)(d + n + 16) = w;
|
|
|
|
} while (n >= 32);
|
|
|
|
} else {
|
|
|
|
i = 0;
|
|
|
|
do {
|
|
|
|
v = *(const xmm_t *)(s + i);
|
|
|
|
w = *(const xmm_t *)(s + i + 16);
|
|
|
|
*(xmm_t *)(d + i) = v;
|
|
|
|
*(xmm_t *)(d + i + 16) = w;
|
|
|
|
} while ((i += 32) + 32 <= n);
|
|
|
|
d += i;
|
|
|
|
s += i;
|
|
|
|
n -= i;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
2021-10-08 15:11:51 +00:00
|
|
|
if (n) {
|
|
|
|
if (n >= 16) {
|
|
|
|
v = *(const xmm_t *)s;
|
|
|
|
w = *(const xmm_t *)(s + n - 16);
|
|
|
|
*(xmm_t *)d = v;
|
|
|
|
*(xmm_t *)(d + n - 16) = w;
|
|
|
|
} else if (n >= 8) {
|
2021-09-28 05:58:51 +00:00
|
|
|
__builtin_memcpy(&a, s, 8);
|
|
|
|
__builtin_memcpy(&b, s + n - 8, 8);
|
|
|
|
__builtin_memcpy(d, &a, 8);
|
|
|
|
__builtin_memcpy(d + n - 8, &b, 8);
|
2021-10-08 15:11:51 +00:00
|
|
|
} else if (n >= 4) {
|
2021-09-28 05:58:51 +00:00
|
|
|
__builtin_memcpy(&a, s, 4);
|
|
|
|
__builtin_memcpy(&b, s + n - 4, 4);
|
|
|
|
__builtin_memcpy(d, &a, 4);
|
|
|
|
__builtin_memcpy(d + n - 4, &b, 4);
|
2021-10-08 15:11:51 +00:00
|
|
|
} else if (n >= 2) {
|
2021-09-28 05:58:51 +00:00
|
|
|
__builtin_memcpy(&a, s, 2);
|
2021-10-08 15:11:51 +00:00
|
|
|
__builtin_memcpy(&b, s + n - 2, 2);
|
2021-09-28 05:58:51 +00:00
|
|
|
__builtin_memcpy(d, &a, 2);
|
2021-10-08 15:11:51 +00:00
|
|
|
__builtin_memcpy(d + n - 2, &b, 2);
|
|
|
|
} else {
|
|
|
|
*d = *s;
|
|
|
|
}
|
2021-09-28 05:58:51 +00:00
|
|
|
}
|
2023-05-02 02:43:59 +00:00
|
|
|
|
2021-10-08 15:11:51 +00:00
|
|
|
return dst;
|
2021-09-28 05:58:51 +00:00
|
|
|
}
|
|
|
|
}
|
2021-10-14 00:27:13 +00:00
|
|
|
|
Release Cosmopolitan v3.3
This change upgrades to GCC 12.3 and GNU binutils 2.42. The GNU linker
appears to have changed things so that only a single de-duplicated str
table is present in the binary, and it gets placed wherever the linker
wants, regardless of what the linker script says. To cope with that we
need to stop using .ident to embed licenses. As such, this change does
significant work to revamp how third party licenses are defined in the
codebase, using `.section .notice,"aR",@progbits`.
This new GCC 12.3 toolchain has support for GNU indirect functions. It
lets us support __target_clones__ for the first time. This is used for
optimizing the performance of libc string functions such as strlen and
friends so far on x86, by ensuring AVX systems favor a second codepath
that uses VEX encoding. It shaves some latency off certain operations.
It's a useful feature to have for scientific computing for the reasons
explained by the test/libcxx/openmp_test.cc example which compiles for
fifteen different microarchitectures. Thanks to the upgrades, it's now
also possible to use newer instruction sets, such as AVX512FP16, VNNI.
Cosmo now uses the %gs register on x86 by default for TLS. Doing it is
helpful for any program that links `cosmo_dlopen()`. Such programs had
to recompile their binaries at startup to change the TLS instructions.
That's not great, since it means every page in the executable needs to
be faulted. The work of rewriting TLS-related x86 opcodes, is moved to
fixupobj.com instead. This is great news for MacOS x86 users, since we
previously needed to morph the binary every time for that platform but
now that's no longer necessary. The only platforms where we need fixup
of TLS x86 opcodes at runtime are now Windows, OpenBSD, and NetBSD. On
Windows we morph TLS to point deeper into the TIB, based on a TlsAlloc
assignment, and on OpenBSD/NetBSD we morph %gs back into %fs since the
kernels do not allow us to specify a value for the %gs register.
OpenBSD users are now required to use APE Loader to run Cosmo binaries
and assimilation is no longer possible. OpenBSD kernel needs to change
to allow programs to specify a value for the %gs register, or it needs
to stop marking executable pages loaded by the kernel as mimmutable().
This release fixes __constructor__, .ctor, .init_array, and lastly the
.preinit_array so they behave the exact same way as glibc.
We no longer use hex constants to define math.h symbols like M_PI.
2024-02-20 19:12:09 +00:00
|
|
|
/**
|
|
|
|
* Copies memory.
|
|
|
|
*
|
|
|
|
* memmove n=0 661 picoseconds
|
|
|
|
* memmove n=1 661 ps/byte 1,476 mb/s
|
|
|
|
* memmove n=2 330 ps/byte 2,952 mb/s
|
|
|
|
* memmove n=3 330 ps/byte 2,952 mb/s
|
|
|
|
* memmove n=4 165 ps/byte 5,904 mb/s
|
|
|
|
* memmove n=7 141 ps/byte 6,888 mb/s
|
|
|
|
* memmove n=8 82 ps/byte 11 GB/s
|
|
|
|
* memmove n=15 44 ps/byte 21 GB/s
|
|
|
|
* memmove n=16 41 ps/byte 23 GB/s
|
|
|
|
* memmove n=31 32 ps/byte 29 GB/s
|
|
|
|
* memmove n=32 31 ps/byte 30 GB/s
|
|
|
|
* memmove n=63 21 ps/byte 45 GB/s
|
|
|
|
* memmove n=64 15 ps/byte 61 GB/s
|
|
|
|
* memmove n=127 13 ps/byte 73 GB/s
|
|
|
|
* memmove n=128 31 ps/byte 30 GB/s
|
|
|
|
* memmove n=255 20 ps/byte 45 GB/s
|
|
|
|
* memmove n=256 19 ps/byte 49 GB/s
|
|
|
|
* memmove n=511 16 ps/byte 56 GB/s
|
|
|
|
* memmove n=512 17 ps/byte 54 GB/s
|
|
|
|
* memmove n=1023 18 ps/byte 52 GB/s
|
|
|
|
* memmove n=1024 13 ps/byte 72 GB/s
|
|
|
|
* memmove n=2047 9 ps/byte 96 GB/s
|
|
|
|
* memmove n=2048 9 ps/byte 98 GB/s
|
|
|
|
* memmove n=4095 8 ps/byte 112 GB/s
|
|
|
|
* memmove n=4096 8 ps/byte 109 GB/s
|
|
|
|
* memmove n=8191 7 ps/byte 124 GB/s
|
|
|
|
* memmove n=8192 7 ps/byte 125 GB/s
|
|
|
|
* memmove n=16383 7 ps/byte 134 GB/s
|
|
|
|
* memmove n=16384 7 ps/byte 134 GB/s
|
|
|
|
* memmove n=32767 13 ps/byte 72 GB/s
|
|
|
|
* memmove n=32768 13 ps/byte 72 GB/s
|
|
|
|
* memmove n=65535 13 ps/byte 68 GB/s
|
|
|
|
* memmove n=65536 14 ps/byte 67 GB/s
|
|
|
|
* memmove n=131071 14 ps/byte 65 GB/s
|
|
|
|
* memmove n=131072 14 ps/byte 64 GB/s
|
|
|
|
* memmove n=262143 15 ps/byte 63 GB/s
|
|
|
|
* memmove n=262144 15 ps/byte 63 GB/s
|
|
|
|
* memmove n=524287 15 ps/byte 61 GB/s
|
|
|
|
* memmove n=524288 15 ps/byte 61 GB/s
|
|
|
|
* memmove n=1048575 15 ps/byte 61 GB/s
|
|
|
|
* memmove n=1048576 15 ps/byte 61 GB/s
|
|
|
|
* memmove n=2097151 19 ps/byte 48 GB/s
|
|
|
|
* memmove n=2097152 27 ps/byte 35 GB/s
|
|
|
|
* memmove n=4194303 28 ps/byte 33 GB/s
|
|
|
|
* memmove n=4194304 28 ps/byte 33 GB/s
|
|
|
|
* memmove n=8388607 28 ps/byte 33 GB/s
|
|
|
|
* memmove n=8388608 28 ps/byte 33 GB/s
|
|
|
|
*
|
|
|
|
* DST and SRC may overlap.
|
|
|
|
*
|
|
|
|
* @param dst is destination
|
|
|
|
* @param src is memory to copy
|
|
|
|
* @param n is number of bytes to copy
|
|
|
|
* @return dst
|
|
|
|
* @asyncsignalsafe
|
|
|
|
*/
|
|
|
|
void *memmove(void *dst, const void *src, size_t n) {
|
|
|
|
return __memmove(dst, src, n);
|
|
|
|
}
|
|
|
|
|
2023-06-06 10:30:37 +00:00
|
|
|
__weak_reference(memmove, memcpy);
|
2023-05-15 08:51:29 +00:00
|
|
|
|
|
|
|
#endif /* __aarch64__ */
|