Fennix/Kernel
1
0
Fork 0
mirror of https://github.com/Fennix-Project/Kernel.git synced 2025-07-11 07:19:20 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Update kernel

Browse Source
This commit is contained in:
EnderIce2
2024-01-19 06:47:42 +02:00
parent fd15592608
commit 96daa43d38
282 changed files with 25486 additions and 15700 deletions
Download Patch File Download Diff File Expand all files Collapse all files

325
library/simd_memcpy.cpp
View File

@ -24,172 +24,257 @@
/*
TODO: Replace these functions with even more optimized versions.
The current versions are fast but not as fast as they could be and also we need implementation for avx, not only sse.
The current versions are fast but not as fast as they could be and also we need implementation for avx, not only sse.
*/
EXTERNC void *memcpy_sse(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movaps (%0), %%xmm0\n"
"movaps %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movaps (%0), %%xmm0\n"
"movaps %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movups (%0), %%xmm0\n"
"movups %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}
EXTERNC void *memcpy_sse2(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqu (%0), %%xmm0\n"
"movdqu %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}
EXTERNC void *memcpy_sse3(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0x7) == 0)
{
size_t num_vectors = n / 8;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movq (%0), %%xmm0\n"
"movddup %%xmm0, %%xmm1\n"
"movq %%xmm1, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0", "xmm1");
d += 8;
s += 8;
}
n -= num_vectors * 8;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqu (%0), %%xmm0\n"
"movdqu %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}
EXTERNC void *memcpy_ssse3(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa 16(%0), %%xmm1\n"
"palignr $8, %%xmm0, %%xmm1\n"
"movdqa %%xmm1, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0", "xmm1");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("lddqu (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("lddqu (%0), %%xmm0\n"
"movdqu %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}
EXTERNC void *memcpy_sse4_1(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
// movntdqa
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqu (%0), %%xmm0\n"
"movdqu %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}
EXTERNC void *memcpy_sse4_2(void *dest, const void *src, size_t n)
{
#if defined(a64)
char *d = (char *)dest;
const char *s = (const char *)src;
char *d = (char *)dest;
const char *s = (const char *)src;
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"pcmpistri $0, (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
if ((((uintptr_t)d | (uintptr_t)s) & 0xF) == 0)
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqa (%0), %%xmm0\n"
"movdqa %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
else
{
size_t num_vectors = n / 16;
for (size_t i = 0; i < num_vectors; i++)
{
asmv("movdqu (%0), %%xmm0\n"
"movdqu %%xmm0, (%1)\n"
:
: "r"(s), "r"(d)
: "xmm0");
d += 16;
s += 16;
}
n -= num_vectors * 16;
}
memcpy_unsafe(d, s, n);
memcpy_unsafe(d, s, n);
#endif // defined(a64)
return dest;
return dest;
}