2023-06-05 07:37:25 +00:00
|
|
|
#if defined(__x86_64__) && !(__ASSEMBLER__ + __LINKER__ + 0)
|
|
|
|
|
#ifndef _X86GPRINTRIN_H_INCLUDED
|
|
|
|
|
# error "Never use <ia32intrin.h> directly; include <x86gprintrin.h> instead."
|
2023-04-27 09:56:41 +00:00
|
|
|
#endif
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bsfd (int __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ctz (__X);
|
|
|
|
|
}
|
|
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bsrd (int __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_bsrsi (__X);
|
|
|
|
|
}
|
|
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bswapd (int __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_bswap32 (__X);
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
|
|
|
|
#ifndef __iamcu__
|
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
|
|
|
#ifndef __CRC32__
|
2023-04-27 09:56:41 +00:00
|
|
|
#pragma GCC push_options
|
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
|
|
|
#pragma GCC target("crc32")
|
|
|
|
|
#define __DISABLE_CRC32__
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__crc32b (unsigned int __C, unsigned char __V)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_crc32qi (__C, __V);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__crc32w (unsigned int __C, unsigned short __V)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_crc32hi (__C, __V);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__crc32d (unsigned int __C, unsigned int __V)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_crc32si (__C, __V);
|
2023-04-27 09:56:41 +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
|
|
|
#ifdef __DISABLE_CRC32__
|
|
|
|
|
#undef __DISABLE_CRC32__
|
2023-04-27 09:56:41 +00:00
|
|
|
#pragma GCC pop_options
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
#endif
|
|
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__popcntd (unsigned int __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_popcount (__X);
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
#ifndef __iamcu__
|
|
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rdpmc (int __S)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rdpmc (__S);
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
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
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rdtsc (void)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rdtsc ();
|
|
|
|
|
}
|
2023-04-27 09:56:41 +00:00
|
|
|
#ifndef __iamcu__
|
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
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rdtscp (unsigned int *__A)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rdtscp (__A);
|
|
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
extern __inline unsigned char
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rolb (unsigned char __X, int __C)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rolqi (__X, __C);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned short
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rolw (unsigned short __X, int __C)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rolhi (__X, __C);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rold (unsigned int __X, int __C)
|
|
|
|
|
{
|
2023-04-27 09:56:41 +00:00
|
|
|
__C &= 31;
|
|
|
|
|
return (__X << __C) | (__X >> (-__C & 31));
|
|
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline unsigned char
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rorb (unsigned char __X, int __C)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rorqi (__X, __C);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned short
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rorw (unsigned short __X, int __C)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_rorhi (__X, __C);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rord (unsigned int __X, int __C)
|
|
|
|
|
{
|
2023-04-27 09:56:41 +00:00
|
|
|
__C &= 31;
|
|
|
|
|
return (__X >> __C) | (__X << (-__C & 31));
|
|
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline void
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__pause (void)
|
|
|
|
|
{
|
|
|
|
|
__builtin_ia32_pause ();
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
|
|
|
|
#ifdef __x86_64__
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bsfq (long long __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ctzll (__X);
|
|
|
|
|
}
|
|
|
|
|
extern __inline int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bsrq (long long __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_bsrdi (__X);
|
|
|
|
|
}
|
|
|
|
|
extern __inline long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__bswapq (long long __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_bswap64 (__X);
|
2023-04-27 09:56:41 +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
|
|
|
#ifndef __CRC32__
|
2023-04-27 09:56:41 +00:00
|
|
|
#pragma GCC push_options
|
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
|
|
|
#pragma GCC target("crc32")
|
|
|
|
|
#define __DISABLE_CRC32__
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__crc32q (unsigned long long __C, unsigned long long __V)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_crc32di (__C, __V);
|
2023-04-27 09:56:41 +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
|
|
|
#ifdef __DISABLE_CRC32__
|
|
|
|
|
#undef __DISABLE_CRC32__
|
2023-04-27 09:56:41 +00:00
|
|
|
#pragma GCC pop_options
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
extern __inline long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__popcntq (unsigned long long __X)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_popcountll (__X);
|
|
|
|
|
}
|
|
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rolq (unsigned long long __X, int __C)
|
|
|
|
|
{
|
2023-04-27 09:56:41 +00:00
|
|
|
__C &= 63;
|
|
|
|
|
return (__X << __C) | (__X >> (-__C & 63));
|
|
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__rorq (unsigned long long __X, int __C)
|
|
|
|
|
{
|
2023-04-27 09:56:41 +00:00
|
|
|
__C &= 63;
|
|
|
|
|
return (__X >> __C) | (__X << (-__C & 63));
|
|
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline unsigned long long
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__readeflags (void)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_readeflags_u64 ();
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline void
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__writeeflags (unsigned long long __X)
|
|
|
|
|
{
|
|
|
|
|
__builtin_ia32_writeeflags_u64 (__X);
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
2023-06-05 07:37:25 +00:00
|
|
|
#define _bswap64(a) __bswapq(a)
|
2023-04-27 09:56:41 +00:00
|
|
|
#define _popcnt64(a) __popcntq(a)
|
|
|
|
|
#else
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline unsigned int
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__readeflags (void)
|
|
|
|
|
{
|
|
|
|
|
return __builtin_ia32_readeflags_u32 ();
|
|
|
|
|
}
|
|
|
|
|
extern __inline void
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
__writeeflags (unsigned int __X)
|
|
|
|
|
{
|
|
|
|
|
__builtin_ia32_writeeflags_u32 (__X);
|
2023-04-27 09:56:41 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
#ifdef __LP64__
|
2023-06-05 07:37:25 +00:00
|
|
|
#define _lrotl(a,b) __rolq((a), (b))
|
|
|
|
|
#define _lrotr(a,b) __rorq((a), (b))
|
2023-04-27 09:56:41 +00:00
|
|
|
#else
|
2023-06-05 07:37:25 +00:00
|
|
|
#define _lrotl(a,b) __rold((a), (b))
|
|
|
|
|
#define _lrotr(a,b) __rord((a), (b))
|
2023-04-27 09:56:41 +00:00
|
|
|
#endif
|
|
|
|
|
#define _bit_scan_forward(a) __bsfd(a)
|
|
|
|
|
#define _bit_scan_reverse(a) __bsrd(a)
|
2023-06-05 07:37:25 +00:00
|
|
|
#define _bswap(a) __bswapd(a)
|
|
|
|
|
#define _popcnt32(a) __popcntd(a)
|
2023-04-27 09:56:41 +00:00
|
|
|
#ifndef __iamcu__
|
2023-06-05 07:37:25 +00:00
|
|
|
#define _rdpmc(a) __rdpmc(a)
|
2023-04-27 09:56:41 +00:00
|
|
|
#define _rdtscp(a) __rdtscp(a)
|
2023-06-05 07:37:25 +00:00
|
|
|
#endif
|
|
|
|
|
#define _rdtsc() __rdtsc()
|
|
|
|
|
#define _rotwl(a,b) __rolw((a), (b))
|
|
|
|
|
#define _rotwr(a,b) __rorw((a), (b))
|
|
|
|
|
#define _rotl(a,b) __rold((a), (b))
|
|
|
|
|
#define _rotr(a,b) __rord((a), (b))
|
|
|
|
|
#endif
|