2023-06-05 07:37:25 +00:00
|
|
|
#if defined(__x86_64__) && !(__ASSEMBLER__ + __LINKER__ + 0)
|
|
|
|
|
#ifndef _IMMINTRIN_H_INCLUDED
|
|
|
|
|
#error "Never use <avx512bf16intrin.h> directly; include <immintrin.h> instead."
|
|
|
|
|
#endif
|
|
|
|
|
#ifndef _AVX512BF16INTRIN_H_INCLUDED
|
|
|
|
|
#define _AVX512BF16INTRIN_H_INCLUDED
|
2024-07-23 10:16:17 +00:00
|
|
|
#if !defined (__AVX512BF16__) || defined (__EVEX512__)
|
2023-06-05 07:37:25 +00:00
|
|
|
#pragma GCC push_options
|
2024-07-23 10:16:17 +00:00
|
|
|
#pragma GCC target("avx512bf16,no-evex512")
|
2023-06-05 07:37:25 +00:00
|
|
|
#define __DISABLE_AVX512BF16__
|
|
|
|
|
#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 float
|
|
|
|
|
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
|
2024-07-23 10:16:17 +00:00
|
|
|
_mm_cvtsbh_ss (__bf16 __A)
|
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
|
|
|
{
|
2024-07-23 10:16:17 +00:00
|
|
|
return __builtin_ia32_cvtbf2sf (__A);
|
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
|
|
|
}
|
2024-07-23 10:16:17 +00:00
|
|
|
#ifdef __DISABLE_AVX512BF16__
|
|
|
|
|
#undef __DISABLE_AVX512BF16__
|
|
|
|
|
#pragma GCC pop_options
|
|
|
|
|
#endif
|
|
|
|
|
#if !defined (__AVX512BF16__) || !defined (__EVEX512__)
|
|
|
|
|
#pragma GCC push_options
|
|
|
|
|
#pragma GCC target("avx512bf16,evex512")
|
|
|
|
|
#define __DISABLE_AVX512BF16_512__
|
|
|
|
|
#endif
|
|
|
|
|
typedef __bf16 __v32bf __attribute__ ((__vector_size__ (64)));
|
|
|
|
|
typedef __bf16 __m512bh __attribute__ ((__vector_size__ (64), __may_alias__));
|
2023-06-05 07:37:25 +00:00
|
|
|
extern __inline __m512bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_cvtne2ps_pbh (__m512 __A, __m512 __B)
|
|
|
|
|
{
|
2024-07-23 10:16:17 +00:00
|
|
|
return (__m512bh)__builtin_ia32_cvtne2ps2bf16_v32bf(__A, __B);
|
2023-06-05 07:37:25 +00:00
|
|
|
}
|
|
|
|
|
extern __inline __m512bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_mask_cvtne2ps_pbh (__m512bh __A, __mmask32 __B, __m512 __C, __m512 __D)
|
|
|
|
|
{
|
2024-07-23 10:16:17 +00:00
|
|
|
return (__m512bh)__builtin_ia32_cvtne2ps2bf16_v32bf_mask(__C, __D, __A, __B);
|
2023-06-05 07:37:25 +00:00
|
|
|
}
|
|
|
|
|
extern __inline __m512bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_maskz_cvtne2ps_pbh (__mmask32 __A, __m512 __B, __m512 __C)
|
|
|
|
|
{
|
2024-07-23 10:16:17 +00:00
|
|
|
return (__m512bh)__builtin_ia32_cvtne2ps2bf16_v32bf_maskz(__B, __C, __A);
|
2023-06-05 07:37:25 +00:00
|
|
|
}
|
|
|
|
|
extern __inline __m256bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_cvtneps_pbh (__m512 __A)
|
|
|
|
|
{
|
|
|
|
|
return (__m256bh)__builtin_ia32_cvtneps2bf16_v16sf(__A);
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m256bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_mask_cvtneps_pbh (__m256bh __A, __mmask16 __B, __m512 __C)
|
|
|
|
|
{
|
|
|
|
|
return (__m256bh)__builtin_ia32_cvtneps2bf16_v16sf_mask(__C, __A, __B);
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m256bh
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_maskz_cvtneps_pbh (__mmask16 __A, __m512 __B)
|
|
|
|
|
{
|
|
|
|
|
return (__m256bh)__builtin_ia32_cvtneps2bf16_v16sf_maskz(__B, __A);
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m512
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_dpbf16_ps (__m512 __A, __m512bh __B, __m512bh __C)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)__builtin_ia32_dpbf16ps_v16sf(__A, __B, __C);
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m512
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_mask_dpbf16_ps (__m512 __A, __mmask16 __B, __m512bh __C, __m512bh __D)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)__builtin_ia32_dpbf16ps_v16sf_mask(__A, __C, __D, __B);
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m512
|
|
|
|
|
__attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_maskz_dpbf16_ps (__mmask16 __A, __m512 __B, __m512bh __C, __m512bh __D)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)__builtin_ia32_dpbf16ps_v16sf_maskz(__B, __C, __D, __A);
|
|
|
|
|
}
|
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 __m512
|
|
|
|
|
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_cvtpbh_ps (__m256bh __A)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)_mm512_castsi512_ps ((__m512i)_mm512_slli_epi32 (
|
|
|
|
|
(__m512i)_mm512_cvtepi16_epi32 ((__m256i)__A), 16));
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m512
|
|
|
|
|
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_maskz_cvtpbh_ps (__mmask16 __U, __m256bh __A)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)_mm512_castsi512_ps ((__m512i) _mm512_slli_epi32 (
|
|
|
|
|
(__m512i)_mm512_maskz_cvtepi16_epi32 (
|
|
|
|
|
(__mmask16)__U, (__m256i)__A), 16));
|
|
|
|
|
}
|
|
|
|
|
extern __inline __m512
|
|
|
|
|
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
|
|
|
|
|
_mm512_mask_cvtpbh_ps (__m512 __S, __mmask16 __U, __m256bh __A)
|
|
|
|
|
{
|
|
|
|
|
return (__m512)_mm512_castsi512_ps ((__m512i)(_mm512_mask_slli_epi32 (
|
|
|
|
|
(__m512i)__S, (__mmask16)__U,
|
|
|
|
|
(__m512i)_mm512_cvtepi16_epi32 ((__m256i)__A), 16)));
|
|
|
|
|
}
|
2024-07-23 10:16:17 +00:00
|
|
|
#ifdef __DISABLE_AVX512BF16_512__
|
|
|
|
|
#undef __DISABLE_AVX512BF16_512__
|
2023-06-05 07:37:25 +00:00
|
|
|
#pragma GCC pop_options
|
|
|
|
|
#endif
|
|
|
|
|
#endif
|
|
|
|
|
#endif
|