derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
455 lines
17 KiB
C
455 lines
17 KiB
C
/*===-------- avxvnniint8intrin.h - AVXVNNIINT8 intrinsics -----------===
|
|
*
|
|
* Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
* See https://llvm.org/LICENSE.txt for license information.
|
|
* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
*
|
|
*===-----------------------------------------------------------------------===
|
|
*/
|
|
#ifndef __IMMINTRIN_H
|
|
#error \
|
|
"Never use <avxvnniint8intrin.h> directly; include <immintrin.h> instead."
|
|
#endif
|
|
|
|
#ifndef __AVXVNNIINT8INTRIN_H
|
|
#define __AVXVNNIINT8INTRIN_H
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbssd_epi32(__m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSSD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := SignExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j])
|
|
/// tmp2.word := SignExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := SignExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := SignExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbssd_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbssd128((__v4si)(__W), (__v16qi)(__A), \
|
|
(__v16qi)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbssd_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSSD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := SignExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j])
|
|
/// tmp2.word := SignExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := SignExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := SignExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbssd_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbssd256((__v8si)(__W), (__v32qi)(__A), \
|
|
(__v32qi)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbssds_epi32( __m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSSDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := SignExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j])
|
|
/// tmp2.word := SignExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := SignExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := SignExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := SIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbssds_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbssds128((__v4si)(__W), (__v16qi)(__A), \
|
|
(__v16qi)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding signed 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbssds_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSSDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := SignExtend16(__A.byte[4*j]) * SignExtend16(__B.byte[4*j])
|
|
/// tmp2.word := SignExtend16(__A.byte[4*j+1]) * SignExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := SignExtend16(__A.byte[4*j+2]) * SignExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := SignExtend16(__A.byte[4*j+3]) * SignExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := SIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbssds_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbssds256((__v8si)(__W), (__v32qi)(__A), \
|
|
(__v32qi)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbsud_epi32(__m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSUD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := Signed(SignExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j]))
|
|
/// tmp2.word := Signed(SignExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1]))
|
|
/// tmp3.word := Signed(SignExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2]))
|
|
/// tmp4.word := Signed(SignExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3]))
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbsud_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbsud128((__v4si)(__W), (__v16qi)(__A), \
|
|
(__v16qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbsud_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSUD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := Signed(SignExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j]))
|
|
/// tmp2.word := Signed(SignExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1]))
|
|
/// tmp3.word := Signed(SignExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2]))
|
|
/// tmp4.word := Signed(SignExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3]))
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbsud_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbsud256((__v8si)(__W), (__v32qi)(__A), \
|
|
(__v32qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbsuds_epi32( __m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSUDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := Signed(SignExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j]))
|
|
/// tmp2.word := Signed(SignExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1]))
|
|
/// tmp3.word := Signed(SignExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2]))
|
|
/// tmp4.word := Signed(SignExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3]))
|
|
/// dst.dword[j] := SIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbsuds_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbsuds128((__v4si)(__W), (__v16qi)(__A), \
|
|
(__v16qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbsuds_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBSUDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := Signed(SignExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j]))
|
|
/// tmp2.word := Signed(SignExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1]))
|
|
/// tmp3.word := Signed(SignExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2]))
|
|
/// tmp4.word := Signed(SignExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3]))
|
|
/// dst.dword[j] := SIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbsuds_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbsuds256((__v8si)(__W), (__v32qi)(__A), \
|
|
(__v32qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbuud_epi32(__m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBUUD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := ZeroExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j])
|
|
/// tmp2.word := ZeroExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := ZeroExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := ZeroExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbuud_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbuud128((__v4si)(__W), (__v16qu)(__A), \
|
|
(__v16qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W, and store the packed 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbuud_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBUUD instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := ZeroExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j])
|
|
/// tmp2.word := ZeroExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := ZeroExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := ZeroExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := __W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbuud_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbuud256((__v8si)(__W), (__v32qu)(__A), \
|
|
(__v32qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in \a __A with
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm_dpbuuds_epi32( __m128i __W, __m128i __A, __m128i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBUUDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \param __B
|
|
/// A 128-bit vector of [16 x unsigned char].
|
|
/// \returns
|
|
/// A 128-bit vector of [4 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 3
|
|
/// tmp1.word := ZeroExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j])
|
|
/// tmp2.word := ZeroExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := ZeroExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := ZeroExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := UNSIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:128] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm_dpbuuds_epi32(__W, __A, __B) \
|
|
((__m128i)__builtin_ia32_vpdpbuuds128((__v4si)(__W), (__v16qu)(__A), \
|
|
(__v16qu)(__B)))
|
|
|
|
// clang-format off
|
|
/// corresponding unsigned 8-bit integers in \a __B, producing 4 intermediate
|
|
/// signed 16-bit results. Sum these 4 results with the corresponding
|
|
/// 32-bit integer in \a __W with signed saturation, and store the packed
|
|
/// 32-bit results in \a dst.
|
|
///
|
|
/// \headerfile <x86intrin.h>
|
|
///
|
|
/// \code
|
|
/// _mm256_dpbuuds_epi32(__m256i __W, __m256i __A, __m256i __B);
|
|
/// \endcode
|
|
///
|
|
/// This intrinsic corresponds to the \c VPDPBUUDS instruction.
|
|
///
|
|
/// \param __A
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \param __B
|
|
/// A 256-bit vector of [32 x unsigned char].
|
|
/// \returns
|
|
/// A 256-bit vector of [8 x int].
|
|
///
|
|
/// \code{.operation}
|
|
/// FOR j := 0 to 7
|
|
/// tmp1.word := ZeroExtend16(__A.byte[4*j]) * ZeroExtend16(__B.byte[4*j])
|
|
/// tmp2.word := ZeroExtend16(__A.byte[4*j+1]) * ZeroExtend16(__B.byte[4*j+1])
|
|
/// tmp3.word := ZeroExtend16(__A.byte[4*j+2]) * ZeroExtend16(__B.byte[4*j+2])
|
|
/// tmp4.word := ZeroExtend16(__A.byte[4*j+3]) * ZeroExtend16(__B.byte[4*j+3])
|
|
/// dst.dword[j] := UNSIGNED_DWORD_SATURATE(__W.dword[j] + tmp1 + tmp2 + tmp3 + tmp4)
|
|
/// ENDFOR
|
|
/// dst[MAX:256] := 0
|
|
/// \endcode
|
|
// clang-format on
|
|
#define _mm256_dpbuuds_epi32(__W, __A, __B) \
|
|
((__m256i)__builtin_ia32_vpdpbuuds256((__v8si)(__W), (__v32qu)(__A), \
|
|
(__v32qu)(__B)))
|
|
|
|
#endif // __AVXVNNIINT8INTRIN_H
|