Google Git
Sign in
third-party-mirror / eigen / fuchsia / 43872d846644009af7b03b582a11f975485e197d / . / unsupported / Eigen / CXX11 / src / FixedPoint / MatMatProductAVX2.h
blob: bf8d9405c36405088bffe35e8c61fe30b3267361 [file] [log] [blame]
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2015 Benoit Steiner <benoit.steiner.goog@gmail.com>
// Copyright (C) 2015 Matthew Sarett <msarett@google.com>
// Copyright (C) 2016 Nishant Patil <nishantpatil@google.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_CXX11_FIXED_POINT_MAT_MAT_PRODUCT_AVX2_H
#define EIGEN_CXX11_FIXED_POINT_MAT_MAT_PRODUCT_AVX2_H
namespace Eigen {
namespace internal {
// AVX2 optimized implementation of Mat-Mat product.
// LHS is encoded using signed 16-bit integers.
// RHS is encoded using signed 16-bit integers.
#ifdef EIGEN_USE_OPTIMIZED_INT16_INT16_MAT_MAT_PRODUCT
// Define quantized traits
template <bool _ConjLhs, bool _ConjRhs>
class gebp_traits<QInt16, QInt16, _ConjLhs, _ConjRhs> {
public:
typedef QInt16 LhsScalar;
typedef QInt16 RhsScalar;
typedef QInt32 ResScalar;
enum {
// Define register blocking scheme.
nr = 16,
mr = 16,
kr = 4,
// Ignore progress tracking per loop iteration.
LhsProgress = -1,
RhsProgress = -1
};
};
// Specialized blocking for quantized implementations.
// Used by TensorContractionThreadPool, inputs must have dimensions that are
// multiples of 32.
template <typename Index, int ShardingType>
class TensorContractionBlocking<QInt16, QInt16, Index, ShardingType> {
public:
TensorContractionBlocking(Index k, Index m, Index n, Index num_threads = 1)
: kc_(((k + 15) / 16) * 16),
mc_(((m + 15) / 16) * 16),
nc_(((n + 15) / 16) * 16) {
eigen_assert(mc_ % 16 == 0);
eigen_assert(kc_ % 16 == 0);
if (!k || !m || !n) {
return;
}
if (ShardingType == ShardByCol) {
eigen_assert(nc_ % 16 == 0);
nc_ = (((nc_ / num_threads) + 15) / 16) * 16;
} else {
eigen_assert(nc_ % 16 == 0);
mc_ = (((mc_ / num_threads) + 15) / 16) * 16;
}
}
EIGEN_ALWAYS_INLINE Index kc() const { return kc_; }
EIGEN_ALWAYS_INLINE Index mc() const { return mc_; }
EIGEN_ALWAYS_INLINE Index nc() const { return nc_; }
private:
Index kc_;
Index mc_;
Index nc_;
};
// Specialized blocking for quantized implementations.
// Used by TensorContraction and GeneralMatrixMatrix, inputs are padded to
// multiples of 32.
template <int MaxRows, int MaxCols, int MaxDepth, int KcFactor>
class gemm_blocking_space<ColMajor, QInt16, QInt16, MaxRows, MaxCols, MaxDepth,
KcFactor, false>
: public level3_blocking<QInt16, QInt16> {
DenseIndex m_sizeA;
DenseIndex m_sizeB;
public:
gemm_blocking_space(DenseIndex rows, DenseIndex cols, DenseIndex depth,
DenseIndex /*num_threads*/, bool /*l3_blocking*/) {
this->m_mc = ((rows + 15) / 16) * 16;
this->m_nc = ((cols + 15) / 16) * 16;
this->m_kc = ((depth + 15) / 16) * 16;
m_sizeA = this->m_mc * this->m_kc;
m_sizeB = this->m_kc * this->m_nc;
}
void allocateA() {
if (this->m_blockA == 0) this->m_blockA = aligned_new<QInt16>(m_sizeA);
}
void allocateB() {
if (this->m_blockB == 0) this->m_blockB = aligned_new<QInt16>(m_sizeB);
}
void allocateAll() {
allocateA();
allocateB();
}
~gemm_blocking_space() {
aligned_delete(this->m_blockA, m_sizeA);
aligned_delete(this->m_blockB, m_sizeB);
}
};
// Below are the fully optimized versions that are correct only for sizes that
// are multiple of 16. It is about a 10% performance benefit to keep these
// implementations separate.
// Arrange a block of the left input matrix in contiguous memory.
//
// Given column major input (A0 beside A1 in memory):
// A0 B0 C0 D0 E0 F0 G0 H0 ...
// A1 B1 C1 D1 E1 F1 G1 H1 ...
// A2 B2 C2 D2 E2 F2 G2 H2 ...
// A3 B3 C3 D3 E3 F3 G3 H3 ...
// A4 B4 C4 D4 E4 F4 G4 H4 ...
// A5 B5 C5 D5 E5 F5 G5 H5 ...
// A6 B6 C6 D6 E6 F6 G6 H6 ...
// A7 B7 C7 D7 E7 F7 G7 H7 ...
// A8 ...
// ...
//
// Packing with m = 8 yields row major output (A0 beside B0 in memory):
// A0 B0
// A1 B1
// A2 B2
// A3 B3
// A4 B4
// A5 B5
// A6 B6
// A7 B7
// ...
//
// The purpose is to collect m rows of size k. Two elements of the same
// row are arranged contiguously because madd performs an adjacent addition
// in the kernel.
template <typename Index, typename DataMapper, int Pack1, int Pack2,
bool Conjugate, bool PanelMode>
struct gemm_pack_lhs<QInt16, Index, DataMapper, Pack1, Pack2, ColMajor,
Conjugate, PanelMode> {
EIGEN_DONT_INLINE void operator()(QInt16* blockA, const DataMapper& lhs,
Index depth, Index rows, Index stride = 0,
Index offset = 0);
};
template <typename Index, typename DataMapper, int Pack1, int Pack2,
bool Conjugate, bool PanelMode>
EIGEN_DONT_INLINE void gemm_pack_lhs<QInt16, Index, DataMapper, Pack1, Pack2,
ColMajor, Conjugate, PanelMode>::
operator()(QInt16* blockA, const DataMapper& lhs, Index depth, Index rows,
Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Use alternate function for weird sizes
if (rows % 16 != 0 || depth % 16 != 0) {
assert(false && "only depths and rows that are a multiple of 16 are currently supported");
// gemm_pack_lhs_any<QInt16, Index, DataMapper, Pack1, Pack2, ColMajor,
// Conjugate, PanelMode> lhs_pack;
// return lhs_pack(blockA, lhs, depth, rows, stride, offset);
}
// Get vector pointer
__m256i* blockA_256 = reinterpret_cast<__m256i*>(blockA);
// Pack rows in sets of 16
for (Index m = 0; m < rows; m += 16) {
// Pack depth in sets of 4
for (Index k = 0; k < depth; k += 4) {
// Load vectors
__m256i L_A = lhs.loadPacket(m, k);
__m256i L_B = lhs.loadPacket(m, k + 1);
__m256i L_C = lhs.loadPacket(m, k + 2);
__m256i L_D = lhs.loadPacket(m, k + 3);
// Rearrange the inputs as required by the kernel
__m256i L_AB0_AB7 = _mm256_unpacklo_epi16(L_A, L_B);
__m256i L_AB8_AB15 = _mm256_unpackhi_epi16(L_A, L_B);
__m256i L_CD0_CD7 = _mm256_unpacklo_epi16(L_C, L_D);
__m256i L_CD8_CD15 = _mm256_unpackhi_epi16(L_C, L_D);
__m256i L_AD0 = _mm256_permute2x128_si256(L_AB0_AB7, L_AB8_AB15, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
__m256i L_AD8 = _mm256_permute2x128_si256(L_CD0_CD7, L_CD8_CD15, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
__m256i L_AD16 = _mm256_permute2x128_si256(L_AB0_AB7, L_AB8_AB15, 0x31);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_CD0_CD7, L_CD8_CD15, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
}
}
}
// Arrange a block of the right input matrix in contiguous memory.
//
// Given column major input (A0 beside A1 in memory):
// A0 B0 C0 D0 E0 F0 G0 H0 ...
// A1 B1 C1 D1 E1 F1 G1 H1 ...
// A2 B2 C2 D2 E2 F2 G2 H2 ...
// A3 B3 C3 D3 E3 F3 G3 H3 ...
// A4 B4 C4 D4 E4 F4 G4 H4 ...
// A5 B5 C5 D5 E5 F5 G5 H5 ...
// A6 B6 C6 D6 E6 F6 G6 H6 ...
// A7 B7 C7 D7 E7 F7 G7 H7 ...
// A8 ...
// ...
// Packing yields row major output (A0 beside A1 in memory):
// A0 A1 A2 A3 A4 A5 A6 A7
// B0 B1 B2 B3 B4 B5 B6 B7
// ...
//
// At least two elements of the same col are arranged contiguously because
// maddubs and madd both perform an adjacent addition in the kernel. We can
// save work by leaving 4 adjacent elements because kr = 4.
// The purpose is to collect n cols of size k. Two elements of the same
// col are arranged contiguously because madd performs an adjacent addition
// in the kernel.
template <typename Index, typename DataMapper, int nr, bool Conjugate,
bool PanelMode>
struct gemm_pack_rhs<QInt16, Index, DataMapper, nr, ColMajor, Conjugate,
PanelMode> {
EIGEN_DONT_INLINE void operator()(QInt16* blockB, const DataMapper& rhs,
Index depth, Index cols, Index stride = 0,
Index offset = 0);
};
template <typename Index, typename DataMapper, int nr, bool Conjugate,
bool PanelMode>
EIGEN_DONT_INLINE void
gemm_pack_rhs<QInt16, Index, DataMapper, nr, ColMajor, Conjugate, PanelMode>::
operator()(QInt16* blockB, const DataMapper& rhs, Index depth, Index cols,
Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Use alternate function for weird sizes
if (cols % 16 != 0 || depth % 16 != 0) {
assert(false && "only depths and cols that are a multiple of 16 are currently supported");
// gemm_pack_rhs_any<QInt16, Index, DataMapper, nr, ColMajor, Conjugate,
// PanelMode> rhs_pack;
// return rhs_pack(blockB, rhs, depth, cols, stride, offset);
}
// Get vector pointer
__m256i* blockB_256 = reinterpret_cast<__m256i*>(blockB);
// Perform a step of the packing for 4 columns
__m256i R_AB_L, R_AB_H, R_CD_L, R_CD_H, R_AD_0, R_AD_4, R_AD_8, R_AD_12;
#define PACK_STEP \
R_AB_L = _mm256_unpacklo_epi64(R_A, R_B); \
R_CD_L = _mm256_unpacklo_epi64(R_C, R_D); \
R_AB_H = _mm256_unpackhi_epi64(R_A, R_B); \
R_CD_H = _mm256_unpackhi_epi64(R_C, R_D); \
R_AD_0 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x20); \
R_AD_8 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x31); \
R_AD_4 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x20); \
R_AD_12 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x31); \
_mm256_store_si256(blockB_256, R_AD_0); \
_mm256_store_si256(blockB_256 + 4, R_AD_4); \
_mm256_store_si256(blockB_256 + 8, R_AD_8); \
_mm256_store_si256(blockB_256 + 12, R_AD_12); \
blockB_256++;
// Pack cols in sets of 16
for (Index n = 0; n < cols; n += 16) {
// Pack depth in sets of 16
for (Index k = 0; k < depth; k += 16) {
__m256i R_A = rhs.loadPacket(k, n);
__m256i R_B = rhs.loadPacket(k, n + 1);
__m256i R_C = rhs.loadPacket(k, n + 2);
__m256i R_D = rhs.loadPacket(k, n + 3);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 4);
R_B = rhs.loadPacket(k, n + 5);
R_C = rhs.loadPacket(k, n + 6);
R_D = rhs.loadPacket(k, n + 7);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 8);
R_B = rhs.loadPacket(k, n + 9);
R_C = rhs.loadPacket(k, n + 10);
R_D = rhs.loadPacket(k, n + 11);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 12);
R_B = rhs.loadPacket(k, n + 13);
R_C = rhs.loadPacket(k, n + 14);
R_D = rhs.loadPacket(k, n + 15);
PACK_STEP;
blockB_256 += 12;
}
}
#undef PACK_STEP
}
// Perform the actual multiplication on packed inputs
template <typename Index, typename DataMapper, int mr, int nr,
bool ConjugateLhs, bool ConjugateRhs>
struct gebp_kernel<QInt16, QInt16, Index, DataMapper, mr, nr, ConjugateLhs,
ConjugateRhs> {
typedef typename DataMapper::LinearMapper LinearMapper;
EIGEN_DONT_INLINE
void operator()(const DataMapper& res, const QInt16* blockA,
const QInt16* blockB, Index rows, Index depth, Index cols,
QInt32 alpha, Index strideA = -1, Index strideB = -1,
Index offsetA = 0, Index offsetB = 0);
};
template <typename Index, typename DataMapper, int mr, int nr,
bool ConjugateLhs, bool ConjugateRhs>
EIGEN_DONT_INLINE void gebp_kernel<QInt16, QInt16, Index, DataMapper, mr, nr,
ConjugateLhs, ConjugateRhs>::
operator()(const DataMapper& res, const QInt16* blockA, const QInt16* blockB,
Index rows, Index depth, Index cols, QInt32 alpha, Index strideA,
Index strideB, Index offsetA, Index offsetB) {
EIGEN_STATIC_ASSERT(!ConjugateLhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
EIGEN_STATIC_ASSERT(!ConjugateRhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
eigen_assert(alpha.value == 1);
eigen_assert(strideA == -1);
eigen_assert(strideB == -1);
eigen_assert(offsetA == 0);
eigen_assert(offsetB == 0);
eigen_assert(rows > 0);
eigen_assert(cols > 0);
eigen_assert(depth > 0);
eigen_assert(blockA);
eigen_assert(blockB);
// Use alternate function for weird sizes
if (rows % 16 != 0 || cols % 16 != 0 || depth % 16 != 0) {
assert(false && "only depths, cols and rows that are a multiple of 16 are currently supported");
// gebp_kernel_any<QInt16, QInt16, Index, DataMapper, mr, nr, ConjugateLhs,
// ConjugateRhs> gebp;
// return gebp(res, blockA, blockB, rows, depth, cols, alpha, strideA,
// strideB, offsetA, offsetB);
}
// Create result block
QInt32* blockO = aligned_new<QInt32>(16 * 16);
memset(blockO, 0, 16 * 16 * sizeof(QInt32));
// Get vectorized pointers
__m256i* blockO_256 = reinterpret_cast<__m256i*>(blockO);
const __m256i* blockA_256 = reinterpret_cast<const __m256i*>(blockA);
const __m256i* blockB_256 = reinterpret_cast<const __m256i*>(blockB);
// Loop over blocks of 16 columns
for (Index n = 0; n < cols; n += 16) {
// Reset index into blockA
Index indexL = 0;
// Loop over blocks of 16 rows
for (Index m = 0; m < rows; m += 16) {
// Reset index into blockB
Index indexR = n / 16 * depth;
// Loop over blocks of 4 on depth
for (Index k = 0; k < depth; k += 4) {
// Load inputs
__m256i L_AD0 = blockA_256[indexL++];
__m256i L_AD8 = blockA_256[indexL++];
__m256i L_EH0 = blockA_256[indexL++];
__m256i L_EH8 = blockA_256[indexL++];
__m256i R_AH0 = blockB_256[indexR++];
__m256i R_AH4 = blockB_256[indexR++];
__m256i R_AH8 = blockB_256[indexR++];
__m256i R_AH12 = blockB_256[indexR++];
// Declare variables used in COMPUTE_STEP
__m256i P_32_A, P_32_B, P_32;
#define COMPUTE_STEP(R_INPUT_A, R_INPUT_B, OFFSET) \
P_32_A = _mm256_madd_epi16(R_INPUT_A, L_AD0); \
P_32_B = _mm256_madd_epi16(R_INPUT_B, L_AD8); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 2 * OFFSET, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 2 * OFFSET), P_32)); \
\
P_32_A = _mm256_madd_epi16(R_INPUT_A, L_EH0); \
P_32_B = _mm256_madd_epi16(R_INPUT_B, L_EH8); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 2 * OFFSET + 1, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 2 * OFFSET + 1), P_32));
// Permute and shuffle to copy a single value across the entire vector
// Then compute the multiplication
// Replicate lower 128-bits of R_AH0 across both lanes
__m256i R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x00);
// Copy first two elements of R_AH0 across entire vector
__m256i R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
// Copy second two elements of R_AH0 across entire vector
__m256i R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 0);
__m256i R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 1);
// Replicate upper 128-bits of R_AH0 across both lanes
R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x11);
__m256i R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
__m256i R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 2);
__m256i R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 3);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 4);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 5);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 6);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 7);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 8);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 9);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 10);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 11);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 12);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 13);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 14);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 15);
#undef COMPUTE_STEP
}
// Transfer the results to the result matrix
Index i = 0;
for (Index j = n; j < n + 16; j++) {
LinearMapper r0 = res.getLinearMapper(m, j);
LinearMapper r1 = res.getLinearMapper(m + 8, j);
r0.storePacket(0, _mm256_add_epi32(blockO_256[i++], r0.loadPacket(0)));
r1.storePacket(0, _mm256_add_epi32(blockO_256[i++], r1.loadPacket(0)));
}
// Zero the result block so it can be reused
memset(blockO, 0, 16 * 16 * sizeof(QInt32));
}
}
aligned_delete(blockO, 16 * 16);
}
#endif
// AVX2 optimized implementation of Mat-Mat product.
// LHS is encoded using signed 8-bit integers.
// RHS is encoded using unsigned 8-bit integers.
#ifdef EIGEN_USE_OPTIMIZED_INT8_UINT8_MAT_MAT_PRODUCT
// Define quantized traits
template<bool _ConjLhs, bool _ConjRhs>
class gebp_traits<QInt8, QUInt8, _ConjLhs, _ConjRhs>
{
public:
typedef QInt8 LhsScalar;
typedef QUInt8 RhsScalar;
typedef QInt32 ResScalar;
enum {
// Define register blocking scheme.
nr = 32,
mr = 32,
kr = 8,
// Ignore progress tracking per loop iteration.
LhsProgress = -1,
RhsProgress = -1
};
};
// Specialized blocking for quantized implementations.
// Used by TensorContractionThreadPool, inputs must have dimensions that are
// multiples of 32.
template <typename Index, int ShardingType>
class TensorContractionBlocking<QInt8, QUInt8, Index, ShardingType> {
public:
TensorContractionBlocking(Index k, Index m, Index n, Index num_threads = 1) :
kc_(((k + 31)/32)*32),
mc_(((m + 31)/32)*32),
nc_(((n + 31)/32)*32)
{
eigen_assert(mc_ % 32 == 0);
eigen_assert(kc_ % 32 == 0);
if (!k || !m || !n) {
return;
}
if (ShardingType == ShardByCol) {
eigen_assert(nc_ % 32 == 0);
nc_ = (((nc_ / num_threads) + 31) / 32) * 32;
}
else {
eigen_assert(nc_ % 32 == 0);
mc_ = (((mc_ / num_threads) + 31) / 32) * 32;
}
}
EIGEN_ALWAYS_INLINE Index kc() const { return kc_; }
EIGEN_ALWAYS_INLINE Index mc() const { return mc_; }
EIGEN_ALWAYS_INLINE Index nc() const { return nc_; }
private:
Index kc_;
Index mc_;
Index nc_;
};
// Specialized blocking for quantized implementations.
// Used by TensorContraction and GeneralMatrixMatrix, inputs are padded to
// multiples of 32.
template <int MaxRows, int MaxCols, int MaxDepth, int KcFactor>
class gemm_blocking_space<ColMajor, QInt8, QInt8, MaxRows, MaxCols, MaxDepth,
KcFactor, false>
: public level3_blocking<QInt8, QInt8> {
DenseIndex m_sizeA;
DenseIndex m_sizeB;
public:
gemm_blocking_space(DenseIndex rows, DenseIndex cols, DenseIndex depth,
DenseIndex /*num_threads*/, bool /*l3_blocking*/) {
this->m_mc = ((rows + 31) / 32) * 32;
this->m_nc = ((cols + 31) / 32) * 32;
this->m_kc = ((depth + 31) / 32) * 32;
m_sizeA = this->m_mc * this->m_kc;
m_sizeB = this->m_kc * this->m_nc;
}
void allocateA() {
if (this->m_blockA == 0) this->m_blockA = aligned_new<QInt8>(m_sizeA);
}
void allocateB() {
if (this->m_blockB == 0) this->m_blockB = aligned_new<QInt8>(m_sizeB);
}
void allocateAll() {
allocateA();
allocateB();
}
~gemm_blocking_space() {
aligned_delete(this->m_blockA, m_sizeA);
aligned_delete(this->m_blockB, m_sizeB);
}
};
template <int MaxRows, int MaxCols, int MaxDepth, int KcFactor>
class gemm_blocking_space<ColMajor, QInt8, QUInt8, MaxRows, MaxCols, MaxDepth,
KcFactor, false>
: public level3_blocking<QInt8, QUInt8> {
DenseIndex m_sizeA;
DenseIndex m_sizeB;
public:
gemm_blocking_space(DenseIndex rows, DenseIndex cols, DenseIndex depth,
DenseIndex /*num_threads*/, bool /*l3_blocking*/) {
this->m_mc = ((rows + 31) / 32) * 32;
this->m_nc = ((cols + 31) / 32) * 32;
this->m_kc = ((depth + 31) / 32) * 32;
m_sizeA = this->m_mc * this->m_kc;
m_sizeB = this->m_kc * this->m_nc;
}
void allocateA() {
if (this->m_blockA == 0) this->m_blockA = aligned_new<QInt8>(m_sizeA);
}
void allocateB() {
if (this->m_blockB == 0) this->m_blockB = aligned_new<QUInt8>(m_sizeB);
}
void allocateAll() {
allocateA();
allocateB();
}
~gemm_blocking_space() {
aligned_delete(this->m_blockA, m_sizeA);
aligned_delete(this->m_blockB, m_sizeB);
}
};
// Alternate templates for any input sizes
template<typename Scalar, typename Index, typename DataMapper, int Pack1, int Pack2, int StorageOrder, bool Conjugate = false, bool PanelMode = false>
struct gemm_pack_lhs_any;
template <typename Index, typename DataMapper, int Pack1, int Pack2, bool Conjugate, bool PanelMode>
struct gemm_pack_lhs_any<QInt8, Index, DataMapper, Pack1, Pack2, ColMajor, Conjugate, PanelMode> {
EIGEN_DONT_INLINE void operator()
(QInt8* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride = 0, Index offset = 0);
};
template<typename Scalar, typename Index, typename DataMapper, int nr, int StorageOrder, bool Conjugate = false, bool PanelMode=false>
struct gemm_pack_rhs_any;
template <typename Index, typename DataMapper, int nr, bool Conjugate, bool PanelMode>
struct gemm_pack_rhs_any<QUInt8, Index, DataMapper, nr, ColMajor, Conjugate, PanelMode> {
EIGEN_DONT_INLINE void operator()
(QUInt8* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0);
};
template<typename LhsScalar, typename RhsScalar, typename Index, typename DataMapper, int mr, int nr, bool ConjugateLhs=false, bool ConjugateRhs=false>
struct gebp_kernel_any;
template<typename Index, typename DataMapper, int mr, int nr, bool ConjugateLhs, bool ConjugateRhs>
struct gebp_kernel_any<QInt8, QUInt8, Index, DataMapper, mr, nr, ConjugateLhs, ConjugateRhs>
{
typedef typename DataMapper::LinearMapper LinearMapper;
EIGEN_DONT_INLINE
void operator()(const DataMapper& res, const QInt8* blockA, const QUInt8* blockB,
Index rows, Index depth, Index cols, QInt32 alpha,
Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0);
};
// Alternate implementations for any input sizes
template <typename Index, typename DataMapper, int Pack1, int Pack2, bool Conjugate, bool PanelMode>
EIGEN_DONT_INLINE void gemm_pack_lhs_any<QInt8, Index, DataMapper, Pack1, Pack2, ColMajor, Conjugate, PanelMode>::
operator()(QInt8* blockA, const DataMapper& lhs, Index depth, Index rows, Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Get vector pointer
__m256i* blockA_256 = reinterpret_cast<__m256i*>(blockA);
// Get even multiples of the dimensions
Index rows_32 = (rows / 32) * 32;
Index depth_8 = (depth / 8) * 8;
// Get padding for when depth is not a multiple of 32
int padding = 0;
if (depth % 32 != 0) {
int depth_32 = (depth / 32) * 32;
int extra_depth = depth - depth_32;
int extra_depth_8 = ((extra_depth + 7) / 8) * 8;
padding = 32 - extra_depth_8;
}
// Pack rows in sets of 32
for (Index m = 0; m < rows_32; m += 32) {
// Pack depth in sets of 8
for (Index k = 0; k < depth_8; k += 8) {
// Load vectors
__m256i L_A = lhs.loadPacket(m, k);
__m256i L_B = lhs.loadPacket(m, k + 1);
// Interleave 8-bit elements
__m256i L_AB0_AB16 = _mm256_unpacklo_epi8(L_A, L_B);
__m256i L_AB8_AB24 = _mm256_unpackhi_epi8(L_A, L_B);
__m256i L_C = lhs.loadPacket(m, k + 2);
__m256i L_D = lhs.loadPacket(m, k + 3);
__m256i L_CD0_CD16 = _mm256_unpacklo_epi8(L_C, L_D);
__m256i L_CD8_CD24 = _mm256_unpackhi_epi8(L_C, L_D);
// Interleave 16-bit elements
__m256i L_AD0_AD16 = _mm256_unpacklo_epi16(L_AB0_AB16, L_CD0_CD16);
__m256i L_AD4_AD20 = _mm256_unpackhi_epi16(L_AB0_AB16, L_CD0_CD16);
// Use permute before we store to cross 128-bit lanes
__m256i L_AD0 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
// Complete packing for 32 x 8 block
__m256i L_AD16 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x31);
__m256i L_AD8_AD24 = _mm256_unpacklo_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD12_AD28 = _mm256_unpackhi_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD8 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
__m256i L_E = lhs.loadPacket(m, k + 4);
__m256i L_F = lhs.loadPacket(m, k + 5);
__m256i L_EF0_EF16 = _mm256_unpacklo_epi8(L_E, L_F);
__m256i L_EF8_EF24 = _mm256_unpackhi_epi8(L_E, L_F);
__m256i L_G = lhs.loadPacket(m, k + 6);
__m256i L_H = lhs.loadPacket(m, k + 7);
__m256i L_GH0_GH16 = _mm256_unpacklo_epi8(L_G, L_H);
__m256i L_GH8_GH24 = _mm256_unpackhi_epi8(L_G, L_H);
__m256i L_EH0_EH16 = _mm256_unpacklo_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH4_EH20 = _mm256_unpackhi_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH0 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x20);
_mm256_store_si256(blockA_256++, L_EH0);
__m256i L_EH16 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x31);
__m256i L_EH8_EH24 = _mm256_unpacklo_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH12_EH28 = _mm256_unpackhi_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH8 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x20);
_mm256_store_si256(blockA_256++, L_EH8);
_mm256_store_si256(blockA_256++, L_EH16);
__m256i L_EH24 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x31);
_mm256_store_si256(blockA_256++, L_EH24);
}
// Finish the k dimension, padding with zeros
if (depth_8 < depth) {
__m256i L_A, L_B, L_C, L_D, L_E, L_F, L_G, L_H;
switch (depth - depth_8) {
case 1:
L_A = lhs.loadPacket(m, depth_8);
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 2:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 3:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = lhs.loadPacket(m, depth_8 + 2);
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 4:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = lhs.loadPacket(m, depth_8 + 2);
L_D = lhs.loadPacket(m, depth_8 + 3);
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 5:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = lhs.loadPacket(m, depth_8 + 2);
L_D = lhs.loadPacket(m, depth_8 + 3);
L_E = lhs.loadPacket(m, depth_8 + 4);
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 6:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = lhs.loadPacket(m, depth_8 + 2);
L_D = lhs.loadPacket(m, depth_8 + 3);
L_E = lhs.loadPacket(m, depth_8 + 4);
L_F = lhs.loadPacket(m, depth_8 + 5);
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
break;
case 7:
L_A = lhs.loadPacket(m, depth_8);
L_B = lhs.loadPacket(m, depth_8 + 1);
L_C = lhs.loadPacket(m, depth_8 + 2);
L_D = lhs.loadPacket(m, depth_8 + 3);
L_E = lhs.loadPacket(m, depth_8 + 4);
L_F = lhs.loadPacket(m, depth_8 + 5);
L_G = lhs.loadPacket(m, depth_8 + 6);
L_H = _mm256_setzero_si256();
break;
}
// Interleave 8-bit elements
__m256i L_AB0_AB16 = _mm256_unpacklo_epi8(L_A, L_B);
__m256i L_AB8_AB24 = _mm256_unpackhi_epi8(L_A, L_B);
__m256i L_CD0_CD16 = _mm256_unpacklo_epi8(L_C, L_D);
__m256i L_CD8_CD24 = _mm256_unpackhi_epi8(L_C, L_D);
// Interleave 16-bit elements
__m256i L_AD0_AD16 = _mm256_unpacklo_epi16(L_AB0_AB16, L_CD0_CD16);
__m256i L_AD4_AD20 = _mm256_unpackhi_epi16(L_AB0_AB16, L_CD0_CD16);
// Use permute before we store to cross 128-bit lanes
__m256i L_AD0 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
// Complete packing
__m256i L_AD16 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x31);
__m256i L_AD8_AD24 = _mm256_unpacklo_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD12_AD28 = _mm256_unpackhi_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD8 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
__m256i L_EF0_EF16 = _mm256_unpacklo_epi8(L_E, L_F);
__m256i L_EF8_EF24 = _mm256_unpackhi_epi8(L_E, L_F);
__m256i L_GH0_GH16 = _mm256_unpacklo_epi8(L_G, L_H);
__m256i L_GH8_GH24 = _mm256_unpackhi_epi8(L_G, L_H);
__m256i L_EH0_EH16 = _mm256_unpacklo_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH4_EH20 = _mm256_unpackhi_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH0 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x20);
_mm256_store_si256(blockA_256++, L_EH0);
__m256i L_EH16 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x31);
__m256i L_EH8_EH24 = _mm256_unpacklo_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH12_EH28 = _mm256_unpackhi_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH8 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x20);
_mm256_store_si256(blockA_256++, L_EH8);
_mm256_store_si256(blockA_256++, L_EH16);
__m256i L_EH24 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x31);
_mm256_store_si256(blockA_256++, L_EH24);
}
blockA_256 += padding;
}
// Finish the m dimension, padding with zeros
if (rows_32 < rows) {
// Pack depth in sets of 8
for (Index k = 0; k < depth_8; k += 8) {
// Load vectors
__m256i L_A = _mm256_setzero_si256();
__m256i L_B = _mm256_setzero_si256();
__m256i L_C = _mm256_setzero_si256();
__m256i L_D = _mm256_setzero_si256();
__m256i L_E = _mm256_setzero_si256();
__m256i L_F = _mm256_setzero_si256();
__m256i L_G = _mm256_setzero_si256();
__m256i L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
QInt8* ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, k);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, k + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, k + 2);
ptr = (QInt8*) &L_D;
ptr[m] = lhs(rows_32 + m, k + 3);
ptr = (QInt8*) &L_E;
ptr[m] = lhs(rows_32 + m, k + 4);
ptr = (QInt8*) &L_F;
ptr[m] = lhs(rows_32 + m, k + 5);
ptr = (QInt8*) &L_G;
ptr[m] = lhs(rows_32 + m, k + 6);
ptr = (QInt8*) &L_H;
ptr[m] = lhs(rows_32 + m, k + 7);
}
// Interleave 8-bit elements
__m256i L_AB0_AB16 = _mm256_unpacklo_epi8(L_A, L_B);
__m256i L_AB8_AB24 = _mm256_unpackhi_epi8(L_A, L_B);
__m256i L_CD0_CD16 = _mm256_unpacklo_epi8(L_C, L_D);
__m256i L_CD8_CD24 = _mm256_unpackhi_epi8(L_C, L_D);
// Interleave 16-bit elements
__m256i L_AD0_AD16 = _mm256_unpacklo_epi16(L_AB0_AB16, L_CD0_CD16);
__m256i L_AD4_AD20 = _mm256_unpackhi_epi16(L_AB0_AB16, L_CD0_CD16);
// Use permute before we store to cross 128-bit lanes
__m256i L_AD0 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
// Complete packing for 32 x 8 block
__m256i L_AD16 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x31);
__m256i L_AD8_AD24 = _mm256_unpacklo_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD12_AD28 = _mm256_unpackhi_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD8 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
__m256i L_EF0_EF16 = _mm256_unpacklo_epi8(L_E, L_F);
__m256i L_EF8_EF24 = _mm256_unpackhi_epi8(L_E, L_F);
__m256i L_GH0_GH16 = _mm256_unpacklo_epi8(L_G, L_H);
__m256i L_GH8_GH24 = _mm256_unpackhi_epi8(L_G, L_H);
__m256i L_EH0_EH16 = _mm256_unpacklo_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH4_EH20 = _mm256_unpackhi_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH0 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x20);
_mm256_store_si256(blockA_256++, L_EH0);
__m256i L_EH16 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x31);
__m256i L_EH8_EH24 = _mm256_unpacklo_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH12_EH28 = _mm256_unpackhi_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH8 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x20);
_mm256_store_si256(blockA_256++, L_EH8);
_mm256_store_si256(blockA_256++, L_EH16);
__m256i L_EH24 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x31);
_mm256_store_si256(blockA_256++, L_EH24);
}
// Finish the k dimension, padding with zeros
if (depth_8 < depth) {
__m256i L_A, L_B, L_C, L_D, L_E, L_F, L_G, L_H;
QInt8* ptr;
switch (depth - depth_8) {
case 1:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
QInt8* ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
}
break;
case 2:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
}
break;
case 3:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, depth_8 + 2);
}
break;
case 4:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, depth_8 + 2);
ptr = (QInt8*) &L_D;
ptr[m] = lhs(rows_32 + m, depth_8 + 3);
}
break;
case 5:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, depth_8 + 2);
ptr = (QInt8*) &L_D;
ptr[m] = lhs(rows_32 + m, depth_8 + 3);
ptr = (QInt8*) &L_E;
ptr[m] = lhs(rows_32 + m, depth_8 + 4);
}
break;
case 6:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, depth_8 + 2);
ptr = (QInt8*) &L_D;
ptr[m] = lhs(rows_32 + m, depth_8 + 3);
ptr = (QInt8*) &L_E;
ptr[m] = lhs(rows_32 + m, depth_8 + 4);
ptr = (QInt8*) &L_F;
ptr[m] = lhs(rows_32 + m, depth_8 + 5);
}
break;
case 7:
L_A = _mm256_setzero_si256();
L_B = _mm256_setzero_si256();
L_C = _mm256_setzero_si256();
L_D = _mm256_setzero_si256();
L_E = _mm256_setzero_si256();
L_F = _mm256_setzero_si256();
L_G = _mm256_setzero_si256();
L_H = _mm256_setzero_si256();
for (Index m = 0; m < rows - rows_32; m++) {
ptr = (QInt8*) &L_A;
ptr[m] = lhs(rows_32 + m, depth_8);
ptr = (QInt8*) &L_B;
ptr[m] = lhs(rows_32 + m, depth_8 + 1);
ptr = (QInt8*) &L_C;
ptr[m] = lhs(rows_32 + m, depth_8 + 2);
ptr = (QInt8*) &L_D;
ptr[m] = lhs(rows_32 + m, depth_8 + 3);
ptr = (QInt8*) &L_E;
ptr[m] = lhs(rows_32 + m, depth_8 + 4);
ptr = (QInt8*) &L_F;
ptr[m] = lhs(rows_32 + m, depth_8 + 5);
ptr = (QInt8*) &L_G;
ptr[m] = lhs(rows_32 + m, depth_8 + 6);
}
break;
}
// Interleave 8-bit elements
__m256i L_AB0_AB16 = _mm256_unpacklo_epi8(L_A, L_B);
__m256i L_AB8_AB24 = _mm256_unpackhi_epi8(L_A, L_B);
__m256i L_CD0_CD16 = _mm256_unpacklo_epi8(L_C, L_D);
__m256i L_CD8_CD24 = _mm256_unpackhi_epi8(L_C, L_D);
// Interleave 16-bit elements
__m256i L_AD0_AD16 = _mm256_unpacklo_epi16(L_AB0_AB16, L_CD0_CD16);
__m256i L_AD4_AD20 = _mm256_unpackhi_epi16(L_AB0_AB16, L_CD0_CD16);
// Use permute before we store to cross 128-bit lanes
__m256i L_AD0 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
// Complete packing
__m256i L_AD16 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x31);
__m256i L_AD8_AD24 = _mm256_unpacklo_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD12_AD28 = _mm256_unpackhi_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD8 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
__m256i L_EF0_EF16 = _mm256_unpacklo_epi8(L_E, L_F);
__m256i L_EF8_EF24 = _mm256_unpackhi_epi8(L_E, L_F);
__m256i L_GH0_GH16 = _mm256_unpacklo_epi8(L_G, L_H);
__m256i L_GH8_GH24 = _mm256_unpackhi_epi8(L_G, L_H);
__m256i L_EH0_EH16 = _mm256_unpacklo_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH4_EH20 = _mm256_unpackhi_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH0 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x20);
_mm256_store_si256(blockA_256++, L_EH0);
__m256i L_EH16 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x31);
__m256i L_EH8_EH24 = _mm256_unpacklo_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH12_EH28 = _mm256_unpackhi_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH8 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x20);
_mm256_store_si256(blockA_256++, L_EH8);
_mm256_store_si256(blockA_256++, L_EH16);
__m256i L_EH24 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x31);
_mm256_store_si256(blockA_256++, L_EH24);
}
}
}
template <typename Index, typename DataMapper, int nr, bool Conjugate, bool PanelMode>
EIGEN_DONT_INLINE void gemm_pack_rhs_any<QUInt8, Index, DataMapper, nr, ColMajor, Conjugate, PanelMode>::
operator()(QUInt8* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Get vector pointer
__m256i* blockB_256 = reinterpret_cast<__m256i*>(blockB);
// Get even multiples of the dimensions
Index cols_32 = (cols / 32) * 32;
Index depth_32 = (depth / 32) * 32;
// Perform a step of the packing for 4 columns
__m256i R_AB_L, R_AB_H, R_CD_L, R_CD_H, R_AD_0, R_AD_8, R_AD_16, R_AD_24;
#define PACK_STEP \
R_AB_L = _mm256_unpacklo_epi64(R_A, R_B); \
R_CD_L = _mm256_unpacklo_epi64(R_C, R_D); \
R_AB_H = _mm256_unpackhi_epi64(R_A, R_B); \
R_CD_H = _mm256_unpackhi_epi64(R_C, R_D); \
R_AD_0 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x20); \
R_AD_16 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x31); \
R_AD_8 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x20); \
R_AD_24 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x31); \
_mm256_store_si256(blockB_256, R_AD_0); \
_mm256_store_si256(blockB_256 + 8, R_AD_8); \
_mm256_store_si256(blockB_256 + 16, R_AD_16); \
_mm256_store_si256(blockB_256 + 24, R_AD_24); \
blockB_256++;
// Pack cols in sets of 32
for (Index n = 0; n < cols_32; n += 32) {
// Pack depth in sets of 32
for (Index k = 0; k < depth_32; k += 32) {
__m256i R_A = rhs.loadPacket(k, n);
__m256i R_B = rhs.loadPacket(k, n + 1);
__m256i R_C = rhs.loadPacket(k, n + 2);
__m256i R_D = rhs.loadPacket(k, n + 3);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 4);
R_B = rhs.loadPacket(k, n + 5);
R_C = rhs.loadPacket(k, n + 6);
R_D = rhs.loadPacket(k, n + 7);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 8);
R_B = rhs.loadPacket(k, n + 9);
R_C = rhs.loadPacket(k, n + 10);
R_D = rhs.loadPacket(k, n + 11);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 12);
R_B = rhs.loadPacket(k, n + 13);
R_C = rhs.loadPacket(k, n + 14);
R_D = rhs.loadPacket(k, n + 15);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 16);
R_B = rhs.loadPacket(k, n + 17);
R_C = rhs.loadPacket(k, n + 18);
R_D = rhs.loadPacket(k, n + 19);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 20);
R_B = rhs.loadPacket(k, n + 21);
R_C = rhs.loadPacket(k, n + 22);
R_D = rhs.loadPacket(k, n + 23);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 24);
R_B = rhs.loadPacket(k, n + 25);
R_C = rhs.loadPacket(k, n + 26);
R_D = rhs.loadPacket(k, n + 27);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 28);
R_B = rhs.loadPacket(k, n + 29);
R_C = rhs.loadPacket(k, n + 30);
R_D = rhs.loadPacket(k, n + 31);
PACK_STEP;
blockB_256 += 24;
}
if (depth_32 < depth) {
QUInt8* ptr;
__m256i R_A = _mm256_setzero_si256();
__m256i R_B = _mm256_setzero_si256();
__m256i R_C = _mm256_setzero_si256();
__m256i R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 1);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 2);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 3);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 4);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 5);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 6);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 7);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 8);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 9);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 10);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 11);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 12);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 13);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 14);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 15);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 16);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 17);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 18);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 19);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 20);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 21);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 22);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 23);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 24);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 25);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 26);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 27);
}
PACK_STEP;
R_A = _mm256_setzero_si256();
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n + 28);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 29);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 30);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 31);
}
PACK_STEP;
blockB_256 += 24;
}
}
// Finish packing cols
if (cols_32 < cols) {
// Pack depth in sets of 32
for (Index k = 0; k < depth_32; k += 32) {
__m256i R_A, R_B, R_C, R_D;
Index n;
for (n = cols_32; n < cols; n += 4) {
switch (cols - n) {
case 1:
R_A = rhs.loadPacket(k, n);
R_B = _mm256_setzero_si256();
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
PACK_STEP;
break;
case 2:
R_A = rhs.loadPacket(k, n);
R_B = rhs.loadPacket(k, n + 1);
R_C = _mm256_setzero_si256();
R_D = _mm256_setzero_si256();
PACK_STEP;
break;
case 3:
R_A = rhs.loadPacket(k, n);
R_B = rhs.loadPacket(k, n + 1);
R_C = rhs.loadPacket(k, n + 2);
R_D = _mm256_setzero_si256();
PACK_STEP;
break;
default:
R_A = rhs.loadPacket(k, n);
R_B = rhs.loadPacket(k, n + 1);
R_C = rhs.loadPacket(k, n + 2);
R_D = rhs.loadPacket(k, n + 3);
PACK_STEP;
break;
}
}
// Increment the block pointer.
// We must pad if cols is not a multiple of 32.
blockB_256 += 32 - (n - cols_32) / 4;
}
if (depth_32 < depth) {
for (Index n = cols_32; n < cols; n += 4) {
QUInt8* ptr;
__m256i R_A = _mm256_setzero_si256();
__m256i R_B = _mm256_setzero_si256();
__m256i R_C = _mm256_setzero_si256();
__m256i R_D = _mm256_setzero_si256();
switch (cols - n) {
case 1:
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n);
}
PACK_STEP;
break;
case 2:
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 1);
}
PACK_STEP;
break;
case 3:
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 1);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 2);
}
PACK_STEP;
break;
default:
for (Index k = depth_32; k < depth; k++) {
ptr = (QUInt8*) &R_A;
ptr[k - depth_32] = rhs(k, n);
ptr = (QUInt8*) &R_B;
ptr[k - depth_32] = rhs(k, n + 1);
ptr = (QUInt8*) &R_C;
ptr[k - depth_32] = rhs(k, n + 2);
ptr = (QUInt8*) &R_D;
ptr[k - depth_32] = rhs(k, n + 3);
}
PACK_STEP;
break;
}
}
}
}
#undef PACK_STEP
}
template<typename Index, typename DataMapper, int mr, int nr, bool ConjugateLhs, bool ConjugateRhs>
EIGEN_DONT_INLINE
void gebp_kernel_any<QInt8, QUInt8, Index, DataMapper, mr, nr, ConjugateLhs, ConjugateRhs>
::operator()(const DataMapper& res, const QInt8* blockA, const QUInt8* blockB,
Index rows, Index depth, Index cols, QInt32 alpha,
Index strideA, Index strideB, Index offsetA, Index offsetB)
{
EIGEN_STATIC_ASSERT(!ConjugateLhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
EIGEN_STATIC_ASSERT(!ConjugateRhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
eigen_assert(alpha.value == 1);
eigen_assert(strideA == -1);
eigen_assert(strideB == -1);
eigen_assert(offsetA == 0);
eigen_assert(offsetB == 0);
eigen_assert(rows > 0);
eigen_assert(cols > 0);
eigen_assert(depth > 0);
eigen_assert(blockA);
eigen_assert(blockB);
Index rows_32 = ((rows + 31) / 32) * 32;
Index cols_32 = ((cols + 31) / 32) * 32;
Index depth_32 = ((depth + 31) / 32) * 32;
// Create result block
ei_declare_aligned_stack_constructed_variable(QInt32, blockO, 32 * 32, 0);
memset(blockO, 0, 32 * 32 * sizeof(QInt32));
// Get vectorized pointers
__m256i* blockO_256 = reinterpret_cast<__m256i*>(blockO);
const __m256i* blockA_256 = reinterpret_cast<const __m256i*>(blockA);
const __m256i* blockB_256 = reinterpret_cast<const __m256i*>(blockB);
// Loop over blocks of 32 columns
for (Index n = 0; n < cols_32; n += 32) {
// Reset index into blockA
Index indexL = 0;
// Loop over blocks of 32 rows
for (Index m = 0; m < rows_32; m += 32) {
// Reset index into blockB
Index indexR = n / 32 * depth_32;
// Loop over blocks of 8 on depth
for (Index k = 0; k < depth_32; k += 8) {
// Load inputs
__m256i L_AD0 = blockA_256[indexL++];
__m256i L_AD8 = blockA_256[indexL++];
__m256i L_AD16 = blockA_256[indexL++];
__m256i L_AD24 = blockA_256[indexL++];
__m256i L_EH0 = blockA_256[indexL++];
__m256i L_EH8 = blockA_256[indexL++];
__m256i L_EH16 = blockA_256[indexL++];
__m256i L_EH24 = blockA_256[indexL++];
__m256i R_AH0 = blockB_256[indexR++];
__m256i R_AH4 = blockB_256[indexR++];
__m256i R_AH8 = blockB_256[indexR++];
__m256i R_AH12 = blockB_256[indexR++];
__m256i R_AH16 = blockB_256[indexR++];
__m256i R_AH20 = blockB_256[indexR++];
__m256i R_AH24 = blockB_256[indexR++];
__m256i R_AH28 = blockB_256[indexR++];
// This constant is used with madd to convert 16 bit to 32 bit
const __m256i ONE = _mm256_set1_epi32(0x00010001);
// Declare variables used in COMPUTE_STEP
__m256i P_16_A, P_16_B, P_32_A, P_32_B, P_32;
#define COMPUTE_STEP(R_INPUT_A, R_INPUT_B, OFFSET) \
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD0); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH0); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD8); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH8); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 1, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 1), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD16); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH16); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 2, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 2), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD24); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH24); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 3, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 3), P_32));
// Permute and shuffle to copy a single value across the entire vector
// Then compute the multiplication
__m256i R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x00);
__m256i R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
__m256i R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 0);
__m256i R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 1);
R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x11);
__m256i R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
__m256i R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 2);
__m256i R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 3);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 4);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 5);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 6);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 7);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 8);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 9);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 10);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 11);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 12);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 13);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 14);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 15);
R_AH0_ = _mm256_permute2x128_si256(R_AH16, R_AH16, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 16);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 17);
R_AH0_ = _mm256_permute2x128_si256(R_AH16, R_AH16, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 18);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 19);
R_AH0_ = _mm256_permute2x128_si256(R_AH20, R_AH20, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 20);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 21);
R_AH0_ = _mm256_permute2x128_si256(R_AH20, R_AH20, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 22);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 23);
R_AH0_ = _mm256_permute2x128_si256(R_AH24, R_AH24, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 24);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 25);
R_AH0_ = _mm256_permute2x128_si256(R_AH24, R_AH24, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 26);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 27);
R_AH0_ = _mm256_permute2x128_si256(R_AH28, R_AH28, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 28);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 29);
R_AH0_ = _mm256_permute2x128_si256(R_AH28, R_AH28, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 30);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 31);
#undef COMPUTE_STEP
}
// Transfer the results to the result matrix.
if (m + 32 <= rows && n + 32 <= cols) {
Index i = 0;
for (Index j = n; j < n + 32; j++) {
LinearMapper r0 = res.getLinearMapper(m, j);
LinearMapper r1 = res.getLinearMapper(m + 8, j);
LinearMapper r2 = res.getLinearMapper(m + 16, j);
LinearMapper r3 = res.getLinearMapper(m + 24, j);
r0.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r0.loadPacket(0)));
r1.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r1.loadPacket(0)));
r2.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r2.loadPacket(0)));
r3.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r3.loadPacket(0)));
}
}
else {
for (Index j = n; j < cols; j++) {
for (Index i = m; i < rows; i++) {
res(i, j) = blockO[(j - n) * 32 + (i - m)];
}
}
}
// Zero the result block so it can be reused
memset(blockO, 0, 32 * 32 * sizeof(QInt32));
}
}
}
// Below are the fully optimized versions that are correct only for sizes that
// are multiple of 32. It is about a 10% performance benefit to keep these
// implementations separate.
// Arrange a block of the left input matrix in contiguous memory.
//
// Given column major input (A0 beside A1 in memory):
// A0 B0 C0 D0 E0 F0 G0 H0 ...
// A1 B1 C1 D1 E1 F1 G1 H1 ...
// A2 B2 C2 D2 E2 F2 G2 H2 ...
// A3 B3 C3 D3 E3 F3 G3 H3 ...
// A4 B4 C4 D4 E4 F4 G4 H4 ...
// A5 B5 C5 D5 E5 F5 G5 H5 ...
// A6 B6 C6 D6 E6 F6 G6 H6 ...
// A7 B7 C7 D7 E7 F7 G7 H7 ...
// A8 ...
// ...
//
// Packing yields output (A0 beside B0 in memory):
// A0 B0 C0 D0
// A1 B1 C1 D1
// A2 B2 C2 D2
// A3 B3 C3 D3
// A4 B4 C4 D4
// A5 B5 C5 D5
// A6 B6 C6 D6
// A7 B7 C7 D7
// ...
// A31 B31 C31 D31
// E0 F0 G0 H0
// E1 F1 G1 H1
// E2 F2 G2 H2
// E3 F3 G3 H3
// E4 F4 G4 H4
// E5 F5 G5 H5
// E6 F6 G6 H6
// E7 F7 G7 H7
// ...
//
// Four elements of the same row are arranged contiguously because maddubs and
// madd both perform an adjacent addition in the kernel.
template <typename Index, typename DataMapper, int Pack1, int Pack2,
bool Conjugate, bool PanelMode>
struct gemm_pack_lhs<QInt8, Index, DataMapper, Pack1, Pack2, ColMajor,
Conjugate, PanelMode> {
EIGEN_DONT_INLINE void operator()(QInt8* blockA, const DataMapper& lhs,
Index depth, Index rows, Index stride = 0,
Index offset = 0);
};
template <typename Index, typename DataMapper, int Pack1, int Pack2,
bool Conjugate, bool PanelMode>
EIGEN_DONT_INLINE void gemm_pack_lhs<QInt8, Index, DataMapper, Pack1, Pack2,
ColMajor, Conjugate, PanelMode>::
operator()(QInt8* blockA, const DataMapper& lhs, Index depth, Index rows,
Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Use alternate function for weird sizes
if (rows % 32 != 0 || depth % 32 != 0) {
gemm_pack_lhs_any<QInt8, Index, DataMapper, Pack1, Pack2, ColMajor, Conjugate, PanelMode> lhs_pack;
return lhs_pack(blockA, lhs, depth, rows, stride, offset);
}
// Get vector pointer
__m256i* blockA_256 = reinterpret_cast<__m256i*>(blockA);
// Pack rows in sets of 32
for (Index m = 0; m < rows; m += 32) {
// Pack depth in sets of 8
for (Index k = 0; k < depth; k += 8) {
// Load vectors
__m256i L_A = lhs.loadPacket(m, k);
__m256i L_B = lhs.loadPacket(m, k + 1);
// Interleave 8-bit elements
__m256i L_AB0_AB16 = _mm256_unpacklo_epi8(L_A, L_B);
__m256i L_AB8_AB24 = _mm256_unpackhi_epi8(L_A, L_B);
__m256i L_C = lhs.loadPacket(m, k + 2);
__m256i L_D = lhs.loadPacket(m, k + 3);
__m256i L_CD0_CD16 = _mm256_unpacklo_epi8(L_C, L_D);
__m256i L_CD8_CD24 = _mm256_unpackhi_epi8(L_C, L_D);
// Interleave 16-bit elements
__m256i L_AD0_AD16 = _mm256_unpacklo_epi16(L_AB0_AB16, L_CD0_CD16);
__m256i L_AD4_AD20 = _mm256_unpackhi_epi16(L_AB0_AB16, L_CD0_CD16);
// Use permute before we store to cross 128-bit lanes
__m256i L_AD0 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x20);
_mm256_store_si256(blockA_256++, L_AD0);
// Complete packing for 32 x 8 block
__m256i L_AD16 = _mm256_permute2x128_si256(L_AD0_AD16, L_AD4_AD20, 0x31);
__m256i L_AD8_AD24 = _mm256_unpacklo_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD12_AD28 = _mm256_unpackhi_epi16(L_AB8_AB24, L_CD8_CD24);
__m256i L_AD8 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x20);
_mm256_store_si256(blockA_256++, L_AD8);
_mm256_store_si256(blockA_256++, L_AD16);
__m256i L_AD24 = _mm256_permute2x128_si256(L_AD8_AD24, L_AD12_AD28, 0x31);
_mm256_store_si256(blockA_256++, L_AD24);
__m256i L_E = lhs.loadPacket(m, k + 4);
__m256i L_F = lhs.loadPacket(m, k + 5);
__m256i L_EF0_EF16 = _mm256_unpacklo_epi8(L_E, L_F);
__m256i L_EF8_EF24 = _mm256_unpackhi_epi8(L_E, L_F);
__m256i L_G = lhs.loadPacket(m, k + 6);
__m256i L_H = lhs.loadPacket(m, k + 7);
__m256i L_GH0_GH16 = _mm256_unpacklo_epi8(L_G, L_H);
__m256i L_GH8_GH24 = _mm256_unpackhi_epi8(L_G, L_H);
__m256i L_EH0_EH16 = _mm256_unpacklo_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH4_EH20 = _mm256_unpackhi_epi16(L_EF0_EF16, L_GH0_GH16);
__m256i L_EH0 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x20);
_mm256_store_si256(blockA_256++, L_EH0);
__m256i L_EH16 = _mm256_permute2x128_si256(L_EH0_EH16, L_EH4_EH20, 0x31);
__m256i L_EH8_EH24 = _mm256_unpacklo_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH12_EH28 = _mm256_unpackhi_epi16(L_EF8_EF24, L_GH8_GH24);
__m256i L_EH8 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x20);
_mm256_store_si256(blockA_256++, L_EH8);
_mm256_store_si256(blockA_256++, L_EH16);
__m256i L_EH24 = _mm256_permute2x128_si256(L_EH8_EH24, L_EH12_EH28, 0x31);
_mm256_store_si256(blockA_256++, L_EH24);
}
}
}
// Arrange a block of the right input matrix in contiguous memory.
//
// Given column major input (A0 beside A1 in memory):
// A0 B0 C0 D0 E0 F0 G0 H0 ...
// A1 B1 C1 D1 E1 F1 G1 H1 ...
// A2 B2 C2 D2 E2 F2 G2 H2 ...
// A3 B3 C3 D3 E3 F3 G3 H3 ...
// A4 B4 C4 D4 E4 F4 G4 H4 ...
// A5 B5 C5 D5 E5 F5 G5 H5 ...
// A6 B6 C6 D6 E6 F6 G6 H6 ...
// A7 B7 C7 D7 E7 F7 G7 H7 ...
// A8 ...
// ...
//
// Packing yields row major output (A0 beside A1 in memory):
// A0 A1 A2 A3 A4 A5 A6 A7
// B0 B1 B2 B3 B4 B5 B6 B7
// ...
//
// At least four elements of the same col are arranged contiguously because
// maddubs and madd both perform an adjacent addition in the kernel. We can
// save work by leaving 8 adjacent elements because kr = 8.
template <typename Index, typename DataMapper, int nr, bool Conjugate,
bool PanelMode>
struct gemm_pack_rhs<QUInt8, Index, DataMapper, nr, ColMajor, Conjugate,
PanelMode> {
EIGEN_DONT_INLINE void operator()(QUInt8* blockB, const DataMapper& rhs,
Index depth, Index cols, Index stride = 0,
Index offset = 0);
};
template <typename Index, typename DataMapper, int nr, bool Conjugate,
bool PanelMode>
EIGEN_DONT_INLINE void gemm_pack_rhs<QUInt8, Index, DataMapper, nr, ColMajor,
Conjugate, PanelMode>::
operator()(QUInt8* blockB, const DataMapper& rhs, Index depth, Index cols,
Index stride, Index offset) {
eigen_assert(stride == 0);
eigen_assert(offset == 0);
// Use alternate function for weird sizes
if (cols % 32 != 0 || depth % 32 != 0) {
gemm_pack_rhs_any<QUInt8, Index, DataMapper, nr, ColMajor, Conjugate, PanelMode> rhs_pack;
return rhs_pack(blockB, rhs, depth, cols, stride, offset);
}
// Get vector pointer
__m256i* blockB_256 = reinterpret_cast<__m256i*>(blockB);
// Perform a step of the packing for 4 columns
__m256i R_AB_L, R_AB_H, R_CD_L, R_CD_H, R_AD_0, R_AD_8, R_AD_16, R_AD_24;
#define PACK_STEP \
R_AB_L = _mm256_unpacklo_epi64(R_A, R_B); \
R_CD_L = _mm256_unpacklo_epi64(R_C, R_D); \
R_AB_H = _mm256_unpackhi_epi64(R_A, R_B); \
R_CD_H = _mm256_unpackhi_epi64(R_C, R_D); \
R_AD_0 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x20); \
R_AD_16 = _mm256_permute2x128_si256(R_AB_L, R_CD_L, 0x31); \
R_AD_8 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x20); \
R_AD_24 = _mm256_permute2x128_si256(R_AB_H, R_CD_H, 0x31); \
_mm256_store_si256(blockB_256, R_AD_0); \
_mm256_store_si256(blockB_256 + 8, R_AD_8); \
_mm256_store_si256(blockB_256 + 16, R_AD_16); \
_mm256_store_si256(blockB_256 + 24, R_AD_24); \
blockB_256++;
// Pack cols in sets of 32
for (Index n = 0; n < cols; n += 32) {
// Pack depth in sets of 32
for (Index k = 0; k < depth; k += 32) {
__m256i R_A = rhs.loadPacket(k, n);
__m256i R_B = rhs.loadPacket(k, n + 1);
__m256i R_C = rhs.loadPacket(k, n + 2);
__m256i R_D = rhs.loadPacket(k, n + 3);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 4);
R_B = rhs.loadPacket(k, n + 5);
R_C = rhs.loadPacket(k, n + 6);
R_D = rhs.loadPacket(k, n + 7);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 8);
R_B = rhs.loadPacket(k, n + 9);
R_C = rhs.loadPacket(k, n + 10);
R_D = rhs.loadPacket(k, n + 11);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 12);
R_B = rhs.loadPacket(k, n + 13);
R_C = rhs.loadPacket(k, n + 14);
R_D = rhs.loadPacket(k, n + 15);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 16);
R_B = rhs.loadPacket(k, n + 17);
R_C = rhs.loadPacket(k, n + 18);
R_D = rhs.loadPacket(k, n + 19);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 20);
R_B = rhs.loadPacket(k, n + 21);
R_C = rhs.loadPacket(k, n + 22);
R_D = rhs.loadPacket(k, n + 23);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 24);
R_B = rhs.loadPacket(k, n + 25);
R_C = rhs.loadPacket(k, n + 26);
R_D = rhs.loadPacket(k, n + 27);
PACK_STEP;
R_A = rhs.loadPacket(k, n + 28);
R_B = rhs.loadPacket(k, n + 29);
R_C = rhs.loadPacket(k, n + 30);
R_D = rhs.loadPacket(k, n + 31);
PACK_STEP;
blockB_256 += 24;
}
}
#undef PACK_STEP
}
// Perform the actual multiplication on packed inputs
template<typename Index, typename DataMapper, int mr, int nr, bool ConjugateLhs, bool ConjugateRhs>
struct gebp_kernel<QInt8, QUInt8, Index, DataMapper, mr, nr, ConjugateLhs, ConjugateRhs>
{
typedef typename DataMapper::LinearMapper LinearMapper;
EIGEN_DONT_INLINE
void operator()(const DataMapper& res, const QInt8* blockA, const QUInt8* blockB,
Index rows, Index depth, Index cols, QInt32 alpha,
Index strideA=-1, Index strideB=-1, Index offsetA=0, Index offsetB=0);
};
template<typename Index, typename DataMapper, int mr, int nr, bool ConjugateLhs, bool ConjugateRhs>
EIGEN_DONT_INLINE
void gebp_kernel<QInt8, QUInt8, Index, DataMapper, mr, nr, ConjugateLhs, ConjugateRhs>
::operator()(const DataMapper& res, const QInt8* blockA, const QUInt8* blockB,
Index rows, Index depth, Index cols, QInt32 alpha,
Index strideA, Index strideB, Index offsetA, Index offsetB)
{
EIGEN_STATIC_ASSERT(!ConjugateLhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
EIGEN_STATIC_ASSERT(!ConjugateRhs, YOU_MADE_A_PROGRAMMING_MISTAKE);
eigen_assert(alpha.value == 1);
eigen_assert(strideA == -1);
eigen_assert(strideB == -1);
eigen_assert(offsetA == 0);
eigen_assert(offsetB == 0);
eigen_assert(rows > 0);
eigen_assert(cols > 0);
eigen_assert(depth > 0);
eigen_assert(blockA);
eigen_assert(blockB);
// Use alternate function for weird sizes
if (rows % 32 != 0 || cols % 32 != 0 || depth % 32 != 0) {
gebp_kernel_any<QInt8, QUInt8, Index, DataMapper, mr, nr, ConjugateLhs, ConjugateRhs> gebp;
return gebp(res, blockA, blockB, rows, depth, cols, alpha, strideA, strideB, offsetA, offsetB);
}
// Create result block
QInt32* blockO = aligned_new<QInt32>(32 * 32);
// Allocating the result block is about 5-10% faster than declaring stack
// space. It is unclear why this is the case.
// ei_declare_aligned_stack_constructed_variable(QInt32, blockO, 32 * 32, 0);
memset(blockO, 0, 32 * 32 * sizeof(QInt32));
// Get vectorized pointers
__m256i* blockO_256 = reinterpret_cast<__m256i*>(blockO);
const __m256i* blockA_256 = reinterpret_cast<const __m256i*>(blockA);
const __m256i* blockB_256 = reinterpret_cast<const __m256i*>(blockB);
// Loop over blocks of 32 columns
for (Index n = 0; n < cols; n += 32) {
// Reset index into blockA
Index indexL = 0;
// Loop over blocks of 32 rows
for (Index m = 0; m < rows; m += 32) {
// Reset index into blockB
Index indexR = n / 32 * depth;
// Loop over blocks of 8 on depth
for (Index k = 0; k < depth; k += 8) {
// Load inputs
__m256i L_AD0 = blockA_256[indexL++];
__m256i L_AD8 = blockA_256[indexL++];
__m256i L_AD16 = blockA_256[indexL++];
__m256i L_AD24 = blockA_256[indexL++];
__m256i L_EH0 = blockA_256[indexL++];
__m256i L_EH8 = blockA_256[indexL++];
__m256i L_EH16 = blockA_256[indexL++];
__m256i L_EH24 = blockA_256[indexL++];
__m256i R_AH0 = blockB_256[indexR++];
__m256i R_AH4 = blockB_256[indexR++];
__m256i R_AH8 = blockB_256[indexR++];
__m256i R_AH12 = blockB_256[indexR++];
__m256i R_AH16 = blockB_256[indexR++];
__m256i R_AH20 = blockB_256[indexR++];
__m256i R_AH24 = blockB_256[indexR++];
__m256i R_AH28 = blockB_256[indexR++];
// This constant is used with madd to convert 16 bit to 32 bit
const __m256i ONE = _mm256_set1_epi32(0x00010001);
// Declare variables used in COMPUTE_STEP
__m256i P_16_A, P_16_B, P_32_A, P_32_B, P_32;
#define COMPUTE_STEP(R_INPUT_A, R_INPUT_B, OFFSET) \
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD0); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH0); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD8); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH8); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 1, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 1), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD16); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH16); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 2, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 2), P_32)); \
\
P_16_A = _mm256_maddubs_epi16(R_INPUT_A, L_AD24); \
P_32_A = _mm256_madd_epi16(P_16_A, ONE); \
P_16_B = _mm256_maddubs_epi16(R_INPUT_B, L_EH24); \
P_32_B = _mm256_madd_epi16(P_16_B, ONE); \
P_32 = _mm256_add_epi32(P_32_A, P_32_B); \
_mm256_store_si256( \
blockO_256 + 4 * OFFSET + 3, \
_mm256_add_epi32(_mm256_load_si256(blockO_256 + 4 * OFFSET + 3), P_32));
// Permute and shuffle to copy a single value across the entire vector
// Then compute the multiplication
__m256i R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x00);
__m256i R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
__m256i R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 0);
__m256i R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 1);
R_AH0_ = _mm256_permute2x128_si256(R_AH0, R_AH0, 0x11);
__m256i R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
__m256i R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 2);
__m256i R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
__m256i R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 3);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 4);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 5);
R_AH0_ = _mm256_permute2x128_si256(R_AH4, R_AH4, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 6);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 7);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 8);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 9);
R_AH0_ = _mm256_permute2x128_si256(R_AH8, R_AH8, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 10);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 11);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 12);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 13);
R_AH0_ = _mm256_permute2x128_si256(R_AH12, R_AH12, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 14);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 15);
R_AH0_ = _mm256_permute2x128_si256(R_AH16, R_AH16, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 16);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 17);
R_AH0_ = _mm256_permute2x128_si256(R_AH16, R_AH16, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 18);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 19);
R_AH0_ = _mm256_permute2x128_si256(R_AH20, R_AH20, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 20);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 21);
R_AH0_ = _mm256_permute2x128_si256(R_AH20, R_AH20, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 22);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 23);
R_AH0_ = _mm256_permute2x128_si256(R_AH24, R_AH24, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 24);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 25);
R_AH0_ = _mm256_permute2x128_si256(R_AH24, R_AH24, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 26);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 27);
R_AH0_ = _mm256_permute2x128_si256(R_AH28, R_AH28, 0x00);
R_AD0 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH0 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD0, R_EH0, 28);
R_AD1 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH1 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD1, R_EH1, 29);
R_AH0_ = _mm256_permute2x128_si256(R_AH28, R_AH28, 0x11);
R_AD2 = _mm256_shuffle_epi32(R_AH0_, 0x00);
R_EH2 = _mm256_shuffle_epi32(R_AH0_, 0x55);
COMPUTE_STEP(R_AD2, R_EH2, 30);
R_AD3 = _mm256_shuffle_epi32(R_AH0_, 0xAA);
R_EH3 = _mm256_shuffle_epi32(R_AH0_, 0xFF);
COMPUTE_STEP(R_AD3, R_EH3, 31);
#undef COMPUTE_STEP
}
// Transfer the results to the result matrix
Index i = 0;
for (Index j = n; j < n + 32; j++) {
LinearMapper r0 = res.getLinearMapper(m, j);
LinearMapper r1 = res.getLinearMapper(m + 8, j);
LinearMapper r2 = res.getLinearMapper(m + 16, j);
LinearMapper r3 = res.getLinearMapper(m + 24, j);
r0.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r0.loadPacket(0)));
r1.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r1.loadPacket(0)));
r2.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r2.loadPacket(0)));
r3.storePacket(
0, _mm256_add_epi32(blockO_256[i++], r3.loadPacket(0)));
}
// Zero the result block so it can be reused
memset(blockO, 0, 32 * 32 * sizeof(QInt32));
}
}
aligned_delete(blockO, 32 * 32);
}
#endif // EIGEN_USE_OPTIMIZED_INT8_UINT8_MAT_MAT_PRODUCT
} // namespace internal
} // namespace Eigen
#endif // EIGEN_CXX11_FIXED_POINT_MAT_MAT_PRODUCT_AVX2_H