Implement MatMul Function and Operator with Floating Point support for CPU

- Implements MatMul function and operator for floating point datatype FP16/FP32
- Includes support for transposing dynamic tensors prior to matrix multiplication.
- Adds tests for 2D/3D/4D+ tensors in MatMul with F32/F16 datatype (with all combinations of transposed/not-transposed tensors)
- Updates fixture to allow for testing fused activation in MatMul
- Adds tests for matmul with and without fused activation

Resolved: [COMPMID-5898]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: Iefa84b26dd723c9a51e6c3f91023152c6c31ace2
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9411
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

3 files changed, 374 insertions, 1 deletions

diff --git a/src/cpu/operators/CpuMatMul.cpp b/src/cpu/operators/CpuMatMul.cpp
new file mode 100644
index 0000000000..b5359e51af
--- /dev/null
+++ b/src/cpu/operators/CpuMatMul.cpp
@@ -0,0 +1,226 @@
+/*
+ * Copyright (c) 2023 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "src/cpu/operators/CpuMatMul.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/experimental/Types.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "arm_compute/runtime/NEON/NEScheduler.h"
+#include "arm_compute/runtime/NEON/functions/NEMatMul.h"
+#include "src/common/utils/Log.h"
+#include "src/core/CPP/Validate.h"
+#include "src/core/helpers/AutoConfiguration.h"
+#include "src/core/helpers/MemoryHelpers.h"
+#include "src/cpu/utils/CpuAuxTensorHandler.h"
+
+using namespace arm_compute::experimental;
+
+namespace arm_compute
+{
+namespace cpu
+{
+CpuMatMul::CpuMatMul()
+    : _transpose_kernel_lhs(), _transpose_kernel_rhs(), _asm_glue(), _lhs_transposed(), _rhs_transposed(), _original_lhs_shape(), _original_rhs_shape(), _original_dst_shape()
+{
+}
+
+Status CpuMatMul::validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &info, const CpuMatMulSettings &settings)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(lhs, rhs);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(lhs, 1, DataType::F32, DataType::F16);
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs->are_values_constant(), "LHS Tensor must be dynamic.");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(rhs->are_values_constant(), "RHS Tensor must be dynamic.");
+    ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(lhs);
+    ARM_COMPUTE_RETURN_ERROR_ON_CPU_BF16_UNSUPPORTED(lhs);
+
+    const auto adj_lhs = info.adj_lhs();
+    const auto adj_rhs = info.adj_rhs();
+
+    const ITensorInfo *lhs_to_use = lhs;
+    const ITensorInfo *rhs_to_use = rhs;
+    TensorInfo         lhs_transposed{};
+    TensorInfo         rhs_transposed{};
+
+    auto gemm_info            = AsmGemmInfo();
+    gemm_info.activation_info = info.fused_activation();
+    gemm_info.fast_mode       = settings.fast_math();
+
+    // Validate and then permute a/b
+    if(adj_lhs)
+    {
+        auto_init_if_empty(lhs_transposed, lhs->clone()->set_tensor_shape(misc::shape_calculator::compute_transposed_shape(*lhs)));
+        ARM_COMPUTE_RETURN_ON_ERROR(cpu::kernels::CpuTransposeKernel::validate(lhs_to_use, &lhs_transposed));
+        // Assign lhs_to_use pointer to use transposed TensorInfo
+        lhs_to_use = &lhs_transposed;
+    }
+    if(adj_rhs)
+    {
+        auto_init_if_empty(rhs_transposed, rhs->clone()->set_tensor_shape(misc::shape_calculator::compute_transposed_shape(*rhs)));
+        ARM_COMPUTE_RETURN_ON_ERROR(cpu::kernels::CpuTransposeKernel::validate(rhs_to_use, &rhs_transposed));
+        // Assign rhs_to_use pointer to use transposed TensorInfo
+        rhs_to_use = &rhs_transposed;
+    }
+
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs_to_use->dimension(0) != rhs_to_use->dimension(1),
+                                    "The product AB is defined only if the number of columns in A is equal to the number of rows in B (after transpose)");
+
+    if(lhs_to_use->num_dimensions() > 2)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs_to_use->num_dimensions() != rhs_to_use->num_dimensions(), "Broadcasting in Batch dimension is unsupported by this operator.");
+    }
+
+    // Iterate over dimensions to be collapsed in operator - check dimensions are equivelent between tensors
+    for(unsigned int i = 2; i < lhs_to_use->num_dimensions(); i++)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs_to_use->dimension(i) != rhs_to_use->dimension(i), "Broadcasting in Batch dimension is unsupported by this operator.");
+    }
+
+    cpu::CpuGemmAssemblyDispatch::validate(lhs_to_use, rhs_to_use, nullptr, dst, gemm_info);
+
+    return Status{};
+}
+
+void CpuMatMul::configure(ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &info, const CpuMatMulSettings &settings)
+{
+    ARM_COMPUTE_ERROR_ON_NULLPTR(lhs, rhs, dst);
+    ARM_COMPUTE_LOG_PARAMS(lhs, rhs, dst, info, settings);
+    ARM_COMPUTE_ERROR_THROW_ON(CpuMatMul::validate(lhs, rhs, dst, info, settings));
+
+    _adj_lhs   = info.adj_lhs();
+    _adj_rhs   = info.adj_rhs();
+    _fast_math = settings.fast_math();
+
+    // 1. Create and reshape tensors
+    // ------------------------------------------------------
+    // a. Clone TensorInfo to prevent changing original tensor values during setup
+    // b. Change shape of lhs/dst to [x, y, 1, collapsed(z)] to match assembly kernel configuration
+    // c. For rhs collapse all dimensions larger than 3 to z dimension
+    TensorInfo lhs_to_use = *lhs->clone();
+    TensorInfo dst_to_use = *dst->clone();
+    TensorInfo rhs_to_use = *rhs->clone();
+
+    // Save starting shape of tensors
+    _original_lhs_shape = lhs_to_use.tensor_shape();
+    _original_dst_shape = dst_to_use.tensor_shape();
+    _original_rhs_shape = rhs_to_use.tensor_shape();
+
+    // Reshape lhs for use with assembly kernels.
+    lhs_to_use.set_tensor_shape(TensorShape(_original_lhs_shape.x(), _original_lhs_shape.y(), 1, _original_lhs_shape.collapsed_from(2).z()));
+    dst_to_use.set_tensor_shape(TensorShape(_original_dst_shape.x(), _original_dst_shape.y(), 1, _original_dst_shape.collapsed_from(2).z()));
+    rhs_to_use.set_tensor_shape(_original_rhs_shape.collapsed_from(2));
+
+    // 2.  Configuration for transpose of lhs/rhs
+    // ------------------------------------------------------
+    // Initialise transposed TensorInfo class for aux tensors (intermediary tensors)
+    if(_adj_lhs)
+    {
+        // Setup transpose LHS
+        _transpose_kernel_lhs = std::make_unique<cpu::kernels::CpuTransposeKernel>();
+        _transpose_kernel_lhs->configure(&lhs_to_use, &_lhs_transposed);
+    }
+
+    if(_adj_rhs)
+    {
+        // Setup transpose RHS
+        _transpose_kernel_rhs = std::make_unique<cpu::kernels::CpuTransposeKernel>();
+        _transpose_kernel_rhs->configure(&rhs_to_use, &_rhs_transposed);
+    }
+
+    // 3. Configure assembly kernel using transposed tensors.
+    // -----------------------------------------------------
+    // Use transposed tensors if the corresponding transpose flags are set
+    // Fill AsmGemmInfo class object before configuration
+    _gemm_info.activation_info = info.fused_activation();
+    _gemm_info.fast_mode       = settings.fast_math();
+
+    lhs_to_use = (_adj_lhs) ? _lhs_transposed : lhs_to_use;
+    rhs_to_use = (_adj_rhs) ? _rhs_transposed : rhs_to_use;
+
+    // Configure Asm Kernel
+    _asm_glue = std::make_unique<cpu::CpuGemmAssemblyDispatch>();
+    _asm_glue->configure(&lhs_to_use, &rhs_to_use, nullptr, &dst_to_use, _gemm_info); // c is nullptr as bias not supported in MatMul
+
+    // Specify memory requirements for intermediate tensors
+    auto asm_mem_req = _asm_glue->workspace();
+    // Specify memory required by gemm kernel
+    int idx = 0;
+    for(const auto &aux : asm_mem_req)
+    {
+        _aux_mem[idx] = aux;
+        idx++;
+    }
+    // Memory requirements for transposed tensors
+    _aux_mem[TransposeLHS] = MemoryInfo(offset_int_vec(TransposeLHS), MemoryLifetime::Temporary, lhs->total_size());
+    _aux_mem[TransposeRHS] = MemoryInfo(offset_int_vec(TransposeRHS), MemoryLifetime::Temporary, rhs->total_size());
+}
+
+void CpuMatMul::run(ITensorPack &tensors)
+{
+    // Retrieve tensors from tensor pack
+    auto lhs = tensors.get_tensor(ACL_SRC_0);
+    auto rhs = tensors.get_const_tensor(ACL_SRC_1);
+    auto dst = tensors.get_tensor(ACL_DST);
+
+    // Reshape LHS and DST to ensure compatibility with GEMM asm kernel (Batch dimensions is 4th for lhs and dst within asm)
+    // Collapse RHS (necessary to support dimensions larger than 3 in gemm assembly)
+    lhs->info()->set_tensor_shape(TensorShape(_original_lhs_shape.x(), _original_lhs_shape.y(), 1, _original_lhs_shape.collapsed_from(2).z())); // Collapsed 3+ dimensions into z
+    dst->info()->set_tensor_shape(TensorShape(_original_dst_shape.x(), _original_dst_shape.y(), 1, _original_dst_shape.collapsed_from(2).z())); // Collapsed 3+ dimensions into z
+    rhs->info()->set_tensor_shape(_original_rhs_shape.collapsed_from(2));
+
+    // Initialise object to handle stored transposed tensors in auxillary memory
+    CpuAuxTensorHandler lhs_transposed(offset_int_vec(TransposeLHS), _lhs_transposed, tensors, true);
+    CpuAuxTensorHandler rhs_transposed(offset_int_vec(TransposeRHS), _rhs_transposed, tensors, true);
+
+    // Create tensor pack for asm kernel
+    ITensorPack asm_tensors(tensors);
+
+    // Run transpose lhs if necessary
+    if(_adj_lhs)
+    {
+        ITensorPack lhs_transpose_pack = { { TensorType::ACL_SRC, lhs }, { TensorType::ACL_DST, lhs_transposed.get() } };
+        NEScheduler::get().schedule_op(_transpose_kernel_lhs.get(), Window::DimY, _transpose_kernel_lhs->window(), lhs_transpose_pack);
+        asm_tensors.add_const_tensor(TensorType::ACL_SRC_0, lhs_transposed.get());
+    }
+    // Run transpose rhs if necessary
+    if(_adj_rhs)
+    {
+        ITensorPack rhs_transpose_pack = { { TensorType::ACL_SRC, rhs }, { TensorType::ACL_DST, rhs_transposed.get() } };
+        NEScheduler::get().schedule_op(_transpose_kernel_rhs.get(), Window::DimY, _transpose_kernel_rhs->window(), rhs_transpose_pack);
+        asm_tensors.add_const_tensor(TensorType::ACL_SRC_1, rhs_transposed.get());
+    }
+    // Run asm kernel
+    _asm_glue->run(asm_tensors);
+
+    // Undo reshape of tensors
+    dst->info()->set_tensor_shape(_original_dst_shape);
+    lhs->info()->set_tensor_shape(_original_lhs_shape);
+    rhs->info()->set_tensor_shape(_original_rhs_shape);
+}
+
+experimental::MemoryRequirements CpuMatMul::workspace() const
+{
+    return _aux_mem;
+}
+} // namespace cpu
+} // namespace arm_compute
diff --git a/src/cpu/operators/CpuMatMul.h b/src/cpu/operators/CpuMatMul.h
new file mode 100644
index 0000000000..ae6345141e
--- /dev/null
+++ b/src/cpu/operators/CpuMatMul.h
@@ -0,0 +1,115 @@
+/*
+ * Copyright (c) 2023 Arm Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef SRC_CPU_OPERATORS_CPUMATMUL
+#define SRC_CPU_OPERATORS_CPUMATMUL
+
+#include "arm_compute/core/TensorInfo.h"
+#include "src/core/common/Macros.h"
+#include "src/cpu/ICpuOperator.h"
+#include "src/cpu/kernels/CpuTransposeKernel.h"
+#include "src/cpu/operators/internal/CpuGemmAssemblyDispatch.h"
+
+namespace arm_compute
+{
+// Forward Declarations
+class MatMulInfo;
+class CpuMatMulSettings;
+
+namespace cpu
+{
+/** Function to execute MatMul Operation. This function calls the following functions/kernels:
+ *
+ * If adjoint/adj flag is enabled for either input lhs or rhs (or both) :
+ *  -# @ref cpu::kernels::CpuTransposeKernel
+ * Then :
+ *  -# @ref cpu::CpuGemmAssemblyDispatch
+ */
+class CpuMatMul : public ICpuOperator
+{
+public:
+    /* Constructor */
+    CpuMatMul();
+    /* Destructor */
+    ~CpuMatMul() = default;
+
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuMatMul);
+    /** Configure operator for a given list of arguments
+     *
+     * Note: Check documentation of @ref NEMatMul for a list of supported datatypes and layouts
+     *
+     *
+     * @param[in]  lhs      Source tensor info.
+     * @param[in]  rhs      Source tensor info.
+     * @param[out] dst      Destination tensor info. Data types supported: same as @p lhs / @p rhs.
+     * @param[in]  info     Contains MatMul operation information described in @ref MatMulInfo.
+     * @param[in]  settings The settings for matmul operation (i.e fast math)
+     */
+    void configure(ITensorInfo *lhs, ITensorInfo *rhs, ITensorInfo *dst, const MatMulInfo &info, const CpuMatMulSettings &settings);
+    /** Static function to check if given info will lead to a valid configuration
+     *
+     * Similar to CpuMatMul::configure()
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *lhs, const ITensorInfo *rhs, const ITensorInfo *dst, const MatMulInfo &info, const CpuMatMulSettings &settings);
+
+    // Inherited methods overridden:
+    void run(ITensorPack &tensors) override;
+    experimental::MemoryRequirements workspace() const override;
+
+private:
+    enum InternalTensorIdx
+    {
+        AsmGemmWorkspace = 0, // Pre-allocate workspace tensors for CpuGemmAssemblyDispatch
+        PretransposeRHS,      // Pre-allocate workspace tensors for CpuGemmAssemblyDispatch
+        TransposeLHS,
+        TransposeRHS,
+        Count
+    };
+
+    // Define unique pointers to kernels/operators used by matmul
+    std::unique_ptr<kernels::CpuTransposeKernel> _transpose_kernel_lhs{ nullptr };
+    std::unique_ptr<kernels::CpuTransposeKernel> _transpose_kernel_rhs{ nullptr };
+    std::unique_ptr<CpuGemmAssemblyDispatch>     _asm_glue{ nullptr };
+
+    // TensorInfo for tensors stored in auxillary memory
+    TensorInfo _lhs_transposed{};
+    TensorInfo _rhs_transposed{};
+
+    // Original tensor shapes prior to reshaping tensors and collapsing dimensions
+    TensorShape _original_lhs_shape{};
+    TensorShape _original_rhs_shape{};
+    TensorShape _original_dst_shape{};
+
+    // Note : adj_lhs means the same as transposing lhs
+    bool                             _adj_lhs{ false };
+    bool                             _adj_rhs{ false };
+    bool                             _fast_math{ false };
+    AsmGemmInfo                      _gemm_info{};
+    experimental::MemoryRequirements _aux_mem{ Count };
+};
+}
+}
+
+#endif /* SRC_CPU_OPERATORS_CPUMATMUL */
diff --git a/src/cpu/operators/internal/CpuGemmAssemblyDispatch.h b/src/cpu/operators/internal/CpuGemmAssemblyDispatch.h
index 0c51c92359..588c45294a 100644
--- a/src/cpu/operators/internal/CpuGemmAssemblyDispatch.h
+++ b/src/cpu/operators/internal/CpuGemmAssemblyDispatch.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2022 Arm Limited.
+ * Copyright (c) 2018-2023 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -82,6 +82,38 @@ public:
 public:
     /** If supported create a Compute Library function else fallback to the arm_gemm function.
      *
+     * @note Configuring "batches"
+     * The shapes of @p a @p b and @p d are arranged as follows:
+     *     Lowest dimension <-> Highest dimension
+     * a: [K, M, Batch, Multi]
+     * b: [N, K, Multi]
+     * d: [N, M, Batch, Multi]
+     *
+     * The "Batch" refers to where "Batch" number of MxK slices of tensor a multiplies with a single KxN slice of b
+     * The "Multi" refers to where "Multi" number of individual multiplication of a with b
+     *
+     * E.g. the following are some example input shape configurations
+     *
+     * (1) Normal 2D gemm
+     * a: [K=3, M=4]
+     * b: [N=5, K=3]
+     * d: [N=5, M=4]
+     *
+     * (2) Batches of a sharing b (e.g. gemm-based batched convolution where b is the shared )
+     * a: [K=3, M=4, Batch=9]
+     * b: [N=5, K=3]
+     * d: [N=5, M=4, Batch=9]
+     *
+     * (3) "Batches" of independent gemm (e.g. batched matmul)
+     * a: [K=3, M=4, Batch=1, Multi=7]
+     * b: [N=5, K=3, Multi=7]
+     * d: [N=5, M=4, Batch=1, Multi=7]
+     *
+     * (4) "Batches" of independent gemm where b is also shared
+     * a: [K=3, M=4, Batch=4, Multi=7]
+     * b: [N=5, K=3, Multi=7]
+     * d: [N=5, M=4, Batch=4, Multi=7]
+     *
      * @param[in]  a    Input tensor (Matrix A)
      * @param[in]  b    Input tensor (Matrix B)
      * @param[in]  c    Input tensor (Matrix C) used to pass the bias for quantized calculations