fb335f6a5f
CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
180 lines
7.3 KiB
Plaintext
180 lines
7.3 KiB
Plaintext
/***************************************************************************************************
|
|
* Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without modification, are permitted
|
|
* provided that the following conditions are met:
|
|
* * Redistributions of source code must retain the above copyright notice, this list of
|
|
* conditions and the following disclaimer.
|
|
* * Redistributions in binary form must reproduce the above copyright notice, this list of
|
|
* conditions and the following disclaimer in the documentation and/or other materials
|
|
* provided with the distribution.
|
|
* * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used
|
|
* to endorse or promote products derived from this software without specific prior written
|
|
* permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
|
|
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
|
|
* FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
|
|
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
|
* STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*
|
|
**************************************************************************************************/
|
|
/*! \file
|
|
|
|
\brief Unit tests for thread-level GEMM
|
|
*/
|
|
#include "cutlass/arch/wmma.h"
|
|
|
|
#if defined(CUTLASS_ARCH_WMMA_SM72_ENABLED)
|
|
|
|
#include "../../common/cutlass_unit_test.h"
|
|
|
|
#include "cutlass/aligned_buffer.h"
|
|
#include "cutlass/half.h"
|
|
|
|
#include "cutlass/gemm/warp/default_mma_wmma_tensor_op.h"
|
|
|
|
#include "cutlass/core_io.h"
|
|
#include "cutlass/util/host_tensor.h"
|
|
#include "cutlass/util/tensor_view_io.h"
|
|
|
|
#include "cutlass/util/reference/host/tensor_fill.h"
|
|
#include "cutlass/util/reference/host/tensor_compare.h"
|
|
#include "cutlass/util/reference/host/gemm.h"
|
|
|
|
#include "testbed.h"
|
|
|
|
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
|
|
///////////////////////////////////////////// Integer wmma.mma ////////////////////////////////////////////////
|
|
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
|
|
// TODO: FIXME SM75 should SM72, but the compilation breaks as SM72 shows up and runs on VOLTA
|
|
TEST(SM75_warp_wmma_row_col_s8, 16x16x16_16x16x16_16x16x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<16, 16, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<16, 16, 16>;
|
|
using ElementA = int8_t;
|
|
using ElementB = int8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<16, 16, 16> >().run();
|
|
}
|
|
|
|
TEST(SM75_warp_wmma_row_col_s8, 32x8x16_32x8x16_32x8x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<32, 8, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<32, 8, 16>;
|
|
using ElementA = int8_t;
|
|
using ElementB = int8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<32, 8, 16> >().run();
|
|
}
|
|
|
|
TEST(SM75_warp_wmma_row_col_s8, 8x32x16_8x32x16_8x32x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<8, 32, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<8, 32, 16>;
|
|
using ElementA = int8_t;
|
|
using ElementB = int8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<8, 32, 16> >().run();
|
|
}
|
|
|
|
TEST(SM75_warp_wmma_row_col_u8, 16x16x16_16x16x16_16x16x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<16, 16, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<16, 16, 16>;
|
|
using ElementA = uint8_t;
|
|
using ElementB = uint8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<16, 16, 16> >().run();
|
|
}
|
|
|
|
TEST(SM75_warp_wmma_row_col_u8, 32x8x16_32x8x16_32x8x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<32, 8, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<32, 8, 16>;
|
|
using ElementA = uint8_t;
|
|
using ElementB = uint8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<32, 8, 16> >().run();
|
|
}
|
|
|
|
TEST(SM75_warp_wmma_row_col_u8, 8x32x16_8x32x16_8x32x16) {
|
|
// Threadblock and warp with just one native WMMA operation (most basic unit test)
|
|
using WarpShape = cutlass::gemm::GemmShape<8, 32, 16>;
|
|
using InstructionShape = cutlass::gemm::GemmShape<8, 32, 16>;
|
|
using ElementA = uint8_t;
|
|
using ElementB = uint8_t;
|
|
using ElementC = int32_t;
|
|
using LayoutA = cutlass::layout::RowMajor;
|
|
using LayoutB = cutlass::layout::ColumnMajor;
|
|
using LayoutC = cutlass::layout::RowMajor;
|
|
|
|
using WmmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOpWmma<
|
|
WarpShape,
|
|
InstructionShape,
|
|
ElementA, LayoutA,
|
|
ElementB, LayoutB,
|
|
ElementC, LayoutC>::Type;
|
|
|
|
test::gemm::warp::Testbed<WmmaTensorOp, cutlass::gemm::GemmShape<8, 32, 16> >().run();
|
|
}
|
|
#endif //CUTLASS_ARCH_WMMA_SM72_ENABLED
|