skip to main content


Title: UNIT: Unifying Tensorized Instruction Compilation
Because of the increasing demand for intensive computation in deep neural networks, researchers have developed both hardware and software mechanisms to reduce the compute and memory burden. A widely adopted approach is to use mixed precision data types. However, it is hard to benefit from mixed precision without hardware specialization because of the overhead of data casting. Recently, hardware vendors offer tensorized instructions specialized for mixed-precision tensor operations, such as Intel VNNI, Nvidia Tensor Core, and ARM DOT. These instructions involve a new computing idiom, which reduces multiple low precision elements into one high precision element. The lack of compilation techniques for this emerging idiom makes it hard to utilize these instructions. In practice, one approach is to use vendor-provided libraries for computationally-intensive kernels, but this is inflexible and prevents further optimizations. Another approach is to manually write hardware intrinsics, which is error-prone and difficult for programmers. Some prior works tried to address this problem by creating compilers for each instruction. This requires excessive efforts when it comes to many tensorized instructions. In this work, we develop a compiler framework, UNIT, to unify the compilation for tensorized instructions. The key to this approach is a unified semantics abstraction which makes the integration of new instructions easy, and the reuse of the analysis and transformations possible. Tensorized instructions from different platforms can be compiled via UNIT with moderate effort for favorable performance. Given a tensorized instruction and a tensor operation, UNIT automatically detects the applicability of the instruction, transforms the loop organization of the operation, and rewrites the loop body to take advantage of the tensorized instruction. According to our evaluation, UNIT is able to target various mainstream hardware platforms. The generated end-to-end inference model achieves 1.3 x speedup over Intel oneDNN on an x86 CPU, 1.75x speedup over Nvidia cuDNN on an Nvidia GPU, and 1.13x speedup over a carefully tuned TVM solution for ARM DOT on an ARM CPU.  more » « less
Award ID(s):
1751400
NSF-PAR ID:
10225723
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
2021 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)
Page Range / eLocation ID:
77 to 89
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Various hardware accelerators have been developed for energy-efficient and real-time inference of neural networks on edge devices. However, most training is done on high-performance GPUs or servers, and the huge memory and computing costs prevent training neural networks on edge devices. This paper proposes a novel tensor-based training framework, which offers orders-of-magnitude memory reduction in the training process. We propose a novel rank-adaptive tensorized neural network model, and design a hardware-friendly low-precision algorithm to train this model. We present an FPGA accelerator to demonstrate the benefits of this training method on edge devices. Our preliminary FPGA implementation achieves 59× speedup and 123× energy reduction compared to embedded CPU, and 292× memory reduction over a standard full-size training. 
    more » « less
  2. The engineering samples of the NVIDIA Grace CPU Superchip and NVIDIA Grace Hopper Superchips were tested using different benchmarks and scientific applications. The benchmarks include HPCC and HPCG. The real application-based benchmark includes AI-Benchmark-Alpha (a TensorFlow benchmark), Gromacs, OpenFOAM, and ROMS. The performance was compared to multiple Intel, AMD, ARM CPUs and several x86 with NVIDIA GPU systems. A brief energy efficiency estimate was performed based on TDP values. We found that in HPCC benchmark tests, the per-core performance of Grace is similar to or faster than AMD Milan cores, and the high core count often allows NVIDIA Grace CPU Superchip to have per-node performance similar to Intel Sapphire Rapids with High Bandwidth Memory: slower in matrix multiplication (by 17%) and FFT (by 6%), faster in Linpack (by 9%)). In scientific applications, the NVIDIA Grace CPU Superchip performance is slower by 6% to 18% in Gromacs, faster by 7% in OpenFOAM, and right between HBM and DDR modes of Intel Sapphire Rapids in ROMS. The combined CPU-GPU performance in Gromacs is significantly faster (by 20% to 117% faster) than any tested x86-NVIDIA GPU system. Overall, the new NVIDIA Grace Hopper Superchip and NVIDIA Grace CPU Superchip Superchip are high-performance and most likely energy-efficient solutions for HPC centers. 
    more » « less
  3. Abstract

    Cloud microphysics is one of the most time‐consuming components in a climate model. In this study, we port the cloud microphysics parameterization in the Community Atmosphere Model (CAM), known as Parameterization of Unified Microphysics Across Scales (PUMAS), from CPU to GPU to seek a computational speedup. The directive‐based methods (OpenACC and OpenMP target offload) are determined as the best fit specifically for our development practices, which enable a single version of source code to run either on the CPU or GPU, and yield a better portability and maintainability. Their performance is first examined in a PUMAS stand‐alone kernel and the directive‐based methods can outperform a CPU node as long as there is enough computational burden on the GPU. A consistent behavior is observed when we run PUMAS on the GPU in a practical CAM simulation. A 3.6× speedup of the PUMAS execution time, including data movement between CPU and GPU, is achieved at a coarse horizontal resolution (8 NVIDIA V100 GPUs against 36 Intel Skylake CPU cores). This speedup further increases up to 5.4× at a high resolution (24 NVIDIA V100 GPUs against 108 Intel Skylake CPU cores), which highlights the fact that GPU favors larger problem size. This study demonstrates that using GPU in a CAM simulation can save noticeable computational costs even with a small portion of code being GPU‐enabled. Therefore, we are encouraged to port more parameterizations to GPU to take advantage of its computational benefit.

     
    more » « less
  4. Due to the recent announcement of the Frontier supercomputer, many scientific application developers are working to make their applications compatible with AMD (CPU-GPU) architectures, which means moving away from the traditional CPU and NVIDIA-GPU systems. Due to the current limitations of profiling tools for AMD GPUs, this shift leaves a void in how to measure application performance on AMD GPUs. In this article, we design an instruction roofline model for AMD GPUs using AMD’s ROCProfiler and a benchmarking tool, BabelStream (the HIP implementation), as a way to measure an application’s performance in instructions and memory transactions on new AMD hardware. Specifically, we create instruction roofline models for a case study scientific application, PIConGPU, an open source particle-in-cell simulations application used for plasma and laser-plasma physics on the NVIDIA V100, AMD Radeon Instinct MI60, and AMD Instinct MI100 GPUs. When looking at the performance of multiple kernels of interest in PIConGPU we find that although the AMD MI100 GPU achieves a similar, or better, execution time compared to the NVIDIA V100 GPU, profiling tool differences make comparing performance of these two architectures hard. When looking at execution time, GIPS, and instruction intensity, the AMD MI60 achieves the worst performance out of the three GPUs used in this work. 
    more » « less
  5. Arbitrary-precision integer multiplication is the core kernel of many applications including scientific computing, cryptographic algorithms, etc. Existing acceleration of arbitrary-precision integer multiplication includes CPUs, GPUs, FPGAs, and ASICs. To leverage the hardware intrinsics low-bit function units (32/64-bit), arbitrary-precision integer multiplication can be calculated using Karatsuba decomposition, and Schoolbook decomposition by decomposing the two large operands into several small operands, generating a set of low-bit multiplications that can be processed either in a spatial or sequential manner on the low-bit function units, e.g., CPU vector instructions, GPU CUDA cores, FPGA digital signal processing (DSP) blocks. Among these accelerators, reconfigurable computing, e.g., FPGA accelerators are promised to provide both good energy efficiency and flexibility. We implement the state-of-the-art (SOTA) FPGA accelerator and compare it with the SOTA libraries on CPUs and GPUs. Surprisingly, in terms of energy efficiency, we find that the FPGA has the lowest energy efficiency, i.e., 0.29x of the CPU and 0.17x of the GPU with the same generation fabrication. Therefore, key questions arise: Where do the energy efficiency gains of CPUs and GPUs come from? Can reconfigurable computing do better? If can, how to achieve that? We first identify that the biggest energy efficiency gains of the CPUs and GPUs come from the dedicated vector units, i.e., vector instruction units in CPUs and CUDA cores in GPUs. FPGA uses DSPs and lookup tables (LUTs) to compose the needed computation, which incurs overhead when compared to using vector units directly. New reconfigurable computing, e.g., “FPGA+vector units” is a novel and feasible solution to improve energy efficiency. In this paper, we propose to map arbitrary-precision integer multiplication onto such a “FPGA+vector units” platform, i.e., AMD/Xilinx Versal ACAP architecture, a heterogeneous reconfigurable computing platform that features 400 AI engine tensor cores (AIE) running at 1 GHz, FPGA programmable logic (PL), and a general-purpose CPU in the system fabricated with the TSMC 7nm technology. Designing on Versal ACAP incurs several challenges and we propose AIM: Arbitrary-precision Integer Multiplication on Versal ACAP to automate and optimize the design. AIM accelerator is composed of AIEs, PL, and CPU. AIM framework includes analytical models to guide design space exploration and AIM automatic code generation to facilitate the system design and on-board design verification. We deploy the AIM framework on three different applications, including large integer multiplication (LIM), RSA, and Mandelbrot, on the AMD/Xilinx Versal ACAP VCK190 evaluation board. Our experimental results show that compared to existing accelerators, AIM achieves up to 12.6x, and 2.1x energy efficiency gains over the Intel Xeon Ice Lake 6346 CPU, and NVidia A5000 GPU respectively, which brings reconfigurable computing the most energy-efficient platform among CPUs and GPUs. 
    more » « less