Micro base stations, with limited antennas and extensive deployment, require scaled-down hardware. Software-defined radio solutions (e.g., CPU, many-core systems, GPU) offer flexibility but incur high area and power costs, while traditional DSP lacks efficient acceleration for smaller configurations. The key challenge for micro base stations is achieving minimal area and power overhead while meeting 5G requirements. This article presents a hardware-software co-designed architecture, Sayram, which minimizes overhead for 5G physical layer processing. Sayram integrates an instruction fusion mechanism, along with the compiler for simplified programming, a Vector Indirect Addressing Memory (VIAM) to minimize memory access cycles, and an improved vector register design to accelerate small-scale matrix computation, thereby improving overall processor efficiency. Operating at 1 GHz, Sayram achieves 158 GOPS with a 1.18 mm \(^2\) area, supporting 2T2R and 4T4R Physical Uplink Shared Channel (PUSCH) processing in single-core and dual-core modes, respectively. Evaluations show that Sayram’s area efficiency is 3× and 9× higher than traditional DSP and CGRA architectures, respectively, with power efficiency improvements of 44× and 6×. Sayram’s energy and area efficiency surpass CPU solutions by orders of magnitude.

Neural network processors and accelerators are domain-specific architectures deployed to solve the high computational requirements of deep learning algorithms. This article proposes a new instruction set extension for tensor computing, TCX, using Reduced Instruction Set Computer (RISC) instructions enhanced with variable length tensor extensions. It features a multi-dimensional register file, dimension registers, and fully generic tensor instructions. It can be seamlessly integrated into existing RISC Instruction Set Architectures and provides software compatibility for scalable hardware implementations. We present a tensor accelerator implementation of the tensor extensions using an out-of-order RISC microarchitecture. The tensor accelerator is scalable in computation units from several hundred to tens of thousands. An optimized register renaming mechanism is described that allows for many physical tensor registers without requiring architectural support for large tensor register names. We describe new tensor load and store instructions that reduce bandwidth requirements using tensor dimension registers. Implementations may balance data bandwidth and computation utilization for different types of tensor computations such as element-wise, depthwise, and matrix-multiplication. We characterize the computation precision of tensor operations to balance area, generality, and accuracy loss for several well-known neural networks. The TCX processor runs at 1 GHz and sustains 8.2 Tera operations per second using a 4,096 multiply-accumulate compute unit. It consumes 12.8 mm 2 while dissipating 0.46W/TOPs in TSMC 28-nm technology.

Neural network processors and accelerators are domain-specific architectures deployed to solve the high computational requirements of deep learning algorithms. This paper proposes a new instruction set extension for tensor computing, TCX, with RISC-style instructions and variable length tensor extensions. It features a multidimensional register file, dimension registers, and fully generic tensor instructions. It can be seamlessly integrated into existing RISC ISAs and provides software compatibility for scalable hardware implementations. We present an implementation of the TCX tensor computing accelerator using an out-of-order microarchitecture implementation. The tensor accelerator is scalable in computation units from several hundred to tens of thousands. An optimized register renaming mechanism is described which allows for many physical tensor registers without requiring architectural support for large tensor register names. We describe new tensor load and store instructions that reduce bandwidth requirements based on tensor dimensions. Implementations may balance data bandwidth and computation utilization for different types of tensor computations such as element-wise, depth-wise, and matrix-multiplication. We characterize the computation precision of tensor operations to balance area, generality, and accuracy loss for several well-known neural networks. The TCX processor runs at 1 GHz and sustains 8.2 Tera operations per second using a 4096 multiplication-accumulation compute unit with up to 98.83% MAC utilization. It consumes 12.8 square millimeters while dissipating 0.46 Watts per TOP in TSMC 28nm technology.