基于定制指令的椭圆曲线标量乘处理器

An Elliptic Curve Scalar Multiplication Processor Based on Custom Instruction

椭圆曲线标量乘法的计算量大，对处理器的运算能力要求高，是整个椭圆曲线密码体制实现的关键；标量乘运算依赖于下层的点运算和最底层的有限域乘法、加法、求逆的运算速度，因此设计高效的标量乘和有限域运算硬件结构是提升椭圆曲线标量乘性能的关键．采用指令方式实现Montg：omen模逆运算，相比于传统实现减少了近1／3的延迟；提出一种混合点运算的数据流水排布模式，相比于顺序执行点加、倍点运算减少了近一半的延迟，具有较高的并行度和运算部件利用率；设计并实现了一种基于定制指令的椭圆曲线标量乘处理器，它在Virtex6VSX475T FPGA上完成一次标量乘运算需要0．275ms，吞吐量为18180次／秒，相比于相关工作。可以提升约45．9％-97．6％的运算延迟．

The elliptic curve scalar multiplication is the key in ECC system, which has a large quantityof calculation and requires a considerable processing performance. The efficiency of scalar multiplication depends on the speed of the finite field arithmetic such as multiplication, addition and inversion. So the keyto improving ECC efficiency is the high-efficiency architecture. This paper firstly proposed ainstruction-basedMontgomery inversion implementation, which reducing the latency nearly down to 1/3 compared with the tra- ditional implementation; secondly described a hybrid data-pipeline implementation, which owning higher degree of parallelism and better arithmetic unit utilization while reducing the latency by 50% compared with the point operation in-order;finally proposed a fast 256-bit ECC scalar multiplication processor on prime field based on the custom instruction. The scalar multiplication processor can implement a scalar multiplication in 0.275 and achieve a throughput of 18180 per second on Virtex6VSX475T FPGA, which improving the latency performance nearly by up between 45.9% and 97.6% compared to current work.