This work proposes a novel quad-issue VLIW architecture, called Caliburn, for directly executing legacy MIPS32 binary programs. To schedule and pack legacy MIPS32 binary codes on-the-fly, Caliburn has an integrated novel hardware instruction morphing mechanism that converts legacy MIPS32 binary instructions into a VLIW instruction bundles without the intervention of software compilers. The performance enhancement of Caliburn with a pipelined MIPS32 processor is evaluated. The Caliburn VLIW processor is implemented using Bluespec SystemVerilog HDL and synthesized using Synopsys Design Compiler. The experimental result reveals that the Caliburn processor achieves 3.08X speedup, and can be operated at a frequency of 425 MHz by the fabrication of TSMC 40nm technology library. Streszczenie. W artykule przedstawiono propozycje nowej struktury VLIW na potrzeby wykonywania programow w architekturze MIPS32. W rozwiązaniu zastosowano technike morfingu, w celu eliminacji programowych kompilatorow. Wykonano badania eksperymentalne na procesorze MIPS32, potwierdzające efektywnośc i szybkośc opracowanej architektury. (Caliburn - procesor VLIW MIPS32 ze sprzetowym mechanizmem morfingu).
[1]
Todd M. Austin,et al.
Cyclone: a broadcast-free dynamic instruction scheduler with selective replay
,
2003,
ISCA '03.
[2]
Harsh Sharangpani,et al.
Itanium Processor Microarchitecture
,
2000,
IEEE Micro.
[3]
Sumedh W. Sathaye,et al.
Dynamic rescheduling: a technique for object code compatibility in VLIW architectures
,
1995,
Proceedings of the 28th Annual International Symposium on Microarchitecture.
[4]
Alberto Ferreira de Souza,et al.
Dynamically Scheduling VLIW Instructions
,
2000,
J. Parallel Distributed Comput..
[5]
J. LaFountain.
Inc.
,
2013,
American Art.