Journal Journal: The world's first Decimal floating-point Arithmetic IPs
-PLDesignLine; By Clive Maxfield-
As you may or may not recall, way back in the mists of time that we used to call "2007", I penned an article called BCD coming to an FPGA near you. In this humble offering, I noted that we require four items in order to create decimal floating-point High-Performance Computing (HPC) solutions, and that the only missing item was a Decimal Arithmetic IP Library.
Well, I just heard from Ahmed Sami, the Marketing Director at SilMinds. Ahmed tells me that the folks at SilMinds have developed a complete library of IP Cores implementing decimal floating-point algebraic operations. Ahmed says that SilMinds is the first company to provide such cores (other than IBM which is not an IP provider). Decimal floating-point arithmetic is required for a variety of tasks, including financial databases and banking applications. Oracle, for example, is already supporting the decimal floating point operations present in the IBM Power 6 Processor.
The problem is that software implementations for floating-point arithmetic can be painfully slow. By comparison, Ahmed says that hardware implementations can provide speed increases of 100X to 1000X.
Furthermore, Ahmed notes that the SilMinds library is fully compliant with the IEEE754-2008 standard (it seems that the SilMinds CTO, Prof. Hossam Fahmy, is a member of the reviewing and editing committee for this standard).
SilMinds is currently working on tailoring this library for use in FPGA-based accelerators. Click Here for more information on the SilMinds Floating-Point Library.
Well, I just heard from Ahmed Sami, the Marketing Director at SilMinds. Ahmed tells me that the folks at SilMinds have developed a complete library of IP Cores implementing decimal floating-point algebraic operations. Ahmed says that SilMinds is the first company to provide such cores (other than IBM which is not an IP provider). Decimal floating-point arithmetic is required for a variety of tasks, including financial databases and banking applications. Oracle, for example, is already supporting the decimal floating point operations present in the IBM Power 6 Processor.
The problem is that software implementations for floating-point arithmetic can be painfully slow. By comparison, Ahmed says that hardware implementations can provide speed increases of 100X to 1000X.
Furthermore, Ahmed notes that the SilMinds library is fully compliant with the IEEE754-2008 standard (it seems that the SilMinds CTO, Prof. Hossam Fahmy, is a member of the reviewing and editing committee for this standard).
SilMinds is currently working on tailoring this library for use in FPGA-based accelerators. Click Here for more information on the SilMinds Floating-Point Library.
