Intel: Should I use Native Floating point DSP or Floating Point IP to implement Multiply Accumulate on Arria® 10?
Arria
DSP/Filter
Category: DSP
Tools: Quartus® Prime
Device: Arria® 10
Both Multiply Accumulate have corresponding modes, so depending on the IP settings, both may produce the same result.
Each has the following differences.
[Arria 10 Native Floating Point DSP IP]
- Specialized for implementation of 1 DSP block (resources are not ALM assuming 1 DSP block), only single precision floating point numbers are supported
- Fine-tuning of the Hardware Floating Point DSP block
- A calculation engine that partially parallelizes large-scale Multiply Accumulate calculations can be used as a material for users to design HDL.
[Floating Point IP (ALTERA_FP_FUNCTIONS)]
- Supports other families such as the V series, supports double precision floating point numbers and other precisions
- May result in no use of DSP blocks (specify non-use)
Experienced FAE
Free consultation is available.
From specific product specifications to parts selection, the Company FAE will answer your technical concerns free of charge. Please feel free to contact us.