Super fast, accurate, reconfigurable DSP
Super fast, accurate, reconfigurable DSP using small silicon area
InferX delivers high DSP performance from small silicon area and is reconfigurable (in microseconds for advanced nodes).
InferX has flexibility similar to a vector processor but with performance near hard-wired ASIC.
InferX is soft IP for the TPUs which are controlled by EFLX eFPGA which is hard IP.
Ask about benchmarks on 40/28/22/16/12/7/6/5/3/18A and for other DSP operators.
High Level DSP Programming
High Level DSP Programming
Program InferX using Matlab Simulink, like you do FPGA. Use the GUI to model your DSP data flow: select InferX DSP operators for the vector-oriented computations and EFLX eFPGA operators for the bit-manipulation operations. Do anything from a simple FFT or FIR filter to a OFDM SDR (below).
The output of Simulink is a high level netlist processed by the InferX Compiler with the EFLX Compiler in a heterogenous way and mapped on to a hardware instantiation. The InferX Compiler will determine the throughput and the latency, which is deterministic, for the target node.
InferX can be programmed in streaming mode (maximum throughput) or packet mode (minimum silicon).
InferX Operators and Precision
InferX Operators and Precision
InferX optimized operators include FFT, IFFT, FIR, Polyphase Channelizer, Matrix Multiplication, Matrix Inversion, Matrix/Vector multiplication and more.
Flexible data precision: inputs may be Real int16, Complex int16, Real int8 or Complex Int8. All accumulation is Int40 for high accuracy. Output precision may be Real int32, Complex int32, Real int16 or Complex int16.
Softlogic DSP Operators
Softlogic DSP Operators
We provide the DSP operators for InferX. Our team is very experienced with fast, dense Verilog coding and the InferX architecture. We check all of our operators in simulation to ensure they are cycle accurate and confirm accuracy by comparing to FP32. Below is a simulation segment of FFT:
We have all of the common operators across a wide range of sizes with Real and Complex versions.
