A High-Performance Significand BCD Adder with IEEE 754-2008 Decimal Rounding

We present a new method and architecture to merge efficiently IEEE 754­-2008 decimal rounding with significand BCD addition and subtraction. This is a key component to improve several decimal floating-point operations such as addition, multiplication and fused multiply-add. The decimal rounding unit is based on a direct implementation of the IEEE 754­2008 round­ing modes. We show that the resultant implementations for IEEE 754-2008 Decimal64 (16 precision digits) and Decimal128 (34 precision digits) formats reduce significantly the area and latency required for significand BCD addition/subtraction and decimal rounding in previous high-performance decimal floating-point adders.

[1]  G. Sutter,et al.  Decimal addition in FPGA , 2009, 2009 5th Southern Conference on Programmable Logic (SPL).

[2]  Eric M. Schwarz,et al.  Decimal floating-point support on the IBM System z10 processor , 2009, IBM J. Res. Dev..

[3]  Peter-Michael Seidel,et al.  Delay-optimized implementation of IEEE floating-point addition , 2004, IEEE Transactions on Computers.

[4]  R. K. Richards,et al.  Arithmetic operations in digital computers , 2013 .

[5]  Neil Burgess.  Prenormalization rounding in IEEE floating-point operations using a flagged prefix adder , 2005, IEEE Transactions on Very Large Scale Integration (VLSI) Systems.

[6]  Eric M. Schwarz,et al.  IBM POWER6 accelerators: VMX and DFU , 2007, IBM J. Res. Dev..

[7]  Álvaro Vázquez Álvarez.  High-performance decimal floating point units , 2009 .

[8]  Michael F. Cowlishaw,et al.  Decimal floating-point: algorism for computers , 2003, Proceedings 2003 16th IEEE Symposium on Computer Arithmetic.

[9]  Michael J. Schulte,et al.  Decimal Floating-Point Adder and Multifunction Unit with Injection-Based Rounding , 2007, 18th IEEE Symposium on Computer Arithmetic (ARITH '07).

[10]  James Demmel,et al.  IEEE Standard for Floating-Point Arithmetic , 2008 .

[11]  Michael J. Schulte,et al.  A 64-bit decimal floating-point adder , 2004, IEEE Computer Society Annual Symposium on VLSI.