# US 2019/0317726 A1 · Normalization of a product on a datapath

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0317726 A1
- Document ID: US20190317726A1
- Status: Published application
- Kind: A1
- Application number: 16/453,695
- Filed: 2019-06-26
- Published: 2019-10-17
- Earliest claimed: 2017-07-17
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Klaus M. Kroener; Cedric Lichtenau; Silvia M. Mueller; Andreas Wagner
- Web page: https://patentgrep.com/patent/US20190317726A1

## CPC

- G06F 5/01: for shifting, e.g. justifying, scaling, normalising {(digital stores in which the information is moved stepwise, e.g. shift-registers G11C19/00; digital stores in which the information circulates G11C21/00)}
- G06F 7/483: Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers {(G06F7/4806, G06F7/4824, G06F7/49, G06F7/491, G06F7/544 take precedence)}
- G06F 7/5443: {Sum of products (for applications thereof, see the relevant places, e.g. G06F17/10, H03H17/00)}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A unit operates on a sum term and a carry term separated into a high part and a low part of a product and performs a method that includes iteratively computing a carry save product and separating the carry save product into the high part and the low part: an intermediate product. The unit generates an intermediate wide result by performing a wide addition of the intermediate product to generate an unrounded sum for the high part (i.e., a fused-multiply-add high part) and the low part (i.e., a fused-multiply-add high part). The unit pre-aligns the intermediate wide result on two fixed length shifters such that the fused-multiply-add high part and the fused-multiply-add low part are pre-aligned to each fit on one shifter of the two fixed length shifters.

## Claims

### Claim 1

A computer-implemented method comprising:
computing a carry save product;
separating the carry save product into the high part and the low part, wherein the high part and the low part comprise an intermediate product;
generating an intermediate wide result by performing a wide addition of the intermediate product to generate an unrounded sum for the high part and the low part, wherein the unrounded sum for the high part comprises a fused-multiply-add high part and the unrounded sum for the low part comprises a fused-multiply-add low part; and
pre-aligning the intermediate wide result on two fixed length shifters.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the pre-aligning comprising pre-aligning such that the fused-multiply-add high part and the fused-multiply-add low part each fit on a respective one shifter of the two fixed length shifters.

### Claim 3 (depends on 1)

The computer-implemented of claim 1, wherein the pre-aligning comprises balancing a width of the fused-multiply-add high part and a width of the fused-multiply-add low part to minimize a required width of the two fixed length shifters.

### Claim 4 (depends on 1)

The computer-implemented of claim 1, wherein the pre-aligning comprises reducing a wider width of the fused-multiply-add high part or the fused-multiply-add low part.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein the pre-aligning comprises:
determining if the fused-multiply-add high part is zero; and
based on determining that the fused-multiply-add high part is zero, pre-aligning the intermediate wide result by performing a default normalization on the fused-multiply-add low part, wherein the normalized fused-multiply-add low part fits on one of the two fixed length shifters.

### Claim 6 (depends on 5)

The computer-implemented method of claim 5, wherein the performing a default normalization comprises:
pre-aligning the fused-multiply-add low part by performing a fixed right shift;
feeding the fused-multiply-add low part down to a normalizer;
detecting leading zeros in the fused-multiply-add low part; and
utilizing a number of leading zeros to shift the fused-multiply-add low part via a default small normalizer.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein the pre-aligning comprises:
determining if the fused-multiply-add high part is zero; and
based on determining that the fused-multiply-add high part is not zero:
performing parallel shifts of the fused-multiply-add high part and the fused-multiply-add low part;
merging results of the parallel shifting to create a merged result; and
rounding the merged result.
