MOTIVATION: Multiple sequence alignment (MSA) is a fundamental operation in bioinformatics, yet existing MSA tools are struggling to keep up with the speed and volume of incoming data. This is because the runtimes and memory requirements of current MSA tools become untenable when processing large numbers of long input sequences, and they also fail to fully harness the parallelism provided by modern CPUs and GPUs. RESULTS: We present Tall and Wide Alignments at High Throughput (TWILIGHT), a novel MSA tool optimized for speed, accuracy, scalability, and memory constraints, with both CPU and GPU support. TWILIGHT incorporates innovative parallelization and memory-efficiency strategies that enable it to build ultralarge alignments at high speed even on memory-constrained devices. On challenging datasets, TWILIGHT outperformed all other tools in speed and accuracy. It scaled beyond the limits of existing tools and performed an alignment of 1 million RNASim sequences within 30 min while utilizing <16 GB of memory. TWILIGHT is the first tool to align over 8 million publicly available SARS-CoV-2 sequences, setting a new standard for large-scale genomic alignment and data analysis. AVAILABILITY AND IMPLEMENTATION: TWILIGHT's code is freely available under the MIT license at https://github.com/TurakhiaLab/TWILIGHT. The test datasets and experimental results, including our alignment of 8 million SARS-CoV-2 sequences, are available at https://zenodo.org/records/14722035.

Pairwise sequence alignment is one of the most fundamental and computationally intensive steps in genome analysis. With the improving costs and throughput of third-generation sequencing technologies and the growing availability of whole-genome datasets, longer alignments are becoming more common in the field of bioinformatics. However, the high memory demands of long alignments create significant obstacles to hardware acceleration. Banding techniques allow recovering high-quality alignments with lower memory, but they also require more memory for long alignments than what is typically available on-chip in hardware accelerators. Recently, tiling-based hardware accelerators have made remarkable strides in accelerating sequence alignment, achieving three to four orders of magnitude improvement in alignment throughput over software tools without any restrictions on alignment length. However, it is crucial to note that existing tiling heuristics can cause the alignment quality to degrade, which is a critical concern for the wider adoption of accelerators in the field of bioinformatics. To address this issue, this paper describes TALCO - a novel method for tiling long sequence alignments, that, similar to prior tiling techniques, maintains a constant memory footprint during the acceleration step independent of alignment length. However, unlike previous techniques, TALCO also ensures optimal alignments under banding constraints. TALCO does this by leveraging the convergence of traceback paths beyond a tile to a single point on the boundary of that tile - a strategy that generalizes well to a broad set of sequence alignment algorithms. We demonstrate the advantages of TALCO by applying it to two different and widely-used banded sequence alignment algorithms, X-Drop and WFA-Adapt. To the best of our knowledge, this is the first time that a tiling technique is being applied to a non-classical algorithm for sequence alignment, such as WFA-Adapt. The TALCO tiling strategy is beneficial to both software and hardware. When implemented in software, the TALCO strategy reduces the memory requirements for X-Drop and WFA-Adapt algorithms by up to 39 × and 57 ×, respectively, and when implemented as ASIC accelerator, it provides up to 1,900 × and 2,000 × improvement in alignment throughput/watt over CPU baselines implementing the same algorithms. Compared to state-of-the-art GPU and ASIC baselines implementing tiling heuristics, TALCO provides up to 50 × and 1.1 × improvement in alignment throughput, respectively, while also maintaining a higher alignment quality. Code availability: https://github.com/TurakhiaLab/TALCO.

This brief compares quantized float-point representation in posit and fixed-posit formats for a wide variety of pre-trained deep neural networks (DNNs). We observe that fixed-posit representation is far more suitable for DNNs as it results in a faster and low-power computation circuit. We show that accuracy remains within the range of 0.3% and 0.57% of top-1 accuracy for posit and fixed-posit quantization. We further show that the posit-based multiplier requires higher power-delay-product (PDP) and area, whereas fixed-posit reduces PDP and area consumption by 71% and 36%, respectively, compared to (Devnathet al., 2020) for the same bit-width.

In this paper, we propose the design of a power-efficient floating point configurable approximate multiplier (FPCAM) suitable for error resilient applications. FPCAM allows systematic approximation, and the amount of approximation can be configured at run-time, depending on the error-tolerance of the applications. We show that compared to the existing state of the art multipliers, FPCAM on average consumes 62% lesser power and has 69% less power delay product. FPCAM also has 66% less area as compared to state of the art approximate multipliers. We have analyzed FPCAM for three different multimedia applications and random inputs to show that we achieve similar output quality as compared to existing multipliers while benefiting in power, area, and power delay product.