Hardware vs. Software Implementation of Warp-Level Features in Vortex RISC-V GPU

TL;DR

This paper compares hardware and software methods for implementing warp-level features in RISC-V GPUs, demonstrating significant performance gains with hardware and viable alternatives with software.

Contribution

It introduces and evaluates both hardware and software approaches for warp-level feature support in RISC-V GPUs, highlighting their trade-offs.

Findings

Hardware implementation achieves up to 4x IPC speedup.

Software solutions are viable in area-constrained scenarios.

Warp-level features enhance GPU efficiency.

Abstract

RISC-V GPUs present a promising path for supporting GPU applications. Traditionally, GPUs achieve high efficiency through the SPMD (Single Program Multiple Data) programming model. However, modern GPU programming increasingly relies on warp-level features, which diverge from the conventional SPMD paradigm. In this paper, we explore how RISC-V GPUs can support these warp-level features both through hardware implementation and via software-only approaches. Our evaluation shows that a hardware implementation achieves up to 4 times geomean IPC speedup in microbenchmarks, while software-based solutions provide a viable alternative for area-constrained scenarios.