Herdora offers an automated platform for identifying and optimizing performance inefficiencies in accelerator code. It scans kernels, benchmarks performance variants, and automatically applies patches to resolve bottlenecks, maximizing compute resource utilization.
Funding
Funding not disclosed
Founders
Product
Problem
Accelerator code often contains underutilized performance potential, leading to inefficient compute resource allocation. Identifying and rectifying these bottlenecks manually is a time-consuming and complex process, hindering optimal hardware utilization.
Solution
Herdora provides an automated platform designed to identify and optimize performance inefficiencies within accelerator code. The system scans kernels to detect performance regressions and opportunities for improvement. It facilitates cross-architecture porting by analyzing code compatibility and suggesting necessary modifications. Herdora benchmarks various performance variants and automatically applies patches to address identified bottlenecks. This comprehensive approach enables users to maximize the productive output of their compute resources.
Target Audience
The primary target audience includes developers and engineers working with high-performance computing (HPC) environments and those responsible for optimizing accelerator code performance.
Features
- Automated kernel scanning for performance bottleneck identification
- Cross-architecture code porting analysis and recommendation engine
- Performance benchmarking suite for evaluating code variants
- Automated bottleneck patching and optimization capabilities
- Support for identifying and resolving underutilized compute resources