1. Porting of GCC, LLVM and other frameworks - Expertise in building cleanroom ports of GCC and associated developer tools and libraries such as binutils, GDB, GLibC, uClibc, newlib, etc., as well as implementing Link Time Optimizations (LTO) in LLVM and GCC
2. Developing AI/ML compilers based on the MLIR/LLVM framework -  We can compile fully connected networks, convolutional networks, and recurrent networks (LSTM) for any AI architecture. Currently, we are working on the compilation of transformers and exploring SSM/Mamba. We have identified the manner in which the compilation in the model can be mapped to custom AI hardware.
3. Development of back-end optimizations -   Our team specializes in designing and developing compiler back-ends for various processor architectures. We work with frameworks such as LLVM, GCC, Open64, LCC, SDCC, and proprietary solutions. Leveraging our expertise in SUIF and SPARK frameworks, we seamlessly integrate with front-ends like EDG and Clang. Our focus lies in architecture optimization, achieved through the implementation of multi-passes within the LLVM/GCC framework. This approach ensures efficient code generation and performance enhancements. Additionally, our experience extends to dynamic binary translators and instrumentation tools, enabling us to address complex challenges in the back-end domain.
4. Designing Runtime Architecture (ABI) - We help design application binary interface (ABI) for VLIW, Vector and RISC processors. We have designed linking models including but not limited to the following: Dynamic Linking;  Static Linking ; Static Linking for Bare Metal systems ; Static Linking without the C Runtime ; 1 x 1 Threads with or without TLS ; M x N Threads with or without TLS
 We actively contribute to the design of Application Binary Interfaces (ABIs) for various processor types: VLIW (Very Long Instruction Word), Vector, RISC (Reduced Instruction Set Computer).  Our expertise encompasses diverse linking models, including: Dynamic Linking, Static Linking, Static Linking for Bare Metal Systems, Static Linking without the C Runtime, 1 x 1 Threads (with or without TLS), M x N Threads (with or without TLS). By carefully crafting ABIs, we ensure seamless communication between software components, efficient memory management, and optimal performance.
5. Parsers and Linting Tools - Our proficiency extends to building robust lexical analyzers and parsers for various programming languages: Java, C, VHDL. Additionally, we handle proprietary domain-specific languages. Furthermore, we create linting tools that identify and flag special constructs, ensuring code quality and adherence to best practices.
6. Custom Libraries - We work closely with our partners in developing custom libraries for their specialized development platforms, such as, Fixed Point Math libraries, Fixed Point Graphics libraries, C/C++ libraries without OS support, Floating Point Math libraries for Fixed Point systems
.
7. Performance Analysis - We evaluate performance bottlenecks for benchmarks such as SPECINT, SPECFP, EEMBC, and GeekBench in the context of the compiler toolchain. Our goal is to identify optimization opportunities to overcome these bottlenecks. Additionally, we customize benchmarks to run under various constraints and modified environments, such as without an operating system or a filesystem, while ensuring that the essence of the benchmark remains unchanged.
8. Testing and Certification - We set up test benches to regress compilers and other tools. We collaborate with clients to build a detailed test plan and develop test suites tailored to specific needs. Additionally, we certify compilers and developer tools using test suites such as Perennial, PlumHall, and the GCC Torture suite. Our goal is to improve compliance by fixing defects in the toolchain.
9. Support and Maintenance - We undertake both feature upgrades and defect fixes to maintain an existing toolchain code base according to the Service Level Agreement (SLA). Additionally, we provide code security and hardening for compiler codebases.

 Read about some customer implementations.

 The Helprack team will work with you to create a compiler that best fits your needs. We would love to hear from you. Let's connect.