LLVM’s intermediate representation (IR) has two deferred undefined behavior (UB) values: undef and poison. The existence of these two values has been a persistent source of bugs. Reasoning about the correctness of analyses and optimizations for the regular cases is already tricky; ensuring that these are sound for all UB cases is highly non-trivial. Undef values, in particular, are one of the most misunderstood concepts of LLVM IR. On paper, the definition is simple: they represent an arbitrary value of the underlying type, and can yield a different value each time they are observed. However, this property makes even simple algebraic rewrites, such as replacing 2 × y with y + y , unsound in LLVM. Because reasoning about undef is hard, and the benefits of having it are limited, we have set a roadmap to eliminate it altogether from LLVM IR. The last remaining use of undef is the value of uninitialized memory, which has implications on the lowering of bitfields, as well as raw data copies and comparisons. In this paper, we propose an extension to LLVM IR that includes a raw memory value type and a freezing load. We show that these two constructs are sufficient to replace the remaining uses of undef in LLVM. Our implementation shows that these changes have minimal impact on both run-time and compile-time performance. By removing the final hurdle to eliminating undef from LLVM IR, this work paves the way for a simpler semantic model and easier reasoning about the soundness of IR analyses and optimizations.

Advances in computational capabilities and large volumes of experimental data have established computer simulations of brain tissue models as an important pillar in modern neuroscience. Alongside, a variety of domain specific languages (DSLs) have been developed to succinctly express properties of these models, ensure their portability to different platforms, and provide an abstraction that allows scientists to work in their comfort zone of mathematical equations, delegating concerns about performance optimizations to downstream compilers. One of the popular DSLs in modern neuroscience is the NEURON MODeling Language (NMODL). Until now, its compilation process has been split into first transpiling NMODL to C++ and then using a C++ toolchain to emit the efficient machine code. This approach has several drawbacks including the reliance on different programming models to target heterogeneous hardware, maintainability of multiple compiler back-ends and the lack of flexibility to use the domain information for C++ code optimization. To overcome these limitations, we present MOD2IR, a new open-source code generation pipeline for NMODL. MOD2IR leverages the LLVM toolchain to target multiple CPU and GPU hardware platforms. Generating LLVM IR allows the vector extensions of modern CPU architectures to be targeted directly, producing optimized SIMD code. Additionally, this gives MOD2IR significant potential for further optimizations based on the domain information available when LLVM IR code is generated. We present experiments showing that MOD2IR is able to produce on-par execution performance using a single compiler back-end implementation compared to code generated via state-of-the-art C++ compilers, and can even surpass them by up to 1.26×. Moreover, MOD2IR supports JIT-execution of NMODL, yielding both efficient code and an on-the-fly execution workflow.

Graph databases (GDBs) are crucial in academic and industry applications. The key challenges in developing GDBs are achieving high performance, scalability, programmability, and portability. To tackle these challenges, we harness established practices from the HPC landscape to build a system that outperforms all past GDBs presented in the literature by orders of magnitude, for both OLTP and OLAP workloads. For this, we first identify and crystallize performance-critical building blocks in the GDB design, and abstract them into a portable and programmable API specification, called the Graph Database Interface (GDI), inspired by the best practices of MPI. We then use GDI to design a GDB for distributed-memory RDMA architectures. Our implementation harnesses onesided RDMA communication and collective operations, and it offers architecture-independent theoretical performance guarantees. The resulting design achieves extreme scales of more than a hundred thousand cores. Our work will facilitate the development of next-generation extreme-scale graph databases.