AMD has announced the Boltzmann Initiative, plan to change the approach to the market sector of HPC (High Performance Computing), among others by making improvements in performance calculations and software compatibility. One of the main objectives of AMD is to reduce the distance to the main competitor, NVIDIA, as evidence they have redesigned its HPC ecosystem in order to close the gap with NVIDIA and offer an environment competitive (and compatible) with CUDA.
AMD announced Boltzmann Initiative: C++ and CUDA Compilers for AMD GPUs
The first of the main goals of the initiative is a strong change in drivers. AMD has created a dedicated, 64-bit driver for the Linux platform, so it can provide the best possible performance, as well as guarantee the full implementation of the HSA solutions for applications as AMD referred to “HSA +”. This solution allows AMD to use the current standard of HSA, which ensured the extension to be prepared for primarily about FirePro graphics cards series.
The second part of Boltzmann Initiative, in addition to the drivers is a new compiler for the High Performance Computing sector, called HCC (Heterogeneous Compute Compiler), with support for C ++. OpenCL as a standard is very important for AMD since its inception the company has pinned high hopes on it. NVIDIA on the other hand, provides support for OpenCL 1.2, even though a newer version is already available. This encourages AMD OpenCL activities in the field. That’s why AMD wants to get closer to the software developers and provide them with easier use of AMD hardware and using a compiler for the tasks allocated to the shoulders of the GPU and the CPU.
The purpose of the HCC will be to allow developers to write CPU and/or GPU code using a single compiler, in a single language, inside a single source file. The end result is something that resembles Microsoft’s C++ AMP, with developers simply making parallel calls within a C++ program as they see fit.
Perhaps the most important element of the new Logo Initiative is the third and last position, namely the use of CUDA to compile their own software ecosystem. AMD hopes to create an environment that directly competes with NVIDIA environment. Perhaps realizing that for NVIDIA CUDA the environment is already well developed and popular in the sector of HPC, AMD decided to create the developers tools that have the ability to create application code for AMD Graphics processors in terms similar to coding for CUDA. AMD HIP (Heterogeneous-Compute interface for portability) code conversion is to enable CUDA to AMD HIP code, providing translation between CUDA and HIP. It is worth noting that this is not directly on the GPU to run applications compiled for AMD. This means that developers have to simplify a way of using graphics processors from AMD, even if you are currently accustomed to using the CUDA ecosystem.
It is difficult to determine whether the application layer of the translator of the CUDA ecosystem will allow the company to effectively combat the dominance of NVIDIA in the sector of HPC, but you can certainly see this as a quite significant action to improve the condition of its platform, especially since the company is also looking forward to significantly improve the Linux drivers, making it even easier to use HSA.
Source: AMD, ed. incl., Anandtech