The HyperDrive technology

15. The HyperDrive technology

HyperDrive is a core library including several time-critical functions required by VEGA ZZ for high speed computing. The highly optimized and and parallel code, especially designed for the modern CPUs, allows to speed-up the programs and make faster the development without deep skills in programming. Moreover, the library offers features that are useful not only in developing of molecular modelling software, but also of generic application. In particular, the key features are here summarized:

Hardware independent.
You can develop the same application for Linux (ARM, x86 and x64) and Windows (x86 and x64) without changes in the source code.
Same software for single or multiprocessor systems.
The library checks the number of CPUs and automatically switches itself from sequential to parallel mode.
OpenCL support
Some routines (such as virtual LogP calculation) are written to run on the GPU thanks to OpenCL abstraction layer.
Simultaneous multithreading (SMT) ready
It uses the full power of modern multiscalar CPUs with hardware multithreading such as Intel i5, i6, i9 and AMD Ryzen 5, 7, Threadripper.
Multi core CPU ready
The latest multi core CPUs provided by AMD and Intel are detected and the parallel execution is automatically enabled.
SIMD optimization
The functions that are more frequently called, are written in assembly and optimized by using SSE SIMD instruction set.
No special compiler required.
You can use your preferred C/C++ compiler. C++ envelops are included in the headers.

HyperDrive requires an initialization phase that is executed by the host application when it starts. The host application can choose the appropriate HyperDrive version for the installed microprocessor and the HyperDrive detects the number of CPUs switching itself to run in sequential mode (one CPU) or in parallel mode:

HyperDrive initialization

After the initialization phase, the host application can call the HyperDrive functions in transparent mode as a normal library: the HyperDrive library select internally the most appropriate code for that hardware/software system and if it can run more than one thread at the same time at hardware level such as for multicore or SMT CPUs, the code is executed in parallel:

HyperDrive parallelism

HyperDrive library includes several functions that are shown in the following table according to their application field:

Molecular modelling functions		Generic functions to build molecules. Functions to manage ATDL templates and assign the atom types. Bond management, connectivity build and chirality detection. Functions to check and assign the partial atomic charges. Lipophilicity prediction. Molecular mechanics. Surface calculation (SSE and OpenCL accelerated).
Mathematical and statistical functions		Discrete Fourier transformation. Linear regression. Matrix manipulation Receiver Operating Characteristic curve calculation. Generic statistics functions.
Low level		Accelerated memory management. Argument parser. Base64 encoding/decoding. Cyclic redundancy check (CRC16, CRC32, CRC64 and MD5). Directory manipulation. Endian management. File management. Interchange File Format (IFF) management. Memory lists. Memory streams. Multithreading (threads, mutex, semaphore barriers and CPU detection). Network (sockets, HTTP client, WOL client). OpenCL helper (to simplify the OpenCL programming). Process management . Pseudo-random number generator. Secure Socket Layer (SSL). Sort (Insert sort, quick sort and fast parallel quick sort for large data sets). Internal spreadsheet implementation. String manipulation. Time functions.