Article information
2024 , Volume 29, ¹ 4, p.95-109
Giatsintov A.M., Mamrosenko K.A.
Testing graphics processing operations with graphics accelerator emulation
Nowadays, an increasing number of problems require the use of hardware accelerators to reduce the time required for calculations as well as for displaying their results. An important aspect of working with graphics operations is to achieve correctness for execution of operations and debugging
of both subsystems and drivers involved in displaying of graphics information. This paper presents the architecture of an emulator of the graphics accelerator designed to diagnose and correct hard-to-detect errors within components of the graphics subsystem. Unlike the known solutions, the developed architecture allows using the emulator as a part of system emulators such as a QEMU and as an attached library for a driver of the graphics accelerator. The architecture of the developed emulator can be divided into several layers: the upper level provides API for application interaction with the emulator (initialization, programming of registers of the virtual device, etc.); the middle level performs service functions necessary for correct work of the emulator (memory management, thread pooling, checking if the emulated device has the requested register, etc.); the lower level performs emulation of a particular graphics chip. The virtual chip is controlled by reading and writing to the device registers — the emulator provides separate functions for these operations. One
of the main advantages is the ability to run the graphics accelerator driver on a tool machine with capability for step-by-step debugging. The emulator allows detecting some intermittent errors that were hard to reproduce on real hardware.
Keywords: visualization, graphics system, rendering, operating system, emulation, QEMU
doi: 10.25743/ICT.2024.29.4.007
Author(s): Giatsintov Alexander Michaylovich PhD. Position: Senior Research Scientist Office: Federal State Institution Scientific Research Institute for System Analysis of the Russian Academy of Sciences Address: 117218, Russia, Moscow, 36-1, Nakhimovsky prospect
E-mail: giatsintov@niisi.ras.ru SPIN-code: 1359-4413Mamrosenko Kirill Anatolievich PhD. Position: Director Office: Federal State Institution Scientific Research Institute for System Analysis of the Russian Academy of Sciences Address: 117218, Russia, Moscow, 36-1, Nakhimovsky prospect
E-mail: mamrosenko_k@niisi.ras.ru SPIN-code: 5596-7877 Bibliography link: Giatsintov A.M., Mamrosenko K.A. Testing graphics processing operations with graphics accelerator emulation // Computational technologies. 2024. V. 29. ¹ 4. P. 95-109
|