Intel ya wallafa sakin aikin oidn 2.0 (Buɗe Hoton Denoise), wanda ke haɓaka tarin abubuwan tacewa don cire hayaniya daga hotunan da aka shirya ta amfani da tsarin samar da haske. Buɗe Hoton Denoise ana haɓaka shi azaman wani ɓangare na babban aikin, kayan aikin API Rendering guda ɗaya, da nufin haɓaka kayan aikin gani na software don lissafin kimiyya (SDVis (Software Defined Visualization), gami da ɗakin karatu na Embree ray, tsarin sarrafa hoto na GLuRay, an rarraba OSPRay. dandalin binciken ray da tsarin rasterization software na OpenSWR An rubuta lambar a cikin C++ kuma an buga ta ƙarƙashin lasisin Apache 2.0.
Manufar aikin ita ce samar da ingantacciyar inganci, inganci, da sauƙin amfani da fasalulluka waɗanda za a iya amfani da su don haɓaka ingancin sakamakon gano hasken. Abubuwan tacewa da aka tsara suna ba da izini, dangane da sakamakon gajeriyar zagayowar gano hasashe, don samun matakin inganci na ƙarshe mai kwatankwacin sakamakon mafi tsada da tsari mai ɗaukar lokaci na yin cikakken bayani.
Buɗe Hoton Denoise yana kawar da hayaniyar bazuwar da ke faruwa, alal misali, a cikin binciken ray na Monte Carlo RT (MCRT). Don cimma ma'ana mai inganci a cikin irin waɗannan algorithms, wajibi ne a bi diddigin haskoki masu yawa, in ba haka ba kayan tarihi masu ban sha'awa a cikin sigar bazuwar amo suna bayyana a cikin hoton da aka samu.
Amfani da Buɗe Hoto Denoise yana ba ku damar rage adadin ƙididdiga masu mahimmanci yayin ƙididdige kowane pixel ta umarni da yawa na girma. Sakamakon haka, zaku iya ƙirƙirar hoto mai hayaniya da sauri da sauri, amma sai ku kawo shi zuwa ingantaccen inganci ta amfani da algorithms rage amo cikin sauri. Idan kuna da kayan aikin da suka dace, ana iya amfani da kayan aikin da aka tsara don gano ma'amalar hasken haske tare da cire amo kan-da-tashi.
Ana iya amfani da ɗakin karatu akan nau'ikan na'urori daban-daban, daga kwamfutar tafi-da-gidanka da kwamfutoci zuwa nodes a cikin gungu. An inganta aiwatar da azuzuwan daban-daban na 64-bit Intel CPUs tare da tallafi don SSE4, AVX2, AVX-512 da XMX (Xe Matrix Extensions) umarnin, Apple Silicon chips da tsarin tare da Intel Xe GPUs (Arc, Flex da Max jerin), NVIDIA (tushen Volta, Turing, Ampere, Ada Lovelace da Hopper architectures) da AMD (dangane da gine-ginen RDNA2 (Navi 21) da RDNA3 (Navi 3x). An bayyana goyan bayan SSE4.1 a matsayin ƙaramin buƙatu.
Manyan canje-canje a cikin Buɗe Hoton Denoise 2.0:
- Taimako don haɓaka ayyukan rage amo ta amfani da GPU. An aiwatar da tallafi don ƙaddamar da lissafin zuwa gefen GPU ta amfani da tsarin SYCL, CUDA da HIP, waɗanda za a iya amfani da su tare da GPUs dangane da Intel Xe, AMD RDNA2, AMD RDNA3, NVIDIA Volta, NVIDIA Turing, NVIDIA Ampere, NVIDIA Ada Lovelace da NVIDIA Hopper architecture.
- An ƙara sabon API ɗin sarrafa buffer, yana ba ku damar zaɓar nau'in ajiya, kwafin bayanai daga mai masaukin, da shigo da buffer na waje daga APIs masu zane kamar Vulkan da Direct3D 12.
- Ƙara goyon baya don yanayin aiwatar da asynchronous (oidnExecuteFilterAsync da ayyukan oidnSyncDevice).
- An ƙara API don aika buƙatun zuwa na'urorin zahiri waɗanda ke cikin tsarin.
- Ƙara aikin oidnNewDeviceByID don ƙirƙirar sabuwar na'ura dangane da ID na na'urar jiki, kamar UUID ko adireshin PCI.
- Ƙara ayyuka don ɗaukakawa tare da SYCL, CUDA da HIP.
- An ƙara sabbin sigogin duba na'ura (SystemMemorySupported, sarrafaMemorySupported, externalMemoryTypes).
- Ƙara ma'auni don saita ingancin matakin tacewa.
source: budenet.ru