Ua hoʻopuka ʻo Intel i ka hoʻokuʻu ʻana o ka oidn 2.0 (Open Image Denoise) papahana, kahi e hoʻomohala ai i kahi hōʻuluʻulu o nā kānana no ka hōʻole ʻana i nā kiʻi i hoʻomākaukau ʻia me ka hoʻohana ʻana i nā ʻōnaehana hoʻohālikelike ray tracing. Ke hoʻomohala ʻia nei ka Open Image Denoise ma ke ʻano he ʻāpana o kahi papahana ʻo API Rendering Toolkit nui e pili ana i ka hoʻomohala ʻana i nā hāmeʻa ʻike polokalamu ʻepekema (SDVis (Software Defined Visualization)), me ka hale waihona puke Embree ray tracing, ka GLuRay photorealistic rendering system, ka OSPRay distributed ray tracing platform , a me ka OpenSWR software rasterization system Ua kākau ʻia ke code ma C++ a paʻi ʻia ma lalo o ka laikini Apache 2.0.
ʻO ka pahuhopu o ka papahana ʻo ia ka hāʻawi ʻana i nā hiʻohiʻona denoising kiʻekiʻe, kūpono, a maʻalahi hoʻi e hoʻohana i hiki ke hoʻohana ʻia no ka hoʻomaikaʻi ʻana i ka maikaʻi o nā hualoaʻa ray tracing. ʻAe nā kānana i manaʻo ʻia, ma muli o ka hopena o ka pōʻaiapuni o ka hihi hihihihi i pōkole, e loaʻa i kahi pae maikaʻi hope loa e like me ka hopena o kahi kaʻina hana ʻoi aku ka nui o ke kumu kūʻai a hoʻopau manawa.
Kānā ʻo Open Image Denoise i nā walaʻau like ʻole e like me Monte Carlo Numerical Integration (MCRT) ray tracing. No ka hoʻokō ʻana i ka hoʻohālikelike kiʻekiʻe ma ia mau algorithm, pono ke ʻimi ʻana i kahi helu nui o nā kukuna, inā ʻaʻole e ʻike ʻia nā mea kiʻi i ʻike ʻia i ke ʻano o ka walaʻau ʻole i ke kiʻi hopena.
ʻO ka hoʻohana ʻana iā Open Image Denoise hiki iā ʻoe ke hōʻemi i ka helu o nā helu e pono ai e kekahi mau kauoha o ka nui i ka helu ʻana i kēlā me kēia pika. ʻO ka hopena, hiki ke hoʻohua i kahi kiʻi walaʻau mua i ʻoi aku ka wikiwiki, akā lawe mai ia i kahi maikaʻi e ʻae ʻia me ka hoʻohana ʻana i nā algorithms hoʻemi wikiwiki. Me nā lako kūpono, hiki ke hoʻohana ʻia nā mea hana i manaʻo ʻia no ka huli ʻana o ka ray interactive me ka hoʻopau ʻana i ka leo lele.
Hiki ke hoʻohana ʻia ka waihona ma nā papa like ʻole, mai nā laptops a me nā PC a i nā nodes i nā pūʻulu. Hoʻolālā ʻia ka hoʻokō ʻana no nā papa like ʻole o 64-bit Intel CPUs me ke kākoʻo no SSE4, AVX2, AVX-512 a me XMX (Xe Matrix Extensions), nā ʻāpana Apple Silicon a me nā ʻōnaehana me Intel Xe GPUs (Arc, Flex a me Max series), NVIDIA (ma muli o Volta, Turing, Ampere, Ada Lovelace a me Hopper architectures) a me AMD (ma muli o RDNA2 (Navi 21) a me RDNA3 (Navi 3x) architectures). Hoʻolaha ʻia ke kākoʻo no SSE4.1 ma ke ʻano he koi haʻahaʻa.
Nā hoʻololi nui i ka hoʻokuʻu ʻana o Open Image Denoise 2.0:
- Kākoʻo no ka hoʻokō ʻana i nā hana hoʻohaʻahaʻa walaʻau me ka GPU. Hoʻokomo ʻia ke kākoʻo no ka hoʻouka ʻana o GPU me ka hoʻohana ʻana i nā ʻōnaehana SYCL, CUDA, a me HIP i hiki ke hoʻohana ʻia me nā GPU e pili ana i ka papa hana Intel Xe, AMD RDNA2, AMD RDNA3, NVIDIA Volta, NVIDIA Turing, NVIDIA Ampere, NVIDIA Ada Lovelace, a me NVIDIA Hopper.
- Ua hoʻohui ʻia kahi API hoʻokele buffer hou, e ʻae iā ʻoe e koho i ke ʻano o ka waiho ʻana, kope i ka ʻikepili mai ka mea hoʻokipa, a hoʻokomo i nā buffer waho mai nā API kiʻi e like me Vulkan a me Direct3D 12.
- Hoʻohui i ke kākoʻo no ke ʻano hoʻokō asynchronous (hana oidnExecuteFilterAsync a me oidnSyncDevice).
- Hoʻohui i kahi API no ka hoʻouna ʻana i nā noi i nā mea kino i loaʻa i ka ʻōnaehana.
- Hoʻohui ʻia ka hana oidnNewDeviceByID e hana i kahi mea hou e pili ana i ka ID hāmeʻa kino, e like me ka UUID a i ʻole ka helu PCI.
- Hoʻohui i nā hiʻohiʻona no ka portability me SYCL, CUDA a me HIP.
- Hoʻohui ʻia nā koho scan mea hou (systemMemorySupported, managedMemorySupported, externalMemoryTypes).
- Hoʻohui ʻia kahi ʻāpana e hoʻonohonoho i ka pae maikaʻi o nā kānana.
Source: opennet.ru