Ukukhishwa kwe-Mesa 20.1.0, ukuqaliswa kwamahhala kwe-OpenGL ne-Vulkan

Kuthunyelwe ngu- ukukhululwa kokuqaliswa kwamahhala kwe-OpenGL ne-Vulkan API - Mesa 20.1.0. Ukukhishwa kokuqala kwegatsha le-Mesa 20.1.0 kunesimo sokuhlola - ngemva kokuqiniswa kokugcina kwekhodi, inguqulo ezinzile 20.1.1 izokhishwa. Ku-Mesa 20.1 kwenziwe usekelo olugcwele lwe-OpenGL 4.6 lwe-Intel (i965, iris) ne-AMD (radeonsi) GPUs, usekelo lwe-OpenGL 4.5 lwe-AMD (r600) kanye ne-NVIDIA (nvc0) GPUs, i-OpenGL 4.3 ye-virgl (virtual GPU I-Virgil3D ye-QEMU/KVM), kanye nokusekelwa kwe-Vulkan 1.2 kwamakhadi e-Intel ne-AMD.

Phakathi kwe izinguquko:

• Kwengeziwe Isendlalelo sokukhetha idivayisi esisebenzayo se-Vulkan API kumasistimu anama-GPU amaningi anikwe amandla e-Vulkan, asebenza ngokufanayo ne-DRI_PRIME ye-OpenGL. Ukuze ukhethe umshayeli osebenzayo kanye ne-GPU, i-MESA_VK_DEVICE_SELECT imvelo eguquguqukayo inikezwa (uma ingafakiwe, i-DRI_PRIME iyasetshenziswa).
• Ukusekelwa kwama-chips okulindeleke ngonyaka ozayo ngokusekelwe ku-microarchitecture entsha yengezwe ku-i965 kanye nabashayeli be-iris be-Intel GPUs. IChibi Lerocket.
• Umshayeli we-ANV Vulkan wenzelwe i-Intel GPUs kwengezwe ukulungiselelwa kwama-chips ngokusekelwe ku-Icelake microarchitecture (Gen11), okuvumela ukusetshenziswa kwemibala emsulwa lapho kuthunyelwa umbhalo. Lapho ihlolwa ku-Dota2, lolu shintsho lwehlise inani lemisebenzi yokuguqulwa kombala ngo-95% futhi lwakhuphula ukusebenza ngo-3.5%.
• Kumshayeli we-Vulkan ANV anda ukusebenza kahle kokusetshenziswa kwenqolobane kumasistimu ane-Intel Ivybridge nama-Haswell chips. Ukusebenzisa ukuhlolwa kokusebenza kwe-Vulkan compute kusuka ku-Geekbench 5 kubonise ukwenyuka kokusebenza okungu-330% kuhadiwe ye-Haswell GT3 (ukwanda ngenxa yokuthi ngaphambilini inqolobane yayingasetshenziswa kwezinye izimo).
• Abashayeli be-Intel GPUs (I965, iris) kungezwe Imodi “yembobo emnyama” (isandiso se-OpenGL INTEL_blackhole_render), esikhubaza yonke imisebenzi yokunikezela ehanjiswa yi-GPU, kodwa igcine ukucutshungulwa kwemisebenzi ye-OpenGL.
• Ukusekelwa kwe-Vectorization okwengezwe ngaphambilini kuma-chips e-AMD kufakwe kuma-Intel graphics chips Nir, ukumelwa okuphakathi okungabhala (IR) kwamashadi okuhloswe ukusebenza ezingeni eliphansi kakhulu, ngaphansi kwe-GLSL IR kanye ne-IR yangaphakathi ye-Mesa. Ngasohlangothini olusebenzayo, ngenxa yokulungiselelwa okungcono kwama-shader, ushintsho lwenze kwaba nokwenzeka ukukhulisa ukusebenza kwe-OpenGL ne-Vulkan emidlalweni eminingi kumasistimu ane-Intel GPUs. Ngokwesibonelo, emdlalweni
  Rise of the Tomb Raider kuphawuliwe ukusebenza kwenyuke ngo-3%, kwathi eShadow of the Tomb Raider ngo-10%.

• Ngemuva kokuhlanganisa ama-shaders "Aco“, ethuthukiswa iValve njengenye indlela ye-LLVM shader compiler, ukwesekwa kohlobo lwe-shaderInt9 kwengezwe ku-GFX16+ GPU, okuvumela ukusetshenziswa kwezinombolo ezingu-16-bit kukhodi ye-shader. Ngoba
  I-AMD Navi GPU (GFX10) ivikelekile ukusetshenziswa kwezinjini ze-NGG (Next-Gen Geometry) lapho usebenza nama-vertex nama-tessellation shader.

• Okwe-AMD Navi 12 kanye ne-Navi 14 GPUs ifakiwe ukusekelwa kwemodi ye-DCC (Delta Color Compression) ebonisiwe, eqinisekisa ukusebenza ngedatha yombala ecindezelwe lapho uhlela ukuphuma kwesibonisi.
• Kwengeziwe ukusekelwa kokuhlola kwe-NIR komshayeli wakudala we-Gallium3D u-R600 (AMD Radeon HD 2000-6000) ngokusekelwa kwejometri, ucezu, i-vertex kanye i-tessellation imithunzi.
• Umshayeli we-Vulkan RADV kungezwe Isiqeshana esithuthukisa ukusebenza kwemidlalo ye-Id Tech kumasistimu ane-AMD APU ngokuthuthukisa ukuphathwa kwenkumbulo.
• Kumshayeli wePanfrost kwenziwe ukusekelwa kokuhlola kwe-OpenGL ES 3.0 kanye enikeziwe Usekelo lokunikeza lwe-3D lwe-Bifrost GPU (Mali G31). Ukuqaliswa kokuqala kokuhlanganiswa kwe-shader kulungiselelwe okusekela isethi ye-Bifrost GPU-specific yemiyalelo yangaphakathi.
• Umshayeli we-Vulkan TURNIP, owenzelwe i-Qualcomm Adreno GPUs, kwengezwe ukusekelwa kwe-geometry shaders kanye Adreno 650 chips.
• Kumshayeli we-Gallium3D u-LLVMpipe, ohlinzeka ngokuhlinzekwa kwesofthiwe, kwavela ukusekelwa kwama-tessellation shader.
• Kwethulwe большая ingxenye ukulungiselelwa ku-glthread (ukuqaliswa okucushwe okuningi kwe-OpenGL). Ngemva kokwenza izinguquko, ukusebenza kwesifanisi somjaho we-Torcs kukhuphuke ngo-16% ekucushweni okuzenzakalelayo futhi ngo-40% lapho i-glthread inikwe amandla.
• Kwengeziwe allow_draw_out_of_order option (inikwe amandla nge-driconf) ukunika amandla ukulungiselelwa ukuze kusheshiswe imisebenzi yokudweba ngaphandle kwe-oda eqondene ne-CAD. Uma le nketho inikwe amandla, ukusheshisa okungu-11% kubonakala ekuhlolweni kwe-Viewperf7 Catia.
• Kwengezwe izandiso ezintsha ze-OpenGL:
  • GL_ARB_compute_variable_group_size ngo i965.
  • GL_EXT_depth_bounds_test ngo-Iris.
  • GL_EXT_texture_shadow_lod okwe-radeonsi kanye ne-nvc0.
  • GL_EXT_draw_instanced kwe gles2.
  • GL_NV_alpha_to_coverage_dither_control ngoba radeonsi
  • GL_NV_copy_image kubo bonke abashayeli be-gallium.
  • GL_NV_pixel_buffer_object kubo bonke abashayeli be-gallium, kanye ne-i915, i-i965 ne-swrast.
  • GL_NV_viewport_array2 kwe-nvc0 (GM200+).
  • GL_NV_viewport_swizzle kwe-nvc0 (GM200+).
• Kungezwe izandiso kumshayeli we-RADV Vulkan (wamakhadi e-AMD):
  • VK_AMD_memory_overallocation_behavior
  • VK_KHR_shader_non_semantic_info
  • VK_EXT_robustness2
  • VK_KHR_8bit_storage kumakhadi e-GFX8+ uma usebenzisa isipele se-“ACO” sokuhlanganiswa kwe-shader
  • VK_KHR_16bit_storage kumakhadi e-GFX8+ uma usebenzisa i-backend ye-“ACO” shader yokuhlanganiswa (ngaphandle kosekelo lwe-storeInputOutput16)
  • VK_KHR_shader_float16_int8 kumakhadi e-GFX8+ uma usebenzisa i-backend ye-“ACO” shader yokuhlanganiswa (ngaphandle kosekelo lwe-storeInputOutput16)
• Kungezwe izandiso kumshayeli we-ANV Vulkan (wamakhadi we-Intel):
  • VK_KHR_shader_non_semantic_info
  • VK_EXT_robustness2

  Source: opennet.ru