Abalawuli beMesa kukhona isoftware ye-Linux yomthombo ovulekile itholakalela i-AMD, i-NVIDIA ne-Intel hardware. Iphrojekthi kaMesa iqale njengokusetshenziswa komthombo ovulekile kokucaciswa kwe-OpenGL (uhlelo lokunikeza ihluzo ze-3D ezisebenzisanayo).
Kuyo yonke le minyaka, iphrojekthi yakhula yasebenzisa ama-API wehluzo amaningi, kufaka phakathi i-OpenGL ES (izinhlobo 1, 2, 3), i-OpenCL, i-OpenMAX, i-VDPAU, i-VA API, i-XvMC, ne-Vulkan. Izilawuli ezahlukahlukene yamadivayisi ivumela imitapo yolwazi yeMesa ukuthi isetshenziswe ezindaweni eziningi ezahlukahlukene, kusuka ekulingisweni kwesoftware ukuqedela ukusheshisa kwehadiwe kwama-GPU anamuhla.
IMesa isebenzisa isendlalelo sokuhumusha esizimele somthengisi phakathi kwe-API yehluzo efana ne-OpenGL kanye nezishayeli zemidwebo ku-kernel yohlelo lokusebenza.
Yini okusha kuMesa 19.3.0?
Kwethulwe muva nje kwethulwa iThebula 19.3.0 Le yinguqulo yokuqala yegatsha leMesa elinesimo sokuhlola, okuthi ngemuva ukuzinza kokugcina kwekhodi, inguqulo ezinzile izokhishwa kuMesa 19.3.1.
Ithebula 19.3.0 lihlinzeka ukuxhaswa okugcwele kwe-OpenGL 4.6 yama-Intel GPUs (i965, abashayeli be-iris), ukusekelwa kwe-OpenGL 4.5 kwe- I-AMD (r600, radeonsi) ne-NVIDIA (nvc0) GPUs, kanye Ukusekelwa kweVulkan 1.1 yamakhadi we-Intel ne-AMD.
para I-RADV (Umshayeli weVulkan wama-chip we-AMD) i-backend entsha yokuhlanganisa izithunzi »ACO», okuyi ithuthukisa iValve njengenye indlela yokuhlanganisa i-LLVM shader. Inhloso ye-backend ukuqinisekisa ukuthi ikhodi inikezwa ngendlela engcono kakhulu yezinsiza zohlelo lokusebenza lomdlalo nokufeza isivinini sokuhlanganiswa okuphezulu kakhulu.
I-ACO ibhalwe ku-C ++ futhi iyakhiwa ngenhloso yokuthi isetshenziselwe ukuhlanganiswa kwe-JIT futhi isebenzisa izakhiwo zedatha esheshayo ukwehlisa, ukugwema izakhiwo ezenzelwe ukukhomba.
Ukunikezwa kwekhodi okuphakathi kususelwa ngokuphelele ku-SSA (Static Single Assignment) futhi kuvumela ukwabiwa kwerekhodi, ukufaka irekhodi kwangaphambi kokusebenza kwerekhodi ngokuya nge-shader I-ACO ingenziwa isebenze ngeVega 8, Vega 9, Vega 10, Vega 20 kanye neNavi 10 GPUs ukusetha ukuguquguquka kwemvelo "RADV_PERFTEST = aco";
Umshayeli weZink Gallium3D ufakiwe kwisisekelo sekhodi, esebenzisa i-OpenGL API ngaphezulu kweVulkan. I-Zink ikuvumela ukuthi uthole i-OpenGL esheshayo yehardware uma kukhona abashayeli ohlelweni abakhawulelwe ukusekela i-Vulkan API kuphela.
Umshayeli we-ANV Vulkan kanye nomshayeli we-OpenGL iris banikela ukwesekwa kokuqala kwesizukulwane se-12 sama-Intel chips (iTiger Lake, gen12). Ku-kernel ye-Linux, izingxenye zokusekelwa kweTiger Lake zifakiwe kusukela ku-5.4;
I-Los abalawuli be-i965 kanye ne-iris bahlinzekile ukusekelwa kwe- isethulo esiphakathi se izithunzi ze-SPIR-V, okwenze ukuthi kutholakale ukusekelwa okugcwele kwe-OpenGL 4.6 kulaba bashayeli.
Kungeziwe Ukuxhaswa kwe-AMD Navi 14 GPU kumshayeli weRadeonSI futhi kuthuthukiswe amathuluzi okusheshisa wokumisa amavidiyo, ngokwesibonelo, ukungezwa kokukhishwa kwamavidiyo we-8K kumafomethi we-H.265 ne-VP9 kungeziwe;
La hlanganisa ukusekelwa kwemodi protected yengezwe kusilawuli I-Vulkan nge-RADV, lapho ukulandelana okuqhutshwa ukuhlanganisa ama-shader kuhlukaniswa kusetshenziswa indlela ye-seccomp.
Abashayeli bama-chip we-AMD basebenzisa isikhombimsebenzisi sohlelo ukusetha kabusha i-GPU ebonakale kwimodyuli eyinhloko ye-AMDGPU.
Isebenzile thuthukisa ukusebenza ezinhlelweni ezinama-AMD Radeon APU. Iphinde yathuthukisa ukusebenza komshayeli weGallium3D Iris kuma-Intel GPUs;
Kokusha Izandiso ze-OpenGL kungezwe okugqanyisiwe esikhangisweni esingasithola:
- I-GL_ARB_gl_spirv ye-i965
- I-GL_ARB_spirv _ ye-i965
- GL_EXT_demote_to_helper_invocation ye-i965
Izandiso zeVulkan RADV zamakhadi we-AMD:
- I-VK_ANDROID_external_memory_android_hardware_buffer
- VK_EXT_shader_demote_to_helper_vocation
- I-VK_KHR_shader_clock
- VK_KHR_shader_float_controls
- VK_KHR_spirv_1_4
- Isikhathi se-VK_KHR_semaphore
- VK_EXT_texel_buffer_ ukuqondanisa
Izandiso ze-ANV Vulkan zamakhadi we-Intel:
- VK_INTEL_performance_query
- VK_KHR_vulkan_memory_model
- I-VK_EXT_shader_subgroup_ballot
- VK_EXT_shader_subgroup_vote
- VK_KHR_spirv_1_4
- I-VK_KHR_shader_clock
- VK_KHR_shader_float_controls
Ngaphezu kwalokho, singabona ukushicilelwa yi-AMD kwemibhalo ekwakhiweni kwemishini ye-APU "Vega" 7nm ngokususelwa ku-microarchitecture ye-GCN (Graphics Core Next).
Ekugcineni, linda le nguqulo entsha yabashayeli beMesa ukuthi iqale ukusatshalaliswa ekusatshalalisweni okuyinhloko kweLinux ezinsukwini ezimbalwa ezizayo.
Sawubona. Ngifunde ezindaweni eziningi mayelana nokusetha ukuguquguquka kwesistimu ye-RADV_PERFTEST = aco kepha akekho ochazayo ukuthi angakwenza kanjani lokho ... kungaba yimininingwane yokuthi othile aphawule ngokuthi angakusebenzisa kanjani noma akumise kanjani lokho kuguquguquka. Ngiyabonga
I-ACO ingumhlanganisi wezithunzi ezithi ngombono "zinikwe amandla ngokwakhona" (okusho ukuthi, akukho okumele kwenziwe) kufanele ube nohlobo lwakamuva lwabashayeli beMesa noma uma kwenzeka uSteam ungaphoqa ukusetshenziswa kwe umhlanganisi engeza kusiqalisi segeyimu inketho `RADV_PERFTEST = aco% command%`
Noma ukwenza inketho kuhlelo lonke kumane wengeze ukuguquguquka kwemvelo RADV_PERFTEST = aco kufayela le- ~ / .profile.
Kumele ucabangele ukuthi akuwona wonke ama-GPU ahambisanayo ngakho-ke awunakukwazi ukuqala lokhu kuhlanganisa.
Kuwebhu kunemininingwane eminingi yokuthi ungakulungisa kanjani okuguqukayo ekusatshalalisweni kwe-Linux ngakunye noma mayelana nezinye izinhlelo zokusebenza ezinjengeSteam, iLutris, njll.
Ukubingelela!