I-AMD ibivele isebenza kuLinux Kernel 4.20 eseqale ngayo ukuhambisa imigqa eminingi yekhodi ku-Linux Kernel entsha ngaphezu kokuqhubeka nokunikela kusoftware yamahhala.
Uma ucabanga ngakho konke ukunikwa amandla kwe-hardware okusha okufike ku-Linux kernel muva nje, kanye nokukhulula i-AMDGPU DDX 18.1 nokukhululwa okukhulu kwe-ROCm 1.9, iqoqo lakho lokugcina lezinguquko zezinsizakusebenza nalo lathola i-DRM-Okulandelayo ngaphambi komjikelezo weLinux Kernel 4.20 ~ 5.0.
Le kuzoba enye inguqulo ethokozisayo yabasebenzisi beRadeon Linux.
Ngaphandle kwalokhu UMichael larabel wasePhoronix wenze ukuhlaziywa lapho anikeze khona ama-akhawunti womsebenzi owenziwe kubanikeli abahlukahlukene, ikakhulukazi abakhiqizi be-GPUS.
I-AMD inikela ngekhodi elingu-8.5x ngaphezulu ku-kernel ye-Linux kune-NVIDIA
Noma kunjalo akufanele kube isimanga esikhulu, kucatshangelwa wonke umsebenzi owenziwe yi-AMD kusitaki sayo semithombo evulekile emisha muva nje.
Ngesikhathi I-NVIDIA isanikela okuncane kakhulu (ngikhuluma kancane) maqondana ne-Linux kernel, i-AMD iza nolayini abangaphezulu kokuphindwe kasishiyagalombili kwekhodi engezwe ku-kernel ngokuqhathaniswa ne-NVIDIA.
Abathuthukisi be-AMD banikele ngama-2.168. Imigqa engu-104 yekhodi i-Linux kernel, imigqa engu-414.761 kule nqubo, noma inzuzo ephelele yemigqa yekhodi engu-1.753.343.
I-NVIDIA, nokho, inikele inani eliphelele lemigqa engama-303.180 yekhodi ku-kernel futhi yasusa i-97.197, noma inzuzo ephelele yemigqa yekhodi engama-205.983 kuphela.
Iminikelo ye-AMD eKernel faka phakathi ama-AMD CPU / chipset ahlukahlukene asebenza kanye izingxenye zalo kanye Umthombo ovulekile umshayeli wehluzo we-AMDKFD / AMDGPU / Radeon futhi isebenza ngokuphathelene nokuphathwa kwememori ye-TTM, i-DRM enkulu, njll.
I-AMD ilungiselela izinsiza eziningi zeKernel 4.20
Umsebenzi othunyelwe ngu-Alex Deucher ovela ku-AMD ufaka:
Ukusekelwa kwasekuqaleni kwemidwebo ye-APU Picasso kulinde ukukhishwa.
Ama-AMD APUs asanda kuvula amathokheni awo we-AMDGPU ala ma-APU amasha asekelwe eVega / GFX9 ekuqaleni kwesonto.
Siyaqhubeka nokusebenza ekwakheni iVega 20 ethokozisayo, okumele yenziwe ngaphambi kokuphela konyaka.
Kubukeka sengathi ngalokhu kukhishwa kwe-kernel okuzayo, ukuxhaswa kweVega 20 kuzobhekwa konke futhi kube nokuma okuhle ngesikhathi sokwethulwa. Kulindeleke kabanzi ukuthi le Vega 20 GPU yokuqala ingumkhiqizo wesiteshi sokusebenzela esingu-7nm.
Okuhlobene ne-Vega 20, I-AMDGPU manje inokusekelwa kokuqala kwe-xGMI ngaphakathi komshayeli we-kernel.
I-XGMI ukuxhumana okusha iVega 20 namaGPUs esizayo azokuxhasa njengenye indlela ye-PCI Express 3.0 / 4.0.
Ikhodi ye-AMDKFD ihlanganiswa ne-AMDGPU ukuze ibe yimodyuli ye-kernel.
Umsebenzi obonakala ungenakho ukuphathwa kwamandla wokugcina uyaqhubeka nalo mjikelezo, wenza ngcono i-ACP yokuvuselelwa kwamandla nezinye izinguquko.
Kukhona nenani elihle lokuxhaswa kokubonisa, kufaka phakathi ukuxhaswa kwe-ABGR / XBGR, ukuxhaswa kwe-DisplayPort YCbCr ngekhodi Ukuboniswa kwe-DC, ukusekelwa kwe-LVDS kukhodi ye-DC, izici ezintsha zokulungisa amaphutha nomunye umsebenzi.
- Ukusekelwa kokuchazwa kabusha okuncane kwe-shaders ngokubulala ukubulala kwe-GFX / Compute.
- Ukusekelwa kwenjini ye-VCN JPEG yama-Raven Ridge APU nokusha. Futhi ngaphambili kweRaven kukhona isitendi sokulayisha i-DMCU firmware.
- IRaven Ridge APU manje futhi inokwesekwa kwe-GFXOFF ukucisha injini yehluzo lapho ingadingeki futhi usekele imodi yokungingiza.
- Ukusekelwa kokulinganiswa komthwalo ngaphakathi kwesihleli se-AMDGPU / DRM nokuyisebenzisela uhlelo lokuhlela.
- Ukuthuthuka kokusebenza kwememori ebonakalayo ye-GPUVM. Kukhona nokunye ukuphatha ukuthuthuka kokusebenza kahle kwi-GPUVM LRU.
Kukhona nokunye ukulungiswa kwe-TTM nokusekelwa kokunyakaza okukhulu ngaphakathi kwaleyo khodi yokuphathwa kwememori.
Sekukonke, bekungumjikelezo omatasatasa wabathuthukisi be-AMD ukulungiselela ukusekelwa kweVega 20 nezinsizakusebenza ezihlobene, ezifana ne-xGMI, ukuthola ukwesekwa kwemidwebo yokuqala yeRaven2 nePicasso, kusathuthukiswa ukuphathwa kwamandla ne-AMDGPU DC, ngaphezu kokunye ukusebenza.