I would like to add that the kms_swrast is now working already with softpipe so if understood correctly the gist from Mauro's email the problems were with swrast and not specifically llvmpipe.
However, all the kms_swrast related patches are not yet in mesa3d upstream so that might make things a bit harder. The patches are in our fork that is used with the Replicant 9 work. Joonas Kylmälä: > Hi dllud, > > thanks for the research! As we can see in > https://github.com/maurossi/mesa/commit/f7c2a171a4320a5a713861440f5f3df2611d7b1e > the change in Android build files to make llvmpipe work might be > relatively small. Because this doesn't appear a to be huge amount of > work and maintanance burden I would try this first and if things don't > work out then we could try SwiftShader. > > Joonas > > dllud: >> Hi everybody, >> >> Mauro Rossi from the Android-x86 project had the kindness to give a detailed >> reply to my inquiry about llvmpipe's status on Android builds. >> >> Since all this is public information I am forwarding Mauro's email bellow >> for future reference. >> >> Summing it up: llvmpipe is not building for Android and there's not a single >> project using it. Replicant will/would be the only project using llvmpipe on >> Android. The burden of maintaining it would thus come down to Replicant for >> the foreseeable future. >> >> It is worth considering having all software rendering done just through >> Google's SwiftShader. Keeping Mesa around just to use Lima or Panfrost. >> Some people, me included, are not comfortable with relying solely on a >> Google project. Past experience (libAGL) told us that it can quickly get >> deprecated. >> I will rise this discussion on IRC. > _______________________________________________ > Replicant mailing list > Replicant@osuosl.org > https://lists.osuosl.org/mailman/listinfo/replicant > _______________________________________________ Replicant mailing list Replicant@osuosl.org https://lists.osuosl.org/mailman/listinfo/replicant
