baggins183
baggins183
Thanks for working on this! The readconstbuffer thing is weird, because IIRC when reading indirect user data/SRTs only S_LOAD_DWORD variants should be used, not S_BUFFER_LOAD_DWORD. Also I dont think bloodborne...
/*0000000002f8*/ s_buffer_load_dwordx8 s[0:7], s[16:19], 0xc ... /*000000000334*/ s_branch .L1308_0 .L824_0: /*000000000338*/ v_mul_lo_i32 v0, v13, 3 /*000000000340*/ s_load_dwordx4 s[0:3], s[6:7], 0xc I looked at the shader and it looks like the...
Ive had random problems debugging using gdb/lldb when compiling with clang, and better luck with gcc. So I vote for having a gcc action
> I think just a job to verify it compiles would be enough, and we can skip packaging and uploading for those jobs and remove from pre-release dependencies. This sounds...
> Seems like a lot of normal games are running the SRT code this is intentional. Any shader that uses indirection of a pointer stored in user data sgprs will...
Thanks for testing
the mov edi, dword ptr [rdi+30h] looks wrong (dst should be rdi). also the rest of the movs should be dword, but these are probably harmess Xbyak is doing the...
last commit might have fixed the edi thing
nah i dont mind. Its definitely appreciated
~there might be a system dependent issue with xbyak for moving a 64 bit immediate to a register because the initial PushPtr works on my os/cpu and should match the...