On 15/05/2015 14:43, Richard Henderson wrote: > On 05/15/2015 05:09 AM, Leon Alrae wrote: >> On 14/05/2015 20:12, Richard Henderson wrote: >>> /* We know both pages are present and writable. */ >>> if (eaddr == baddr + 15) { >>> /* Consecutive pages in RAM. */ >>> memcpy(baddr, register, 16); >>> } else { >>> /* Someone's doing an MSA store to device memory. */ >>> for (i = 0; i < 2; ++i) { >>> helper_ret_stq_mmu(env, vaddr + i*8, register.d[0], >>> make_memop_idx(MO_UNALN | MO_TEQ, mmu_idx), >>> GETRA()); >>> } >>> } >> >> We would additionally need to take care of vector elements' endianness >> before using this code. Therefore always using helper_ret_st[bwlq]_mmu >> (depending on data format) in a loop probably is simpler. > > I suppose. I'd thought one of the patches defined functions to do the > swapping, so I assumed that would still be used.
IIUC they were required because of the byte access workaround, but now with the MO_UNALN support in TCG (thanks for this!) neither the workaround nor the swapping is needed for MSA I think. Leon