>>> On 20.01.15 at 18:57, <andrew.coop...@citrix.com> wrote: > On 20/01/15 11:07, Jan Beulich wrote: >> @@ -1761,15 +1761,15 @@ int nvmx_handle_vmwrite(struct cpu_user_ >> vmcs_encoding = reg_read(regs, decode.reg2); >> __set_vvmcs(nvcpu->nv_vvmcx, vmcs_encoding, operand); >> >> - switch ( vmcs_encoding ) >> + switch ( vmcs_encoding & ~VMCS_HIGH(0) ) > > While this is functionally fine, it is quite odd to read. One option > would be to use > > case IO_BITMAP_A: case VMCS_HIGH(IO_BITMAP_A): > > but I am not sure whether a compiler could optimise that as well as > masking the bottom bit out and halfing the number of case statements.
A good compiler certainly could, but I prefer to make such optimizations explicit (and hence compiler independent) in the source code. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
