Two days old build breakage on i386.
marlowsd at gmail.com
Fri Jun 27 11:27:50 UTC 2014
On 27/06/2014 12:23, Johan Tibell wrote:
> On Fri, Jun 27, 2014 at 1:14 PM, Simon Marlow <marlowsd at gmail.com> wrote:
>> The problem is that this instruction requires three separate registers, but
>> cmpxchgl already reads and writes %eax leaving only two free registers (%ecx
>> and %edx).
>> You'll need to arrange to not use the complicated addressing modes with
>> cmpxchg on i386, and keep the number of free regs required <= 2.
> Where's the best place to arrange for that? If I switch from using
> getAmode to computing the address into a temp register in StgCmmPrim
> will that ensure that the address is simple, or could some
> optimization replace my temp register computation with a complex
> address again?
There aren't any optimisations that happen on the instructions after
codegen, and if there were, they would have to respect the same rule.
So using a register is the right thing, yes.
More information about the ghc-devs