> On June 28, 2013, 7:51 a.m., Nilay Vaish wrote:
> > src/arch/x86/isa/insts/x87/control/save_and_restore_x87_environment.py, 
> > line 106
> > <http://reviews.gem5.org/r/1917/diff/2/?file=36103#file36103line106>
> >
> >     Why do we need these eret instructions?

If I understand things correctly, you need to execute eret to return from a 
microcode sequence (if it's a separate microcode sequence in the microcode 
rom). There is a new version of this patch that doesn't use microcode rom (I 
suspected there were issues related to exceptions in microcode, but I'm unsure) 
in my fixes-x87 branch on GitHub 
(https://github.com/andysan/gem5/tree/fixes-x87).


- Andreas


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/1917/#review4483
-----------------------------------------------------------


On June 13, 2013, 6:24 a.m., Andreas Sandberg wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/1917/
> -----------------------------------------------------------
> 
> (Updated June 13, 2013, 6:24 a.m.)
> 
> 
> Review request for Default and Gabe Black.
> 
> 
> Repository: gem5
> 
> 
> Description
> -------
> 
> Changeset 9760:f7df7b862186
> ---------------------------
> x86: Add support for FLDENV & FNSTENV
> 
> 
> Diffs
> -----
> 
>   src/arch/x86/isa/microasm.isa 9df73385c878 
>   src/arch/x86/isa/insts/x87/control/save_and_restore_x87_environment.py 
> 9df73385c878 
>   src/arch/x86/isa/decoder/x87.isa 9df73385c878 
> 
> Diff: http://reviews.gem5.org/r/1917/diff/
> 
> 
> Testing
> -------
> 
> Quick regressions pass. Minor stat differences in full regressions (to be 
> expected since some instructions aren't no-ops anymore). Solves a bug 
> triggered by switching between a simulated CPU that tries to save the x87 
> state and a virtualized CPU that restores the state (which is completely 
> bogus unless FNSTENV is implemented).
> 
> 
> Thanks,
> 
> Andreas Sandberg
> 
>

_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to