[llvm-commits] [PATCH, RFC] Fix PR13891 (AliasChain not properly maintained in ScheduleDAGInstrs::buildSchedGraph())

[Sergei Larin]

Bill,

  Sorry it is taking me a while to review this. At the first glance I am somewhat skeptical, but then, I need more time to look into it.
I have a surgery scheduled for later today, so I still might not get a chance to look at it more carefully.

Maybe Andy can spare some cycles in the mean time...? Sorry for the delay again...

Sergei

---
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation


> -----Original Message-----
> From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-
> bounces at cs.uiuc.edu] On Behalf Of William J. Schmidt
> Sent: Friday, September 21, 2012 3:35 PM
> To: llvm-commits at cs.uiuc.edu
> Subject: [llvm-commits] [PATCH, RFC] Fix PR13891 (AliasChain not
> properly maintained in ScheduleDAGInstrs::buildSchedGraph())
> 
> PR13891 identifies a test case where a load and store that reference
> the same memory location may be improperly interchanged.  Investigation
> showed this was due to a lost aliasing dependence.  If the only
> dependence between two instructions is a may-alias via the AliasChain,
> that dependence may be missed because the head of the AliasChain is not
> always properly updated.  This patch fixes that.
> 
> There was one regression failure in test/CodeGen/ARM/vstlane.ll.  I
> found that the code gen was identical before and after the patch,
> modulo one change in assigned registers.  I modified the test to be
> less dogmatic about which register is assigned in this case.
> 
> test/CodeGen/PowerPC/pr13891.ll is a new test that verifies the fix.
> 
> No additional regressions in llvm/tests or in llvm/projects/test-suite.
> Is this OK to commit?
> 
> Thanks,
> Bill
> 
> --
> Bill Schmidt, Ph.D.
> IBM Advance Toolchain for PowerLinux
> IBM Linux Technology Center
> wschmidt at linux.vnet.ibm.com
> wschmidt at us.ibm.com