Page MenuHomePhabricator

dean (Dean De Leo)
User

Projects

User does not belong to any projects.

User Details

User Since
Jul 6 2015, 7:42 AM (334 w, 4 d)

Recent Activity

Jan 13 2016

dean added a comment to D14900: [mips] SelectionDAGISel subclasses now follow the optimization level..

not sure, has this functionality been merged? If not, is not it anymore needed?

Jan 13 2016, 7:24 AM

Nov 12 2015

dean added a comment to D14111: Use "_$" prefix instead of "$" for dynamic checker function inserted by LLDB during expression evaluation.

Hi,
is there any update related to this patch? I don't see any activity from the last two weeks.

Nov 12 2015, 4:20 AM

Oct 13 2015

dean added a comment to D13511: [mips] Use correct frame register for DWARF info when dynamically realigning the stack..

Hi, I actually confirm the patch solves the issue. Apologies for my previous wrong comment, the failure was due to my test. Though a different question, in the dwarf info the matches between the line numbers in the sources and the addresses in the assembly are not much contiguous, with the statements in the source interspersed at different places in the assembly. In general, the assembly looks somewhat optimised. This makes a bit harder to debug the code. Is there something that can be mitigated to obtain a more direct and contiguous correspondence between the original source and the assembly?

Oct 13 2015, 5:06 AM
dean added a comment to D13511: [mips] Use correct frame register for DWARF info when dynamically realigning the stack..

Hi, sorry I was off yesterday. You've got me, I was assuming you altered the frame base for the function to the $sp. Though I recall I tried with the debugger to print the values and still I've got some of them incorrect. I'll check today what is going wrong.

Oct 13 2015, 12:43 AM

Oct 9 2015

dean added a comment to D13511: [mips] Use correct frame register for DWARF info when dynamically realigning the stack..

Hi, thank you for looking into this issue. I've just checked the patch with the tip of llvm but doesn't solve the reported bug. The output is the same as the one reported in the original bug. Is it expected that it requires some additional changes?

Oct 9 2015, 9:21 AM

Sep 23 2015

dean added a comment to D13104: Mips - Mark the section .eh_frame as writeable for pic.

@danalbert: thanks for the clarification. I've tested this with the bfd. In Android, the mclinker is currently invoked by the renderscript runtime, will not this be the case in the future?

Sep 23 2015, 10:34 AM · Restricted Project
dean added a comment to D13104: Mips - Mark the section .eh_frame as writeable for pic.

yes, I also see it as a workaround until the problem will not be properly addressed. I do not get whether this must be a responsibility of the linker, I was expecting with PIC there should be no absolute addresses in the code in the first place, can you confirm this is not the case? Even if it was, this should be resolved somewhere else in the compiler, my aim is to have a way to proceed until is not.

Sep 23 2015, 10:28 AM · Restricted Project
dean updated the diff for D13104: Mips - Mark the section .eh_frame as writeable for pic.

Source indentation

Sep 23 2015, 9:31 AM · Restricted Project
dean retitled D13104: Mips - Mark the section .eh_frame as writeable for pic from to Mips - Mark the section .eh_frame as writeable for pic.
Sep 23 2015, 9:24 AM · Restricted Project

Sep 18 2015

dean retitled D12966: Renderscript plugin support for aarch64 from to Renderscript plugin support for aarch64.
Sep 18 2015, 8:46 AM

Aug 4 2015

dean added a comment to D11270: Expr evaluation - Avoid to check when the memory handle is zero.

Hi, as friendly *bump*, can you please provide any feedback regarding this patch? Let me know whether further details or info are required...

Aug 4 2015, 6:49 AM

Jul 16 2015

dean retitled D11270: Expr evaluation - Avoid to check when the memory handle is zero from to Expr evaluation - Avoid to check when the memory handle is zero.
Jul 16 2015, 9:57 AM