https://bugs.kde.org/show_bug.cgi?id=386945
Mark Wielaard <m...@klomp.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #116593|0 |1
is obsolete| |
--- Comment #53 from Mark Wielaard <m...@klomp.org> ---
Created attachment 116742
--> https://bugs.kde.org/attachment.cgi?id=116742&action=edit
Option1: Implement ldbrx as 64-bit load and Iop_Reverse8sIn64_x1
(In reply to Julian Seward from comment #52)
> Personally I'd vote for (1) because (a) it's easier and (b) I am not sure
> that Memcheck will work correctly when presented with opposite-from-the-host
> endian loads/stores.
Lets go with option 1. That is the code that this patch implements.
Changes from the previous version are:
1) Added a comment that there is an earlier check that ldbrx is only used in
64bit mode.
2) Added a comment explaining what we could do if we went with option 2.
3) Add a vassert(mode64) for the usage of Iop_Reverse8sIn64_x1 in
host_ppc_isel.c.
4) Simplified the last part of Iop_Reverse8sIn64_x1 by removing the unnecessary
masking as suggested by Julian.
This make the ldbrx-aligned testcase work as expected.
--
You are receiving this mail because:
You are watching all bug changes.
