- User Since
- Jan 1 2019, 11:11 PM (54 w, 3 d)
@MaskRay Did you merge it to LLVM 10 branch?
Thu, Jan 16
Wed, Jan 15
Mon, Jan 13
Mon, Jan 6
A general comment. All the tests have been done with -malign-branch-prefix-size=5. I don't see any explicit reason to change the default prefix size from 5 to 4.
@ruiu LLVM 10.0.0 will be branched on Jan. 15. I would expect that the LLD CET support can be included in this release. If you have concern with current implementation, let's have an offline discussion and move it forward quickly.
Fri, Dec 20
Thu, Dec 19
To be concrete, I propose:
".autopad", ".noautopad": allow/disallow the assembler to emit padding via inserting a nop or prefix before any instruction, as needed.
".align_branch_boundary [N,] [instruction,]": Enable branch-boundary padding (per previous description).
Dec 17 2019
Dec 16 2019
Thanks for coordinating the meeting and having a clear summary. It helps a lot to accelerate the patch review. I really appreciate it!
Annita will refresh current patch with two key changes. 1) Drop prefix support and simplify and 2) drop clang driver support for now. Desire is to minimize cycle time before next iteration so that feedback on approach can be given while reviewers are still around.
More performance data was posted on http://lists.llvm.org/pipermail/llvm-dev/2019-December/137609.html and http://lists.llvm.org/pipermail/llvm-dev/2019-December/137610.html. Let's move on based on the data.
Dec 10 2019
Dec 9 2019
I have a general comment. I think the macro-fused jcc instructions should be taken into account even for NOP padding. From our experience, it's better to take the macro-fused instructions as an integrated instruction, and ensure it not cross or against 32-byte boundary by padding if necessary.
The point is that we have explicit requirement at the start and we have a lowering into 16-byte sequence that we need to be preserved exactly as it is.
Essentially what we need is a "protection" for this sequence from any changes by machinery that generates the binary code.
How can we protect a particular byte sequence from being changed by this branch aligner?
Dec 7 2019
Dec 6 2019
Third, I have not see a justification for why complexity for instruction prefix padding is necessary. All the effected CPUs support multi-byte nops, so we're talking about a *single micro op* difference between the nop form and prefix form. Can anyone point to a performance delta due to this? If not, I'd suggest we should start with the nop form, and then build the prefix form in a generic manner for all alignment varieties.
+1. Starting from just NOP padding sounds a simple and good first step. We can explore segment override prefixes in the future.
Dec 3 2019
Nov 14 2019
Nov 8 2019
Sounds good. Thank you!
Oct 21 2019
Oct 15 2019
I added a comment to x86-64 ABI google group link. I also posted it below.
May 28 2019
The current problem is the 2nd PLT ABI was in GCC side for more than 2 years. I think we need to work with GCC community to have a consistent ABI so we won't break each other. Let's have more discussion in x86_64 ABI site.