- User Since
- Sep 26 2016, 7:58 AM (63 w, 1 d)
Thu, Dec 7
Wed, Dec 6
Tue, Dec 5
Mon, Dec 4
Added some more tests.
Added a test case that checks that the build_pair -> combine load transform is done during ISel.
Sun, Dec 3
Thu, Nov 30
Tue, Nov 28
Mon, Nov 27
Sun, Nov 26
A BUNDLE instruction is supposed to describe the combined properties of the bundled instructions (at least when it comes to machine operands). Properties like hasSideEffects/mayLoad/mayStore are static, so we can't update those properties for the BUNDLE instruction depending on the properties of the bundled instructions. So maybe it is a good idea to have the defensive approach of letting the BUNDLE instruction have hasSideEffects=1.
Sat, Nov 25
Fri, Nov 24
I have no further comments. And it looks good from our out-of-tree-targets perspective.
In many situations I think this is NFC. For example the consecutive load/store optimizations have other checks verifying that the involved types are byte sized, so the end result will be the same.
But I haven't really reviewed the MIPS/Hexagon specific parts, and there are no test cases showing that something is more correct (I'm not sure all changes are NFC).
Thu, Nov 23
Nov 6 2017
Nov 2 2017
Nov 1 2017
Oct 26 2017
I get errors when compiling due to this commit:
Oct 25 2017
Oct 24 2017
Oct 22 2017
Add a FIXME as requested by Eli.
Oct 18 2017
Oct 17 2017
Oct 16 2017
If nobody remembers why this very old patch never was landed, then I guess I'll make a new attempt as it would avoid the asserts that I've seen.
And it is also one step towards getting rid of the default arguments in DataLayout::getPointerSizeInBits(unsigned AS = 0) , marked as a FIXME in DataLayout.h.
Oct 12 2017
Eli, I think this looks good.
Oct 11 2017
Does anyone remember why this has been "abandoned" (still not finished after four years being in need for review)?
Oct 9 2017
Oct 4 2017
Oct 3 2017
Oct 2 2017
Sep 28 2017
Sep 27 2017
Narrowed the test case even further by stopping already after virtregrewriter.
Sep 26 2017
Added REQUIRES asserts to the test case.
Added a second RUN line to also verify the MIR.
Sep 25 2017
Sep 21 2017
I'd still like to contribute by landing this bugfix. We are using it for our out-of-tree target and the problem seems to exist for other big-endian targets as well, as shown by the test case for powerpc.
Sep 19 2017
Nobody else seems to bother. At least no objections, so I think we can land this now.
Sep 13 2017
Some fixes related to comments from Adrian Prantl:
- Strip some attributes from the test case.
- Stop test after livedebugvalues (that made it possible to relate checks to positions in the fixedStack instead of load instructions).
- Reformulated the code comment that did not take stack growth direction into consideration.
- Removed the assert comparing frame index with for the "other" BUILD_PAIR operand. I think we can live without it, as it got a little bit more complicated when taking stack growth into account.
- Factored out the "LowAddressOp" number to enhance readability.
Sep 12 2017
This patch looks good to me now.
But I'm currently working in the same out-of-tree project as the author, so it would be nice if someone else could have a look as well and accept this.