- User Since
- Nov 21 2016, 2:12 PM (230 w, 1 d)
Mar 3 2021
Mar 2 2021
Mar 1 2021
Follow-up patch is https://reviews.llvm.org/D97686
Some comments below, in addition to these questions
Feb 24 2021
Cool, I have a patch to update to the right specifiers, and to update the reference output where it has changed (005 and 007 have unsigned numbers being printed as signed, and showing up in the current output as negative).
Feb 23 2021
@afd Some nits about changing the printfs. I have so far presumed that where a number is unsigned, you want me to print it as unsigned (using "%u" or equivalent via inttypes.h), and where it is signed, you want me to print it using "%i" (or equivalent). It raised some questions, below.
I'll commit a fix that uses the inttypes macros.
Feb 19 2021
Yeah, I can land this for you, I'll try and do that today.
Feb 15 2021
Feb 13 2021
I did a quick look for what had caused this, and it seems it's related to the switch to the new pass manager, and a commit that was marked NFC and wasn't. I've asked the relevant part of the commit to be reverted.
Jan 26 2021
@luismarques @asb @mundaym Please can one of you take over this patch if lowRISC intends to keep working on it? I do realise that we weren't sure what the scheduling/performance payoff would be from this work, and it was showing some pretty obscure bugs.
Jan 19 2021
Jan 18 2021
LGTM. I'll land this for you today.
Jan 15 2021
Sorry, I meant to accept this before it landed.
Address @jrtc27's Feedback:
Jan 14 2021
- Add Variadic Testcase
- Provide correct datalayout
- Add clang support and tests
- Modify fp reservation code so we don't always use a frame pointer, even if we conservatively reserve it when using ilp32e + D extension.
There's one warning to fix (which should be trivial), and then I'm happy for this to land. Given the warning, the fix will hopefully be obvious?
The reasoning behind this seems solid, but I'd like someone from the webassembly backend to chime in.
We've found some issues in testing, so for the moment I've reverted this patch in rG7c9c2a2ea5e3760d7310309c96c9a4ce41fa4d9b.
Jan 13 2021
Jan 12 2021
- Add comments in line with review feedback.
I'm going to look into removing more of the xori … 1 which are used by the selects in some of these cases.
For a moment I thought this was doing what D94535 is, but more elegantly, but I now see it's for SELECT, not brcond.
@arsenm Please may you clarify why you think the scope is still useful to represent this late?