[DAG] SDPatternMatch - add m_SpecificFP matcher #167438
Conversation
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
from
bcd3afc to
485284a
Compare
There was a problem hiding this comment.
You need to use m_c_BinOp if you want to test for commutation - otherwise you need to swap the operand matchers (or you can add another test and do both!)
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
2 times, most recently
from
1229219 to
0c9d220
Compare
🐧 Linux x64 Test Results
|
There was a problem hiding this comment.
| if (ConstantFPSDNode *C = isConstOrConstSplatFP(V, /*AllowUndefs=*/true)) |
There was a problem hiding this comment.
Sorry, did get the question. I guess you forgot to link the code snippet
There was a problem hiding this comment.
Sorry, I meant why would I use m_SpecificFP(SDValue) and not just m_Specific(SDValue)?
There was a problem hiding this comment.
Ahh. Ok, I understood. Ok, I removed m_SpecificFP(SDValue)
There was a problem hiding this comment.
TBH I'd prefer to use a APFloat for matching instead of converting to/from double
@mshockwave thoughts?
There was a problem hiding this comment.
yeah we can provide a double version of m_SpecificFP (which you did) but for the matcher class I prefer using APFloat as well. SpecificInt_match also uses APInt.
(also, why std::optional<double>?)
There was a problem hiding this comment.
yeah we can provide a double version of m_SpecificFP (which you did) but for the matcher class I prefer using APFloat as well.
Done
(also, why std::optional?)
removed. That was an artifact from a previous impl.
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
from
0c9d220 to
7771c1b
Compare
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
4 times, most recently
from
b185199 to
b7fc120
Compare
|
@RKSimon @mshockwave could you have a second look? (The re-review button doesn't show up in my case) |
RKSimon
left a comment
RKSimon
left a comment
There was a problem hiding this comment.
Add some EXPECT_FALSE negative tests and m_SpecificFP(double) test coverage
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
from
b7fc120 to
119fb77
Compare
ravil-mobile
force-pushed
the
ravil/sd-specific-fp
branch
from
119fb77 to
55f6ac8
Compare
RKSimon
changed the title
|
@RKSimon thanks for the review. Do we have to wait for @mshockwave or can I just merge? |
|
Ok to merge. Cheers |
3 participants
This patch introduces SpecificFP matcher for SelectionDAG nodes.
This includes:
Adding SpecificFP_match() in SDPatternMatch.h.
Adding test coverage in SelectionDAGPatternMatchTest.cpp.
Closes #165566