 Matt Arsenault
|
b05e4778f0
Introduce getSelect usage and use more getSelectCC
|
12 years ago |
 Benjamin Kramer
|
e4fae84b0b
Remove double semicolons.
|
12 years ago |
 Andrew Trick
|
ac6d9bec67
Track IR ordering of SelectionDAG nodes 2/4.
|
12 years ago |
|
Matt Arsenault
|
225ed7069c
Add LLVMContext argument to getSetCCResultType
|
12 years ago |
|
Matt Arsenault
|
798925bac7
Fix vselect when getSetCCResultType returns a different type from the operands
|
12 years ago |
 Nadav Rotem
|
d99a5a3ab4
SelectionDAG compile time improvement.
|
12 years ago |
 Michael Liao
|
eedff3547d
Fix PR15267
|
12 years ago |
 Preston Gurd
|
ea387fc3b8
This patch aims to reduce compile time in LegalizeTypes by using SmallDenseMap,
|
12 years ago |
|
Benjamin Kramer
|
4dc478308f
When lowering an inreg sext first shift left, then right arithmetically.
|
12 years ago |
|
Nadav Rotem
|
66de2af815
PPC: Implement efficient lowering of sign_extend_inreg.
|
12 years ago |
 Patrik Hagglund
|
319bb39923
Change TargetLowering::getTypeToPromoteTo to take and return MVTs,
|
12 years ago |
|
Patrik Hagglund
|
88ef514cc6
Change TargetLowering::getTruncStoreAction to take MVTs, instead of EVTs.
|
12 years ago |
|
Patrik Hagglund
|
34525f9ac0
Revert EVT->MVT changes, r169836-169851, due to buildbot failures.
|
12 years ago |
|
Patrik Hagglund
|
bb2543bb0e
Change TargetLowering::getTypeToPromoteTo to take and return MVTs,
|
12 years ago |
|
Patrik Hagglund
|
3166283ac1
Change TargetLowering::getTruncStoreAction to take MVTs, instead of EVTs.
|
12 years ago |
 Eli Friedman
|
43147afd71
Mark FP_EXTEND form v2f32 to v2f64 as "expand" for ARM NEON. Patch by Pete Couperus.
|
12 years ago |
|
Eli Friedman
|
846ce8ea67
Mark FP_ROUND for converting NEON v2f64 to v2f32 as expand. Add a missing
|
12 years ago |
|
Nadav Rotem
|
ee77da6b28
Fix a typo.
|
13 years ago |
|
Nadav Rotem
|
f55ef64544
Generate better select code by allowing the target to use scalar select, and not sign-extend.
|
13 years ago |
 Pete Cooper
|
d906017c1a
Only legalise a VSELECT in to bitwise operations if the vector mask bool is zeros or all ones. A vector bool with just ones isn't suitable for masking with.
|
13 years ago |
|
Nadav Rotem
|
e757f00446
|
13 years ago |
 Craig Topper
|
6b1e1d8b3d
Add FMA to switch statement in VectorLegalizer::LegalizeOp so that it can be expanded when it isn't legal.
|
13 years ago |
 Jim Grosbach
|
926dc168c8
'Promote' vector [su]int_to_fp should widen elements.
|
13 years ago |
 Nadav Rotem
|
3ab32ea49e
When emulating vselect using OR/AND/XOR make sure to bitcast the result back to the original type.
|
13 years ago |
|
Craig Topper
|
5e25ee8a1f
Convert assert(0) to llvm_unreachable
|
13 years ago |
 Chandler Carruth
|
63974b2144
Initial CodeGen support for CTTZ/CTLZ where a zero input produces an
|
13 years ago |
|
Nadav Rotem
|
815af82b74
Improve code generation for vselect on SSE2:
|
14 years ago |
|
Nadav Rotem
|
fbf19ef186
Fix a bug in the legalization of vector anyext-load and trunc-store. Mem Index starts with zero.
|
14 years ago |
 Duncan Sands
|
17001ce25c
Fix a bunch of unused variable warnings when doing a release
|
14 years ago |
 Chad Rosier
|
60655413ce
Removed set, but unused variable.
|
14 years ago |
