Commits · 1af37c2fca9faa904e53c2d8a9273733e5578eeb · Roger Ferrer / llvm-epi-0.8

Nov 19, 2008
- Add support for rematerialization in pre-alloc-splitting. · 1af37c2f
  Owen Anderson authored Nov 19, 2008
```
llvm-svn: 59587
```
  1af37c2f
Nov 18, 2008

Fix a bug introduced by my previous patch. With this change, SPEC is now... · 896512af

Owen Anderson authored Nov 18, 2008

Fix a bug introduced by my previous patch.  With this change, SPEC is now clean with prealloc splitting enabled.

llvm-svn: 59544

896512af

Nov 13, 2008
- Don't allow the restore point to be placed after terminators. With this change, · d36b4c1b
  Owen Anderson authored Nov 13, 2008
```
MultiSource/Applications is clean with the prealloc splitter.  Some failures
remain in SPEC.

llvm-svn: 59267
```
  d36b4c1b
Nov 11, 2008
- Don't walk into predecessors in which the vreg is not live when doing shrinkwrapping. · fd395eb8
  Owen Anderson authored Nov 11, 2008
```
This lets several failing tests get farther along, but doesn't completely fix any of them.

llvm-svn: 59086
```
  fd395eb8
Nov 05, 2008
- Use the new predicate to control when we do prealloc splitting. Fix a small bug. · 26c10f1b
  Owen Anderson authored Nov 05, 2008
```
llvm-svn: 58738
```
  26c10f1b
Nov 04, 2008

First pass at checking for the creation of a new join point when doing... · a926ed4b

Owen Anderson authored Nov 04, 2008

First pass at checking for the creation of a new join point when doing pre-alloc splitting.  This is not turned on yet.

llvm-svn: 58726

a926ed4b

Nov 03, 2008
- Revert my last patch until I consult with Evan about it. · b0678438
  Owen Anderson authored Nov 03, 2008
```
llvm-svn: 58591
```
  b0678438
Nov 02, 2008

Don't do pre-splitting if doing so would create a value join that did not · 2bb2aaa7

Owen Anderson authored Nov 02, 2008

exist before.  Updating the live intervals in that care is tricky in the general
case.

Evan, if you see a tighter guard condition for this, let me know.

llvm-svn: 58560

2bb2aaa7

Oct 31, 2008
- Add a fixme. · d1f8c9f4
  Evan Cheng authored Oct 31, 2008
```
llvm-svn: 58514
```
  d1f8c9f4
Oct 29, 2008
- - More pre-split fixes: spill slot live interval computation bug; restore point bug. · ce3ccc1e
  Evan Cheng authored Oct 29, 2008
```
- If a def is spilt, remember its spill index to allow its reuse.

llvm-svn: 58375
```
  ce3ccc1e
- - Rewrite code that update register live interval that's split. · 938e2015
  Evan Cheng authored Oct 29, 2008
```
- Create and update spill slot live intervals.
- Lots of bug fixes.

llvm-svn: 58367
```
  938e2015
Oct 28, 2008
- If def is in the same mbb as the barrier, spilt the value after the last use before the barrier. · 5e88d903
  Evan Cheng authored Oct 28, 2008
```
llvm-svn: 58314
```
  5e88d903
- Add command line option to limit the number splits to help debugging. · 9f854697
  Evan Cheng authored Oct 28, 2008
```
llvm-svn: 58312
```
  9f854697
- Avoid putting a split past the end of the live range; always shrink wrap live... · fab31680
  Evan Cheng authored Oct 28, 2008
```
Avoid putting a split past the end of the live range; always shrink wrap live interval in the barrier mbb.

llvm-svn: 58309
```
  fab31680
- Silence a bogus compile time warning. · e67ad47a
  Evan Cheng authored Oct 27, 2008
```
llvm-svn: 58297
```
  e67ad47a
- Remove val# defined by a remat'ed def that is now dead. · f46642ad
  Evan Cheng authored Oct 27, 2008
```
llvm-svn: 58294
```
  f46642ad
Oct 27, 2008

For now, don't split live intervals around x87 stack register barriers.... · f7137229

Evan Cheng authored Oct 27, 2008

For now, don't split live intervals around x87 stack register barriers. FpGET_ST0_80 must be right after a call instruction (and ADJCALLSTACKUP) so we need to find a way to prevent reload of x87 registers between them.

llvm-svn: 58230

f7137229

Oct 26, 2008
- Do not shrink wrap live interval in a mbb if it's livein any of its successor... · ed033ede
  Evan Cheng authored Oct 26, 2008
```
Do not shrink wrap live interval in a mbb if it's livein any of its successor blocks. The mbb can be revisited again after all of the successors are processed.

llvm-svn: 58184
```
  ed033ede
- Handle cases where there aren't uses in the barrier mbb. · f48367b8
  Evan Cheng authored Oct 25, 2008
```
llvm-svn: 58174
```
  f48367b8
Oct 25, 2008

If val# def is ~0U, meaning it's defined by a PHI, and it's previously split,... · 85d71d45

Evan Cheng authored Oct 25, 2008

If val# def is ~0U, meaning it's defined by a PHI, and it's previously split, spill before the barrier because it's impossible to determine if all the defs are spilled in the same spill slot.

llvm-svn: 58129

85d71d45

Oct 24, 2008
- Fix a pasto. · cd160f98
  Evan Cheng authored Oct 24, 2008
```
llvm-svn: 58102
```
  cd160f98
- Fix a end() dereference; remove an abort() that wasn't meant to be left in. · f8b09d42
  Evan Cheng authored Oct 24, 2008
```
llvm-svn: 58072
```
  f8b09d42
- Avoid splitting an interval multiple times; avoid splitting re-materializable val# (for now). · 4bac4d0a
  Evan Cheng authored Oct 24, 2008
```
llvm-svn: 58068
```
  4bac4d0a
Oct 23, 2008

Committing a good chunk of the pre-register allocation live interval splitting... · d0d0317b

Evan Cheng authored Oct 23, 2008

Committing a good chunk of the pre-register allocation live interval splitting pass. It's handling simple cases and appear to do good things. Next: avoid splitting an interval multiple times; renumber registers when possible; record stack slot live intervals for coloring; rematerialize defs when possible.

llvm-svn: 58044

d0d0317b

Oct 20, 2008
- Add skeleton for the pre-register allocation live interval splitting pass. · 7e721ecd
  Evan Cheng authored Oct 20, 2008
```
llvm-svn: 57847
```
  7e721ecd