Commits · 7417348a7e01e8ec063e39a8d22e40b1767eebc6 · Roger Ferrer / llvm-epi-0.8

Apr 12, 2008
- 80 col fix · 7417348a
  Nate Begeman authored Apr 12, 2008
```
llvm-svn: 49569
```
  7417348a
Apr 10, 2008
- add a note, this is actually not too bad to implement. · aeb23a8a
  Chris Lattner authored Apr 10, 2008
```
llvm-svn: 49466
```
  aeb23a8a
- move the x86-32 part of PR2108 here. · c6921880
  Chris Lattner authored Apr 10, 2008
```
llvm-svn: 49465
```
  c6921880
- Fix the x86-64 side of PR2108 by adding a v2f64 version of · ad753024
  Chris Lattner authored Apr 10, 2008
```
MOVZQI2PQIrr.  This would be better handled as a dag combine 
(with the goal of eliminating the bitconvert) but I don't know
how to do that safely.  Thoughts welcome.

llvm-svn: 49463
```
  ad753024
Apr 09, 2008
- Make isVectorClearMaskLegal's operand list const. · 33b33001
  Dan Gohman authored Apr 09, 2008
```
llvm-svn: 49446
```
  33b33001
- Add XMM1 as a second return value register for f32 and f64 on x86-64. This · 3d074a31
  Dan Gohman authored Apr 09, 2008
```
is needed for the x86-64-ABI handling of structs that contain floating-point
members that are returned by value.

llvm-svn: 49441
```
  3d074a31
- Add DX as a second return value register for i16 on x86. · cbf87313
  Dan Gohman authored Apr 09, 2008
```
llvm-svn: 49440
```
  cbf87313
Apr 08, 2008
- Handle the situation in 2008-01-25-EmptyFunction.ll · fe767621
  Dale Johannesen authored Apr 08, 2008
```
correctly when unwind info is being generated.

llvm-svn: 49366
```
  fe767621
- Implement new llc flag -disable-required-unwind-tables. · 344aec29
  Dale Johannesen authored Apr 08, 2008
```
Corresponds to -fno-unwind-tables (usually default in gcc).

llvm-svn: 49361
```
  344aec29
Apr 07, 2008
- Rename MemOperand to MachineMemOperand. This was suggested by · 3bc3ddd6
  Dan Gohman authored Apr 07, 2008
```
review feedback from Chris quite a while ago. No functionality
change.

llvm-svn: 49348
```
  3bc3ddd6
- Re-commit of the r48822, where the infinite looping problem discovered · 51f532f9
  Roman Levenstein authored Apr 07, 2008
```
by Dan Gohman is fixed.

llvm-svn: 49330
```
  51f532f9
Apr 06, 2008

API changes for class Use size reduction, wave 1. · e9ecc68d

Gabor Greif authored Apr 06, 2008

Specifically, introduction of XXX::Create methods
for Users that have a potentially variable number of
Uses.

llvm-svn: 49277

e9ecc68d

Apr 05, 2008
- Favors pshufd over shufps when shuffling elements from one vector. pshufd is faster than shufps. · f77b5ef3
  Evan Cheng authored Apr 05, 2008
```
llvm-svn: 49244
```
  f77b5ef3
Apr 03, 2008
- Re-enable SSE4. · 6c66bd36
  Evan Cheng authored Apr 03, 2008
```
llvm-svn: 49158
```
  6c66bd36
- Fix x86-64 encoding bug. REX prefix must always follow 0x0F prefix. For... · 6db4b4cc
  Evan Cheng authored Apr 03, 2008
```
Fix x86-64 encoding bug. REX prefix must always follow 0x0F prefix. For example, extractps in 64bit mode: 66 REX 0F 3A 17, not 66 0F 3A REX 17.

llvm-svn: 49157
```
  6db4b4cc
- Cosmetic · d9129d1d
  Evan Cheng authored Apr 03, 2008
```
llvm-svn: 49156
```
  d9129d1d
- Temporarily disabling SSE4 until we fix the encoding issues. · 3063c554
  Evan Cheng authored Apr 03, 2008
```
llvm-svn: 49129
```
  3063c554
- Backing out 48222 temporarily. · 025cea11
  Evan Cheng authored Apr 03, 2008
```
llvm-svn: 49124
```
  025cea11
Apr 02, 2008

Cosmetic changes per EH patch review feedback. · 8780ecbb
Dale Johannesen authored Apr 02, 2008
```
llvm-svn: 49096
```
8780ecbb
Add new CC lowering rule: provide a list of registers, which can be 'shadowed', · 20c9e4cb
Anton Korobeynikov authored Apr 02, 2008
```
when some another register is used for argument passing.
Currently is used on Win64.

llvm-svn: 49079
```
20c9e4cb

Recommitting EH patch; this should answer most of the · fd967cf3

Dale Johannesen authored Apr 02, 2008

review feedback.
-enable-eh is still accepted but doesn't do anything.
EH intrinsics use Dwarf EH if the target supports that,
and are handled by LowerInvoke otherwise.
The separation of the EH table and frame move data is,
I think, logically figured out, but either one still
causes full EH info to be generated (not sure how to
split the metadata correctly).
MachineModuleInfo::needsFrameInfo is no longer used and
is removed.

llvm-svn: 49064

fd967cf3

ReMat of load from stub in pic mode extends the life of pic base. Currently... · b86595fb

Evan Cheng authored Apr 01, 2008

ReMat of load from stub in pic mode extends the life of pic base. Currently spiller doesn't do a good job of estimating the impact. Disable for now.

llvm-svn: 49059

b86595fb

Apr 01, 2008
- Remove unnecessary and non-deterministic checking code. Re-enable remat of load from gv stub. · 19a6dd9f
  Evan Cheng authored Apr 01, 2008
```
llvm-svn: 49054
```
  19a6dd9f
- Don't use __bzero for memset if the second argument isn't zero. · cb9f8f6e
  Dan Gohman authored Apr 01, 2008
```
llvm-svn: 49050
```
  cb9f8f6e
- Speculatively micro-optimize memory-zeroing calls on Darwin 10. · 980d7200
  Dan Gohman authored Apr 01, 2008
```
llvm-svn: 49048
```
  980d7200
- Revert 49006 for the moment. · 5e4e051c
  Dale Johannesen authored Apr 01, 2008
```
llvm-svn: 49046
```
  5e4e051c
- Disabling remat of load from gv stub (temporarily) again to fix llvmgcc bootstrap miscompare. · 306e3dcf
  Evan Cheng authored Apr 01, 2008
```
llvm-svn: 49037
```
  306e3dcf
- Accept 'y' constraint (MMX) in inline asm. · efa81a69
  Dale Johannesen authored Apr 01, 2008
```
llvm-svn: 49011
```
  efa81a69
- Emit exception handling info for functions which are · 7d02cf3c
  Dale Johannesen authored Mar 31, 2008
```
not marked nounwind, or for all functions when -enable-eh
is set, provided the target supports Dwarf EH.

llvm-gcc generates nounwind in the right places; other FEs
will need to do so also.  Given such a FE, -enable-eh should
no longer be needed.

llvm-svn: 49006
```
  7d02cf3c
- It's not safe to fold a load from GV stub or constantpool into a two-address use. · e4f77c69
  Evan Cheng authored Mar 31, 2008
```
llvm-svn: 49002
```
  e4f77c69
Mar 31, 2008
- Move reMaterialize() from TargetRegisterInfo to TargetInstrInfo. · ed6e34fe
  Evan Cheng authored Mar 31, 2008
```
llvm-svn: 48995
```
  ed6e34fe
- Re-apply 48911. · 1973a46c
  Evan Cheng authored Mar 31, 2008
```
llvm-svn: 48977
```
  1973a46c
Mar 29, 2008
- Fix a tokenfactor node to use the load chain rather than the · fd2eb00c
  Dan Gohman authored Mar 28, 2008
```
load value. This fixes PR2177.

llvm-svn: 48932
```
  fd2eb00c
Mar 28, 2008
- Backing out 48911 for now. It's breaking stuff. · b8654202
  Evan Cheng authored Mar 28, 2008
```
llvm-svn: 48922
```
  b8654202
- New entry. · 81e0c9a3
  Evan Cheng authored Mar 28, 2008
```
llvm-svn: 48912
```
  81e0c9a3
- Load from stub is already re-materializable. · 9ae4d7b7
  Evan Cheng authored Mar 28, 2008
```
llvm-svn: 48911
```
  9ae4d7b7
Mar 27, 2008
- Code clean up. · 308e5646
  Evan Cheng authored Mar 27, 2008
```
llvm-svn: 48856
```
  308e5646
- Allow certain lea instructions to be rematerialized. · 29e62a59
  Evan Cheng authored Mar 27, 2008
```
llvm-svn: 48855
```
  29e62a59
- Remove an unused command line option. · 4fb07c65
  Evan Cheng authored Mar 27, 2008
```
llvm-svn: 48854
```
  4fb07c65
Mar 26, 2008

Use a linked data structure for the uses lists of an SDNode, just like · 358e04a1

Roman Levenstein authored Mar 26, 2008

LLVM Value/Use does and MachineRegisterInfo/MachineOperand does.
This allows constant time for all uses list maintenance operations.

The idea was suggested by Chris. Reviewed by Evan and Dan.
Patch is tested and approved by Dan.

On normal use-cases compilation speed is not affected. On very big basic
blocks there are compilation speedups in the range of 15-20% or even better. 

llvm-svn: 48822

358e04a1