Commits · 5a6199f3872af8dd9eecea7529dd7cb12ee59958 · Roger Ferrer / llvm-epi-0.8

Sep 30, 2005
- Change this code ot pass register classes into the stack slot spiller/reloader · 5a6199f3
  Chris Lattner authored Sep 30, 2005
```
code.  PrologEpilogInserter hasn't been updated yet though, so targets cannot
use this info.

llvm-svn: 23536
```
  5a6199f3
- Use the 32-bit version for now · 08f157c5
  Chris Lattner authored Sep 30, 2005
```
llvm-svn: 23534
```
  08f157c5
- Add a bunch of patterns for F64 FP ops, add some more integer ops · 027a2671
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23533
```
  027a2671
- Remove code for patterns that are autogenerated · 1de5706e
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23532
```
  1de5706e
- begining alpha subtarget support · a7a83b92
  Andrew Lenharth authored Sep 29, 2005
```
llvm-svn: 23531
```
  a7a83b92
- tblgen autogens this pattern now · 0a1cd715
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23530
```
  0a1cd715
Sep 29, 2005
- copy and paste error · bae1f9d7
  Andrew Lenharth authored Sep 29, 2005
```
llvm-svn: 23528
```
  bae1f9d7
- now that tblgen is smarter, this pattern is not needed. Also, tblgen · a748e3ae
  Chris Lattner authored Sep 29, 2005
```
now inverts commuted versions of ANDC/ORC with the current .td file.

llvm-svn: 23527
```
  a748e3ae
- Insert stores after phi nodes in the normal dest. This fixes · a554c947
  Chris Lattner authored Sep 29, 2005
```
LowerInvoke/2005-08-03-InvokeWithPHI.ll

llvm-svn: 23525
```
  a554c947
- consistency with other cases, no functionality change · 02d3ba3d
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23524
```
  02d3ba3d
- Make the JIT default to the DAG isel instead of the pattern isel, like LLC. · eca4f566
  Chris Lattner authored Sep 29, 2005
```
The Pattern isel has some strange memory corruption issues going on. :(

This should have been converted over anyway, but it got forgotten somehow
when switching to the dag isel.

llvm-svn: 23523
```
  eca4f566
- Fix two bugs in my patch earlier today that broke int->fp conversion on X86. · 5b2be1f8
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23522
```
  5b2be1f8
- Fold isascii into a simple comparison. This speeds up 197.parser by 7.4%, · 87ef943a
  Chris Lattner authored Sep 29, 2005
```
bringing the LLC time down to the CBE time.

llvm-svn: 23521
```
  87ef943a
- remove a bunch of unneeded stuff, or self evident comments · 5f6035fe
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23519
```
  5f6035fe
- Implement a couple of memcmp folds from the todo list · c244e7c1
  Chris Lattner authored Sep 29, 2005
```
llvm-svn: 23517
```
  c244e7c1
- Silence VC++ redeclaration warnings. · b01a41a0
  Jeff Cohen authored Sep 29, 2005
```
llvm-svn: 23516
```
  b01a41a0
- Never rely on ReplaceAllUsesWith when selecting, use CodeGenMap instead. · 08c319fb
  Chris Lattner authored Sep 29, 2005
```
ReplaceAllUsesWith does not replace scalars SDOperand floating around on
the stack, permitting things to be selected multiple times.

llvm-svn: 23515
```
  08c319fb
- Codegen ADD X, IMM -> addis/addi if needed. · d4e9e8b7
  Chris Lattner authored Sep 28, 2005
```
This implements PowerPC/fold-li.ll

llvm-svn: 23514
```
  d4e9e8b7
- Autogen MUL, move FP cases together · b9b2e772
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23512
```
  b9b2e772
- disentangle FP from INT versions of div/mul · 5769311c
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23511
```
  5769311c
- Use the autogenerated matcher for ADD/SUB · 585131ba
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23510
```
  585131ba
- add a patter for SUBFIC · f023b2cd
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23509
```
  f023b2cd
- Mark int binops as int-only, add FP binops. Mark FADD/FMUL as commutative but · 21551ea5
  Chris Lattner authored Sep 28, 2005
```
not associative.  Add [SU]REM.

llvm-svn: 23508
```
  21551ea5
- wrap a long line · cd002b24
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23507
```
  cd002b24
- Add FP versions of the binary operators, keeping the int and fp worlds seperate. · d3ea19b5
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23506
```
  d3ea19b5
- Add FP versions of the binary operators, keeping the int and fp worlds seperate. · 0815dcae
  Chris Lattner authored Sep 28, 2005
```
Though I have done extensive testing, it is possible that this will break
things in configs I can't test.  Please let me know if this causes a problem
and I'll fix it ASAP.

llvm-svn: 23505
```
  0815dcae
- Add FP versions of the binary operators, keeping the int and fp worlds seperate. · 6f3b577e
  Chris Lattner authored Sep 28, 2005
```
Though I have done extensive testing, it is possible that this will break
things in configs I can't test.  Please let me know if this causes a problem
and I'll fix it ASAP.

llvm-svn: 23504
```
  6f3b577e
Sep 28, 2005
- Mark associative nodes as associative · 7fe6734d
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23503
```
  7fe6734d
- Nate pointed out that mulh[us] are commutative as well. Thanks! · b97b054b
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23500
```
  b97b054b
- expose commutativity information · 89d168ce
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23498
```
  89d168ce
- All (xor *) cases are autogenerated now · fab48b32
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23497
```
  fab48b32
- add support for missed eqv tests · 037d69a4
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23496
```
  037d69a4
- Implement PowerPC/eqv-andc-orc-nor.ll:EQV3 · 33f8e08c
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23494
```
  33f8e08c
- learn to codegen not as NOR instead of xoris/xori · 8cd7b88a
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23490
```
  8cd7b88a
- These nodes are all autogenerated · bb5939a4
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23489
```
  bb5939a4
- Constant fold llvm.sqrt · ea7214b2
  Chris Lattner authored Sep 28, 2005
```
llvm-svn: 23487
```
  ea7214b2
- add a note about a way to improve this code further, that I won't be getting · 3b63bb37
  Chris Lattner authored Sep 27, 2005
```
to right now.

llvm-svn: 23485
```
  3b63bb37
- Fix a regression in my previous patch, fixing GlobalOpt/2005-09-27-Crash.ll · eb953f0e
  Chris Lattner authored Sep 27, 2005
```
and PR632.

llvm-svn: 23484
```
  eb953f0e
- Darwin, like many BSD systems, has a setjmp/longjmp which saves the signal mask · a028e7a3
  Chris Lattner authored Sep 27, 2005
```
on setjmp calls and restores it on longjmp calls (both of which require syscalls).

This makes the calls REALLY slow.  Use _setjmp/_longjmp instead.  This speeds up
hexxagon from 120.31s to 15.68s: from 5.53x slower than GCC to 28% faster than GCC.

llvm-svn: 23482
```
  a028e7a3
- If the target prefers it, use _setjmp/_longjmp should be used instead of... · 0fd8f9fb
  Chris Lattner authored Sep 27, 2005
```
If the target prefers it, use _setjmp/_longjmp should be used instead of setjmp/longjmp for llvm.setjmp/llvm.longjmp.

llvm-svn: 23481
```
  0fd8f9fb