RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 03:33:20 +01:00
Code Issues Projects Releases Wiki Activity
7,086 Commits 4 Branches 5 Tags 704 MiB
54e47afae3
Commit Graph

7086 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Chris Lattner
54e47afae3 Sort the file lists 
llvm-svn: 7638
 2003-08-06 16:02:50 +00:00
Chris Lattner
4c4bbb5acf Loop unroller doesn't necessarily need to be profile driven 
llvm-svn: 7637
 2003-08-06 15:40:37 +00:00
Chris Lattner
07a484e6f6 Completely eliminate the isVoid TSFlag, shifting over all other fields 
llvm-svn: 7636
 2003-08-06 15:32:20 +00:00
Chris Lattner
83831c66fc add a pattern for RET, immediates no longer need to be explicitly typed 
llvm-svn: 7635
 2003-08-06 15:31:35 +00:00
Chris Lattner
2d20d75416 Start adding usefulness to the DAG node definitions, add a new Expander 
class

llvm-svn: 7634
 2003-08-06 15:31:02 +00:00
Chris Lattner
4671977130 Add error checking code to the node type parser. Start the instruction pattern 
reader

llvm-svn: 7632
 2003-08-06 06:16:35 +00:00
Chris Lattner
6a87b317a2 Initial support for an instruction selector emitter 
llvm-svn: 7631
 2003-08-06 05:42:05 +00:00
Chris Lattner
47f18bb3a1 All good classes with virtual functions should have virtual dtors... 
llvm-svn: 7630
 2003-08-06 05:39:03 +00:00
Chris Lattner
51baaaa269 Add an instruction selector emitter skeleton 
llvm-svn: 7629
 2003-08-06 04:47:56 +00:00
Chris Lattner
0e4fda4b71 convert over to using TableGen backends 
llvm-svn: 7628
 2003-08-06 04:36:35 +00:00
Chris Lattner
2e7568698e Switch code over to being a TableGenBackend 
llvm-svn: 7627
 2003-08-06 04:32:07 +00:00
Chris Lattner
81a4f321fb Add more helper methods 
llvm-svn: 7626
 2003-08-06 04:31:26 +00:00
Chris Lattner
a93ef47190 New common interface for backends to use 
llvm-svn: 7625
 2003-08-06 04:23:04 +00:00
Sumant Kowshik
7f7a4a1472 Added POOLFLAGS option to run pool allocation 
llvm-svn: 7624
 2003-08-06 01:03:28 +00:00
Chris Lattner
506e6abb67 "fix" coding style stuff 
Change some <>'s into &lt;&gt;'s

llvm-svn: 7623
 2003-08-05 22:54:23 +00:00
Chris Lattner
138a3fc9ab Completely eliminate the per-machine-instruction regsUsed set. 
This substantially shrinks the size of each machine instruction, which should
make allocation faster and the cache footprint of the machine code lighter.

Here are some timings for code generation of the larger benchmarks we have.
This are timings of code generation phases of the X86 JIT, when compiled in
debug mode:

		Before	After	Diff
164.gzip:
  InstSel	0.0878	0.0722	-21.6%
  RegAlloc	0.2031	0.1757	-15.6%
  TOTAL		0.5585	0.4999	-11.7%
Ptrdist-bc:
  InstSel	0.0878	0.0722	-21.6%
  RegAlloc	0.2070	0.1933	- 7.1%
  TOTAL		0.6972	0.6464	- 7.9%
197.parser:
  InstSel	0.2148	0.2148	- 0.0%
  RegAlloc	0.4941	0.4277	-15.5%
  TOTAL		1.3749	1.2851	- 7.0%
175.vpr:
  InstSel	0.2519	0.2109	-19.4%
  RegAlloc	0.5976	0.5663	- 5.5%
  TOTAL		1.6933	1.6347	- 3.5%
254.gap:
  InstSel	1.1328	0.9921	-14.2%
  RegAlloc	2.6933	2.4804	- 8.6%
  TOTAL		7.7871	7.2499	- 7.4%

llvm-svn: 7622
 2003-08-05 22:39:13 +00:00
Chris Lattner
5a778ec126 Use a new local data structure instead of the MachineInstr::regsUsed set 
llvm-svn: 7621
 2003-08-05 22:11:13 +00:00
Chris Lattner
937abf2262 Add a map 
llvm-svn: 7620
 2003-08-05 22:09:31 +00:00
Chris Lattner
a38b57808a Minor cleanups 
llvm-svn: 7619
 2003-08-05 22:03:27 +00:00
Chris Lattner
18b0b92033 Physical registers no longer live in the regsUsed set for each machine instr 
llvm-svn: 7618
 2003-08-05 21:55:58 +00:00
Chris Lattner
a826c8d485 Do not insert physical regsiters into the regsUsed set 
llvm-svn: 7617
 2003-08-05 21:55:20 +00:00
John Criswell
bc2e14bd15 Reverted back to using OR for cmp/mv operations for lex/yacc output. 
The shell AND/OR operators short-circuit on command success/failure, which is
the inverse of exit status (i.e. 0 means success, non-zero means failure).

llvm-svn: 7616
 2003-08-05 21:38:28 +00:00
John Criswell
a40b9262ec Switched from using diff to cmp for two reasons: 
o Not all versions of diff have the -q option
	o The cmp program is probably faster than diff
Fixed the logic that only copies the file over if no differences are found.

llvm-svn: 7615
 2003-08-05 21:21:58 +00:00
Tanya Lattner
6fbe85279f Fixed minor bug in SafeToHoist and made some changes suggested by Chris. 
llvm-svn: 7614
 2003-08-05 20:39:02 +00:00
Tanya Lattner
9d218f68d1 Added LICM test cases to: 
1) Check that trapping instructionns that are not guaranteed to execute are not hoisted.
2) Check that trapping instructions that are guaranteed to execute are hoisted.

llvm-svn: 7613
 2003-08-05 18:52:42 +00:00
Tanya Lattner
48745a598c Fixed LICM bug that hoists trapping instructions that are not guaranteed to execute. 
llvm-svn: 7612
 2003-08-05 18:45:46 +00:00
Chris Lattner
629c78afe3 Minor changes: 
* Expand most tabs into spaces
  * Move #define DEBUG_TYPE to top of file to avoid warning

llvm-svn: 7611
 2003-08-05 18:44:12 +00:00
Chris Lattner
e2ee62c91c Add more verbose comment 
llvm-svn: 7610
 2003-08-05 18:38:37 +00:00
Chris Lattner
bd872cdaaf Add a comment to the method decl 
llvm-svn: 7609
 2003-08-05 18:38:16 +00:00
Chris Lattner
8da1b00db6 Remove unused method 
llvm-svn: 7608
 2003-08-05 17:09:08 +00:00
Sumant Kowshik
9d60ea4810 Added declaration of mergeInGlobalsGraph 
llvm-svn: 7607
 2003-08-05 17:06:18 +00:00
Sumant Kowshik
1a4fce3511 Added function mergeInGlobalsGraph which merges in the entire globals graph with the graph of a function 
llvm-svn: 7606
 2003-08-05 17:04:41 +00:00
Sumant Kowshik
18c49a775f Major bug fixes including a memory leak and tracking some exceptional conditions. Also added support for including global and indirect call information in the DS graphs used by the pool allocation 
llvm-svn: 7605
 2003-08-05 17:01:54 +00:00
Chris Lattner
7e9e83e817 Specify DEBUG_TYPE's for the JIT debug messages 
llvm-svn: 7604
 2003-08-05 17:00:32 +00:00
Chris Lattner
54c2758e95 This method has now been changed to preserve flags for us! 
llvm-svn: 7603
 2003-08-05 16:59:24 +00:00
Chris Lattner
b047a0d449 All callers of these methods actually wanted them to preserve the flags, 
so get rid of the def/use parameters that were getting passed in.

**** This now changes the semantics of these methods to preserve the flags,
     not clobber them!

llvm-svn: 7602
 2003-08-05 16:58:46 +00:00
Sumant Kowshik
d0be623358 Added the declaration of InlineIndirectCalls 
llvm-svn: 7601
 2003-08-05 16:56:59 +00:00
Chris Lattner
6488844c2e Factor shared code 
llvm-svn: 7600
 2003-08-05 16:34:44 +00:00
Chris Lattner
488b0e52a8 Fix bug: SimplifyCFG/2003-08-05-InvokeCrash.ll 
Fix bug: SimplifyCFG/2003-08-05-MishandleInvoke.ll

llvm-svn: 7599
 2003-08-05 16:27:44 +00:00
Chris Lattner
e465467534 new testcases that simplifycfg breaks 
llvm-svn: 7598
 2003-08-05 16:11:36 +00:00
Misha Brukman
8cedc4ffa6 * Removed `using' declaration, now use full namespace qualifier std::string 
* Simplified code by using an inline function instead of copy-pasted code

llvm-svn: 7597
 2003-08-05 16:01:50 +00:00
Chris Lattner
cd4179f054 If we're debugging the SimplifyCFG pass, we _REALLY_ don't want to use it for 
narrowing, no matter what.

llvm-svn: 7596
 2003-08-05 15:51:05 +00:00
Chris Lattner
b44a9376d1 Implement TODO: print out short form of Invoke if possible 
llvm-svn: 7595
 2003-08-05 15:34:45 +00:00
Misha Brukman
65bebf03f0 Only test the vector of functions if it is non-empty. 
llvm-svn: 7594
 2003-08-05 15:26:21 +00:00
Misha Brukman
7452a521b3 Stop special-casing annul and predict bits (which are Sparc-specific anyway) 
since those bits are now hard-coded in Sparc*.td files.

llvm-svn: 7593
 2003-08-05 14:35:35 +00:00
Misha Brukman
c666ae831c * Set annul bit to be 0, because the Sparc backend currently does not use it. 
* Use the name of the predict field instead of just the const 1 in the
  Instruction.

llvm-svn: 7592
 2003-08-05 14:34:38 +00:00
Chris Lattner
297f7d2451 Fix bugs handling ESP in alloca references 
llvm-svn: 7591
 2003-08-05 04:13:58 +00:00
Chris Lattner
8f8fbd7106 No functional changes, comment the fix I just put in 
llvm-svn: 7590
 2003-08-05 03:59:01 +00:00
Chris Lattner
d8cbdc4c67 The CodeEmitterGenerator used to consider ANY uninitialized field as being an 
operand (unless it's annul or predict).  Now we only consider fields to be
operands if they are uninitialized AND used in the "Inst" field.

llvm-svn: 7589
 2003-08-05 03:53:04 +00:00
Chris Lattner
6fece1bf61 Revert previous change, and be really anal about what physical registers can do. 
llvm-svn: 7588
 2003-08-05 00:49:09 +00:00