1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-21 12:02:58 +02:00
Commit Graph

73441 Commits

Author SHA1 Message Date
Benjamin Kramer
2d266249a6 PR10267: Don't combine an equality compare with an AND into an inequality compare when the AND has more than one use.
This can pessimize code, inequalities are generally more expensive.

llvm-svn: 134379
2011-07-04 20:16:36 +00:00
Roman Divacky
08f7d5cf91 Remove accidentaly left node from previous iteration of the patch.
Noticed by Benjamin Kramer!

llvm-svn: 134376
2011-07-04 15:42:45 +00:00
Rafael Espindola
29113212a6 Move early tail duplication earlier.
This fixes the issue noted in PR10251 where early tail dup of bbs with
indirectbr would cause a bb to be duplicated into a loop preheader
and then into its predecessors, creating phi nodes with identical
operands just before register allocation.

This helps with jsinterp.o size (__TEXT goes from 163568 to 126656)
and a bit with performance 1.005x faster on sunspider (jits still enabled).

The result on webkit with the jit disabled is more significant: 1.021x faster.

llvm-svn: 134372
2011-07-04 04:54:22 +00:00
Rafael Espindola
962773db64 Move most of the pre BB code to TailDuplicateAndUpdate. Change the
HasIndirectbr variable to be just that. No functionality change.

llvm-svn: 134371
2011-07-04 01:21:42 +00:00
Rafael Espindola
ce4f4ff705 Reduce indentation and fix the count of how many PHIs we have inserted.
llvm-svn: 134370
2011-07-04 00:13:36 +00:00
Jakob Stoklund Olesen
9950d41b39 Fix PR10244.
A split point inserted in a block with a landing pad successor may be
hoisted above the call to ensure that it dominates all successors. The
code that handles the rest of the basic block must take this into
account.

I am not including a test case, it would be very fragile. PR10244 comes
from building clang with exceptions enabled.

llvm-svn: 134369
2011-07-04 00:05:28 +00:00
Roman Divacky
eae7daa59b Make the i64 and f64 be 64bit ABI aligned in the target description.
This is what both the ABI and clang says.

llvm-svn: 134367
2011-07-03 16:24:07 +00:00
NAKAMURA Takumi
c0837d703b test/CodeGen/X86/lsr-nonaffine.ll: Relax expressions for Win64 CC to appease Win32 hosts.
llvm-svn: 134366
2011-07-03 09:26:14 +00:00
Rafael Espindola
f04e6b50ca Fix an easy fixme.
llvm-svn: 134364
2011-07-03 05:26:42 +00:00
Chandler Carruth
e07bb36a9e FileCheck-ize another test. Reduces the llc invocations from 8 to 1, and
makes one of the tests actually mean something (as the string 'add' will
always appear in the output of this file).

llvm-svn: 134358
2011-07-02 21:34:52 +00:00
Chandler Carruth
941beb26e9 Avoid writing to an arbitrary filename during the test run by writing to
a file descriptor.

llvm-svn: 134355
2011-07-02 20:43:18 +00:00
Chandler Carruth
78b12b3ed4 FileCheck-ize another X86 test, making it more precisely verify the
desired result based on the comments in the file.

llvm-svn: 134354
2011-07-02 20:43:16 +00:00
Chandler Carruth
1de252c70a Quote two greps which contain parentheses.
llvm-svn: 134353
2011-07-02 20:43:13 +00:00
Chandler Carruth
1926e141f1 FileCheck-ize and simplify RUN lines.
llvm-svn: 134352
2011-07-02 20:43:11 +00:00
Chandler Carruth
5de1d825e4 FileCheck-ize
llvm-svn: 134351
2011-07-02 20:43:08 +00:00
Chandler Carruth
01e8f9314e FileCheck-ize and tighten up assertions to only check the relevant sections.
llvm-svn: 134350
2011-07-02 20:43:04 +00:00
Chandler Carruth
500b05b1bb FileCheck-ize and cleanup IR.
llvm-svn: 134349
2011-07-02 20:43:01 +00:00
Chandler Carruth
c674fb38ef FileCheck-ize
llvm-svn: 134348
2011-07-02 20:42:59 +00:00
Chandler Carruth
341ed5f0a0 Remove a grep that is already checked with FileCheck.
llvm-svn: 134346
2011-07-02 20:42:56 +00:00
Chandler Carruth
88e183829b FileCheck-ize
llvm-svn: 134345
2011-07-02 20:42:53 +00:00
Chandler Carruth
7a0f51e003 FileCheck-ize and modernize IR.
llvm-svn: 134344
2011-07-02 20:42:50 +00:00
Chandler Carruth
4af34fe339 FileCheck-ize and simplify RUNs.
llvm-svn: 134343
2011-07-02 20:42:48 +00:00
Chandler Carruth
9e114fc3ee FileCheck-ize and modernize the RUN line.
llvm-svn: 134342
2011-07-02 20:42:44 +00:00
Chandler Carruth
df1690a113 FileCheck-ize, tightening checks and avoiding a temporary file.
llvm-svn: 134341
2011-07-02 20:42:42 +00:00
Chandler Carruth
a5b1de166b FileCheck-ize, tightening checks and avoiding a temporary file.
llvm-svn: 134340
2011-07-02 20:42:39 +00:00
Chandler Carruth
c041ee0766 FileCheck-ize
llvm-svn: 134339
2011-07-02 20:42:36 +00:00
Chandler Carruth
4f82b948fd FileCheck-ize
llvm-svn: 134338
2011-07-02 20:42:33 +00:00
Chandler Carruth
e344d9c676 FileCheck-ize a test, avoiding a temporary file.
llvm-svn: 134337
2011-07-02 20:42:31 +00:00
Chandler Carruth
d939fba46d FileCheck-ize and simplify this test.
llvm-svn: 134336
2011-07-02 20:42:28 +00:00
Chandler Carruth
b870175dd5 FileCheck-ize
llvm-svn: 134335
2011-07-02 20:42:25 +00:00
Chandler Carruth
d98a57cc5a FileCheck-ize another codegen test.
llvm-svn: 134334
2011-07-02 20:42:22 +00:00
Chandler Carruth
4c7e28777b Partially FileCheck-ize a test to remove a weird quoting situation.
llvm-svn: 134333
2011-07-02 20:42:20 +00:00
Chandler Carruth
0d1da937eb FileCheck-ize another test, and upgrade its syntax a bit.
llvm-svn: 134332
2011-07-02 20:42:17 +00:00
Chandler Carruth
4fd8502d12 FileCheck-ize another codegen test, tightening it up.
llvm-svn: 134331
2011-07-02 20:42:14 +00:00
Chandler Carruth
b74aff3ce8 FileCheck-ize another test, making it much more precise for testing the
individual cases, while hard coding less about registers in use.

llvm-svn: 134330
2011-07-02 20:42:11 +00:00
Chandler Carruth
70fa55f478 FileCheck-ize another test. This one is more clear and runs fewer
commands as a result.

llvm-svn: 134329
2011-07-02 20:42:08 +00:00
Chandler Carruth
72358a4bf8 FileCheck-ize a test, no functionality changed.
llvm-svn: 134328
2011-07-02 20:42:06 +00:00
Jordy Rose
6f7044e2b9 Use subprocess.Popen instead of popen2 to stop a deprecation warning when running lit on OS X
llvm-svn: 134324
2011-07-02 17:28:55 +00:00
Duncan Sands
da72617832 Remove unused array.
llvm-svn: 134323
2011-07-02 16:36:24 +00:00
Duncan Sands
3eb91b3dfd Revert previous commit. It seems that whether casting to void
is valid or not depends on which system you build.

llvm-svn: 134321
2011-07-02 13:14:22 +00:00
Duncan Sands
938be63a66 Supress gcc-4.5 warning about the result not being used.
llvm-svn: 134319
2011-07-02 13:06:23 +00:00
Duncan Sands
c1389003b7 Remove unused constant.
llvm-svn: 134318
2011-07-02 13:05:02 +00:00
Rafael Espindola
cf67208057 Use getVNInfoAt.
llvm-svn: 134312
2011-07-02 07:50:27 +00:00
Jakob Stoklund Olesen
4d72701c7e Consistent diagnostic capitalization and redundant context elimination.
llvm-svn: 134311
2011-07-02 07:23:40 +00:00
Jakob Stoklund Olesen
b94d989634 Better diagnostics when inline asm fails to allocate.
asm.c:2:7: error: ran out of registers during register allocation
  asm(""::"r"(0), "r"(1), "r"(2), "r"(3), "r"(4), "r"(5), "r"(6), "r"(7), "r"(8), "r"(9));
        ^

llvm-svn: 134310
2011-07-02 07:17:37 +00:00
Rafael Espindola
a8c92aa8ef Check the VN of the src register at the two copies, not just the
register number.

llvm-svn: 134309
2011-07-02 05:34:02 +00:00
Eric Christopher
9689f96b1e Be less specific about register allocation ordering.
llvm-svn: 134308
2011-07-02 04:06:41 +00:00
Jakob Stoklund Olesen
c19c47697f Include a source location when complaining about bad inline assembly.
Add a MI->emitError() method that the backend can use to report errors
related to inline assembly. Call it from X86FloatingPoint.cpp when the
constraints are wrong.

This enables proper clang diagnostics from the backend:

$ clang -c pr30848.c
pr30848.c:5:12: error: Inline asm output regs must be last on the x87 stack
  __asm__ ("" : "=u" (d));  /* { dg-error "output regs" } */
           ^
1 error generated.

llvm-svn: 134307
2011-07-02 03:53:34 +00:00
Andrew Trick
ac38d56661 indvars -disable-iv-rewrite: bug fix involving weird geps and related cleanup.
llvm-svn: 134306
2011-07-02 02:34:25 +00:00
Jakob Stoklund Olesen
60871c3ee0 Use a new strategy for preventing eviction loops in RAGreedy.
Every live range is assigned a cascade number the first time it is
involved in an eviction. As the evictor, it gets a new cascade number.
Every evictee is assigned the same cascade number as the evictor.

Eviction is prohibited if the evictor has a lower assigned cascade
number than the evictee.

This means that assigned cascade numbers are monotonically increasing
with every eviction, yet they are bounded by NextCascade which can only
be incremented by new live ranges. Thus, infinite loops cannot happen,
but eviction cascades can still be triggered by new live ranges as we
want.

Thanks to Andy for explaining this to me.

llvm-svn: 134303
2011-07-02 01:37:09 +00:00