39 Commits

Author SHA1 Message Date
Nadav Rotem
133e437c48 Refactor the VectorTargetTransformInfo interface.
Add getCostXXX calls for different families of opcodes, such as casts, arithmetic, cmp, etc.

Port the LoopVectorizer to the new API.

The LoopVectorizer now finds instructions which will remain uniform after vectorization. It uses this information when calculating the cost of these instructions.

llvm-svn: 166836
2012-10-26 23:49:28 +00:00
Nadav Rotem
f73d286571 LoopVectorize: Teach the cost model to query scalar costs as scalar types and not vectors of 1.
llvm-svn: 166715
2012-10-25 21:03:48 +00:00
Nadav Rotem
5635a9350f Add support for additional reduction variables: AND, OR, XOR.
Patch by Paul Redmond <paul.redmond@intel.com>.

llvm-svn: 166649
2012-10-25 00:08:41 +00:00
Nadav Rotem
9d7ba0ef55 Implement a basic cost model for vector and scalar instructions.
llvm-svn: 166642
2012-10-24 23:47:38 +00:00
Nadav Rotem
23bafecedf whitespace
llvm-svn: 166622
2012-10-24 20:58:40 +00:00
Nadav Rotem
05d9e80245 LoopVectorizer: Add a basic cost model which uses the VTTI interface.
llvm-svn: 166620
2012-10-24 20:36:32 +00:00
Micah Villmow
ce5e56a156 Back out r166591, not sure why this made it through since I cancelled the command. Bleh, sorry about this!
llvm-svn: 166596
2012-10-24 17:25:11 +00:00
Micah Villmow
ae5ce80c36 Delete a directory that wasn't supposed to be checked in yet.
llvm-svn: 166591
2012-10-24 17:20:04 +00:00
Nadav Rotem
3deae09579 Use the AliasAnalysis isIdentifiedObj because it also understands mallocs and c++ news.
PR14158.

llvm-svn: 166491
2012-10-23 18:44:18 +00:00
Nadav Rotem
302d4b678a Don't crash if the load/store pointer is not a GEP.
Fix by Shivarama Rao <Shivarama.Rao@amd.com>

llvm-svn: 166427
2012-10-22 18:27:56 +00:00
Nadav Rotem
ea70508da6 Rename a variable.
llvm-svn: 166410
2012-10-22 04:53:05 +00:00
Nadav Rotem
6b56385c1a Vectorizer: optimize the generation of selects. If the condition is uniform, generate a scalar-cond select (i1 as selector).
llvm-svn: 166409
2012-10-22 04:38:00 +00:00
Nadav Rotem
708e5d2fb0 Update the loop vectorizer docs.
llvm-svn: 166408
2012-10-22 03:52:53 +00:00
Anders Carlsson
d04e66ae01 Avoid an extra hash lookup when inserting a value into the widen map.
llvm-svn: 166395
2012-10-21 16:26:35 +00:00
Jakub Staszak
a477da32fc Simplify code. No functionality change.
llvm-svn: 166393
2012-10-21 15:36:03 +00:00
Jakub Staszak
ed5ec60053 Simplify code. No functionality change.
llvm-svn: 166392
2012-10-21 15:29:19 +00:00
Nadav Rotem
380fe201de Fix a bug in the vectorization of wide load/store operations.
We used a SCEV to detect that A[X] is consecutive. We assumed that X was
the induction variable. But X can be any expression that uses the induction
for example: X = i + 2;

llvm-svn: 166388
2012-10-21 06:49:10 +00:00
Nadav Rotem
825cda19d5 Add support for reduction variables that do not start at zero.
This is important for nested-loop reductions such as :

In the innermost loop, the induction variable does not start with zero:

for (i = 0 .. n)
 for (j = 0 .. m)
  sum += ...

llvm-svn: 166387
2012-10-21 05:52:51 +00:00
Nadav Rotem
5ab04af30a Document change. Describe the pass and some papers that inspired the design of the pass.
llvm-svn: 166386
2012-10-21 04:04:25 +00:00
Nadav Rotem
763abacb83 Vectorizer: fix a bug in the classification of induction/reduction phis.
llvm-svn: 166384
2012-10-21 02:38:01 +00:00
Nadav Rotem
2ee8edf34a Fix an infinite loop in the loop-vectorizer.
PR14134.

llvm-svn: 166379
2012-10-20 20:45:01 +00:00
Nadav Rotem
cdd573e703 Vectorize: teach cavVectorizeMemory to distinguish between A[i]+=x and A[B[i]]+=x.
If the pointer is consecutive then it is safe to read and write. If the pointer is non-loop-consecutive then
it is unsafe to vectorize it because we may hit an ordering issue.

llvm-svn: 166371
2012-10-20 08:26:33 +00:00
Nadav Rotem
762317ecc6 Fix a typo
llvm-svn: 166367
2012-10-20 05:03:27 +00:00
Nadav Rotem
4e013454ca Vectorizer: refactor the memory checks to a new function. No functionality change.
llvm-svn: 166366
2012-10-20 04:59:06 +00:00
Nadav Rotem
61a5b018ad LoopVectorize: Keep the IRBuilder on the stack.
llvm-svn: 166354
2012-10-19 23:27:19 +00:00
Nadav Rotem
8fe03aa4c1 Vectorizer: Add support for loop reductions.
For example:

  for (i=0; i<n; i++)
   sum += A[i] +  B[i] + i;

llvm-svn: 166351
2012-10-19 23:05:40 +00:00
Benjamin Kramer
5080891e07 LoopVectorize: Keep the IRBuilder on the stack.
No functionality change.

llvm-svn: 166274
2012-10-19 08:42:02 +00:00
Nadav Rotem
451f76acc3 vectorizer: Add support for reading and writing from the same memory location.
llvm-svn: 166255
2012-10-19 01:24:18 +00:00
Nadav Rotem
fe5d8c8c09 cleanup the comment.
llvm-svn: 166247
2012-10-18 23:21:01 +00:00
Nadav Rotem
ac2f32d807 fix a naming typo
llvm-svn: 166232
2012-10-18 21:45:31 +00:00
Nadav Rotem
b1d0bfd68b Avoid reconstructing the pointer set when searching for duplicated read/write pointers.
llvm-svn: 166205
2012-10-18 18:34:50 +00:00
Nadav Rotem
f1bbc20bb0 When looking for a vector representation of a scalar, do a single lookup. Also, cache the result of the broadcast instruction.
No functionality change.

llvm-svn: 166191
2012-10-18 17:31:49 +00:00
Nadav Rotem
76ff018e87 remove unused variable to fix a warning.
llvm-svn: 166170
2012-10-18 06:09:21 +00:00
Nadav Rotem
67e1c2c770 Remove the use of dominators and AA.
llvm-svn: 166167
2012-10-18 05:33:02 +00:00
Nadav Rotem
fd924ec3c6 Vectorizer: Add support for loops with an unknown count. For example:
for (i=0; i<n; i++){
        a[i] = b[i+1] + c[i+3];
     }

llvm-svn: 166165
2012-10-18 05:29:12 +00:00
NAKAMURA Takumi
8ffefaa778 LoopVectorize.cpp: Fix a warning. [-Wunused-variable]
llvm-svn: 166153
2012-10-17 23:40:15 +00:00
Jakub Staszak
f66a4adca8 Remove redundant SetInsertPoint call.
llvm-svn: 166138
2012-10-17 23:06:37 +00:00
Roman Divacky
e9f1f9a823 Fix some typos and wrong indenting.
llvm-svn: 166128
2012-10-17 21:07:35 +00:00
Nadav Rotem
8303c909c7 Add a loop vectorizer.
llvm-svn: 166112
2012-10-17 18:25:06 +00:00