Commit Graph

147 Commits

Author SHA1 Message Date
Benjamin Kramer cf406756ce LoopVectorize: Clean up ValueMap a bit and avoid double lookups.
No intended functionality change.

llvm-svn: 173809
2013-01-29 17:31:33 +00:00
Renato Golin 1258519674 Vectorization Factor clarification
llvm-svn: 173691
2013-01-28 16:02:45 +00:00
Nadav Rotem 69a040d3eb LoopVectorize: Refactor the code that vectorizes loads/stores to remove duplication.
llvm-svn: 173500
2013-01-25 21:47:42 +00:00
Benjamin Kramer 21e8da5990 LoopVectorize: Simplify code. No functionality change.
llvm-svn: 173475
2013-01-25 19:43:15 +00:00
Nadav Rotem 8e9ca2f8cb LoopVectorizer: Refactor more code to use the IRBuilder.
llvm-svn: 173471
2013-01-25 19:26:23 +00:00
Nadav Rotem c8adf3ff6e Refactor some code to use the IRBuilder.
llvm-svn: 173467
2013-01-25 18:34:09 +00:00
Nadav Rotem ab3e698ee9 Add support for reverse pointer induction variables. These are loops that contain pointers that count backwards.
For example, this is the hot loop in BZIP:

  do {
    m = *--p;
    *p = ( ... );
  } while (--n);

llvm-svn: 173219
2013-01-23 01:35:00 +00:00
Nadav Rotem b2e7e7a0b6 Fix a comment. Induction vars dont need to start at zero.
llvm-svn: 173061
2013-01-21 17:59:18 +00:00
Benjamin Kramer a6e2e2a0a7 LoopVectorize: Fix a C++11 incompatibility.
llvm-svn: 172990
2013-01-20 20:29:52 +00:00
Nadav Rotem da9f2adffd Fix a build error.
llvm-svn: 172971
2013-01-20 09:39:17 +00:00
Nadav Rotem c42f90b1f4 LoopVectorizer: Implement a new heuristics for selecting the unroll factor.
We ignore the cpu frontend and focus on pipeline utilization. We do this because we
don't have a good way to estimate the loop body size at the IR level.

llvm-svn: 172964
2013-01-20 05:24:29 +00:00
Benjamin Kramer d455ed85d1 LoopVectorizer: Emit memory checks into their own basic block.
This separates the check for "too few elements to run the vector loop" from the
"memory overlap" check, giving a lot nicer code and allowing to skip the memory
checks when we're not going to execute the vector code anyways. We still leave
the decision of whether to emit the memory checks as branches or setccs, but it
seems to be doing a good job. If ugly code pops up we may want to emit them as
separate blocks too. Small speedup on MultiSource/Benchmarks/MallocBench/espresso.

Most of this is legwork to allow multiple bypass blocks while updating PHIs,
dominators and loop info.

llvm-svn: 172902
2013-01-19 13:57:58 +00:00
Nadav Rotem d33ce6f100 LoopVectorizer cost model. Honor the user command line flag that selects the vectorization factor even if the target machine does not have any vector registers.
llvm-svn: 172544
2013-01-15 18:25:16 +00:00
Nadav Rotem 40e45eeae2 Fix PR14547. Handle induction variables of small sizes smaller than i32 (i8 and i16).
llvm-svn: 172348
2013-01-13 07:56:29 +00:00
Nadav Rotem 853fe0acb9 ARM Cost Model: We need to detect the max bitwidth of types in the loop in order to select the max vectorization factor.
We don't have a detailed analysis on which values are vectorized and which stay scalars in the vectorized loop so we use
another method. We look at reduction variables, loads and stores, which are the only ways to get information in and out
of loop iterations. If the data types are extended and truncated then the cost model will catch the cost of the vector
zext/sext/trunc operations.

llvm-svn: 172178
2013-01-11 07:11:59 +00:00
Nadav Rotem 6eae65cfac LoopVectorizer: Fix a bug in the vectorization of BinaryOperators. The BinaryOperator can be folded to an Undef, and we don't want to set NSW flags to undef vals.
PR14878

llvm-svn: 172079
2013-01-10 17:34:39 +00:00
Nadav Rotem b1791a75cd ARM Cost model: Use the size of vector registers and widest vectorizable instruction to determine the max vectorization factor.
llvm-svn: 172010
2013-01-09 22:29:00 +00:00
Nadav Rotem b696c36fcd Cost Model: Move the 'max unroll factor' variable to the TTI and add initial Cost Model support on ARM.
llvm-svn: 171928
2013-01-09 01:15:42 +00:00
Nadav Rotem 3c352c0f4a Code cleanup: refactor the switch statements in the generation of reduction variables into an IR builder call.
llvm-svn: 171871
2013-01-08 17:37:45 +00:00
Nadav Rotem 6f6d21a17b Rename the enum members to match the LLVM coding style.
llvm-svn: 171868
2013-01-08 17:23:17 +00:00
Nadav Rotem 5a197c06f3 LoopVectorizer: Add support for floating point reductions
llvm-svn: 171812
2013-01-07 23:13:00 +00:00
Nadav Rotem c60d7d96f5 LoopVectorizer: When we vectorizer and widen loops we process many elements at once. This is a good thing, except for
small loops. On small loops post-loop that handles scalars (and runs slower) can take more time to execute than the
rest of the loop. This patch disables widening of loops with a small static trip count.

llvm-svn: 171798
2013-01-07 21:54:51 +00:00
Chandler Carruth b348328b5d Simplify LoopVectorize to require target transform info and rely on it
being present. Make a member of one of the helper classes a reference as
part of this.

Reformatting goodness brought to you by clang-format.

llvm-svn: 171726
2013-01-07 11:12:29 +00:00
Chandler Carruth b7e60f6844 Merge the unused header file for LoopVectorizer into the source file.
This makes the loop vectorizer match the pattern followed by roughly all
other passses. =]

Notably, this header file was braken in several regards: it contained
a using namespace directive, global #define's that aren't globaly
appropriate, and global constants defined directly in the header file.

As a side benefit, lots of the types in this file become internal, which
will cause the optimizer to chew on this pass more effectively.

llvm-svn: 171723
2013-01-07 10:44:06 +00:00
Chandler Carruth 2109f47d97 Fix the enumerator names for ShuffleKind to match tho coding standards,
and make its comments doxygen comments.

llvm-svn: 171688
2013-01-07 03:20:02 +00:00