Commit Graph

18 Commits

Author SHA1 Message Date
Filipe Cabecinhas 83f4192a47 Make blend tests more specific
Following the lead set by r209324, I'm making these tests match the whole
instruction, so we can be sure we're lowering them correctly.

llvm-svn: 209947
2014-05-31 00:52:23 +00:00
Filipe Cabecinhas 229dc17610 Added tests for shufflevector lowering to blend instrs.
These tests ensure that a change I will propose in clang works as
expected.

Summary:
Added tests for the generation of blend+immediate instructions from a
shufflevector.
These tests were proposed along with a patch that was dropped. I'm
committing the tests anyway to protect against possible regressions in
codegen.

Reviewers: nadav, bkramer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D3600

llvm-svn: 209853
2014-05-29 22:04:42 +00:00
Filipe Cabecinhas 82ac07c283 Convert some X86 blendv* intrinsics into IR.
Summary:
Implemented an InstCombine transformation that takes a blendv* intrinsic
call and translates it into an IR select, if the mask is constant.

This will eventually get lowered into blends with immediates if possible,
or pblendvb (with an option to further optimize if we can transform the
pblendvb into a blend+immediate instruction, depending on the selector).
It will also enable optimizations by the IR passes, which give up on
sight of the intrinsic.

Both the transformation and the lowering of its result to asm got shiny
new tests.

The transformation is a bit convoluted because of blendvp[sd]'s
definition:

Its mask is a floating point value! This forces us to convert it and get
the highest bit. I suppose this happened because the mask has type
__m128 in Intel's intrinsic and v4sf (for blendps) in gcc's builtin.

I will send an email to llvm-dev to discuss if we want to change this or
not.

Reviewers: grosbach, delena, nadav

Differential Revision: http://reviews.llvm.org/D3859

llvm-svn: 209643
2014-05-27 03:42:20 +00:00
Filipe Cabecinhas e15551832c Lower vselects into X86ISD::BLENDI when appropriate.
LowerVSELECT will, if possible, generate a X86ISD::BLENDI DAG node if the
condition is constant and we can emit that instruction, given the
subtarget.

This is not enough for all cases. An additional SELECTCombine optimization
will be committed.

Fixed tests that were expecting variable blends but where a blend+imm can
be generated.
Added test where we can't emit blend+immediate.
Added avx2 blend+imm tests.

llvm-svn: 209043
2014-05-16 22:47:49 +00:00
Andrea Di Biagio 2ea61f17ad [X86] Add extra rules for combining vselect dag nodes into movsd.
This improves the fix committed at revision 199683 adding the
following new target specific combine rules:

1) fold (v4i32: vselect <0,0,-1,-1>, A, B) ->
        (v4i32 (bitcast (movsd (v2i64 (bitcast A)), (v2i64 (bitcast B))) ))

2) fold (v4f32: vselect <0,0,-1,-1>, A, B) ->
        (v4f32 (bitcast (movsd (v2f64 (bitcast A)), (v2f64 (bitcast B))) ))

3) fold (v4i32: vselect <-1,-1,0,0>, A, B) ->
        (v4i32 (bitcast (movsd (v2i64 (bitcast B)), (v2i64 (bitcast A))) ))

4) fold (v4f32: vselect <-1,-1,0,0>, A, B) ->
        (v4f32 (bitcast (movsd (v2i64 (bitcast B)), (v2i64 (bitcast A))) ))

llvm-svn: 200324
2014-01-28 18:14:21 +00:00
Andrea Di Biagio 450d1661be [X86] Teach how to combine a vselect into a movss/movsd
Add target specific rules for combining vselect dag nodes into movss/movsd
when possible.

If the vector type of the vselect dag node in input is either MVT::v4i13 or
MVT::v4f32, then try to fold according to rules:

  1) fold (vselect (build_vector (0, -1, -1, -1)), A, B) -> (movss A, B)
  2) fold (vselect (build_vector (-1, 0, 0, 0)), A, B) -> (movss B, A)

If the vector type of the vselect dag node in input is either MVT::v2i64 or
MVT::v2f64 (and we have SSE2), then try to fold according to rules:

  3) fold (vselect (build_vector (0, -1)), A, B) -> (movsd A, B)
  4) fold (vselect (build_vector (-1, 0)), A, B) -> (movsd B, A)

llvm-svn: 199683
2014-01-20 19:35:22 +00:00
Rafael Espindola ae6000e86d Replace more uses of sse41 with sse4.1.
llc using the host cpu features and *waning* on unknown features is probably
not a good thing :-(

llvm-svn: 189144
2013-08-23 20:39:19 +00:00
Stephen Lin 6f36b45076 Update to more CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change.
All changes were made by the following bash script:

  find test/CodeGen -name "*.ll" | \
  while read NAME; do
    echo "$NAME"
    grep -q "^; *RUN: *llc.*debug" $NAME && continue
    grep -q "^; *RUN:.*llvm-objdump" $NAME && continue
    grep -q "^; *RUN: *opt.*" $NAME && continue
    TEMP=`mktemp -t temp`
    cp $NAME $TEMP
    sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \
    while read FUNC; do
      sed -i '' "s/;\([A-Za-z0-9_-]*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC[:]* *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP
    done
    sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP
    sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP
    sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP
    sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP
    mv $TEMP $NAME
  done

This script catches a superset of the cases caught by the script associated with commit r186280. It initially found some false positives due to unusual constructs in a minority of tests; all such cases were disambiguated first in commit r186621.

llvm-svn: 186624
2013-07-18 22:47:09 +00:00
Nadav Rotem b7bb72e4f3 Remove the "-promote-elements" flag. This flag is now enabled by default.
llvm-svn: 157925
2012-06-04 11:27:21 +00:00
Nadav Rotem 02ef0c3524 When emulating vselect using OR/AND/XOR make sure to bitcast the result back to the original type.
llvm-svn: 154764
2012-04-15 15:08:09 +00:00
Nadav Rotem 6d31bac85e Revert 147426 because it caused pr11696.
llvm-svn: 147485
2012-01-03 22:19:42 +00:00
Nadav Rotem 6c7a0e6c8b Optimize the sequence blend(sign_extend(x)) to blend(shl(x)) since SSE blend instructions only look at the highest bit.
llvm-svn: 147426
2012-01-02 08:05:46 +00:00
Duncan Sands f2b8c854dd Synthesize x86 max/min instructions also for vectors (i.e. produce
maxps and maxpd).  This broke the sse41-blend.ll testcase by causing
maxpd to be produced rather than a cmp+blend pair, which is the reason
I tweaked it.  Gives a small speedup on doduc with dragonegg when the
GCC vectorizer is used.

llvm-svn: 139986
2011-09-17 16:49:39 +00:00
Nadav Rotem d748dbacb0 Add integer promotion support for vselect
llvm-svn: 139692
2011-09-14 14:42:15 +00:00
Bruno Cardoso Lopes bf6e1e2717 Change testcase commandline to be more strict and silence buildbots
llvm-svn: 139554
2011-09-12 22:59:26 +00:00
Bruno Cardoso Lopes 973d2921e8 Revert the wrong part of r139528, and fix testcases.
llvm-svn: 139541
2011-09-12 21:24:07 +00:00
Bruno Cardoso Lopes be7a086f58 Not sure how CMPPS and CMPPD had already ever worked, I guess it didn't.
However with this fix it does now.

Basically the operand order for the x86 target specific node
is not the same as the instruction, but since the intrinsic need that
specific order at the instruction definition, just change the order
during legalization. Also, there were some wrong invertions of condition
codes, such as GE => LE, GT => LT, fix that too. Fix PR10907.

llvm-svn: 139528
2011-09-12 19:30:40 +00:00
Bruno Cardoso Lopes fb113a0051 Add AVX versions of blend vector operations and fix some issues noticed
in Nadav's r139285 and r139287 commits.

1) Rename vsel.ll to a more descriptive name
2) Change the order of BLEND operands to "Op1, Op2, Cond", this is
necessary because PBLENDVB is already used in different places with
this order, and it was being emitted in the wrong way for vselect
3) Add AVX patterns and tests for the same SSE41 instructions

llvm-svn: 139305
2011-09-08 18:05:08 +00:00