[llvm-branch-commits] [llvm-branch] r315198 - Merging r313366:

Sun Oct 8 16:49:30 PDT 2017

Author: ctopper
Date: Sun Oct  8 16:49:29 2017
New Revision: 315198

URL: http://llvm.org/viewvc/llvm-project?rev=315198&view=rev
Log:
Merging r313366:
------------------------------------------------------------------------
r313366 | ctopper | 2017-09-15 10:09:03 -0700 (Fri, 15 Sep 2017) | 9 lines

[X86] Don't create i64 constants on 32-bit targets when lowering v64i1 constant build vectors

When handling a v64i1 build vector of constants on 32-bit targets we were creating an illegal i64 constant that we then bitcasted back to v64i1. We need to instead create two 32-bit constants, bitcast them to v32i1 and concat the result. We should also take care to handle the halves being all zeros/ones after the split.

This patch splits the build vector and then recursively lowers the two pieces. This allows us to handle the all ones and all zeros cases with minimal effort. Ideally we'd just do the split and concat, and let lowering get called again on the new nodes, but getNode has special handling for CONCAT_VECTORS that reassembles the pieces back into a single BUILD_VECTOR. Hopefully the two temporary BUILD_VECTORS we had to create to do this that don't get returned don't cause any issues.

Fixes PR34605.

Differential Revision: https://reviews.llvm.org/D37858
------------------------------------------------------------------------

Added:
    llvm/branches/release_50/test/CodeGen/X86/pr34605.ll
      - copied, changed from r313366, llvm/trunk/test/CodeGen/X86/pr34605.ll
Modified:
    llvm/branches/release_50/   (props changed)
    llvm/branches/release_50/lib/Target/X86/X86ISelLowering.cpp

Propchange: llvm/branches/release_50/
------------------------------------------------------------------------------

--- svn:mergeinfo (original)
+++ svn:mergeinfo Sun Oct  8 16:49:29 2017
@@ -1 +1,2 @@
+/llvm/trunk:313366
 /llvm/trunk:155241,308483-308484,308503,308808,308813,308847,308891,308906,308950,308963,308978,308986,309044,309071,309113,309120,309122,309140,309227,309302,309321,309323,309325,309330,309343,309353,309355,309422,309481,309483,309495,309555,309561,309594,309614,309651,309744,309758,309849,309928,309930,309945,310066,310071,310190,310240-310242,310250,310253,310262,310267,310481,310492,310498,310510,310534,310552,310604,310712,310779,310784,310796,310842,310906,310926,310939,310979,310988,310990-310991,311061,311068,311071,311087,311229,311258,311263,311387,311429,311554,311565,311572,311623,311835,312022,312285,313334:312337

Modified: llvm/branches/release_50/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/release_50/lib/Target/X86/X86ISelLowering.cpp?rev=315198&r1=315197&r2=315198&view=diff
==============================================================================
--- llvm/branches/release_50/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/branches/release_50/lib/Target/X86/X86ISelLowering.cpp Sun Oct  8 16:49:29 2017
@@ -7026,6 +7026,18 @@ X86TargetLowering::LowerBUILD_VECTORvXi1
     return DAG.getTargetConstant(1, dl, VT);
 
   if (ISD::isBuildVectorOfConstantSDNodes(Op.getNode())) {
+    if (VT == MVT::v64i1 && !Subtarget.is64Bit()) {
+      // Split the pieces.
+      SDValue Lower =
+          DAG.getBuildVector(MVT::v32i1, dl, Op.getNode()->ops().slice(0, 32));
+      SDValue Upper =
+          DAG.getBuildVector(MVT::v32i1, dl, Op.getNode()->ops().slice(32, 32));
+      // We have to manually lower both halves so getNode doesn't try to
+      // reassemble the build_vector.
+      Lower = LowerBUILD_VECTORvXi1(Lower, DAG);
+      Upper = LowerBUILD_VECTORvXi1(Upper, DAG);
+      return DAG.getNode(ISD::CONCAT_VECTORS, dl, MVT::v64i1, Lower, Upper);
+    }
     SDValue Imm = ConvertI1VectorToInteger(Op, DAG);
     if (Imm.getValueSizeInBits() == VT.getSizeInBits())
       return DAG.getBitcast(VT, Imm);

Copied: llvm/branches/release_50/test/CodeGen/X86/pr34605.ll (from r313366, llvm/trunk/test/CodeGen/X86/pr34605.ll)
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/release_50/test/CodeGen/X86/pr34605.ll?p2=llvm/branches/release_50/test/CodeGen/X86/pr34605.ll&p1=llvm/trunk/test/CodeGen/X86/pr34605.ll&r1=313366&r2=315198&rev=315198&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/pr34605.ll (original)
+++ llvm/branches/release_50/test/CodeGen/X86/pr34605.ll Sun Oct  8 16:49:29 2017
@@ -19,8 +19,8 @@ define void @pr34605(i8* nocapture %s, i
 ; CHECK-NEXT:    kunpckdq %k2, %k1, %k1
 ; CHECK-NEXT:    kandq %k1, %k0, %k1
 ; CHECK-NEXT:    vmovdqu8 {{\.LCPI.*}}, %zmm0 {%k1} {z}
-; CHECK-NEXT:    vmovdqu32 %zmm0, (%eax)
-; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vmovdqu8 %zmm0, (%eax)
+; CHECK-NEXT:    vpxord %zmm0, %zmm0, %zmm0
 ; CHECK-NEXT:    vmovdqu32 %zmm0, 64(%eax)
 ; CHECK-NEXT:    vmovdqu32 %zmm0, 128(%eax)
 ; CHECK-NEXT:    vmovdqu32 %zmm0, 192(%eax)


