summaryrefslogtreecommitdiff
path: root/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
diff options
context:
space:
mode:
authorBill Schmidt <wschmidt@linux.vnet.ibm.com>2013-02-08 18:19:17 +0000
committerBill Schmidt <wschmidt@linux.vnet.ibm.com>2013-02-08 18:19:17 +0000
commit0f58dbae4ab2397dc310e19d171a4bc60fc6c9a7 (patch)
tree7c7e52d5180abb20a4fe7d79bee1685e75682ecc /lib/Target/PowerPC/PPCTargetTransformInfo.cpp
parent33daeab1bb8df65273fd9ecbf1a261f96733732e (diff)
downloadllvm-0f58dbae4ab2397dc310e19d171a4bc60fc6c9a7.tar.gz
llvm-0f58dbae4ab2397dc310e19d171a4bc60fc6c9a7.tar.bz2
llvm-0f58dbae4ab2397dc310e19d171a4bc60fc6c9a7.tar.xz
Refine fix to bug 15041.
Thanks to help from Nadav and Hal, I have a more reasonable (and even correct!) approach. This specifically penalizes the insertelement and extractelement operations for the performance hit that will occur on PowerPC processors. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@174725 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/PowerPC/PPCTargetTransformInfo.cpp')
-rw-r--r--lib/Target/PowerPC/PPCTargetTransformInfo.cpp35
1 files changed, 17 insertions, 18 deletions
diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
index f57d7643e1..5e9ad347d3 100644
--- a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
+++ b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
@@ -194,24 +194,23 @@ unsigned PPCTTI::getVectorInstrCost(unsigned Opcode, Type *Val,
unsigned Index) const {
assert(Val->isVectorTy() && "This must be a vector type");
- const unsigned Awful = 1000;
-
- // Vector element insert/extract with Altivec is very expensive.
- // Until VSX is available, avoid vectorizing loops that require
- // these operations.
- if (Opcode == ISD::EXTRACT_VECTOR_ELT ||
- Opcode == ISD::INSERT_VECTOR_ELT)
- return Awful;
-
- // We don't vectorize SREM/UREM so well. Constrain the vectorizer
- // for those as well.
- if (Opcode == ISD::SREM || Opcode == ISD::UREM)
- return Awful;
-
- // VSELECT is not yet implemented, leading to use of insert/extract
- // and ISEL, hence not a good idea.
- if (Opcode == ISD::VSELECT)
- return Awful;
+ int ISD = TLI->InstructionOpcodeToISD(Opcode);
+ assert(ISD && "Invalid opcode");
+
+ // Estimated cost of a load-hit-store delay. This was obtained
+ // experimentally as a minimum needed to prevent unprofitable
+ // vectorization for the paq8p benchmark. It may need to be
+ // raised further if other unprofitable cases remain.
+ unsigned LHSPenalty = 12;
+
+ // Vector element insert/extract with Altivec is very expensive,
+ // because they require store and reload with the attendant
+ // processor stall for load-hit-store. Until VSX is available,
+ // these need to be estimated as very costly.
+ if (ISD == ISD::EXTRACT_VECTOR_ELT ||
+ ISD == ISD::INSERT_VECTOR_ELT)
+ return LHSPenalty +
+ TargetTransformInfo::getVectorInstrCost(Opcode, Val, Index);
return TargetTransformInfo::getVectorInstrCost(Opcode, Val, Index);
}