X86: add costs for 64-bit vector ext/trunc & rebalance

The most important part of this is probably adding any cost at all for
operations like zext <8 x i8> to <8 x i32>. Before they were being
recorded as extremely costly (24, I believe) which made LLVM fall back
on a 4-wide vectorisation of a loop.

It also rebalances the values for sext, zext and trunc. Lacking any
other sane metric that might work across CPU microarchitectures I went
for instructions. This seems to be in reasonable accord with the rest
of the table (sitofp, ...) though no doubt at least one value is
sub-optimal for some bizarre reason.

Finally, separate AVX and AVX2 values are provided where appropriate.
The CodeGen is quite different in many cases.

rdar://problem/15981990

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200928 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 58 insertions, 15 deletions

diff --git a/lib/Target/X86/X86TargetTransformInfo.cpp b/lib/Target/X86/X86TargetTransformInfo.cpp
index 781be2fddd..207a7685c5 100644
--- a/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -411,16 +411,58 @@ unsigned X86TTI::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src) const {
     return TargetTransformInfo::getCastInstrCost(Opcode, Dst, Src);
 
   static const TypeConversionCostTblEntry<MVT::SimpleValueType>
+  AVX2ConversionTbl[] = {
+    { ISD::SIGN_EXTEND, MVT::v16i16, MVT::v16i8,  1 },
+    { ISD::ZERO_EXTEND, MVT::v16i16, MVT::v16i8,  1 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i1,   3 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i1,   3 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,   3 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i8,   3 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i16,  1 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i16,  1 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i1,   3 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i1,   3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,   3 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i8,   3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16,  3 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i16,  3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i32,  1 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i32,  1 },
+
+    { ISD::TRUNCATE,    MVT::v4i8,   MVT::v4i64,  2 },
+    { ISD::TRUNCATE,    MVT::v4i16,  MVT::v4i64,  2 },
+    { ISD::TRUNCATE,    MVT::v4i32,  MVT::v4i64,  2 },
+    { ISD::TRUNCATE,    MVT::v8i8,   MVT::v8i32,  2 },
+    { ISD::TRUNCATE,    MVT::v8i16,  MVT::v8i32,  2 },
+    { ISD::TRUNCATE,    MVT::v8i32,  MVT::v8i64,  4 },
+  };
+
+  static const TypeConversionCostTblEntry<MVT::SimpleValueType>
   AVXConversionTbl[] = {
-    { ISD::SIGN_EXTEND, MVT::v16i16, MVT::v16i8, 1 },
-    { ISD::ZERO_EXTEND, MVT::v16i16, MVT::v16i8, 1 },
-    { ISD::SIGN_EXTEND, MVT::v8i32, MVT::v8i16, 1 },
-    { ISD::ZERO_EXTEND, MVT::v8i32, MVT::v8i16, 1 },
-    { ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i32, 1 },
-    { ISD::ZERO_EXTEND, MVT::v4i64, MVT::v4i32, 1 },
-    { ISD::TRUNCATE,    MVT::v4i32, MVT::v4i64, 1 },
-    { ISD::TRUNCATE,    MVT::v8i16, MVT::v8i32, 1 },
-    { ISD::TRUNCATE,    MVT::v16i8, MVT::v16i16, 2 },
+    { ISD::SIGN_EXTEND, MVT::v16i16, MVT::v16i8, 4 },
+    { ISD::ZERO_EXTEND, MVT::v16i16, MVT::v16i8, 4 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i1,  7 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i1,  4 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,  7 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i8,  4 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i16, 4 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i16, 4 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i1,  6 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i1,  4 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,  6 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i8,  4 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16, 6 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i16, 3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i32, 4 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i32, 4 },
+
+    { ISD::TRUNCATE,    MVT::v4i8,  MVT::v4i64,  4 },
+    { ISD::TRUNCATE,    MVT::v4i16, MVT::v4i64,  4 },
+    { ISD::TRUNCATE,    MVT::v4i32, MVT::v4i64,  4 },
+    { ISD::TRUNCATE,    MVT::v8i8,  MVT::v8i32,  4 },
+    { ISD::TRUNCATE,    MVT::v8i16, MVT::v8i32,  5 },
+    { ISD::TRUNCATE,    MVT::v16i8, MVT::v16i16, 4 },
+    { ISD::TRUNCATE,    MVT::v8i32, MVT::v8i64,  9 },
 
     { ISD::SINT_TO_FP,  MVT::v8f32, MVT::v8i1,  8 },
     { ISD::SINT_TO_FP,  MVT::v8f32, MVT::v8i8,  8 },
@@ -450,14 +492,15 @@ unsigned X86TTI::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src) const {
 
     { ISD::FP_TO_SINT,  MVT::v8i8,  MVT::v8f32, 1 },
     { ISD::FP_TO_SINT,  MVT::v4i8,  MVT::v4f32, 1 },
-    { ISD::ZERO_EXTEND, MVT::v8i32, MVT::v8i1,  6 },
-    { ISD::SIGN_EXTEND, MVT::v8i32, MVT::v8i1,  9 },
-    { ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i1,  8 },
-    { ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i8,  6 },
-    { ISD::SIGN_EXTEND, MVT::v4i64, MVT::v4i16, 6 },
-    { ISD::TRUNCATE,    MVT::v8i32, MVT::v8i64, 3 },
   };
 
+  if (ST->hasAVX2()) {
+    int Idx = ConvertCostTableLookup(AVX2ConversionTbl, ISD,
+                                     DstTy.getSimpleVT(), SrcTy.getSimpleVT());
+    if (Idx != -1)
+      return AVX2ConversionTbl[Idx].Cost;
+  }
+
   if (ST->hasAVX()) {
     int Idx = ConvertCostTableLookup(AVXConversionTbl, ISD, DstTy.getSimpleVT(),
                                      SrcTy.getSimpleVT());