[TTI] The cost model should not assume illegal vector casts get completely scalarized

[mkuper]

Summary

If a vector cast gets split, it's quite possible that the resulting casts are legal and cheap.
So, instead of pessimistically assuming scalarization, we use the costs the concrete TTI provides for the split vector.

This looks like it does the right thing for AVX (a lot of overblown costs drop dramatically) - but I'm less sure about ARM.

[jmolloy]

The ARM test changes look vaguely sensible to me.

[arsenm]

test/Analysis/CostModel/AMDGPU/addrspacecast.ll

39 ↗

(On Diff #60426)

Pretty much everything should be scalarized. The vector insert and extracts are supposed to be free (and the cost is reported as 0 for those) so I think adding the one there is inconsistent and should check the extract/insert cost

[mkuper]

test/Analysis/CostModel/AMDGPU/addrspacecast.ll

39 ↗

(On Diff #60426)

Until now, we assumed scalarization, but I think this is actually the rare case in practice. If the platform cares about vectors, I'd expect it to support most vector operations at least at some vector width, so it usually won't scalarize. And if we assume partial splitting instead of scalarization, using the insert/extract costs will be the wrong thing, regardless of how imprecise "1" is for splitting (and it's definitely imprecise, but it's what the generic getTypeLegalizationCost() uses).

We could trace the entire legalization chain, and see whether the end result is a vector or a scalar, and then use either 1 or the getScalarizationOvehead() based on that, but I'm not a huge fan of that.
(Is full scalarization common on AMDGPU, or is this a corner case? If it's common, perhaps we should specialize this for the AMDGPU TTI.)

[arsenm]

test/Analysis/CostModel/AMDGPU/addrspacecast.ll

39 ↗

(On Diff #60426)

There are no vector operations.Vectors are only for loading and storing, every operation is scalar

39 ↗

(On Diff #60426)

There's also no additional scalarization cost since access of a vector element is just access of a subregister

[mkuper]

Could someone on the X86 side verify that this is sane-ish?
Some costs dropped by a very large factor, and the new costs look better, but I'd appreciate another set of eyes.

test/Analysis/CostModel/AMDGPU/addrspacecast.ll

39 ↗

(On Diff #60426)

Ah, I see.

I could add a TTI hook for "split cost", and set that to 0 for AMDGPU. But I'm not entirely happy with that either. How does that sound to you?
Any other suggestions that will make this work correctly for both AMDGPU and platforms that legalize vectors partially?

[mkuper]

Updated to not affect AMDGPU, by specializing the "split cost".

Unfortunately, we can't just plug in getVectorSplitCost() (either this or a more sophisticated version) into getTypeLegalizationCost().

This is because most of the users of getTypeLegalizationCost() apparently don't use the returned value as a "cost". Rather, they seem assume it's just the number of post-legalization pieces, and multiply per-piece costs by that. So that needs a completely separate clean-up first.
AMDGPU itself is also guilty of this - it multiplies by getTypeLegalizationCost().first, so if getTypeLegalizationCost(LegalType) were to return 0 - which it ought to, if we really consider it the "cost" of legalization - things would get seriously out of whack. But this happens across all targets.

[RKSimon]

The SSE2 costs look very high which I guess is due to missing cost entries - that should be fixable in a pre/post patch if you prefer to keep them separate. The changed x86 costs all look reasonable.

test/Analysis/CostModel/X86/sitofp.ll

72 ↗

(On Diff #61313)

This looks very high, codegen looks like it takes 26 ops - I haven't checked the throughput.

[mkuper]

Thanks, Simon!
And you're right about the SSE2 costs, I'll look into that separately.

[arsenm]

In D21251#462734, @mkuper wrote:

Updated to not affect AMDGPU, by specializing the "split cost".

Unfortunately, we can't just plug in getVectorSplitCost() (either this or a more sophisticated version) into getTypeLegalizationCost().

This is because most of the users of getTypeLegalizationCost() apparently don't use the returned value as a "cost". Rather, they seem assume it's just the number of post-legalization pieces, and multiply per-piece costs by that. So that needs a completely separate clean-up first.
AMDGPU itself is also guilty of this - it multiplies by getTypeLegalizationCost().first, so if getTypeLegalizationCost(LegalType) were to return 0 - which it ought to, if we really consider it the "cost" of legalization - things would get seriously out of whack. But this happens across all targets.

I was thinking that was to get the cost of the split. Everything should be cost * NElts * scalar cost, so e.g. <8 x float> -> <4 x float>, <4 x float> = 2 * 4 * cost

[arsenm]

test/Analysis/CostModel/ARM/cast.ll

267 ↗

(On Diff #61313)

LGTM, but it looks to me like this should be adding 0, so not increasing by 1?

[mkuper]

In D21251#465230, @arsenm wrote:

In D21251#462734, @mkuper wrote:

Unfortunately, we can't just plug in getVectorSplitCost() (either this or a more sophisticated version) into getTypeLegalizationCost().

This is because most of the users of getTypeLegalizationCost() apparently don't use the returned value as a "cost". Rather, they seem assume it's just the number of post-legalization pieces, and multiply per-piece costs by that. So that needs a completely separate clean-up first.
AMDGPU itself is also guilty of this - it multiplies by getTypeLegalizationCost().first, so if getTypeLegalizationCost(LegalType) were to return 0 - which it ought to, if we really consider it the "cost" of legalization - things would get seriously out of whack. But this happens across all targets.

I was thinking that was to get the cost of the split. Everything should be cost * NElts * scalar cost, so e.g. <8 x float> -> <4 x float>, <4 x float> = 2 * 4 * cost

What exactly do you mean when you say "cost of the split"?
My expectation was for getTypeLegalizationCost() to return an approximation of the total cost of legalization (cost of INSERT_SUBVECTOR and EXTRACT_SUBVECTOR operations), that is, an additive factor. So, it would make sense to have "legalization cost + NElts * scalar cost". Or, in the more general case, when we're not fully scalarizing, "legalizaiton cost + NPieces * piece cost". Instead, what it's actually used for is to get NPieces. This is even explicitly referred to as the "split factor" in one of the X86 TTI uses:

std::pair<int, MVT> IdxsLT = TLI->getTypeLegalizationCost(DL, IndexVTy);
std::pair<int, MVT> SrcLT = TLI->getTypeLegalizationCost(DL, SrcVTy);
int SplitFactor = std::max(IdxsLT.first, SrcLT.first);

I'll have to go over all the uses of getTypeLegalizationCost(). If everything uses it as a proxy for the split factor, it's probably just a matter of changing the name to indicate what it really does. If not, then we may need two different functions, one to get the split factor, and one to get the *cost* of performing the splits.
But I think that's independent of this patch.

test/Analysis/CostModel/ARM/cast.ll

267 ↗

(On Diff #61313)

Generally, with the new formula, it makes sense to have costs like 2 * 32 + 1, so I didn't pay too much attention to those little changes (what concerned me more were the big drops e.g. 64 -> 11 on line 326). But you're right, I need to verify that this is really reasonable and not some unexpected artifact.
Thanks!

[delena]

Could you, please, add a test for "sext <16 x i32 > to <16 x i64 >" for Skylake-avx512 and see what happens?
The cost should be 2. I know that this case did not work correctly and in many cases prevented vectorization to 16.

[mkuper]

In D21251#465930, @delena wrote:

Could you, please, add a test for "sext <16 x i32 > to <16 x i64 >" for Skylake-avx512 and see what happens?
The cost should be 2. I know that this case did not work correctly and in many cases prevented vectorization to 16.

The cost is now 3 (instead of the old 48). It's not 2 because of the getVectorSplitCost() fudge factor.
We'll probably need to tune this in the future, as I said before, this is just a first approximation. It's possible that 0 is the right value most of the time.

If you don't mind, I'll add the test as a separate patch - we don't only want this specific test, we need tests for a bunch of sexts/zexts, like ARM has.

test/Analysis/CostModel/ARM/cast.ll

267 ↗

(On Diff #61313)

So it mostly makes sense.
This used to be evaluated as fully scalarizing, with a per-element scalarization cost of 6, and cast cost of 10. So, (6 + 10) * 4 = 64.
Now we evaluate it as 1 + 2 * (cost(fptoui <2 x float> to <2 x i64>)). But the 2-wide cast is still considered fully scalarizing (even though the types are now legal), so we get 1 + 2 * (2 * (10 + 6)) = 65.

[llvm-commits]

If you don't mind, I'll add the test as a separate patch - we don't only want

>this specific test, we need tests for a bunch of sexts/zexts, like ARM has.
>

[Demikhovsky, Elena] yes, you can add tests later. You can look at this revision as a test reference for X86

http://reviews.llvm.org/D15604

Intel Israel (74) Limited

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.

---

llvm-commits mailing list
[e-mail]
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits

[mkuper]

So, just to make sure - at this point, everybody involved is "vaguely OK" with this, and there are no objections, right?

llvm/trunk/include/llvm/CodeGen/BasicTTIImpl.h

unsigned getArithmeticInstrCost(
 
     if (!TLI->isOperationExpand(ISD, LT.second)) {
       // If the operation is custom lowered, then assume that the code is twice
       // as expensive.
       return LT.first * 2 * OpCost;
     }
 
     // Else, assume that we need to scalarize this op.
+    // TODO: If one of the types get legalized by splitting, handle this
+    // similarly to what getCastInstrCost() does.
     if (Ty->isVectorTy()) {
       unsigned Num = Ty->getVectorNumElements();
       unsigned Cost = static_cast<T *>(this)
                           ->getArithmeticInstrCost(Opcode, Ty->getScalarType());
       // return the cost of multiple scalar invocation plus the cost of
       // inserting
       // and extracting the values.
       return getScalarizationOverhead(Ty, true, true) + Num * Cost;
if (Dst->isVectorTy() && Src->isVectorTy()) {
 
         // Just check the op cost. If the operation is legal then assume it
         // costs
         // 1 and multiply by the type-legalization overhead.
         if (!TLI->isOperationExpand(ISD, DstLT.second))
           return SrcLT.first * 1;
       }
 
-      // If we are converting vectors and the operation is illegal, or
-      // if the vectors are legalized to different types, estimate the
-      // scalarization costs.
-      // TODO: This is probably a big overestimate. For splits, we should have
-      // something like getTypeLegalizationCost() + 2 * getCastInstrCost().
-      // The same applies to getCmpSelInstrCost() and getArithmeticInstrCost()
+      // If we are legalizing by splitting, query the concrete TTI for the cost
+      // of casting the original vector twice. We also need to factor int the
+      // cost of the split itself. Count that as 1, to be consistent with
+      // TLI->getTypeLegalizationCost().
+      if ((TLI->getTypeAction(Src->getContext(), TLI->getValueType(DL, Src)) ==
+           TargetLowering::TypeSplitVector) ||
+          (TLI->getTypeAction(Dst->getContext(), TLI->getValueType(DL, Dst)) ==
+           TargetLowering::TypeSplitVector)) {
+        Type *SplitDst = VectorType::get(Dst->getVectorElementType(),
+                                         Dst->getVectorNumElements() / 2);
+        Type *SplitSrc = VectorType::get(Src->getVectorElementType(),
+                                         Src->getVectorNumElements() / 2);
+        T *TTI = static_cast<T *>(this);
+        return TTI->getVectorSplitCost() +
+               (2 * TTI->getCastInstrCost(Opcode, SplitDst, SplitSrc));
+      }
+
+      // In other cases where the source or destination are illegal, assume
+      // the operation will get scalarized.
       unsigned Num = Dst->getVectorNumElements();
       unsigned Cost = static_cast<T *>(this)->getCastInstrCost(
           Opcode, Dst->getScalarType(), Src->getScalarType());
 
       // Return the cost of multiple scalar invocation plus the cost of
       // inserting and extracting the values.
       return getScalarizationOverhead(Dst, true, true) + Num * Cost;
     }
unsigned getCmpSelInstrCost(unsigned Opcode, Type *ValTy, Type *CondTy) {
     if (!(ValTy->isVectorTy() && !LT.second.isVector()) &&
         !TLI->isOperationExpand(ISD, LT.second)) {
       // The operation is legal. Assume it costs 1. Multiply
       // by the type-legalization overhead.
       return LT.first * 1;
     }
 
     // Otherwise, assume that the cast is scalarized.
+    // TODO: If one of the types get legalized by splitting, handle this
+    // similarly to what getCastInstrCost() does.
     if (ValTy->isVectorTy()) {
       unsigned Num = ValTy->getVectorNumElements();
       if (CondTy)
         CondTy = CondTy->getScalarType();
       unsigned Cost = static_cast<T *>(this)->getCmpSelInstrCost(
           Opcode, ValTy->getScalarType(), CondTy);
 
       // Return the cost of multiple scalar invocation plus the cost of
-      // inserting
-      // and extracting the values.
+      // inserting and extracting the values.
       return getScalarizationOverhead(ValTy, true, false) + Num * Cost;
     }
 
     // Unknown scalar opcode.
     return 1;
   }
 
   unsigned getVectorInstrCost(unsigned Opcode, Type *Val, unsigned Index) {
unsigned getReductionCost(unsigned Opcode, Type *Ty, bool IsPairwise) {
     // Assume the pairwise shuffles add a cost.
     unsigned ShuffleCost =
         NumReduxLevels * (IsPairwise + 1) *
         static_cast<T *>(this)
             ->getShuffleCost(TTI::SK_ExtractSubvector, Ty, NumVecElts / 2, Ty);
     return ShuffleCost + ArithCost + getScalarizationOverhead(Ty, false, true);
   }
 
+  unsigned getVectorSplitCost() { return 1; }
+
   /// @}
 };
 
 /// \brief Concrete BasicTTIImpl that can be used if no further customization
 /// is needed.
 class BasicTTIImpl : public BasicTTIImplBase<BasicTTIImpl> {
   typedef BasicTTIImplBase<BasicTTIImpl> BaseT;
   friend class BasicTTIImplBase<BasicTTIImpl>;

llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h

int getArithmeticInstrCost(
     TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
     TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
     TTI::OperandValueProperties Opd2PropInfo = TTI::OP_None);
 
   unsigned getCFInstrCost(unsigned Opcode);
 
   int getVectorInstrCost(unsigned Opcode, Type *ValTy, unsigned Index);
   bool isSourceOfDivergence(const Value *V) const;
+
+  unsigned getVectorSplitCost() { return 0; }
 };
 
 } // end namespace llvm
 
 #endif

llvm/trunk/test/Analysis/CostModel/ARM/cast.ll

define i32 @casts() {
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r114 = fptoui <4 x float> undef to <4 x i16>
   %r114 = fptoui <4 x float> undef to <4 x i16>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r115 = fptosi <4 x float> undef to <4 x i16>
   %r115 = fptosi <4 x float> undef to <4 x i16>
   ; CHECK:  Found an estimated cost of 1 for instruction:   %r116 = fptoui <4 x float> undef to <4 x i32>
   %r116 = fptoui <4 x float> undef to <4 x i32>
   ; CHECK:  Found an estimated cost of 1 for instruction:   %r117 = fptosi <4 x float> undef to <4 x i32>
   %r117 = fptosi <4 x float> undef to <4 x i32>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r118 = fptoui <4 x float> undef to <4 x i64>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r118 = fptoui <4 x float> undef to <4 x i64>
   %r118 = fptoui <4 x float> undef to <4 x i64>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r119 = fptosi <4 x float> undef to <4 x i64>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r119 = fptosi <4 x float> undef to <4 x i64>
   %r119 = fptosi <4 x float> undef to <4 x i64>
 
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r120 = fptoui <4 x double> undef to <4 x i1>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r120 = fptoui <4 x double> undef to <4 x i1>
   %r120 = fptoui <4 x double> undef to <4 x i1>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r121 = fptosi <4 x double> undef to <4 x i1>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r121 = fptosi <4 x double> undef to <4 x i1>
   %r121 = fptosi <4 x double> undef to <4 x i1>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r122 = fptoui <4 x double> undef to <4 x i8>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r122 = fptoui <4 x double> undef to <4 x i8>
   %r122 = fptoui <4 x double> undef to <4 x i8>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r123 = fptosi <4 x double> undef to <4 x i8>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r123 = fptosi <4 x double> undef to <4 x i8>
   %r123 = fptosi <4 x double> undef to <4 x i8>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r124 = fptoui <4 x double> undef to <4 x i16>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r124 = fptoui <4 x double> undef to <4 x i16>
   %r124 = fptoui <4 x double> undef to <4 x i16>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r125 = fptosi <4 x double> undef to <4 x i16>
+  ; CHECK:  Found an estimated cost of 33 for instruction:   %r125 = fptosi <4 x double> undef to <4 x i16>
   %r125 = fptosi <4 x double> undef to <4 x i16>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r126 = fptoui <4 x double> undef to <4 x i32>
+  ; CHECK:  Found an estimated cost of 5 for instruction:   %r126 = fptoui <4 x double> undef to <4 x i32>
   %r126 = fptoui <4 x double> undef to <4 x i32>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r127 = fptosi <4 x double> undef to <4 x i32>
+  ; CHECK:  Found an estimated cost of 5 for instruction:   %r127 = fptosi <4 x double> undef to <4 x i32>
   %r127 = fptosi <4 x double> undef to <4 x i32>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r128 = fptoui <4 x double> undef to <4 x i64>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r128 = fptoui <4 x double> undef to <4 x i64>
   %r128 = fptoui <4 x double> undef to <4 x i64>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r129 = fptosi <4 x double> undef to <4 x i64>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r129 = fptosi <4 x double> undef to <4 x i64>
   %r129 = fptosi <4 x double> undef to <4 x i64>
 
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r130 = fptoui <8 x float> undef to <8 x i1>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r130 = fptoui <8 x float> undef to <8 x i1>
   %r130 = fptoui <8 x float> undef to <8 x i1>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r131 = fptosi <8 x float> undef to <8 x i1>
+  ; CHECK:  Found an estimated cost of 65 for instruction:   %r131 = fptosi <8 x float> undef to <8 x i1>
   %r131 = fptosi <8 x float> undef to <8 x i1>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r132 = fptoui <8 x float> undef to <8 x i8>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r132 = fptoui <8 x float> undef to <8 x i8>
   %r132 = fptoui <8 x float> undef to <8 x i8>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r133 = fptosi <8 x float> undef to <8 x i8>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r133 = fptosi <8 x float> undef to <8 x i8>
   %r133 = fptosi <8 x float> undef to <8 x i8>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r134 = fptoui <8 x float> undef to <8 x i16>
   %r134 = fptoui <8 x float> undef to <8 x i16>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r135 = fptosi <8 x float> undef to <8 x i16>
   %r135 = fptosi <8 x float> undef to <8 x i16>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r136 = fptoui <8 x float> undef to <8 x i32>
   %r136 = fptoui <8 x float> undef to <8 x i32>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r137 = fptosi <8 x float> undef to <8 x i32>
   %r137 = fptosi <8 x float> undef to <8 x i32>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r138 = fptoui <8 x float> undef to <8 x i64>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r138 = fptoui <8 x float> undef to <8 x i64>
   %r138 = fptoui <8 x float> undef to <8 x i64>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r139 = fptosi <8 x float> undef to <8 x i64>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r139 = fptosi <8 x float> undef to <8 x i64>
   %r139 = fptosi <8 x float> undef to <8 x i64>
 
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r140 = fptoui <8 x double> undef to <8 x i1>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r140 = fptoui <8 x double> undef to <8 x i1>
   %r140 = fptoui <8 x double> undef to <8 x i1>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r141 = fptosi <8 x double> undef to <8 x i1>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r141 = fptosi <8 x double> undef to <8 x i1>
   %r141 = fptosi <8 x double> undef to <8 x i1>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r142 = fptoui <8 x double> undef to <8 x i8>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r142 = fptoui <8 x double> undef to <8 x i8>
   %r142 = fptoui <8 x double> undef to <8 x i8>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r143 = fptosi <8 x double> undef to <8 x i8>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r143 = fptosi <8 x double> undef to <8 x i8>
   %r143 = fptosi <8 x double> undef to <8 x i8>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r144 = fptoui <8 x double> undef to <8 x i16>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r144 = fptoui <8 x double> undef to <8 x i16>
   %r144 = fptoui <8 x double> undef to <8 x i16>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r145 = fptosi <8 x double> undef to <8 x i16>
+  ; CHECK:  Found an estimated cost of 67 for instruction:   %r145 = fptosi <8 x double> undef to <8 x i16>
   %r145 = fptosi <8 x double> undef to <8 x i16>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r146 = fptoui <8 x double> undef to <8 x i32>
+  ; CHECK:  Found an estimated cost of 11 for instruction:   %r146 = fptoui <8 x double> undef to <8 x i32>
   %r146 = fptoui <8 x double> undef to <8 x i32>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r147 = fptosi <8 x double> undef to <8 x i32>
+  ; CHECK:  Found an estimated cost of 11 for instruction:   %r147 = fptosi <8 x double> undef to <8 x i32>
   %r147 = fptosi <8 x double> undef to <8 x i32>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r148 = fptoui <8 x double> undef to <8 x i64>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r148 = fptoui <8 x double> undef to <8 x i64>
   %r148 = fptoui <8 x double> undef to <8 x i64>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r149 = fptosi <8 x double> undef to <8 x i64>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r149 = fptosi <8 x double> undef to <8 x i64>
   %r149 = fptosi <8 x double> undef to <8 x i64>
 
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r150 = fptoui <16 x float> undef to <16 x i1>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r150 = fptoui <16 x float> undef to <16 x i1>
   %r150 = fptoui <16 x float> undef to <16 x i1>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r151 = fptosi <16 x float> undef to <16 x i1>
+  ; CHECK:  Found an estimated cost of 131 for instruction:   %r151 = fptosi <16 x float> undef to <16 x i1>
   %r151 = fptosi <16 x float> undef to <16 x i1>
- ; CHECK:  Found an estimated cost of 128 for instruction:   %r152 = fptoui <16 x float> undef to <16 x i8>
+ ; CHECK:  Found an estimated cost of 15 for instruction:   %r152 = fptoui <16 x float> undef to <16 x i8>
   %r152 = fptoui <16 x float> undef to <16 x i8>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r153 = fptosi <16 x float> undef to <16 x i8>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r153 = fptosi <16 x float> undef to <16 x i8>
   %r153 = fptosi <16 x float> undef to <16 x i8>
   ; CHECK:  Found an estimated cost of 8 for instruction:   %r154 = fptoui <16 x float> undef to <16 x i16>
   %r154 = fptoui <16 x float> undef to <16 x i16>
   ; CHECK:  Found an estimated cost of 8 for instruction:   %r155 = fptosi <16 x float> undef to <16 x i16>
   %r155 = fptosi <16 x float> undef to <16 x i16>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r156 = fptoui <16 x float> undef to <16 x i32>
   %r156 = fptoui <16 x float> undef to <16 x i32>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r157 = fptosi <16 x float> undef to <16 x i32>
   %r157 = fptosi <16 x float> undef to <16 x i32>
-  ; CHECK:  Found an estimated cost of 256 for instruction:   %r158 = fptoui <16 x float> undef to <16 x i64>
+  ; CHECK:  Found an estimated cost of 263 for instruction:   %r158 = fptoui <16 x float> undef to <16 x i64>
   %r158 = fptoui <16 x float> undef to <16 x i64>
-  ; CHECK:  Found an estimated cost of 256 for instruction:   %r159 = fptosi <16 x float> undef to <16 x i64>
+  ; CHECK:  Found an estimated cost of 263 for instruction:   %r159 = fptosi <16 x float> undef to <16 x i64>
   %r159 = fptosi <16 x float> undef to <16 x i64>
 
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r160 = fptoui <16 x double> undef to <16 x i1>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r160 = fptoui <16 x double> undef to <16 x i1>
   %r160 = fptoui <16 x double> undef to <16 x i1>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r161 = fptosi <16 x double> undef to <16 x i1>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r161 = fptosi <16 x double> undef to <16 x i1>
   %r161 = fptosi <16 x double> undef to <16 x i1>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r162 = fptoui <16 x double> undef to <16 x i8>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r162 = fptoui <16 x double> undef to <16 x i8>
   %r162 = fptoui <16 x double> undef to <16 x i8>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r163 = fptosi <16 x double> undef to <16 x i8>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r163 = fptosi <16 x double> undef to <16 x i8>
   %r163 = fptosi <16 x double> undef to <16 x i8>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r164 = fptoui <16 x double> undef to <16 x i16>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r164 = fptoui <16 x double> undef to <16 x i16>
   %r164 = fptoui <16 x double> undef to <16 x i16>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r165 = fptosi <16 x double> undef to <16 x i16>
+  ; CHECK:  Found an estimated cost of 135 for instruction:   %r165 = fptosi <16 x double> undef to <16 x i16>
   %r165 = fptosi <16 x double> undef to <16 x i16>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r166 = fptoui <16 x double> undef to <16 x i32>
+  ; CHECK:  Found an estimated cost of 23 for instruction:   %r166 = fptoui <16 x double> undef to <16 x i32>
   %r166 = fptoui <16 x double> undef to <16 x i32>
-  ; CHECK:  Found an estimated cost of 128 for instruction:   %r167 = fptosi <16 x double> undef to <16 x i32>
+  ; CHECK:  Found an estimated cost of 23 for instruction:   %r167 = fptosi <16 x double> undef to <16 x i32>
   %r167 = fptosi <16 x double> undef to <16 x i32>
-  ; CHECK:  Found an estimated cost of 256 for instruction:   %r168 = fptoui <16 x double> undef to <16 x i64>
+  ; CHECK:  Found an estimated cost of 263 for instruction:   %r168 = fptoui <16 x double> undef to <16 x i64>
   %r168 = fptoui <16 x double> undef to <16 x i64>
-  ; CHECK:  Found an estimated cost of 256 for instruction:   %r169 = fptosi <16 x double> undef to <16 x i64>
+  ; CHECK:  Found an estimated cost of 263 for instruction:   %r169 = fptosi <16 x double> undef to <16 x i64>
   %r169 = fptosi <16 x double> undef to <16 x i64>
 
   ; CHECK:  Found an estimated cost of 12 for instruction:   %r170 = uitofp <2 x i1> undef to <2 x float>
   %r170 = uitofp <2 x i1> undef to <2 x float>
   ; CHECK:  Found an estimated cost of 12 for instruction:   %r171 = sitofp <2 x i1> undef to <2 x float>
   %r171 = sitofp <2 x i1> undef to <2 x float>
   ; CHECK:  Found an estimated cost of 3 for instruction:   %r172 = uitofp <2 x i8> undef to <2 x float>
   %r172 = uitofp <2 x i8> undef to <2 x float>
; CHECK:  Found an estimated cost of 15 for instruction:   %r152 = fptoui <16 x float> undef to <16 x i8>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r194 = uitofp <4 x i16> undef to <4 x float>
   %r194 = uitofp <4 x i16> undef to <4 x float>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r195 = sitofp <4 x i16> undef to <4 x float>
   %r195 = sitofp <4 x i16> undef to <4 x float>
   ; CHECK:  Found an estimated cost of 1 for instruction:   %r196 = uitofp <4 x i32> undef to <4 x float>
   %r196 = uitofp <4 x i32> undef to <4 x float>
   ; CHECK:  Found an estimated cost of 1 for instruction:   %r197 = sitofp <4 x i32> undef to <4 x float>
   %r197 = sitofp <4 x i32> undef to <4 x float>
-  ; CHECK:  Found an estimated cost of 56 for instruction:   %r198 = uitofp <4 x i64> undef to <4 x float>
+  ; CHECK:  Found an estimated cost of 57 for instruction:   %r198 = uitofp <4 x i64> undef to <4 x float>
   %r198 = uitofp <4 x i64> undef to <4 x float>
-  ; CHECK:  Found an estimated cost of 56 for instruction:   %r199 = sitofp <4 x i64> undef to <4 x float>
+  ; CHECK:  Found an estimated cost of 57 for instruction:   %r199 = sitofp <4 x i64> undef to <4 x float>
   %r199 = sitofp <4 x i64> undef to <4 x float>
 
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r200 = uitofp <4 x i1> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 17 for instruction:   %r200 = uitofp <4 x i1> undef to <4 x double>
   %r200 = uitofp <4 x i1> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r201 = sitofp <4 x i1> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 17 for instruction:   %r201 = sitofp <4 x i1> undef to <4 x double>
   %r201 = sitofp <4 x i1> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r202 = uitofp <4 x i8> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 9 for instruction:   %r202 = uitofp <4 x i8> undef to <4 x double>
   %r202 = uitofp <4 x i8> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r203 = sitofp <4 x i8> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 9 for instruction:   %r203 = sitofp <4 x i8> undef to <4 x double>
   %r203 = sitofp <4 x i8> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r204 = uitofp <4 x i16> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r204 = uitofp <4 x i16> undef to <4 x double>
   %r204 = uitofp <4 x i16> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r205 = sitofp <4 x i16> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r205 = sitofp <4 x i16> undef to <4 x double>
   %r205 = sitofp <4 x i16> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r206 = uitofp <4 x i32> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 5 for instruction:   %r206 = uitofp <4 x i32> undef to <4 x double>
   %r206 = uitofp <4 x i32> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 16 for instruction:   %r207 = sitofp <4 x i32> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 5 for instruction:   %r207 = sitofp <4 x i32> undef to <4 x double>
   %r207 = sitofp <4 x i32> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r208 = uitofp <4 x i64> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 49 for instruction:   %r208 = uitofp <4 x i64> undef to <4 x double>
   %r208 = uitofp <4 x i64> undef to <4 x double>
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r209 = sitofp <4 x i64> undef to <4 x double>
+  ; CHECK:  Found an estimated cost of 49 for instruction:   %r209 = sitofp <4 x i64> undef to <4 x double>
   %r209 = sitofp <4 x i64> undef to <4 x double>
 
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r210 = uitofp <8 x i1> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r210 = uitofp <8 x i1> undef to <8 x float>
   %r210 = uitofp <8 x i1> undef to <8 x float>
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r211 = sitofp <8 x i1> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r211 = sitofp <8 x i1> undef to <8 x float>
   %r211 = sitofp <8 x i1> undef to <8 x float>
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r212 = uitofp <8 x i8> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r212 = uitofp <8 x i8> undef to <8 x float>
   %r212 = uitofp <8 x i8> undef to <8 x float>
-  ; CHECK:  Found an estimated cost of 48 for instruction:   %r213 = sitofp <8 x i8> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 7 for instruction:   %r213 = sitofp <8 x i8> undef to <8 x float>
   %r213 = sitofp <8 x i8> undef to <8 x float>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r214 = uitofp <8 x i16> undef to <8 x float>
   %r214 = uitofp <8 x i16> undef to <8 x float>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r215 = sitofp <8 x i16> undef to <8 x float>
   %r215 = sitofp <8 x i16> undef to <8 x float>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r216 = uitofp <8 x i32> undef to <8 x float>
   %r216 = uitofp <8 x i32> undef to <8 x float>
   ; CHECK:  Found an estimated cost of 2 for instruction:   %r217 = sitofp <8 x i32> undef to <8 x float>
   %r217 = sitofp <8 x i32> undef to <8 x float>
-  ; CHECK:  Found an estimated cost of 112 for instruction:   %r218 = uitofp <8 x i64> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 115 for instruction:   %r218 = uitofp <8 x i64> undef to <8 x float>
   %r218 = uitofp <8 x i64> undef to <8 x float>
-  ; CHECK:  Found an estimated cost of 112 for instruction:   %r219 = sitofp <8 x i64> undef to <8 x float>
+  ; CHECK:  Found an estimated cost of 115 for instruction:   %r219 = sitofp <8 x i64> undef to <8 x float>
   %r219 = sitofp <8 x i64> undef to <8 x float>
 
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r220 = uitofp <8 x i1> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 35 for instruction:   %r220 = uitofp <8 x i1> undef to <8 x double>
   %r220 = uitofp <8 x i1> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r221 = sitofp <8 x i1> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 35 for instruction:   %r221 = sitofp <8 x i1> undef to <8 x double>
   %r221 = sitofp <8 x i1> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r222 = uitofp <8 x i8> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 19 for instruction:   %r222 = uitofp <8 x i8> undef to <8 x double>
   %r222 = uitofp <8 x i8> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r223 = sitofp <8 x i8> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 19 for instruction:   %r223 = sitofp <8 x i8> undef to <8 x double>
   %r223 = sitofp <8 x i8> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r224 = uitofp <8 x i16> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r224 = uitofp <8 x i16> undef to <8 x double>
   %r224 = uitofp <8 x i16> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r225 = sitofp <8 x i16> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r225 = sitofp <8 x i16> undef to <8 x double>
   %r225 = sitofp <8 x i16> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r226 = uitofp <8 x i16> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r226 = uitofp <8 x i16> undef to <8 x double>
   %r226 = uitofp <8 x i16> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 32 for instruction:   %r227 = sitofp <8 x i16> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r227 = sitofp <8 x i16> undef to <8 x double>
   %r227 = sitofp <8 x i16> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r228 = uitofp <8 x i64> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 99 for instruction:   %r228 = uitofp <8 x i64> undef to <8 x double>
   %r228 = uitofp <8 x i64> undef to <8 x double>
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r229 = sitofp <8 x i64> undef to <8 x double>
+  ; CHECK:  Found an estimated cost of 99 for instruction:   %r229 = sitofp <8 x i64> undef to <8 x double>
   %r229 = sitofp <8 x i64> undef to <8 x double>
 
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r230 = uitofp <16 x i1> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r230 = uitofp <16 x i1> undef to <16 x float>
   %r230 = uitofp <16 x i1> undef to <16 x float>
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r231 = sitofp <16 x i1> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r231 = sitofp <16 x i1> undef to <16 x float>
   %r231 = sitofp <16 x i1> undef to <16 x float>
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r232 = uitofp <16 x i8> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r232 = uitofp <16 x i8> undef to <16 x float>
   %r232 = uitofp <16 x i8> undef to <16 x float>
-  ; CHECK:  Found an estimated cost of 96 for instruction:   %r233 = sitofp <16 x i8> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 15 for instruction:   %r233 = sitofp <16 x i8> undef to <16 x float>
   %r233 = sitofp <16 x i8> undef to <16 x float>
   ; CHECK:  Found an estimated cost of 8 for instruction:   %r234 = uitofp <16 x i16> undef to <16 x float>
   %r234 = uitofp <16 x i16> undef to <16 x float>
   ; CHECK:  Found an estimated cost of 8 for instruction:   %r235 = sitofp <16 x i16> undef to <16 x float>
   %r235 = sitofp <16 x i16> undef to <16 x float>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r236 = uitofp <16 x i32> undef to <16 x float>
   %r236 = uitofp <16 x i32> undef to <16 x float>
   ; CHECK:  Found an estimated cost of 4 for instruction:   %r237 = sitofp <16 x i32> undef to <16 x float>
   %r237 = sitofp <16 x i32> undef to <16 x float>
-  ; CHECK:  Found an estimated cost of 224 for instruction:   %r238 = uitofp <16 x i64> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 231 for instruction:   %r238 = uitofp <16 x i64> undef to <16 x float>
   %r238 = uitofp <16 x i64> undef to <16 x float>
-  ; CHECK:  Found an estimated cost of 224 for instruction:   %r239 = sitofp <16 x i64> undef to <16 x float>
+  ; CHECK:  Found an estimated cost of 231 for instruction:   %r239 = sitofp <16 x i64> undef to <16 x float>
   %r239 = sitofp <16 x i64> undef to <16 x float>
 
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r240 = uitofp <16 x i1> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 71 for instruction:   %r240 = uitofp <16 x i1> undef to <16 x double>
   %r240 = uitofp <16 x i1> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r241 = sitofp <16 x i1> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 71 for instruction:   %r241 = sitofp <16 x i1> undef to <16 x double>
   %r241 = sitofp <16 x i1> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r242 = uitofp <16 x i8> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 39 for instruction:   %r242 = uitofp <16 x i8> undef to <16 x double>
   %r242 = uitofp <16 x i8> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r243 = sitofp <16 x i8> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 39 for instruction:   %r243 = sitofp <16 x i8> undef to <16 x double>
   %r243 = sitofp <16 x i8> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r244 = uitofp <16 x i16> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 31 for instruction:   %r244 = uitofp <16 x i16> undef to <16 x double>
   %r244 = uitofp <16 x i16> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r245 = sitofp <16 x i16> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 31 for instruction:   %r245 = sitofp <16 x i16> undef to <16 x double>
   %r245 = sitofp <16 x i16> undef to <16 x double>
-  ; CHECK:  Found an estimated cost of 64 for instruction:   %r246 = uitofp <16 x i16> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 31 for instruction:   %r246 = uitofp <16 x i16> undef to <16 x double>
   %r246 = uitofp <16 x i16> undef to <16 x double>
-  ;  CHECK:  Found an estimated cost of 64 for instruction:   %r247 = sitofp <16 x i16> undef to <16 x double>
+  ; CHECK:  Found an estimated cost of 31 for instruction:   %r247 = sitofp <16 x i16> undef to <16 x double>
   %r247 = sitofp <16 x i16> undef to <16 x double>
-  ; CHECK:   Found an estimated cost of 192 for instruction:   %r248 = uitofp <16 x i64> undef to <16 x double>
+  ; CHECK:   Found an estimated cost of 199 for instruction:   %r248 = uitofp <16 x i64> undef to <16 x double>
   %r248 = uitofp <16 x i64> undef to <16 x double>
-  ; CHECK:   Found an estimated cost of 192 for instruction:   %r249 = sitofp <16 x i64> undef to <16 x double>
+  ; CHECK:   Found an estimated cost of 199 for instruction:   %r249 = sitofp <16 x i64> undef to <16 x double>
   %r249 = sitofp <16 x i64> undef to <16 x double>
 
   ; CHECK:   Found an estimated cost of 0 for instruction:   ret i32 undef
   ret i32 undef
 }
 

llvm/trunk/test/Analysis/CostModel/PowerPC/ext.ll

 ; RUN: opt < %s -cost-model -analyze -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr7 -mattr=+vsx | FileCheck %s
 target datalayout = "E-m:e-i64:64-n32:64"
 target triple = "powerpc64-unknown-linux-gnu"
 
 define void @exts() {
 
   ; CHECK: cost of 1 {{.*}} sext
   %v1 = sext i16 undef to i32
 
   ; CHECK: cost of 1 {{.*}} sext
   %v2 = sext <2 x i16> undef to <2 x i32>
 
   ; CHECK: cost of 1 {{.*}} sext
   %v3 = sext <4 x i16> undef to <4 x i32>
 
-  ; CHECK: cost of 112 {{.*}} sext
+  ; CHECK: cost of 3 {{.*}} sext
   %v4 = sext <8 x i16> undef to <8 x i32>
 
   ret void
 }
 

llvm/trunk/test/Analysis/CostModel/X86/sitofp.ll

define <4 x double> @sitofpv4i8v4double(<4 x i8> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @sitofpv8i8v8double(<8 x i8> %a) {
   ; SSE2-LABEL: sitofpv8i8v8double
   ; SSE2: cost of 80 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i8v8double
-  ; AVX1: cost of 20 {{.*}} sitofp
+  ; AVX1: cost of 7 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i8v8double
-  ; AVX2: cost of 20 {{.*}} sitofp
+  ; AVX2: cost of 7 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i8v8double
   ; AVX512F: cost of 2 {{.*}} sitofp
   %1 = sitofp <8 x i8> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @sitofpv16i8v16double(<16 x i8> %a) {
   ; SSE2-LABEL: sitofpv16i8v16double
   ; SSE2: cost of 160 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i8v16double
-  ; AVX1: cost of 40 {{.*}} sitofp
+  ; AVX1: cost of 15 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i8v16double
-  ; AVX2: cost of 40 {{.*}} sitofp
+  ; AVX2: cost of 15 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i8v16double
-  ; AVX512F: cost of 44 {{.*}} sitofp
+  ; AVX512F: cost of 5 {{.*}} sitofp
   %1 = sitofp <16 x i8> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @sitofpv32i8v32double(<32 x i8> %a) {
   ; SSE2-LABEL: sitofpv32i8v32double
   ; SSE2: cost of 320 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i8v32double
-  ; AVX1: cost of 80 {{.*}} sitofp
+  ; AVX1: cost of 31 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i8v32double
-  ; AVX2: cost of 80 {{.*}} sitofp
+  ; AVX2: cost of 31 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i8v32double
-  ; AVX512F: cost of 88 {{.*}} sitofp
+  ; AVX512F: cost of 11 {{.*}} sitofp
   %1 = sitofp <32 x i8> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @sitofpv2i16v2double(<2 x i16> %a) {
   ; SSE2-LABEL: sitofpv2i16v2double
   ; SSE2: cost of 20 {{.*}} sitofp
   ;
define <4 x double> @sitofpv4i16v4double(<4 x i16> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @sitofpv8i16v8double(<8 x i16> %a) {
   ; SSE2-LABEL: sitofpv8i16v8double
   ; SSE2: cost of 80 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i16v8double
-  ; AVX1: cost of 20 {{.*}} sitofp
+  ; AVX1: cost of 7 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i16v8double
-  ; AVX2: cost of 20 {{.*}} sitofp
+  ; AVX2: cost of 7 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i16v8double
   ; AVX512F: cost of 2 {{.*}} sitofp
   %1 = sitofp <8 x i16> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @sitofpv16i16v16double(<16 x i16> %a) {
   ; SSE2-LABEL: sitofpv16i16v16double
   ; SSE2: cost of 160 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i16v16double
-  ; AVX1: cost of 40 {{.*}} sitofp
+  ; AVX1: cost of 15 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i16v16double
-  ; AVX2: cost of 40 {{.*}} sitofp
+  ; AVX2: cost of 15 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i16v16double
-  ; AVX512F: cost of 44 {{.*}} sitofp
+  ; AVX512F: cost of 5 {{.*}} sitofp
   %1 = sitofp <16 x i16> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @sitofpv32i16v32double(<32 x i16> %a) {
   ; SSE2-LABEL: sitofpv32i16v32double
   ; SSE2: cost of 320 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i16v32double
-  ; AVX1: cost of 80 {{.*}} sitofp
+  ; AVX1: cost of 31 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i16v32double
-  ; AVX2: cost of 80 {{.*}} sitofp
+  ; AVX2: cost of 31 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i16v32double
-  ; AVX512F: cost of 88 {{.*}} sitofp
+  ; AVX512F: cost of 11 {{.*}} sitofp
   %1 = sitofp <32 x i16> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @sitofpv2i32v2double(<2 x i32> %a) {
   ; SSE2-LABEL: sitofpv2i32v2double
   ; SSE2: cost of 20 {{.*}} sitofp
   ;
define <4 x double> @sitofpv4i32v4double(<4 x i32> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @sitofpv8i32v8double(<8 x i32> %a) {
   ; SSE2-LABEL: sitofpv8i32v8double
   ; SSE2: cost of 80 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i32v8double
-  ; AVX1: cost of 20 {{.*}} sitofp
+  ; AVX1: cost of 3 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i32v8double
-  ; AVX2: cost of 20 {{.*}} sitofp
+  ; AVX2: cost of 3 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i32v8double
   ; AVX512F: cost of 1 {{.*}} sitofp
   %1 = sitofp <8 x i32> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @sitofpv16i32v16double(<16 x i32> %a) {
   ; SSE2-LABEL: sitofpv16i32v16double
   ; SSE2: cost of 160 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i32v16double
-  ; AVX1: cost of 40 {{.*}} sitofp
+  ; AVX1: cost of 7 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i32v16double
-  ; AVX2: cost of 40 {{.*}} sitofp
+  ; AVX2: cost of 7 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i32v16double
-  ; AVX512F: cost of 44 {{.*}} sitofp
+  ; AVX512F: cost of 3 {{.*}} sitofp
   %1 = sitofp <16 x i32> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @sitofpv32i32v32double(<32 x i32> %a) {
   ; SSE2-LABEL: sitofpv32i32v32double
   ; SSE2: cost of 320 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i32v32double
-  ; AVX1: cost of 80 {{.*}} sitofp
+  ; AVX1: cost of 15 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i32v32double
-  ; AVX2: cost of 80 {{.*}} sitofp
+  ; AVX2: cost of 15 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i32v32double
-  ; AVX512F: cost of 88 {{.*}} sitofp
+  ; AVX512F: cost of 7 {{.*}} sitofp
   %1 = sitofp <32 x i32> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @sitofpv2i64v2double(<2 x i64> %a) {
   ; SSE2-LABEL: sitofpv2i64v2double
   ; SSE2: cost of 20 {{.*}} sitofp
   ;
define <4 x double> @sitofpv4i64v4double(<4 x i64> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @sitofpv8i64v8double(<8 x i64> %a) {
   ; SSE2-LABEL: sitofpv8i64v8double
   ; SSE2: cost of 80 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i64v8double
-  ; AVX1: cost of 20 {{.*}} sitofp
+  ; AVX1: cost of 21 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i64v8double
-  ; AVX2: cost of 20 {{.*}} sitofp
+  ; AVX2: cost of 21 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i64v8double
   ; AVX512F: cost of 22 {{.*}} sitofp
   %1 = sitofp <8 x i64> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @sitofpv16i64v16double(<16 x i64> %a) {
   ; SSE2-LABEL: sitofpv16i64v16double
   ; SSE2: cost of 160 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i64v16double
-  ; AVX1: cost of 40 {{.*}} sitofp
+  ; AVX1: cost of 43 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i64v16double
-  ; AVX2: cost of 40 {{.*}} sitofp
+  ; AVX2: cost of 43 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i64v16double
-  ; AVX512F: cost of 44 {{.*}} sitofp
+  ; AVX512F: cost of 45 {{.*}} sitofp
   %1 = sitofp <16 x i64> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @sitofpv32i64v32double(<32 x i64> %a) {
   ; SSE2-LABEL: sitofpv32i64v32double
   ; SSE2: cost of 320 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i64v32double
-  ; AVX1: cost of 80 {{.*}} sitofp
+  ; AVX1: cost of 87 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i64v32double
-  ; AVX2: cost of 80 {{.*}} sitofp
+  ; AVX2: cost of 87 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i64v32double
-  ; AVX512F: cost of 88 {{.*}} sitofp
+  ; AVX512F: cost of 91 {{.*}} sitofp
   %1 = sitofp <32 x i64> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x float> @sitofpv2i8v2float(<2 x i8> %a) {
   ; SSE2-LABEL: sitofpv2i8v2float
   ; SSE2: cost of 15 {{.*}} sitofp
   ;
define <8 x float> @sitofpv8i8v8float(<8 x i8> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @sitofpv16i8v16float(<16 x i8> %a) {
   ; SSE2-LABEL: sitofpv16i8v16float
   ; SSE2: cost of 8 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i8v16float
-  ; AVX1: cost of 44 {{.*}} sitofp
+  ; AVX1: cost of 17 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i8v16float
-  ; AVX2: cost of 44 {{.*}} sitofp
+  ; AVX2: cost of 17 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i8v16float
   ; AVX512F: cost of 2 {{.*}} sitofp
   %1 = sitofp <16 x i8> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @sitofpv32i8v32float(<32 x i8> %a) {
   ; SSE2-LABEL: sitofpv32i8v32float
   ; SSE2: cost of 16 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i8v32float
-  ; AVX1: cost of 88 {{.*}} sitofp
+  ; AVX1: cost of 35 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i8v32float
-  ; AVX2: cost of 88 {{.*}} sitofp
+  ; AVX2: cost of 35 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i8v32float
-  ; AVX512F: cost of 92 {{.*}} sitofp
+  ; AVX512F: cost of 5 {{.*}} sitofp
   %1 = sitofp <32 x i8> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @sitofpv2i16v2float(<2 x i16> %a) {
   ; SSE2-LABEL: sitofpv2i16v2float
   ; SSE2: cost of 15 {{.*}} sitofp
   ;
define <8 x float> @sitofpv8i16v8float(<8 x i16> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @sitofpv16i16v16float(<16 x i16> %a) {
   ; SSE2-LABEL: sitofpv16i16v16float
   ; SSE2: cost of 30 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i16v16float
-  ; AVX1: cost of 44 {{.*}} sitofp
+  ; AVX1: cost of 11 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i16v16float
-  ; AVX2: cost of 44 {{.*}} sitofp
+  ; AVX2: cost of 11 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i16v16float
   ; AVX512F: cost of 2 {{.*}} sitofp
   %1 = sitofp <16 x i16> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @sitofpv32i16v32float(<32 x i16> %a) {
   ; SSE2-LABEL: sitofpv32i16v32float
   ; SSE2: cost of 60 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i16v32float
-  ; AVX1: cost of 88 {{.*}} sitofp
+  ; AVX1: cost of 23 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i16v32float
-  ; AVX2: cost of 88 {{.*}} sitofp
+  ; AVX2: cost of 23 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i16v32float
-  ; AVX512F: cost of 92 {{.*}} sitofp
+  ; AVX512F: cost of 5 {{.*}} sitofp
   %1 = sitofp <32 x i16> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @sitofpv2i32v2float(<2 x i32> %a) {
   ; SSE2-LABEL: sitofpv2i32v2float
   ; SSE2: cost of 15 {{.*}} sitofp
   ;
define <8 x float> @sitofpv8i32v8float(<8 x i32> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @sitofpv16i32v16float(<16 x i32> %a) {
   ; SSE2-LABEL: sitofpv16i32v16float
   ; SSE2: cost of 60 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i32v16float
-  ; AVX1: cost of 44 {{.*}} sitofp
+  ; AVX1: cost of 3 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i32v16float
-  ; AVX2: cost of 44 {{.*}} sitofp
+  ; AVX2: cost of 3 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i32v16float
   ; AVX512F: cost of 1 {{.*}} sitofp
   %1 = sitofp <16 x i32> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @sitofpv32i32v32float(<32 x i32> %a) {
   ; SSE2-LABEL: sitofpv32i32v32float
   ; SSE2: cost of 120 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i32v32float
-  ; AVX1: cost of 88 {{.*}} sitofp
+  ; AVX1: cost of 7 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i32v32float
-  ; AVX2: cost of 88 {{.*}} sitofp
+  ; AVX2: cost of 7 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i32v32float
-  ; AVX512F: cost of 92 {{.*}} sitofp
+  ; AVX512F: cost of 3 {{.*}} sitofp
   %1 = sitofp <32 x i32> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @sitofpv2i64v2float(<2 x i64> %a) {
   ; SSE2-LABEL: sitofpv2i64v2float
   ; SSE2: cost of 15 {{.*}} sitofp
   ;
define <4 x float> @sitofpv4i64v4float(<4 x i64> %a) {
   ret <4 x float> %1
 }
 
 define <8 x float> @sitofpv8i64v8float(<8 x i64> %a) {
   ; SSE2-LABEL: sitofpv8i64v8float
   ; SSE2: cost of 60 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i64v8float
-  ; AVX1: cost of 22 {{.*}} sitofp
+  ; AVX1: cost of 21 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i64v8float
-  ; AVX2: cost of 22 {{.*}} sitofp
+  ; AVX2: cost of 21 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i64v8float
   ; AVX512F: cost of 22 {{.*}} sitofp
   %1 = sitofp <8 x i64> %a to <8 x float>
   ret <8 x float> %1
 }
 
 define <16 x float> @sitofpv16i64v16float(<16 x i64> %a) {
   ; SSE2-LABEL: sitofpv16i64v16float
   ; SSE2: cost of 120 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i64v16float
-  ; AVX1: cost of 44 {{.*}} sitofp
+  ; AVX1: cost of 43 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i64v16float
-  ; AVX2: cost of 44 {{.*}} sitofp
+  ; AVX2: cost of 43 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i64v16float
-  ; AVX512F: cost of 46 {{.*}} sitofp
+  ; AVX512F: cost of 45 {{.*}} sitofp
   %1 = sitofp <16 x i64> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @sitofpv32i64v32float(<32 x i64> %a) {
   ; SSE2-LABEL: sitofpv32i64v32float
   ; SSE2: cost of 240 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv32i64v32float
-  ; AVX1: cost of 88 {{.*}} sitofp
+  ; AVX1: cost of 87 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv32i64v32float
-  ; AVX2: cost of 88 {{.*}} sitofp
+  ; AVX2: cost of 87 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv32i64v32float
-  ; AVX512F: cost of 92 {{.*}} sitofp
+  ; AVX512F: cost of 91 {{.*}} sitofp
   %1 = sitofp <32 x i64> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <8 x double> @sitofpv8i1v8double(<8 x double> %a) {
   ; SSE2-LABEL: sitofpv8i1v8double
   ; SSE2: cost of 80 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv8i1v8double
-  ; AVX1: cost of 20 {{.*}} sitofp
+  ; AVX1: cost of 7 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv8i1v8double
-  ; AVX2: cost of 20 {{.*}} sitofp
+  ; AVX2: cost of 7 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv8i1v8double
   ; AVX512F: cost of 4 {{.*}} sitofp
   %cmpres = fcmp ogt <8 x double> %a, zeroinitializer
   %1 = sitofp <8 x i1> %cmpres to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x float> @sitofpv16i1v16float(<16 x float> %a) {
   ; SSE2-LABEL: sitofpv16i1v16float
   ; SSE2: cost of 8 {{.*}} sitofp
   ;
   ; AVX1-LABEL: sitofpv16i1v16float
-  ; AVX1: cost of 44 {{.*}} sitofp
+  ; AVX1: cost of 17 {{.*}} sitofp
   ;
   ; AVX2-LABEL: sitofpv16i1v16float
-  ; AVX2: cost of 44 {{.*}} sitofp
+  ; AVX2: cost of 17 {{.*}} sitofp
   ;
   ; AVX512F-LABEL: sitofpv16i1v16float
   ; AVX512F: cost of 3 {{.*}} sitofp
   %cmpres = fcmp ogt <16 x float> %a, zeroinitializer
   %1 = sitofp <16 x i1> %cmpres to <16 x float>
   ret <16 x float> %1
 }

llvm/trunk/test/Analysis/CostModel/X86/uitofp.ll

define <4 x double> @uitofpv4i8v4double(<4 x i8> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @uitofpv8i8v8double(<8 x i8> %a) {
   ; SSE2-LABEL: uitofpv8i8v8double
   ; SSE2: cost of 80 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv8i8v8double
-  ; AVX1: cost of 20 {{.*}} uitofp
+  ; AVX1: cost of 5 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv8i8v8double
-  ; AVX2: cost of 20 {{.*}} uitofp
+  ; AVX2: cost of 5 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv8i8v8double
   ; AVX512F: cost of 2 {{.*}} uitofp
   %1 = uitofp <8 x i8> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @uitofpv16i8v16double(<16 x i8> %a) {
   ; SSE2-LABEL: uitofpv16i8v16double
   ; SSE2: cost of 160 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i8v16double
-  ; AVX1: cost of 40 {{.*}} uitofp
+  ; AVX1: cost of 11 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i8v16double
-  ; AVX2: cost of 40 {{.*}} uitofp
+  ; AVX2: cost of 11 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i8v16double
-  ; AVX512F: cost of 44 {{.*}} uitofp
+  ; AVX512F: cost of 5 {{.*}} uitofp
   %1 = uitofp <16 x i8> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @uitofpv32i8v32double(<32 x i8> %a) {
   ; SSE2-LABEL: uitofpv32i8v32double
   ; SSE2: cost of 320 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i8v32double
-  ; AVX1: cost of 80 {{.*}} uitofp
+  ; AVX1: cost of 23 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i8v32double
-  ; AVX2: cost of 80 {{.*}} uitofp
+  ; AVX2: cost of 23 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i8v32double
-  ; AVX512F: cost of 88 {{.*}} uitofp
+  ; AVX512F: cost of 11 {{.*}} uitofp
   %1 = uitofp <32 x i8> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @uitofpv2i16v2double(<2 x i16> %a) {
   ; SSE2-LABEL: uitofpv2i16v2double
   ; SSE2: cost of 20 {{.*}} uitofp
   ;
define <4 x double> @uitofpv4i16v4double(<4 x i16> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @uitofpv8i16v8double(<8 x i16> %a) {
   ; SSE2-LABEL: uitofpv8i16v8double
   ; SSE2: cost of 80 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv8i16v8double
-  ; AVX1: cost of 20 {{.*}} uitofp
+  ; AVX1: cost of 5 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv8i16v8double
-  ; AVX2: cost of 20 {{.*}} uitofp
+  ; AVX2: cost of 5 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv8i16v8double
   ; AVX512F: cost of 2 {{.*}} uitofp
   %1 = uitofp <8 x i16> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @uitofpv16i16v16double(<16 x i16> %a) {
   ; SSE2-LABEL: uitofpv16i16v16double
   ; SSE2: cost of 160 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i16v16double
-  ; AVX1: cost of 40 {{.*}} uitofp
+  ; AVX1: cost of 11 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i16v16double
-  ; AVX2: cost of 40 {{.*}} uitofp
+  ; AVX2: cost of 11 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i16v16double
-  ; AVX512F: cost of 44 {{.*}} uitofp
+  ; AVX512F: cost of 5 {{.*}} uitofp
   %1 = uitofp <16 x i16> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @uitofpv32i16v32double(<32 x i16> %a) {
   ; SSE2-LABEL: uitofpv32i16v32double
   ; SSE2: cost of 320 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i16v32double
-  ; AVX1: cost of 80 {{.*}} uitofp
+  ; AVX1: cost of 23 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i16v32double
-  ; AVX2: cost of 80 {{.*}} uitofp
+  ; AVX2: cost of 23 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i16v32double
-  ; AVX512F: cost of 88 {{.*}} uitofp
+  ; AVX512F: cost of 11 {{.*}} uitofp
   %1 = uitofp <32 x i16> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @uitofpv2i32v2double(<2 x i32> %a) {
   ; SSE2-LABEL: uitofpv2i32v2double
   ; SSE2: cost of 20 {{.*}} uitofp
   ;
define <4 x double> @uitofpv4i32v4double(<4 x i32> %a) {
   ret <4 x double> %1
 }
 
 define <8 x double> @uitofpv8i32v8double(<8 x i32> %a) {
   ; SSE2-LABEL: uitofpv8i32v8double
   ; SSE2: cost of 80 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv8i32v8double
-  ; AVX1: cost of 20 {{.*}} uitofp
+  ; AVX1: cost of 13 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv8i32v8double
-  ; AVX2: cost of 20 {{.*}} uitofp
+  ; AVX2: cost of 13 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv8i32v8double
   ; AVX512F: cost of 1 {{.*}} uitofp
   %1 = uitofp <8 x i32> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @uitofpv16i32v16double(<16 x i32> %a) {
   ; SSE2-LABEL: uitofpv16i32v16double
   ; SSE2: cost of 160 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i32v16double
-  ; AVX1: cost of 40 {{.*}} uitofp
+  ; AVX1: cost of 27 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i32v16double
-  ; AVX2: cost of 40 {{.*}} uitofp
+  ; AVX2: cost of 27 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i32v16double
-  ; AVX512F: cost of 44 {{.*}} uitofp
+  ; AVX512F: cost of 3 {{.*}} uitofp
   %1 = uitofp <16 x i32> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @uitofpv32i32v32double(<32 x i32> %a) {
   ; SSE2-LABEL: uitofpv32i32v32double
   ; SSE2: cost of 320 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i32v32double
-  ; AVX1: cost of 80 {{.*}} uitofp
+  ; AVX1: cost of 55 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i32v32double
-  ; AVX2: cost of 80 {{.*}} uitofp
+  ; AVX2: cost of 55 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i32v32double
-  ; AVX512F: cost of 88 {{.*}} uitofp
+  ; AVX512F: cost of 7 {{.*}} uitofp
   %1 = uitofp <32 x i32> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x double> @uitofpv2i64v2double(<2 x i64> %a) {
   ; SSE2-LABEL: uitofpv2i64v2double
   ; SSE2: cost of 20 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv2i64v2double
   ; AVX1: cost of 20 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv2i64v2double
   ; AVX2: cost of 20 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv2i64v2double
   ; AVX512F: cost of 5 {{.*}} uitofp
   ;
-  ; AVX512DQ: uitofpv2i64v2double
+  ; AVX512DQ-LABEL: uitofpv2i64v2double
   ; AVX512DQ: cost of 1 {{.*}} uitofp
   %1 = uitofp <2 x i64> %a to <2 x double>
   ret <2 x double> %1
 }
 
 define <4 x double> @uitofpv4i64v4double(<4 x i64> %a) {
   ; SSE2-LABEL: uitofpv4i64v4double
   ; SSE2: cost of 40 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv4i64v4double
   ; AVX1: cost of 40 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv4i64v4double
   ; AVX2: cost of 40 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv4i64v4double
   ; AVX512F: cost of 12 {{.*}} uitofp
   ;
-  ; AVX512DQ: uitofpv4i64v4double
+  ; AVX512DQ-LABEL: uitofpv4i64v4double
   ; AVX512DQ: cost of 1 {{.*}} uitofp
   %1 = uitofp <4 x i64> %a to <4 x double>
   ret <4 x double> %1
 }
 
 define <8 x double> @uitofpv8i64v8double(<8 x i64> %a) {
   ; SSE2-LABEL: uitofpv8i64v8double
   ; SSE2: cost of 80 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv8i64v8double
-  ; AVX1: cost of 20 {{.*}} uitofp
+  ; AVX1: cost of 81 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv8i64v8double
-  ; AVX2: cost of 20 {{.*}} uitofp
+  ; AVX2: cost of 81 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv8i64v8double
   ; AVX512F: cost of 26 {{.*}} uitofp
   ;
-  ; AVX512DQ: uitofpv8i64v8double
+  ; AVX512DQ-LABEL: uitofpv8i64v8double
   ; AVX512DQ: cost of 1 {{.*}} uitofp
   %1 = uitofp <8 x i64> %a to <8 x double>
   ret <8 x double> %1
 }
 
 define <16 x double> @uitofpv16i64v16double(<16 x i64> %a) {
   ; SSE2-LABEL: uitofpv16i64v16double
   ; SSE2: cost of 160 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i64v16double
-  ; AVX1: cost of 40 {{.*}} uitofp
+  ; AVX1: cost of 163 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i64v16double
-  ; AVX2: cost of 40 {{.*}} uitofp
+  ; AVX2: cost of 163 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i64v16double
-  ; AVX512F: cost of 44 {{.*}} uitofp
+  ; AVX512F: cost of 53 {{.*}} uitofp
   ;
-  ; AVX512DQ: uitofpv16i64v16double
-  ; AVX512DQ: cost of 44 {{.*}} uitofp
+  ; AVX512DQ-LABEL: uitofpv16i64v16double
+  ; AVX512DQ: cost of 3 {{.*}} uitofp
   %1 = uitofp <16 x i64> %a to <16 x double>
   ret <16 x double> %1
 }
 
 define <32 x double> @uitofpv32i64v32double(<32 x i64> %a) {
   ; SSE2-LABEL: uitofpv32i64v32double
   ; SSE2: cost of 320 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i64v32double
-  ; AVX1: cost of 80 {{.*}} uitofp
+  ; AVX1: cost of 327 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i64v32double
-  ; AVX2: cost of 80 {{.*}} uitofp
+  ; AVX2: cost of 327 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i64v32double
-  ; AVX512F: cost of 88 {{.*}} uitofp
+  ; AVX512F: cost of 107 {{.*}} uitofp
   ;
-  ; AVX512DQ: uitofpv32i64v32double
-  ; AVX512DQ: cost of 88 {{.*}} uitofp
+  ; AVX512DQ-LABEL: uitofpv32i64v32double
+  ; AVX512DQ: cost of 2 {{.*}} uitofp
   %1 = uitofp <32 x i64> %a to <32 x double>
   ret <32 x double> %1
 }
 
 define <2 x float> @uitofpv2i8v2float(<2 x i8> %a) {
   ; SSE2-LABEL: uitofpv2i8v2float
   ; SSE2: cost of 15 {{.*}} uitofp
   ;
define <8 x float> @uitofpv8i8v8float(<8 x i8> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @uitofpv16i8v16float(<16 x i8> %a) {
   ; SSE2-LABEL: uitofpv16i8v16float
   ; SSE2: cost of 8 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i8v16float
-  ; AVX1: cost of 44 {{.*}} uitofp
+  ; AVX1: cost of 11 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i8v16float
-  ; AVX2: cost of 44 {{.*}} uitofp
+  ; AVX2: cost of 11 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i8v16float
   ; AVX512F: cost of 2 {{.*}} uitofp
   %1 = uitofp <16 x i8> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @uitofpv32i8v32float(<32 x i8> %a) {
   ; SSE2-LABEL: uitofpv32i8v32float
   ; SSE2: cost of 16 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i8v32float
-  ; AVX1: cost of 88 {{.*}} uitofp
+  ; AVX1: cost of 23 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i8v32float
-  ; AVX2: cost of 88 {{.*}} uitofp
+  ; AVX2: cost of 23 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i8v32float
-  ; AVX512F: cost of 92 {{.*}} uitofp
+  ; AVX512F: cost of 5 {{.*}} uitofp
   %1 = uitofp <32 x i8> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @uitofpv2i16v2float(<2 x i16> %a) {
   ; SSE2-LABEL: uitofpv2i16v2float
   ; SSE2: cost of 15 {{.*}} uitofp
   ;
define <8 x float> @uitofpv8i16v8float(<8 x i16> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @uitofpv16i16v16float(<16 x i16> %a) {
   ; SSE2-LABEL: uitofpv16i16v16float
   ; SSE2: cost of 30 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i16v16float
-  ; AVX1: cost of 44 {{.*}} uitofp
+  ; AVX1: cost of 11 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i16v16float
-  ; AVX2: cost of 44 {{.*}} uitofp
+  ; AVX2: cost of 11 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i16v16float
   ; AVX512F: cost of 2 {{.*}} uitofp
   %1 = uitofp <16 x i16> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @uitofpv32i16v32float(<32 x i16> %a) {
   ; SSE2-LABEL: uitofpv32i16v32float
   ; SSE2: cost of 60 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i16v32float
-  ; AVX1: cost of 88 {{.*}} uitofp
+  ; AVX1: cost of 23 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i16v32float
-  ; AVX2: cost of 88 {{.*}} uitofp
+  ; AVX2: cost of 23 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i16v32float
-  ; AVX512F: cost of 92 {{.*}} uitofp
+  ; AVX512F: cost of 5 {{.*}} uitofp
   %1 = uitofp <32 x i16> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @uitofpv2i32v2float(<2 x i32> %a) {
   ; SSE2-LABEL: uitofpv2i32v2float
   ; SSE2: cost of 15 {{.*}} uitofp
   ;
define <8 x float> @uitofpv8i32v8float(<8 x i32> %a) {
   ret <8 x float> %1
 }
 
 define <16 x float> @uitofpv16i32v16float(<16 x i32> %a) {
   ; SSE2-LABEL: uitofpv16i32v16float
   ; SSE2: cost of 32 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i32v16float
-  ; AVX1: cost of 44 {{.*}} uitofp
+  ; AVX1: cost of 19 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i32v16float
-  ; AVX2: cost of 44 {{.*}} uitofp
+  ; AVX2: cost of 17 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i32v16float
   ; AVX512F: cost of 1 {{.*}} uitofp
   %1 = uitofp <16 x i32> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @uitofpv32i32v32float(<32 x i32> %a) {
   ; SSE2-LABEL: uitofpv32i32v32float
   ; SSE2: cost of 64 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i32v32float
-  ; AVX1: cost of 88 {{.*}} uitofp
+  ; AVX1: cost of 39 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i32v32float
-  ; AVX2: cost of 88 {{.*}} uitofp
+  ; AVX2: cost of 35 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i32v32float
-  ; AVX512F: cost of 92 {{.*}} uitofp
+  ; AVX512F: cost of 3 {{.*}} uitofp
   %1 = uitofp <32 x i32> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <2 x float> @uitofpv2i64v2float(<2 x i64> %a) {
   ; SSE2-LABEL: uitofpv2i64v2float
   ; SSE2: cost of 15 {{.*}} uitofp
   ;
define <4 x float> @uitofpv4i64v4float(<4 x i64> %a) {
   ret <4 x float> %1
 }
 
 define <8 x float> @uitofpv8i64v8float(<8 x i64> %a) {
   ; SSE2-LABEL: uitofpv8i64v8float
   ; SSE2: cost of 60 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv8i64v8float
-  ; AVX1: cost of 22 {{.*}} uitofp
+  ; AVX1: cost of 21 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv8i64v8float
-  ; AVX2: cost of 22 {{.*}} uitofp
+  ; AVX2: cost of 21 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv8i64v8float
   ; AVX512F: cost of 22 {{.*}} uitofp
   %1 = uitofp <8 x i64> %a to <8 x float>
   ret <8 x float> %1
 }
 
 define <16 x float> @uitofpv16i64v16float(<16 x i64> %a) {
   ; SSE2-LABEL: uitofpv16i64v16float
   ; SSE2: cost of 120 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv16i64v16float
-  ; AVX1: cost of 44 {{.*}} uitofp
+  ; AVX1: cost of 43 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv16i64v16float
-  ; AVX2: cost of 44 {{.*}} uitofp
+  ; AVX2: cost of 43 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv16i64v16float
-  ; AVX512F: cost of 46 {{.*}} uitofp
+  ; AVX512F: cost of 45 {{.*}} uitofp
   %1 = uitofp <16 x i64> %a to <16 x float>
   ret <16 x float> %1
 }
 
 define <32 x float> @uitofpv32i64v32float(<32 x i64> %a) {
   ; SSE2-LABEL: uitofpv32i64v32float
   ; SSE2: cost of 240 {{.*}} uitofp
   ;
   ; AVX1-LABEL: uitofpv32i64v32float
-  ; AVX1: cost of 88 {{.*}} uitofp
+  ; AVX1: cost of 87 {{.*}} uitofp
   ;
   ; AVX2-LABEL: uitofpv32i64v32float
-  ; AVX2: cost of 88 {{.*}} uitofp
+  ; AVX2: cost of 87 {{.*}} uitofp
   ;
   ; AVX512F-LABEL: uitofpv32i64v32float
-  ; AVX512F: cost of 92 {{.*}} uitofp
+  ; AVX512F: cost of 91 {{.*}} uitofp
   %1 = uitofp <32 x i64> %a to <32 x float>
   ret <32 x float> %1
 }
 
 define <8 x i32> @fptouiv8f32v8i32(<8 x float> %a) {
   ; AVX512F-LABEL: fptouiv8f32v8i32
   ; AVX512F: cost of 1 {{.*}} fptoui
   %1 = fptoui <8 x float> %a to <8 x i32>

llvm/trunk/test/Transforms/LoopVectorize/X86/gather_scatter.ll

 ;  for (int i=0; i < SIZE; ++i) {
 ;    if (trigger[i] > 0) {
 ;      out[i] = in[index[i]] + (float) 0.5;
 ;    }
 ;  }
 ;}
 
 ;AVX512-LABEL: @foo1
-;AVX512:  llvm.masked.load.v8i32
-;AVX512: llvm.masked.gather.v8f32
-;AVX512: llvm.masked.store.v8f32
+;AVX512: llvm.masked.load.v16i32
+;AVX512: llvm.masked.gather.v16f32
+;AVX512: llvm.masked.store.v16f32
 ;AVX512: ret void
 
 ; Function Attrs: nounwind uwtable
 define void @foo1(float* noalias %in, float* noalias %out, i32* noalias %trigger, i32* noalias %index) {
 entry:
   %in.addr = alloca float*, align 8
   %out.addr = alloca float*, align 8
   %trigger.addr = alloca i32*, align 8