[CostModel][SVE] Add cost model for shuffle reverse with i1 and scalable vector

[CarolineConcatto]

Summary

This patch adds the cost model for experimental.vector.reverse
with scalable vector types: nxv16i1, nxv8i1, nxv4i1 and nxv2i1.
These types are missing from the previous cost model patch D95603.

The cost model for experimental.vector.reverse with 1 bit mask is used by
loop vectorization in the patch D95363

[sdesmalen]

llvm/test/Analysis/CostModel/AArch64/getIntrinsicInstrCost-vector-reverse.ll

50 ↗

(On Diff #327385)

The operand is not scalable, so this mangling of the intrinsic is not correct.

llvm/test/Analysis/CostModel/AArch64/sve-getIntrinsicInstrCost-vector-reverse.ll

20 ↗

(On Diff #327385)

Just an FYI that this file has been removed, so you'll to rebase it and put the changes in llvm/test/Analysis/CostModel/AArch64 /sve-intrinsics.ll.

[CarolineConcatto]

• update the llvm-ir test
• rebase

[CarolineConcatto]

• return file getIntrinsicInstrCost-vector-reverse.ll with the i1 test

[CarolineConcatto]

Thank you Sander!
My mistake not update main. Thank you for pointing it out.
Now the test is based on the latest main repo.

llvm/test/Analysis/CostModel/AArch64/getIntrinsicInstrCost-vector-reverse.ll

50 ↗

(On Diff #327385)

Thank you Sander.
It was a C&P mistake.

[david-arm]

LGTM!

[sdesmalen]

Other than my comment on removing the fixed-width tests, the patch looks good to me.

llvm/test/Analysis/CostModel/AArch64/getIntrinsicInstrCost-vector-reverse.ll

53 ↗

(On Diff #327471)

I now see your code-changes only relate to scalable vectors, so these fixed-width tests are not really covering any changes in your patch. I'd suggest just removing them.

llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

static const CostTblEntry ShuffleTbl[] = {
       { TTI::SK_Reverse, MVT::nxv16i8,  1 },
       { TTI::SK_Reverse, MVT::nxv8i16,  1 },
       { TTI::SK_Reverse, MVT::nxv4i32,  1 },
       { TTI::SK_Reverse, MVT::nxv2i64,  1 },
       { TTI::SK_Reverse, MVT::nxv8f16,  1 },
       { TTI::SK_Reverse, MVT::nxv8bf16, 1 },
       { TTI::SK_Reverse, MVT::nxv4f32,  1 },
       { TTI::SK_Reverse, MVT::nxv2f64,  1 },
+      { TTI::SK_Reverse, MVT::nxv16i1,  1 },
+      { TTI::SK_Reverse, MVT::nxv8i1,   1 },
+      { TTI::SK_Reverse, MVT::nxv4i1,   1 },
+      { TTI::SK_Reverse, MVT::nxv2i1,   1 },
     };
     std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, Tp);
     if (const auto *Entry = CostTableLookup(ShuffleTbl, Kind, LT.second))
       return LT.first * Entry->Cost;
   }
 
   return BaseT::getShuffleCost(Kind, Tp, Index, SubTp);
 }

llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll

 ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %reverse_nxv8if16 = call <vscale x 8 x half> @llvm.experimental.vector.reverse.nxv8f16(<vscale x 8 x half> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %reverse_nxv16f16 = call <vscale x 16 x half> @llvm.experimental.vector.reverse.nxv16f16(<vscale x 16 x half> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %reverse_nxv4f32 = call <vscale x 4 x float> @llvm.experimental.vector.reverse.nxv4f32(<vscale x 4 x float> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %reverse_nxv8f32 = call <vscale x 8 x float> @llvm.experimental.vector.reverse.nxv8f32(<vscale x 8 x float> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %reverse_nxv2f64 = call <vscale x 2 x double> @llvm.experimental.vector.reverse.nxv2f64(<vscale x 2 x double> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %reverse_nxv4f64 = call <vscale x 4 x double> @llvm.experimental.vector.reverse.nxv4f64(<vscale x 4 x double> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %reverse_nxv8bf16 = call <vscale x 8 x bfloat> @llvm.experimental.vector.reverse.nxv8bf16(<vscale x 8 x bfloat> undef)
 ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %reverse_nxv16bf16 = call <vscale x 16 x bfloat> @llvm.experimental.vector.reverse.nxv16bf16(<vscale x 16 x bfloat> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction:   %reverse_nxv16i1 = call <vscale x 16 x i1> @llvm.experimental.vector.reverse.nxv16i1(<vscale x 16 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction:   %reverse_nxv8i1 = call <vscale x 8 x i1> @llvm.experimental.vector.reverse.nxv8i1(<vscale x 8 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction:   %reverse_nxv4i1 = call <vscale x 4 x i1> @llvm.experimental.vector.reverse.nxv4i1(<vscale x 4 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction:   %reverse_nxv2i1 = call <vscale x 2 x i1> @llvm.experimental.vector.reverse.nxv2i1(<vscale x 2 x i1> undef)
+
   %reverse_nxv16i8 = call <vscale x 16 x i8> @llvm.experimental.vector.reverse.nxv16i8(<vscale x 16 x i8> undef)
   %reverse_nxv32i8 = call <vscale x 32 x i8> @llvm.experimental.vector.reverse.nxv32i8(<vscale x 32 x i8> undef)
   %reverse_nxv8i16 = call <vscale x 8 x i16> @llvm.experimental.vector.reverse.nxv8i16(<vscale x 8 x i16> undef)
   %reverse_nxv16i16 = call <vscale x 16 x i16> @llvm.experimental.vector.reverse.nxv16i16(<vscale x 16 x i16> undef)
   %reverse_nxv4i32 = call <vscale x 4 x i32> @llvm.experimental.vector.reverse.nxv4i32(<vscale x 4 x i32> undef)
   %reverse_nxv8i32 = call <vscale x 8 x i32> @llvm.experimental.vector.reverse.nxv8i32(<vscale x 8 x i32> undef)
   %reverse_nxv2i64 = call <vscale x 2 x i64> @llvm.experimental.vector.reverse.nxv2i64(<vscale x 2 x i64> undef)
   %reverse_nxv4i64 = call <vscale x 4 x i64> @llvm.experimental.vector.reverse.nxv4i64(<vscale x 4 x i64> undef)
   %reverse_nxv8if16 = call <vscale x 8 x half> @llvm.experimental.vector.reverse.nxv8f16(<vscale x 8 x half> undef)
   %reverse_nxv16f16 = call <vscale x 16 x half> @llvm.experimental.vector.reverse.nxv16f16(<vscale x 16 x half> undef)
   %reverse_nxv4f32 = call <vscale x 4 x float> @llvm.experimental.vector.reverse.nxv4f32(<vscale x 4 x float> undef)
   %reverse_nxv8f32 = call <vscale x 8 x float> @llvm.experimental.vector.reverse.nxv8f32(<vscale x 8 x float> undef)
   %reverse_nxv2f64 = call <vscale x 2 x double> @llvm.experimental.vector.reverse.nxv2f64(<vscale x 2 x double> undef)
   %reverse_nxv4f64 = call <vscale x 4 x double> @llvm.experimental.vector.reverse.nxv4f64(<vscale x 4 x double> undef)
   %reverse_nxv8bf16 = call <vscale x 8 x bfloat> @llvm.experimental.vector.reverse.nxv8bf16(<vscale x 8 x bfloat> undef)
   %reverse_nxv16bf16 = call <vscale x 16 x bfloat> @llvm.experimental.vector.reverse.nxv16bf16(<vscale x 16 x bfloat> undef)
+  %reverse_nxv16i1 = call <vscale x 16 x i1> @llvm.experimental.vector.reverse.nxv16i1(<vscale x 16 x i1> undef)
+  %reverse_nxv8i1 =  call <vscale x 8 x i1> @llvm.experimental.vector.reverse.nxv8i1(<vscale x 8 x i1> undef)
+  %reverse_nxv4i1 = call <vscale x 4 x i1> @llvm.experimental.vector.reverse.nxv4i1(<vscale x 4 x i1> undef)
+  %reverse_nxv2i1 = call <vscale x 2 x i1> @llvm.experimental.vector.reverse.nxv2i1(<vscale x 2 x i1> undef)
   ret void
 }
 declare <vscale x 16 x i8> @llvm.experimental.vector.reverse.nxv16i8(<vscale x 16 x i8>)
 declare <vscale x 32 x i8> @llvm.experimental.vector.reverse.nxv32i8(<vscale x 32 x i8>)
 declare <vscale x 8 x i16> @llvm.experimental.vector.reverse.nxv8i16(<vscale x 8 x i16>)
 declare <vscale x 16 x i16> @llvm.experimental.vector.reverse.nxv16i16(<vscale x 16 x i16>)
 declare <vscale x 4 x i32> @llvm.experimental.vector.reverse.nxv4i32(<vscale x 4 x i32>)
 declare <vscale x 8 x i32> @llvm.experimental.vector.reverse.nxv8i32(<vscale x 8 x i32>)
 declare <vscale x 2 x i64> @llvm.experimental.vector.reverse.nxv2i64(<vscale x 2 x i64>)
 declare <vscale x 4 x i64> @llvm.experimental.vector.reverse.nxv4i64(<vscale x 4 x i64>)
 declare <vscale x 8 x half> @llvm.experimental.vector.reverse.nxv8f16(<vscale x 8 x half>)
 declare <vscale x 16 x half> @llvm.experimental.vector.reverse.nxv16f16(<vscale x 16 x half>)
 declare <vscale x 4 x float> @llvm.experimental.vector.reverse.nxv4f32(<vscale x 4 x float>)
 declare <vscale x 8 x float> @llvm.experimental.vector.reverse.nxv8f32(<vscale x 8 x float>)
 declare <vscale x 2 x double> @llvm.experimental.vector.reverse.nxv2f64(<vscale x 2 x double>)
 declare <vscale x 4 x double> @llvm.experimental.vector.reverse.nxv4f64(<vscale x 4 x double>)
 declare <vscale x 8 x bfloat> @llvm.experimental.vector.reverse.nxv8bf16(<vscale x 8 x bfloat>)
 declare <vscale x 16 x bfloat> @llvm.experimental.vector.reverse.nxv16bf16(<vscale x 16 x bfloat>)
+declare <vscale x 16 x i1> @llvm.experimental.vector.reverse.nxv16i1(<vscale x 16 x i1>)
+declare <vscale x 8 x i1> @llvm.experimental.vector.reverse.nxv8i1(<vscale x 8 x i1>)
+declare <vscale x 4 x i1> @llvm.experimental.vector.reverse.nxv4i1(<vscale x 4 x i1>)
+declare <vscale x 2 x i1> @llvm.experimental.vector.reverse.nxv2i1(<vscale x 2 x i1>)
 
 attributes #0 = { "target-features"="+sve,+bf16" }