Reapply "[RISCV][TTI] Extract subvector at index zero is free (#81751)"

This reverts commit 834d11c which was
a revert of my 3a62693.  I had failed
to rebase after new tests added overnight by
fc0b67e.

Original commit message follows:

Extracing a subvector at index zero corresponds to a type conversion and
possibly a subregister operation. We will not emit a vslidedown. As such,
they are free.

As an aside, it looks like we're not passing an index in for cases where
the subvec type is scalable. For at least index zero, we probably should be.

Revert "Revert "[RISCV][TTI] Extract subvector at index zero is free (#81751)""

Author: preames

llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp

@@ -432,6 +432,10 @@ InstructionCost RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
     // must be implemented here.
     break;
   case TTI::SK_ExtractSubvector:
+    // Extract at zero is always a subregister extract
+    if (Index == 0)
+      return TTI::TCC_Free;
+
     // Example sequence:
     // vsetivli     zero, 4, e8, mf2, tu, ma (ignored)
     // vslidedown.vi  v8, v9, 2

llvm/test/Analysis/CostModel/RISCV/rvv-shuffle.ll

@@ -52,14 +52,14 @@ define void  @vector_broadcast() {
 
 define void @vector_insert_extract(<vscale x 4 x i32> %v0, <vscale x 16 x i32> %v1, <16 x i32> %v2) {
 ; CHECK-LABEL: 'vector_insert_extract'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %extract_fixed_from_scalable = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> %v0, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %extract_fixed_from_scalable = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> %v0, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %insert_fixed_into_scalable = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> %v0, <16 x i32> %v2, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %extract_scalable_from_scalable = call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> %v1, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert_scalable_into_scalable = call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> %v1, <vscale x 4 x i32> %v0, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SIZE-LABEL: 'vector_insert_extract'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %extract_fixed_from_scalable = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> %v0, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %extract_fixed_from_scalable = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> %v0, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert_fixed_into_scalable = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> %v0, <16 x i32> %v2, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %extract_scalable_from_scalable = call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> %v1, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert_scalable_into_scalable = call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> %v1, <vscale x 4 x i32> %v0, i64 0)

llvm/test/Analysis/CostModel/RISCV/rvv-vectorextract.ll

@@ -11,13 +11,13 @@ define void @vector_extract_nxv128i8_0(<vscale x 128 x i8> %v) {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m2 = call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m4 = call <vscale x 32 x i8> @llvm.vector.extract.nxv32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m8 = call <vscale x 64 x i8> @llvm.vector.extract.nxv64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SIZE-LABEL: 'vector_extract_nxv128i8_0'
@@ -28,13 +28,13 @@ define void @vector_extract_nxv128i8_0(<vscale x 128 x i8> %v) {
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m2 = call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m4 = call <vscale x 32 x i8> @llvm.vector.extract.nxv32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %scalable_m8 = call <vscale x 64 x i8> @llvm.vector.extract.nxv64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %scalable_mf8 = call <vscale x 1 x i8> @llvm.vector.extract.nxv1i8.nxv128i8(<vscale x 128 x i8> %v, i64 0)
@@ -110,23 +110,23 @@ define void @vector_extract_nxv128i8_1(<vscale x 128 x i8> %v) {
 
 define void @vector_extract_v128i8_0(<128 x i8> %v) {
 ; CHECK-LABEL: 'vector_extract_v128i8_0'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8> %v, i64 0)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8> %v, i64 0)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.v128i8(<128 x i8> %v, i64 0)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SIZE-LABEL: 'vector_extract_v128i8_0'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8> %v, i64 0)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf4 = call <4 x i8> @llvm.vector.extract.v4i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_mf2 = call <8 x i8> @llvm.vector.extract.v8i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m1 = call <16 x i8> @llvm.vector.extract.v16i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m2 = call <32 x i8> @llvm.vector.extract.v32i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m4 = call <64 x i8> @llvm.vector.extract.v64i8.v128i8(<128 x i8> %v, i64 0)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %fixed_m8 = call <128 x i8> @llvm.vector.extract.v128i8.v128i8(<128 x i8> %v, i64 0)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %fixed_mf8 = call <2 x i8> @llvm.vector.extract.v2i8.v128i8(<128 x i8> %v, i64 0)