[CostModel][AArch64] Add tests for ctlz, ctpop and cttz. NFC.

[SjoerdMeijer]

llvm/test/Analysis/CostModel/AArch64/ctlz.ll

67

This tests a 8 x 64 = 256 bit vector, which is already bigger than the vector register.

76

And this tests a 512 bit vector. Having this tests here won't hurt anyone, but I don't think it adds much compared to the previous tests, so might as well remove this one.

112

Same for this one.

191

You could consider adding a 32 x i8 case here, i.e. a test/vector that is bigger than the register width.

llvm/test/Analysis/CostModel/AArch64/ctpop.ll

1

And similar remarks here about the vector widths as in the previous file.

4

Nit, typo: intstructions

[RosieSumpter]

Removed some tests where vectors are larger than 128 bits.

[SjoerdMeijer]

llvm/test/Analysis/CostModel/AArch64/ctlz.ll

58

Sorry if that wasn't clear, but you can keep this one.

103

Keep this one.

148

Keep this one.

llvm/test/Analysis/CostModel/AArch64/ctpop.ll

58

Keep this one.

103

Keep this one.

llvm/test/Analysis/CostModel/AArch64/cttz.ll

58

Keep this one.

103

Keep this one.

[SjoerdMeijer]

Sorry, ignore my previous comments, I have spotted them now (with your help).

This looks like a good set of new cost-model tests to me.

llvm/test/Analysis/CostModel/AArch64/ctlz.ll

+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=aarch64 -cost-model -analyze | FileCheck %s
+
+; Verify the cost of scalar ctlz instructions.
+
+define i64 @test_ctlz_i64(i64 %a) {
+; CHECK-LABEL: 'test_ctlz_i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call i64 @llvm.ctlz.i64(i64 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %ctlz
+;
+  %ctlz = call i64 @llvm.ctlz.i64(i64 %a)
+  ret i64 %ctlz
+}
+
+define i32 @test_ctlz_i32(i32 %a) {
+; CHECK-LABEL: 'test_ctlz_i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call i32 @llvm.ctlz.i32(i32 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %ctlz
+;
+  %ctlz = call i32 @llvm.ctlz.i32(i32 %a)
+  ret i32 %ctlz
+}
+
+define i16 @test_ctlz_i16(i16 %a) {
+; CHECK-LABEL: 'test_ctlz_i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call i16 @llvm.ctlz.i16(i16 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %ctlz
+;
+  %ctlz = call i16 @llvm.ctlz.i16(i16 %a)
+  ret i16 %ctlz
+}
+
+define i8 @test_ctlz_i8(i8 %a) {
+; CHECK-LABEL: 'test_ctlz_i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call i8 @llvm.ctlz.i8(i8 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %ctlz
+;
+  %ctlz = call i8 @llvm.ctlz.i8(i8 %a)
+  ret i8 %ctlz
+}
+
+declare i64 @llvm.ctlz.i64(i64)
+declare i32 @llvm.ctlz.i32(i32)
+declare i16 @llvm.ctlz.i16(i16)
+declare i8 @llvm.ctlz.i8(i8)
+
+; Verify the cost of vector ctlz instructions.
+
+define <2 x i64> @test_ctlz_v2i64(<2 x i64> %a) {
+; CHECK-LABEL: 'test_ctlz_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %ctlz = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %ctlz
+;
+  %ctlz = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %a, i1 false)
+  ret <2 x i64> %ctlz
+}
+
+define <4 x i64> @test_ctlz_v4i64(<4 x i64> %a) {
+; CHECK-LABEL: 'test_ctlz_v4i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %ctlz = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %ctlz
+;
+  %ctlz = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %a, i1 false)
+  ret <4 x i64> %ctlz
+}
+
+define <8 x i64> @test_ctlz_v8i64(<8 x i64> %a) {
+; CHECK-LABEL: 'test_ctlz_v8i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %ctlz = call <8 x i64> @llvm.ctlz.v8i64(<8 x i64> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %ctlz
+;
+  %ctlz = call <8 x i64> @llvm.ctlz.v8i64(<8 x i64> %a, i1 false)
+  ret <8 x i64> %ctlz
+}
+
+define <16 x i64> @test_ctlz_v16i64(<16 x i64> %a) {
+; CHECK-LABEL: 'test_ctlz_v16i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %ctlz = call <16 x i64> @llvm.ctlz.v16i64(<16 x i64> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i64> %ctlz
+;
+  %ctlz = call <16 x i64> @llvm.ctlz.v16i64(<16 x i64> %a, i1 false)
+  ret <16 x i64> %ctlz
+}
+
+define <2 x i32> @test_ctlz_v2i32(<2 x i32> %a) {
+; CHECK-LABEL: 'test_ctlz_v2i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %ctlz
+;
+  %ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %a, i1 false)
+  ret <2 x i32> %ctlz
+}
+
+define <4 x i32> @test_ctlz_v4i32(<4 x i32> %a) {
+; CHECK-LABEL: 'test_ctlz_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %ctlz
+;
+  %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %a, i1 false)
+  ret <4 x i32> %ctlz
+}
+
+define <8 x i32> @test_ctlz_v8i32(<8 x i32> %a) {
+; CHECK-LABEL: 'test_ctlz_v8i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %ctlz = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %ctlz
+;
+  %ctlz = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %a, i1 false)
+  ret <8 x i32> %ctlz
+}
+
+define <16 x i32> @test_ctlz_v16i32(<16 x i32> %a) {
+; CHECK-LABEL: 'test_ctlz_v16i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %ctlz = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %ctlz
+;
+  %ctlz = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %a, i1 false)
+  ret <16 x i32> %ctlz
+}
+
+define <2 x i16> @test_ctlz_v2i16(<2 x i16> %a) {
+; CHECK-LABEL: 'test_ctlz_v2i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <2 x i16> @llvm.ctlz.v2i16(<2 x i16> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i16> %ctlz
+;
+  %ctlz = call <2 x i16> @llvm.ctlz.v2i16(<2 x i16> %a, i1 false)
+  ret <2 x i16> %ctlz
+}
+
+define <4 x i16> @test_ctlz_v4i16(<4 x i16> %a) {
+; CHECK-LABEL: 'test_ctlz_v4i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <4 x i16> @llvm.ctlz.v4i16(<4 x i16> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i16> %ctlz
+;
+  %ctlz = call <4 x i16> @llvm.ctlz.v4i16(<4 x i16> %a, i1 false)
+  ret <4 x i16> %ctlz
+}
+
+define <8 x i16> @test_ctlz_v8i16(<8 x i16> %a) {
+; CHECK-LABEL: 'test_ctlz_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %ctlz
+;
+  %ctlz = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %a, i1 false)
+  ret <8 x i16> %ctlz
+}
+
+define <16 x i16> @test_ctlz_v16i16(<16 x i16> %a) {
+; CHECK-LABEL: 'test_ctlz_v16i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %ctlz = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %ctlz
+;
+  %ctlz = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %a, i1 false)
+  ret <16 x i16> %ctlz
+}
+
+define <2 x i8> @test_ctlz_v2i8(<2 x i8> %a) {
+; CHECK-LABEL: 'test_ctlz_v2i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <2 x i8> @llvm.ctlz.v2i8(<2 x i8> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i8> %ctlz
+;
+  %ctlz = call <2 x i8> @llvm.ctlz.v2i8(<2 x i8> %a, i1 false)
+  ret <2 x i8> %ctlz
+}
+
+define <4 x i8> @test_ctlz_v4i8(<4 x i8> %a) {
+; CHECK-LABEL: 'test_ctlz_v4i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <4 x i8> @llvm.ctlz.v4i8(<4 x i8> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i8> %ctlz
+;
+  %ctlz = call <4 x i8> @llvm.ctlz.v4i8(<4 x i8> %a, i1 false)
+  ret <4 x i8> %ctlz
+}
+
+define <8 x i8> @test_ctlz_v8i8(<8 x i8> %a) {
+; CHECK-LABEL: 'test_ctlz_v8i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <8 x i8> @llvm.ctlz.v8i8(<8 x i8> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i8> %ctlz
+;
+  %ctlz = call <8 x i8> @llvm.ctlz.v8i8(<8 x i8> %a, i1 false)
+  ret <8 x i8> %ctlz
+}
+
+define <16 x i8> @test_ctlz_v16i8(<16 x i8> %a) {
+; CHECK-LABEL: 'test_ctlz_v16i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctlz = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %ctlz
+;
+  %ctlz = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %a, i1 false)
+  ret <16 x i8> %ctlz
+}
+
+declare <2 x i64> @llvm.ctlz.v2i64(<2 x i64>, i1)
+declare <4 x i64> @llvm.ctlz.v4i64(<4 x i64>, i1)
+declare <8 x i64> @llvm.ctlz.v8i64(<8 x i64>, i1)
+declare <16 x i64> @llvm.ctlz.v16i64(<16 x i64>, i1)
+
+declare <2 x i32> @llvm.ctlz.v2i32(<2 x i32>, i1)
+declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1)
+declare <8 x i32> @llvm.ctlz.v8i32(<8 x i32>, i1)
+declare <16 x i32> @llvm.ctlz.v16i32(<16 x i32>, i1)
+
+declare <2 x i16> @llvm.ctlz.v2i16(<2 x i16>, i1)
+declare <4 x i16> @llvm.ctlz.v4i16(<4 x i16>, i1)
+declare <8 x i16> @llvm.ctlz.v8i16(<8 x i16>, i1)
+declare <16 x i16> @llvm.ctlz.v16i16(<16 x i16>, i1)
+
+declare <2 x i8> @llvm.ctlz.v2i8(<2 x i8>, i1)
+declare <4 x i8> @llvm.ctlz.v4i8(<4 x i8>, i1)
+declare <8 x i8> @llvm.ctlz.v8i8(<8 x i8>, i1)
+declare <16 x i8> @llvm.ctlz.v16i8(<16 x i8>, i1)

llvm/test/Analysis/CostModel/AArch64/ctpop.ll

+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=aarch64 -cost-model -analyze | FileCheck %s
+
+; Verify the cost of scalar ctpop intstructions.
+
+define i64 @test_ctpop_i64(i64 %a) {
+; CHECK-LABEL: 'test_ctpop_i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i64 @llvm.ctpop.i64(i64 %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %ctpop
+;
+  %ctpop = call i64 @llvm.ctpop.i64(i64 %a)
+  ret i64 %ctpop
+}
+
+define i32 @test_ctpop_i32(i32 %a) {
+; CHECK-LABEL: 'test_ctpop_i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i32 @llvm.ctpop.i32(i32 %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %ctpop
+;
+  %ctpop = call i32 @llvm.ctpop.i32(i32 %a)
+  ret i32 %ctpop
+}
+
+define i16 @test_ctpop_i16(i16 %a) {
+; CHECK-LABEL: 'test_ctpop_i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i16 @llvm.ctpop.i16(i16 %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %ctpop
+;
+  %ctpop = call i16 @llvm.ctpop.i16(i16 %a)
+  ret i16 %ctpop
+}
+
+define i8 @test_ctpop_i8(i8 %a) {
+; CHECK-LABEL: 'test_ctpop_i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call i8 @llvm.ctpop.i8(i8 %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %ctpop
+;
+  %ctpop = call i8 @llvm.ctpop.i8(i8 %a)
+  ret i8 %ctpop
+}
+
+declare i64 @llvm.ctpop.i64(i64)
+declare i32 @llvm.ctpop.i32(i32)
+declare i16 @llvm.ctpop.i16(i16)
+declare i8 @llvm.ctpop.i8(i8)
+
+; Verify the cost of vector ctpop instructions.
+
+define <2 x i64> @test_ctpop_v2i64(<2 x i64> %a) {
+; CHECK-LABEL: 'test_ctpop_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %ctpop
+;
+  %ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %a)
+  ret <2 x i64> %ctpop
+}
+
+define <4 x i64> @test_ctpop_v4i64(<4 x i64> %a) {
+; CHECK-LABEL: 'test_ctpop_v4i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %ctpop
+;
+  %ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %a)
+  ret <4 x i64> %ctpop
+}
+
+define <8 x i64> @test_ctpop_v8i64(<8 x i64> %a) {
+; CHECK-LABEL: 'test_ctpop_v8i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %ctpop = call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %ctpop
+;
+  %ctpop = call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %a)
+  ret <8 x i64> %ctpop
+}
+
+define <16 x i64> @test_ctpop_v16i64(<16 x i64> %a) {
+; CHECK-LABEL: 'test_ctpop_v16i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %ctpop = call <16 x i64> @llvm.ctpop.v16i64(<16 x i64> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i64> %ctpop
+;
+  %ctpop = call <16 x i64> @llvm.ctpop.v16i64(<16 x i64> %a)
+  ret <16 x i64> %ctpop
+}
+
+define <2 x i32> @test_ctpop_v2i32(<2 x i32> %a) {
+; CHECK-LABEL: 'test_ctpop_v2i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %ctpop
+;
+  %ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %a)
+  ret <2 x i32> %ctpop
+}
+
+define <4 x i32> @test_ctpop_v4i32(<4 x i32> %a) {
+; CHECK-LABEL: 'test_ctpop_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %ctpop
+;
+  %ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %a)
+  ret <4 x i32> %ctpop
+}
+
+define <8 x i32> @test_ctpop_v8i32(<8 x i32> %a) {
+; CHECK-LABEL: 'test_ctpop_v8i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %ctpop
+;
+  %ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %a)
+  ret <8 x i32> %ctpop
+}
+
+define <16 x i32> @test_ctpop_v16i32(<16 x i32> %a) {
+; CHECK-LABEL: 'test_ctpop_v16i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %ctpop = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %ctpop
+;
+  %ctpop = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %a)
+  ret <16 x i32> %ctpop
+}
+
+define <2 x i16> @test_ctpop_v2i16(<2 x i16> %a) {
+; CHECK-LABEL: 'test_ctpop_v2i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i16> @llvm.ctpop.v2i16(<2 x i16> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i16> %ctpop
+;
+  %ctpop = call <2 x i16> @llvm.ctpop.v2i16(<2 x i16> %a)
+  ret <2 x i16> %ctpop
+}
+
+define <4 x i16> @test_ctpop_v4i16(<4 x i16> %a) {
+; CHECK-LABEL: 'test_ctpop_v4i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i16> @llvm.ctpop.v4i16(<4 x i16> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i16> %ctpop
+;
+  %ctpop = call <4 x i16> @llvm.ctpop.v4i16(<4 x i16> %a)
+  ret <4 x i16> %ctpop
+}
+
+define <8 x i16> @test_ctpop_v8i16(<8 x i16> %a) {
+; CHECK-LABEL: 'test_ctpop_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %ctpop
+;
+  %ctpop = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %a)
+  ret <8 x i16> %ctpop
+}
+
+define <16 x i16> @test_ctpop_v16i16(<16 x i16> %a) {
+; CHECK-LABEL: 'test_ctpop_v16i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %ctpop = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %ctpop
+;
+  %ctpop = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %a)
+  ret <16 x i16> %ctpop
+}
+
+define <2 x i8> @test_ctpop_v2i8(<2 x i8> %a) {
+; CHECK-LABEL: 'test_ctpop_v2i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i8> %ctpop
+;
+  %ctpop = call <2 x i8> @llvm.ctpop.v2i8(<2 x i8> %a)
+  ret <2 x i8> %ctpop
+}
+
+define <4 x i8> @test_ctpop_v4i8(<4 x i8> %a) {
+; CHECK-LABEL: 'test_ctpop_v4i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %ctpop = call <4 x i8> @llvm.ctpop.v4i8(<4 x i8> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i8> %ctpop
+;
+  %ctpop = call <4 x i8> @llvm.ctpop.v4i8(<4 x i8> %a)
+  ret <4 x i8> %ctpop
+}
+
+define <8 x i8> @test_ctpop_v8i8(<8 x i8> %a) {
+; CHECK-LABEL: 'test_ctpop_v8i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctpop = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i8> %ctpop
+;
+  %ctpop = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %a)
+  ret <8 x i8> %ctpop
+}
+
+define <16 x i8> @test_ctpop_v16i8(<16 x i8> %a) {
+; CHECK-LABEL: 'test_ctpop_v16i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %ctpop = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %a)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %ctpop
+;
+  %ctpop = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %a)
+  ret <16 x i8> %ctpop
+}
+
+declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>)
+declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>)
+declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>)
+declare <16 x i64> @llvm.ctpop.v16i64(<16 x i64>)
+
+declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>)
+declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>)
+declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>)
+declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>)
+
+declare <2 x i16> @llvm.ctpop.v2i16(<2 x i16>)
+declare <4 x i16> @llvm.ctpop.v4i16(<4 x i16>)
+declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16>)
+declare <16 x i16> @llvm.ctpop.v16i16(<16 x i16>)
+
+declare <2 x i8> @llvm.ctpop.v2i8(<2 x i8>)
+declare <4 x i8> @llvm.ctpop.v4i8(<4 x i8>)
+declare <8 x i8> @llvm.ctpop.v8i8(<8 x i8>)
+declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>)

llvm/test/Analysis/CostModel/AArch64/cttz.ll

+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -mtriple=aarch64 -cost-model -analyze | FileCheck %s
+
+; Verify the cost of scalar ctlz instructions.
+
+define i64 @test_cttz_i64(i64 %a) {
+; CHECK-LABEL: 'test_cttz_i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %cttz = call i64 @llvm.cttz.i64(i64 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %cttz
+;
+  %cttz = call i64 @llvm.cttz.i64(i64 %a)
+  ret i64 %cttz
+}
+
+define i32 @test_cttz_i32(i32 %a) {
+; CHECK-LABEL: 'test_cttz_i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %cttz = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %cttz
+;
+  %cttz = call i32 @llvm.cttz.i32(i32 %a)
+  ret i32 %cttz
+}
+
+define i16 @test_cttz_i16(i16 %a) {
+; CHECK-LABEL: 'test_cttz_i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %cttz = call i16 @llvm.cttz.i16(i16 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %cttz
+;
+  %cttz = call i16 @llvm.cttz.i16(i16 %a)
+  ret i16 %cttz
+}
+
+define i8 @test_cttz_i8(i8 %a) {
+; CHECK-LABEL: 'test_cttz_i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %cttz = call i8 @llvm.cttz.i8(i8 %a, i1 false)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %cttz
+;
+  %cttz = call i8 @llvm.cttz.i8(i8 %a)
+  ret i8 %cttz
+}
+
+declare i64 @llvm.cttz.i64(i64)
+declare i32 @llvm.cttz.i32(i32)
+declare i16 @llvm.cttz.i16(i16)
+declare i8 @llvm.cttz.i8(i8)
+
+; Verify the cost of vector cttz instructions.
+
+define <2 x i64> @test_cttz_v2i64(<2 x i64> %a) {
+; CHECK-LABEL: 'test_cttz_v2i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %cttz = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %cttz
+;
+  %cttz = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 true)
+  ret <2 x i64> %cttz
+}
+
+define <4 x i64> @test_cttz_v4i64(<4 x i64> %a) {
+; CHECK-LABEL: 'test_cttz_v4i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %cttz = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %cttz
+;
+  %cttz = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %a, i1 true)
+  ret <4 x i64> %cttz
+}
+
+define <8 x i64> @test_cttz_v8i64(<8 x i64> %a) {
+; CHECK-LABEL: 'test_cttz_v8i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %cttz = call <8 x i64> @llvm.cttz.v8i64(<8 x i64> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %cttz
+;
+  %cttz = call <8 x i64> @llvm.cttz.v8i64(<8 x i64> %a, i1 true)
+  ret <8 x i64> %cttz
+}
+
+define <16 x i64> @test_cttz_v16i64(<16 x i64> %a) {
+; CHECK-LABEL: 'test_cttz_v16i64'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %cttz = call <16 x i64> @llvm.cttz.v16i64(<16 x i64> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i64> %cttz
+;
+  %cttz = call <16 x i64> @llvm.cttz.v16i64(<16 x i64> %a, i1 true)
+  ret <16 x i64> %cttz
+}
+
+define <2 x i32> @test_cttz_v2i32(<2 x i32> %a) {
+; CHECK-LABEL: 'test_cttz_v2i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %cttz = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %cttz
+;
+  %cttz = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %a, i1 true)
+  ret <2 x i32> %cttz
+}
+
+define <4 x i32> @test_cttz_v4i32(<4 x i32> %a) {
+; CHECK-LABEL: 'test_cttz_v4i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %cttz
+;
+  %cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %a, i1 true)
+  ret <4 x i32> %cttz
+}
+
+define <8 x i32> @test_cttz_v8i32(<8 x i32> %a) {
+; CHECK-LABEL: 'test_cttz_v8i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %cttz = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %cttz
+;
+  %cttz = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %a, i1 true)
+  ret <8 x i32> %cttz
+}
+
+define <16 x i32> @test_cttz_v16i32(<16 x i32> %a) {
+; CHECK-LABEL: 'test_cttz_v16i32'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %cttz = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %cttz
+;
+  %cttz = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %a, i1 true)
+  ret <16 x i32> %cttz
+}
+
+define <2 x i16> @test_cttz_v2i16(<2 x i16> %a) {
+; CHECK-LABEL: 'test_cttz_v2i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %cttz = call <2 x i16> @llvm.cttz.v2i16(<2 x i16> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i16> %cttz
+;
+  %cttz = call <2 x i16> @llvm.cttz.v2i16(<2 x i16> %a, i1 true)
+  ret <2 x i16> %cttz
+}
+
+define <4 x i16> @test_cttz_v4i16(<4 x i16> %a) {
+; CHECK-LABEL: 'test_cttz_v4i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %cttz = call <4 x i16> @llvm.cttz.v4i16(<4 x i16> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i16> %cttz
+;
+  %cttz = call <4 x i16> @llvm.cttz.v4i16(<4 x i16> %a, i1 true)
+  ret <4 x i16> %cttz
+}
+
+define <8 x i16> @test_cttz_v8i16(<8 x i16> %a) {
+; CHECK-LABEL: 'test_cttz_v8i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %cttz = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %cttz
+;
+  %cttz = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %a, i1 true)
+  ret <8 x i16> %cttz
+}
+
+define <16 x i16> @test_cttz_v16i16(<16 x i16> %a) {
+; CHECK-LABEL: 'test_cttz_v16i16'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %cttz = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %cttz
+;
+  %cttz = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %a, i1 true)
+  ret <16 x i16> %cttz
+}
+
+define <2 x i8> @test_cttz_v2i8(<2 x i8> %a) {
+; CHECK-LABEL: 'test_cttz_v2i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %cttz = call <2 x i8> @llvm.cttz.v2i8(<2 x i8> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i8> %cttz
+;
+  %cttz = call <2 x i8> @llvm.cttz.v2i8(<2 x i8> %a, i1 true)
+  ret <2 x i8> %cttz
+}
+
+define <4 x i8> @test_cttz_v4i8(<4 x i8> %a) {
+; CHECK-LABEL: 'test_cttz_v4i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %cttz = call <4 x i8> @llvm.cttz.v4i8(<4 x i8> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i8> %cttz
+;
+  %cttz = call <4 x i8> @llvm.cttz.v4i8(<4 x i8> %a, i1 true)
+  ret <4 x i8> %cttz
+}
+
+define <8 x i8> @test_cttz_v8i8(<8 x i8> %a) {
+; CHECK-LABEL: 'test_cttz_v8i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %cttz = call <8 x i8> @llvm.cttz.v8i8(<8 x i8> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i8> %cttz
+;
+  %cttz = call <8 x i8> @llvm.cttz.v8i8(<8 x i8> %a, i1 true)
+  ret <8 x i8> %cttz
+}
+
+define <16 x i8> @test_cttz_v16i8(<16 x i8> %a) {
+; CHECK-LABEL: 'test_cttz_v16i8'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 106 for instruction: %cttz = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %a, i1 true)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %cttz
+;
+  %cttz = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %a, i1 true)
+  ret <16 x i8> %cttz
+}
+
+declare <2 x i64> @llvm.cttz.v2i64(<2 x i64>, i1)
+declare <4 x i64> @llvm.cttz.v4i64(<4 x i64>, i1)
+declare <8 x i64> @llvm.cttz.v8i64(<8 x i64>, i1)
+declare <16 x i64> @llvm.cttz.v16i64(<16 x i64>, i1)
+
+declare <2 x i32> @llvm.cttz.v2i32(<2 x i32>, i1)
+declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1)
+declare <8 x i32> @llvm.cttz.v8i32(<8 x i32>, i1)
+declare <16 x i32> @llvm.cttz.v16i32(<16 x i32>, i1)
+
+declare <2 x i16> @llvm.cttz.v2i16(<2 x i16>, i1)
+declare <4 x i16> @llvm.cttz.v4i16(<4 x i16>, i1)
+declare <8 x i16> @llvm.cttz.v8i16(<8 x i16>, i1)
+declare <16 x i16> @llvm.cttz.v16i16(<16 x i16>, i1)
+
+declare <2 x i8> @llvm.cttz.v2i8(<2 x i8>, i1)
+declare <4 x i8> @llvm.cttz.v4i8(<4 x i8>, i1)
+declare <8 x i8> @llvm.cttz.v8i8(<8 x i8>, i1)
+declare <16 x i8> @llvm.cttz.v16i8(<16 x i8>, i1)