Add PairwiseAdd and ComplexMultiply to SVE microbenchmark

[ylpoonlg]

Adds two SVE benchmarks making use of SVE2 APIs and interleaved load/store.

Performance Results

Run on Neoverse-V2

Method	Size	Mean	Error	StdDev	Median	Min	Max	Allocated
Scalar	15	13.212 ns	0.0076 ns	0.0068 ns	13.214 ns	13.198 ns	13.222 ns	-
Vector128PairwiseAdd	15	6.525 ns	0.0117 ns	0.0109 ns	6.529 ns	6.503 ns	6.534 ns	-
SvePairwiseAdd	15	5.243 ns	0.0067 ns	0.0056 ns	5.244 ns	5.227 ns	5.251 ns	-
Sve2PairwiseAdd	15	5.040 ns	0.0093 ns	0.0087 ns	5.037 ns	5.027 ns	5.053 ns	-
Scalar	127	100.121 ns	0.2256 ns	0.2111 ns	100.175 ns	99.688 ns	100.390 ns	-
Vector128PairwiseAdd	127	24.714 ns	0.0784 ns	0.0612 ns	24.711 ns	24.632 ns	24.820 ns	-
SvePairwiseAdd	127	32.348 ns	0.0650 ns	0.0608 ns	32.385 ns	32.258 ns	32.416 ns	-
Sve2PairwiseAdd	127	29.272 ns	0.0282 ns	0.0263 ns	29.279 ns	29.207 ns	29.296 ns	-
Scalar	527	391.076 ns	0.4614 ns	0.4090 ns	391.211 ns	390.138 ns	391.601 ns	-
Vector128PairwiseAdd	527	89.575 ns	0.1429 ns	0.1267 ns	89.619 ns	89.281 ns	89.726 ns	-
SvePairwiseAdd	527	127.174 ns	0.0962 ns	0.0900 ns	127.148 ns	127.001 ns	127.321 ns	-
Sve2PairwiseAdd	527	119.669 ns	0.1557 ns	0.1457 ns	119.690 ns	119.427 ns	119.920 ns	-
Scalar	10015	7,389.023 ns	66.0503 ns	55.1550 ns	7,381.598 ns	7,333.689 ns	7,513.358 ns	-
Vector128PairwiseAdd	10015	2,328.763 ns	33.0594 ns	30.9238 ns	2,341.812 ns	2,271.219 ns	2,365.432 ns	-
SvePairwiseAdd	10015	2,406.691 ns	2.8076 ns	2.4889 ns	2,406.969 ns	2,401.705 ns	2,411.131 ns	-
Sve2PairwiseAdd	10015	2,202.091 ns	2.4197 ns	2.2634 ns	2,202.101 ns	2,198.067 ns	2,206.706 ns	-

Method	Size	Mean	Error	StdDev	Median	Min	Max	Allocated
Scalar	15	18.665 ns	0.0076 ns	0.0071 ns	18.663 ns	18.657 ns	18.681 ns	-
Vector128ComplexMultiply	15	7.547 ns	0.0015 ns	0.0011 ns	7.547 ns	7.545 ns	7.549 ns	-
SveComplexMultiply	15	5.893 ns	0.0093 ns	0.0087 ns	5.896 ns	5.879 ns	5.904 ns	-
Sve2ComplexMultiply	15	5.033 ns	0.0015 ns	0.0013 ns	5.034 ns	5.031 ns	5.035 ns	-
Scalar	127	148.051 ns	0.0958 ns	0.0850 ns	148.031 ns	147.947 ns	148.224 ns	-
Vector128ComplexMultiply	127	43.807 ns	0.1016 ns	0.0849 ns	43.784 ns	43.683 ns	44.004 ns	-
SveComplexMultiply	127	47.935 ns	0.0558 ns	0.0495 ns	47.945 ns	47.863 ns	48.015 ns	-
Sve2ComplexMultiply	127	37.890 ns	0.0082 ns	0.0068 ns	37.887 ns	37.880 ns	37.904 ns	-
Scalar	527	595.884 ns	0.7120 ns	0.5946 ns	595.605 ns	595.460 ns	597.372 ns	-
Vector128ComplexMultiply	527	179.198 ns	0.4771 ns	0.4463 ns	179.116 ns	178.661 ns	180.130 ns	-
SveComplexMultiply	527	199.454 ns	0.0967 ns	0.0905 ns	199.454 ns	199.283 ns	199.590 ns	-
Sve2ComplexMultiply	527	158.460 ns	0.3893 ns	0.3642 ns	158.689 ns	157.934 ns	158.831 ns	-
Scalar	10015	11,261.033 ns	6.9704 ns	5.4421 ns	11,260.642 ns	11,253.823 ns	11,269.318 ns	-
Vector128ComplexMultiply	10015	3,417.121 ns	6.5946 ns	5.8460 ns	3,416.762 ns	3,410.709 ns	3,426.948 ns	-
SveComplexMultiply	10015	3,775.434 ns	1.9476 ns	1.7265 ns	3,775.040 ns	3,773.440 ns	3,779.061 ns	-
Sve2ComplexMultiply	10015	3,035.012 ns	4.2198 ns	3.9472 ns	3,033.787 ns	3,031.124 ns	3,044.965 ns	-

cc @dotnet/arm64-contrib @SwapnilGaikwad @LoopedBard3

[Copilot]

Pull request overview

Adds two new SVE microbenchmarks under src/benchmarks/micro/sve to measure pairwise addition and complex multiplication, including SVE2-specific implementations, and gates compilation to appropriate TFMs.

Changes:

• Introduce PairwiseAdd benchmark with Scalar/AdvSimd/Sve/Sve2 implementations.
• Introduce ComplexMultiply benchmark with Scalar/AdvSimd/Sve/Sve2 implementations.
• Update MicroBenchmarks.csproj to exclude these SVE2-dependent benchmarks for TFMs older than net10.0.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/benchmarks/micro/sve/PairwiseAdd.cs	New benchmark for pairwise-add across scalar, AdvSimd, SVE, and SVE2 implementations.
src/benchmarks/micro/sve/ComplexMultiply.cs	New benchmark for complex multiplication across scalar, AdvSimd, SVE, and SVE2 implementations.
src/benchmarks/micro/MicroBenchmarks.csproj	Adds TFM-conditional compile removal for the new SVE2-dependent benchmark files on < net10.0.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.