UPSTREAM PR #21540: metal : add XIELU op

[loci-dev]

Note

Source pull request: ggml-org/llama.cpp#21540

Overview

Add Metal backend support for the XIELU (Extended Integral Exponential Linear Unit) operation.

XIELU is a parameterised activation function with four floats (alpha_n, alpha_p, beta, eps). The Metal kernel passes these through the existing ggml_metal_kargs_unary fields (slope, scale, bias, val) and uses select() to keep the implementation compatible with the vectorised (float4/half4) unary pipeline.

Mentioned in #14909

Testing

Tested on M4 Pro:

• test-backend-ops test -o XIELU — 1/1 tests passed

$ ./build/bin/test-backend-ops test -o XIELU
  XIELU(type=f32,ne=[10,5,4,3]):      OK
  1/1 tests passed, Backend MTL0: OK

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: NO

[loci-review]

No meaningful performance changes were detected across 125311 analyzed functions in the following binaries: build.bin.libllama.so, build.bin.llama-tts, build.bin.llama-bench, build.bin.libmtmd.so, build.bin.llama-cvector-generator, build.bin.llama-tokenize, build.bin.libggml-cpu.so, build.bin.libggml.so, build.bin.libggml-base.so, build.bin.llama-gemma3-cli, build.bin.llama-gguf-split, build.bin.llama-llava-cli, build.bin.llama-minicpmv-cli, build.bin.llama-quantize, build.bin.llama-qwen2vl-cli.

💬 Questions? Tag @loci-dev