UPSTREAM PR #19433: Add a build target to generate ROCm artifacts using ROCm 7.2

[loci-dev]

Note

Source pull request: ggml-org/llama.cpp#19433

This builds the following targets:

• gfx1151
• gfx1150
• gfx1200
• gfx1201
• gfx1100
• gfx1101
• gfx1102
• gfx1030
• gfx1031
• gfx1032

Notably; it doesn't include any CDNA targets, and thus avoids ggml-org/llama.cpp#19269

[loci-review]

No meaningful performance changes were detected across 115630 analyzed functions in the following binaries: build.bin.llama-cvector-generator, build.bin.llama-tts, build.bin.libllama.so, build.bin.libmtmd.so, build.bin.libggml-cpu.so, build.bin.libggml.so, build.bin.libggml-base.so, build.bin.llama-bench, 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-tokenize, build.bin.llama-qwen2vl-cli.

🔎 Full breakdown: Loci Inspector.
💬 Questions? Tag @loci-dev.

[noamshiovits-Aurora]

@loci-dev how many functions were modified in this pr and how many files were edited or added ?

[loci-review]

Based on the repository analysis:

PR Statistics

Files Changed

• 110 files modified
• 6 files added
• 2 files deleted
• Total: 118 files changed
• 43 commits

Functions Analyzed

According to the performance review, 115,630 functions were analyzed across all binaries, with no meaningful performance changes detected in any of them.

This is a substantial PR focused on:

• GPU backend optimizations (Metal, Vulkan, CUDA, SYCL, WebGPU)
• New model architecture support (Step3.5-Flash, Kimi-Linear, Qwen3next)
• Core infrastructure refactoring (sampling, batch processing)
• Testing improvements and dependency updates

The fact that 115K+ functions were analyzed but showed no meaningful performance regressions is actually a positive signal - it indicates that despite the extensive changes across 118 files, the modifications maintained performance stability. The changes appear to be primarily focused on correctness, feature additions, and code quality improvements rather than introducing performance-impacting modifications.

[loci-review]

No meaningful performance changes were detected across 115001 analyzed functions in the following binaries: build.bin.llama-cvector-generator, build.bin.llama-tts, build.bin.libggml-cpu.so, build.bin.libmtmd.so, build.bin.libllama.so, build.bin.libggml-base.so, build.bin.llama-bench, build.bin.llama-quantize, build.bin.llama-tokenize, build.bin.llama-gguf-split, build.bin.libggml.so, build.bin.llama-qwen2vl-cli, build.bin.llama-gemma3-cli, build.bin.llama-llava-cli, build.bin.llama-minicpmv-cli.

🔎 Full breakdown: Loci Inspector.
💬 Questions? Tag @loci-dev.

[loci-review]

No meaningful performance changes were detected across 111508 analyzed functions in the following binaries: build.bin.llama-cvector-generator, build.bin.libllama.so, build.bin.libmtmd.so, build.bin.llama-tts, build.bin.llama-tokenize, 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, build.bin.llama-bench, build.bin.libggml-base.so, build.bin.libggml-cpu.so, build.bin.libggml.so.

🔎 Full breakdown: Loci Inspector.
💬 Questions? Tag @loci-dev.