vulkan: add INTEL_PRE_XE2 arch enum and enable coopmat1 on Intel Xe-LPG Plus (1/3, Xe-LPG Plus)

[fish-jiang]

Overview

Co-authors: @jxia4intel, @sliu39

PR 1/3 of the Intel Xe optimization series — see #24408 (mega PR, draft) for the full feature set.

Target platforms: Xe-LPG Plus (Arrow Lake-H iGPU)

• Adds INTEL_PRE_XE2 enum variant to vk_device_architecture
• Adds PTL (Panther Lake) device ID detection for future platform coverage
• Enables cooperative matrix (coopmat1) support on Intel Xe-LPG Plus, which was previously blocked due to performance regressions on discrete Xe1 GPUs (e.g. Arc A770)
• Overrides Vulkan buffer size limitation to 2 GB on affected Intel devices to support large model weights

Dependency: none — standalone platform identification and capability-enable patch.

Performance (Intel Xe1-ARLH + Windows OS)

BEFORE:
C:\base_build\bin\Release>llama-bench.exe -p 1024 -n 0 -r 3 -fa 0,1 --delay 10 -ngl 99 -m C:\kernel\model\Qwen3-0.6B-Q4_K_M.gguf,C:\kernel\model\gpt-oss-20b-Q4_K_M.gguf,C:\kernel\model\qwen3-8b-q4_k_m.gguf
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Arc(TM) 140T GPU (32GB) (Intel Corporation) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none
| model                          |       size |     params | backend    | ngl |  fa |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | --: | --: | --------------: | -------------------: |
| qwen3 0.6B Q4_K - Medium       | 372.65 MiB |   596.05 M | Vulkan     |  99 |   0 |          pp1024 |      2233.68 ± 23.05 |
| qwen3 0.6B Q4_K - Medium       | 372.65 MiB |   596.05 M | Vulkan     |  99 |   1 |          pp1024 |       2281.30 ± 1.61 |
| gpt-oss 20B Q4_K - Medium      |  10.81 GiB |    20.91 B | Vulkan     |  99 |   0 |          pp1024 |        354.07 ± 0.62 |
| gpt-oss 20B Q4_K - Medium      |  10.81 GiB |    20.91 B | Vulkan     |  99 |   1 |          pp1024 |        364.73 ± 1.95 |
| qwen3 8B Q4_K - Medium         |   4.68 GiB |     8.19 B | Vulkan     |  99 |   0 |          pp1024 |        272.01 ± 1.84 |
| qwen3 8B Q4_K - Medium         |   4.68 GiB |     8.19 B | Vulkan     |  99 |   1 |          pp1024 |        274.38 ± 0.52 |

build: d403f00ec (9554)

AFTER:

C:\Inteloptbuild\bin\Release>llama-bench.exe -p 1024 -n 0 -r 3 -fa 0,1 --delay 10 -ngl 99 -m C:\kernel\model\Qwen3-0.6B-Q4_K_M.gguf,C:\kernel\model\gpt-oss-20b-Q4_K_M.gguf,C:\kernel\model\qwen3-8b-q4_k_m.gguf
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Arc(TM) 140T GPU (32GB) (Intel Corporation) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: KHR_coopmat
| model                          |       size |     params | backend    | ngl |  fa |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | --: | --: | --------------: | -------------------: |
| qwen3 0.6B Q4_K - Medium       | 372.65 MiB |   596.05 M | Vulkan     |  99 |   0 |          pp1024 |       3048.68 ± 3.59 |
| qwen3 0.6B Q4_K - Medium       | 372.65 MiB |   596.05 M | Vulkan     |  99 |   1 |          pp1024 |       2815.11 ± 8.00 |
| gpt-oss 20B Q4_K - Medium      |  10.81 GiB |    20.91 B | Vulkan     |  99 |   0 |          pp1024 |        471.18 ± 0.73 |
| gpt-oss 20B Q4_K - Medium      |  10.81 GiB |    20.91 B | Vulkan     |  99 |   1 |          pp1024 |        486.92 ± 3.20 |
| qwen3 8B Q4_K - Medium         |   4.68 GiB |     8.19 B | Vulkan     |  99 |   0 |          pp1024 |        417.30 ± 1.35 |
| qwen3 8B Q4_K - Medium         |   4.68 GiB |     8.19 B | Vulkan     |  99 |   1 |          pp1024 |        408.85 ± 0.64 |

build: 9f484943b (9555)

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES, used claude code, then lots of manual review/tweaking.

[ggml-gh-bot]

Hi @fish-jiang, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• AI-generated content: This project does not accept PRs, descriptions or commit messages that are fully or predominantly AI-generated. If you have used AI to assist you in writing code, please make sure to disclose that explicitly.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.