Eval bug: error: Double type is not supported on this platform.

[paoletto]

Name and Version

llama-server 1d8504338

Operating systems

Linux

GGML backends

OpenCL

Hardware

i7-1165G7

Models

Hermes-3-Llama-3.1-70B-Uncensored.i1-Q4_K_S.gguf

Problem description & steps to reproduce

after pulling the master branch (1d8504338), i discovered there's now a (apparently preferred over sycl) opencl backend. I initially mistakenly enabled adreno kernels, and got a warning that adreno kernels should not be used on non-adreno gpus (but the following error was not happening!). So i rebuilt without those, and started to get a stream of: error: Double type is not supported on this platform.

ggml_opencl: selecting platform: 'Intel(R) OpenCL Graphics'
ggml_opencl: selecting device: 'Intel(R) Iris(R) Xe Graphics'  
ggml_opencl: device OpenCL version: OpenCL 3.0 NEO
ggml_opencl: OpenCL driver: 23.17.26241.33
ggml_opencl: vector subgroup broadcast support: false
ggml_opencl: device FP16 support: true
ggml_opencl: mem base addr align: 1024
ggml_opencl: max mem alloc size: 4095 MB
ggml_opencl: SVM coarse grain buffer support: true
ggml_opencl: SVM fine grain buffer support: false
ggml_opencl: SVM fine grain system support: false
ggml_opencl: SVM atomics support: false
ggml_opencl: flattening quantized weights representation as struct of arrays (GGML_OPENCL_SOA_Q)
ggml_opencl: kernel compile error:


error: Double type is not supported on this platform.
in kernel: 'kernel_gelu'
error: backend compiler failed build.

And btw you guys must add OpenCL backend to the list of backends to report. Pls dont make me file a ticket for that

First Bad Commit

No response

Relevant log output

ggml_opencl: selecting platform: 'Intel(R) OpenCL Graphics'
ggml_opencl: selecting device: 'Intel(R) Iris(R) Xe Graphics'  
ggml_opencl: device OpenCL version: OpenCL 3.0 NEO
ggml_opencl: OpenCL driver: 23.17.26241.33
ggml_opencl: vector subgroup broadcast support: false
ggml_opencl: device FP16 support: true
ggml_opencl: mem base addr align: 1024
ggml_opencl: max mem alloc size: 4095 MB
ggml_opencl: SVM coarse grain buffer support: true
ggml_opencl: SVM fine grain buffer support: false
ggml_opencl: SVM fine grain system support: false
ggml_opencl: SVM atomics support: false
ggml_opencl: flattening quantized weights representation as struct of arrays (GGML_OPENCL_SOA_Q)
ggml_opencl: kernel compile error:


error: Double type is not supported on this platform.
in kernel: 'kernel_gelu'
error: backend compiler failed build.