%global __brp_check_rpaths %{nil}
%global __requires_exclude ^libmpi
%global packname  ggmlR
%global packver   0.7.0
%global rlibdir   /usr/local/lib/R/library

Name:             R-CRAN-%{packname}
Version:          0.7.0
Release:          1%{?dist}%{?buildtag}
Summary:          'GGML' Tensor Operations for Machine Learning

License:          MIT + file LICENSE
URL:              https://cran.r-project.org/package=%{packname}
Source0:          %{url}&version=%{packver}#/%{packname}_%{packver}.tar.gz


BuildRequires:    glslc
BuildRequires:    vulkan-loader-devel
BuildRequires:    R-devel >= 4.1.0
Requires:         R-core >= 4.1.0

%description
Provides 'R' bindings to the 'GGML' tensor library for machine learning,
designed primarily for 'Vulkan' GPU acceleration with full CPU fallback.
Requires 'Vulkan' 1.2+ with legacy pipeline barriers (avoids
'Synchronization2' due to 'RADV' performance issues); supports 'Push
Descriptors' ('VK_KHR_push_descriptor') to eliminate descriptor pool
overhead when available. 'Vulkan' support is auto-detected at build time
on Linux (when 'libvulkan-dev' and 'glslc' are installed) and on Windows
(when 'Vulkan' 'SDK' is installed and 'VULKAN_SDK' environment variable is
set); all operations fall back to CPU transparently when no GPU is
available. Supports tensors up to 5D natively (GGML_MAX_DIMS=5).
Implements tensor operations, neural network layers, 'quantization', and a
'Keras'-like sequential model API for building and training networks.
Includes 'AdamW' (Adam with Weight decay) and 'SGD' (Stochastic Gradient
Descent) optimizers with 'MSE' (Mean Squared Error) and cross-entropy
losses. Also provides a dynamic 'autograd' engine ('PyTorch'-style) with
data-parallel training via 'dp_train()', broadcast arithmetic, 'f16'
(half-precision) support on 'Vulkan' GPU, and a multi-head attention layer
for building Transformer architectures. Supports 'ONNX' model import via
built-in zero-dependency 'protobuf' parser: load 'pretrained' 'ONNX'
models from 'PyTorch', 'TensorFlow', or other frameworks and run inference
on 'Vulkan' GPU or CPU. Covers 50+ 'ONNX' ops including convolutions,
attention primitives, normalization, quantized ops, shape operations,
'ScatterElements' (with 'Vulkan' 'atomicAdd' for GNN scatter-add), and
fused custom ops (RelPosBias2D for 'BoTNet') — sufficient to run
real-world models such as 'RoBERTa', 'BERT', 'GPT-NeoX', 'SqueezeNet',
'Inception v3', 'BAT-ResNeXt', 'BoTNet', and 'MNIST' out of the box. Reads
'GGUF' files natively: load 'pretrained' weights from any
'gguf'-compatible source ('llama.cpp', 'Hugging Face') with automatic
weight conversion and metadata access. Uses a dedicated weight buffer
architecture for zero-overhead repeated inference — weights are loaded to
GPU once and never re-transferred. Serves as backend for 'LLM' (Large
Language Model) inference via 'llamaR' and Stable Diffusion image
generation via 'sd2R'. See <https://github.com/ggml-org/ggml> for more
information about the underlying library.

%prep
%setup -q -c -n %{packname}

# fix end of executable files
find -type f -executable -exec grep -Iq . {} \; -exec sed -i -e '$a\' {} \;
# prevent binary stripping
[ -d %{packname}/src ] && find %{packname}/src -type f -exec \
  sed -i 's@/usr/bin/strip@/usr/bin/true@g' {} \; || true
[ -d %{packname}/src ] && find %{packname}/src/Make* -type f -exec \
  sed -i 's@-g0@@g' {} \; || true
# don't allow local prefix in executable scripts
find -type f -executable -exec sed -Ei 's@#!( )*/usr/local/bin@#!/usr/bin@g' {} \;

%build

%install

mkdir -p %{buildroot}%{rlibdir}
%{_bindir}/R CMD INSTALL -l %{buildroot}%{rlibdir} %{packname}
test -d %{packname}/src && (cd %{packname}/src; rm -f *.o *.so)
rm -f %{buildroot}%{rlibdir}/R.css
# remove buildroot from installed files
find %{buildroot}%{rlibdir} -type f -exec sed -i "s@%{buildroot}@@g" {} \;

%files
%{rlibdir}/%{packname}