Slow compilation

[wallentri88]

Default compile options compile ALL quants:

[ 33%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_k.cu.o
[ 35%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_k_id.cu.o
[ 35%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_k_r4.cu.o
[ 35%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_kl.cu.o
[ 36%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_kl_id.cu.o
[ 36%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_ks.cu.o
[ 36%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_ks_id.cu.o
[ 36%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_kt.cu.o
[ 38%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_kt_id.cu.o
[ 38%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_s.cu.o
[ 38%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_s_id.cu.o
[ 40%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu.o
[ 40%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_xs_id.cu.o
[ 40%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu.o
[ 40%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq2_xxs_id.cu.o
[ 41%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_k.cu.o
[ 41%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_k_id.cu.o
[ 41%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_k_r4.cu.o
[ 43%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_ks.cu.o
[ 43%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_ks_id.cu.o
[ 43%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_kt.cu.o
[ 44%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_kt_id.cu.o
[ 44%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_s.cu.o
[ 44%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_s_id.cu.o
[ 44%] Building CUDA object ggml/src/CMakeFiles/ggml.dir/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu.o

etc etc

Is there's a way to limit it to quants I actually use? I only use IQ4_K and IQ5_K ones. I feel unused quants are just wasting compilation time :(

[ikawrakow]

I know, it is painful when the CUDA mmq and/or mmvq kernels need to be rebuild. There is currently no way to select a subset of the quantization types. You can add a feature request if you like.