whisper : Metal and ggml-alloc support (#1270)

* metal : init

* whisper : factor out graph builds

* whisper : allocate encoder and decoder using ggml-alloc

* whisper : ggml-alloc is now supported

* whisper : CoreML support ggml-alloc

* build : fix ggml-alloc

* ios : update submodule

* extra : update sync-ggml.sh script to also sync ggml-alloc

* ci : see if this is causing the crash

* whisper : refactor ggml-alloc init

* whisper.android : try to fix build

* whisper : initial Metal version

* ci : try to debug vmem issue

* metal : decoder works on GPU!

* metal : add multi-decoder support

* ggml : fix ggml_nbytes (probably temp solution)

* metal : run "cross" step on the GPU

* whisper : remove ggml_repeat in the encoder

* whisper : offload the Encoder to Metal

* ggml : use simpler ggml_bytes() implementation

* ggml-alloc : try to make CI happy by reducing vram to 128GB

* whisper : add whisper_allocr to wrap ggml_allocr

* whisper : factor out alloc init in a function

* cmake : update to support Metal build

* whisper : add <functional> header

* objc : fix build (no Metal yet)

* ios : add Metal support

* swiftui : fix build

* metal : speed-up KQ multiplication

* metal : sync latest llama.cpp kernels

* readme : add Metal info

* ios : update submodule

* coreml : add code to toggle Core ML config (CPU, ANE, GPU)

* bench : fix timings by running a pre-heat

* bench : start benching the decoder

* whisper : add ggml_mul_mat_pad

* bench : fix uninitialized vars

* whisper : add comment for disabling mul-mat padding

* whisper : add description of ggml_mul_mat_pad

* whisper : clean-up ggml_mul_mat_pad

* metal : remove the "concurrent" flag

* bench : variable n_past

* ios : update SPM package

extra/bench-all.sh

@@ -44,27 +44,26 @@ if [ "$encoder_only" -eq 0 ]; then
     printf "\n"
 fi
 
-printf "| CPU | OS | Config | Model | Th | Load | Enc. | Commit |\n"
-printf "| --- | -- | ------ | ----- | -- | ---- | ---- | ------ |\n"
+printf "| %6s | %6s | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "CPU" "OS" "Config" "Model" "Th" "Enc." "Dec." "PP" "Commit"
+printf "| %6s | %6s | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "---" "---" "---" "---" "---" "---" "---" "---" "---"
 
 for model in "${models[@]}"; do
-    # run once to heat-up the cache
-    ./bench -m ./models/ggml-$model.bin -t $n_threads 2>/dev/null 1>/dev/null
-
     # actual run
     # store stderr output in a variable in order to parse it later
     output=$(./bench -m ./models/ggml-$model.bin -t $n_threads 2>&1)
     ret=$?
 
     # parse the output:
-    load_time=$(echo "$output" | grep "load time" | awk '{print $5}')
-    encode_time=$(echo "$output" | grep "encode time" | awk '{print $5}')
+    encode_time=$(echo "$output" | grep "encode time" | awk '{print $11}')
+    decode_time=$(echo "$output" | grep "decode time" | awk '{print $11}')
+    prompt_time=$(echo "$output" | grep "prompt time" | awk '{print $11}')
     system_info=$(echo "$output" | grep "system_info")
     n_threads=$(echo "$output" | grep "system_info" | awk '{print $4}')
 
     # floor to milliseconds
-    load_time=${load_time%.*}
-    encode_time=${encode_time%.*}
+    #encode_time=${encode_time%.*}
+    #decode_time=${decode_time%.*}
+    #prompt_time=${prompt_time%.*}
 
     config=""
 
@@ -87,6 +86,6 @@ for model in "${models[@]}"; do
     commit=$(git rev-parse --short HEAD)
 
     if [ $ret -eq 0 ]; then
-        printf "| <todo> | <todo> | $config | $model | $n_threads | $load_time | $encode_time | $commit |\n"
+        printf "| <todo> | <todo> | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "$config" "$model" "$n_threads" "$encode_time" "$decode_time" "$prompt_time" "$commit"
     fi
 done

extra/sync-ggml.sh

@@ -1,18 +1,20 @@
 #!/bin/bash
 
-cp -rpv ../ggml/src/ggml.c               ./ggml.c
-cp -rpv ../ggml/src/ggml-cuda.h          ./ggml-cuda.h
-cp -rpv ../ggml/src/ggml-cuda.cu         ./ggml-cuda.cu
-cp -rpv ../ggml/src/ggml-opencl.h        ./ggml-opencl.h
-cp -rpv ../ggml/src/ggml-opencl.cpp      ./ggml-opencl.cpp
-cp -rpv ../ggml/src/ggml-metal.h         ./ggml-metal.h
-cp -rpv ../ggml/src/ggml-metal.m         ./ggml-metal.m
-cp -rpv ../ggml/src/ggml-metal.metal     ./ggml-metal.metal
-cp -rpv ../ggml/include/ggml/ggml.h      ./ggml.h
-cp -rpv ../ggml/examples/common.h        ./examples/common.h
-cp -rpv ../ggml/examples/common.cpp      ./examples/common.cpp
-cp -rpv ../ggml/examples/common-ggml.h   ./examples/common-ggml.h
-cp -rpv ../ggml/examples/common-ggml.cpp ./examples/common-ggml.cpp
+cp -rpv ../ggml/src/ggml.c                ./ggml.c
+cp -rpv ../ggml/src/ggml-alloc.c          ./ggml-alloc.c
+cp -rpv ../ggml/src/ggml-cuda.h           ./ggml-cuda.h
+cp -rpv ../ggml/src/ggml-cuda.cu          ./ggml-cuda.cu
+cp -rpv ../ggml/src/ggml-opencl.h         ./ggml-opencl.h
+cp -rpv ../ggml/src/ggml-opencl.cpp       ./ggml-opencl.cpp
+cp -rpv ../ggml/src/ggml-metal.h          ./ggml-metal.h
+cp -rpv ../ggml/src/ggml-metal.m          ./ggml-metal.m
+cp -rpv ../ggml/src/ggml-metal.metal      ./ggml-metal.metal
+cp -rpv ../ggml/include/ggml/ggml.h       ./ggml.h
+cp -rpv ../ggml/include/ggml/ggml-alloc.h ./ggml-alloc.h
+cp -rpv ../ggml/examples/common.h         ./examples/common.h
+cp -rpv ../ggml/examples/common.cpp       ./examples/common.cpp
+cp -rpv ../ggml/examples/common-ggml.h    ./examples/common-ggml.h
+cp -rpv ../ggml/examples/common-ggml.cpp  ./examples/common-ggml.cpp
 
 cp -rpv ../ggml/examples/whisper/whisper.h    ./whisper.h
 cp -rpv ../ggml/examples/whisper/whisper.cpp  ./whisper.cpp