whispercpp 1.3.3 → 1.3.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/README.md +60 -43
- data/ext/extconf.rb +2 -2
- data/ext/ruby_whisper.c +14 -2
- data/ext/ruby_whisper.h +39 -0
- data/ext/ruby_whisper_context.c +22 -22
- data/ext/ruby_whisper_model.c +12 -12
- data/ext/ruby_whisper_params.c +79 -25
- data/ext/ruby_whisper_segment.c +84 -19
- data/ext/ruby_whisper_token.c +351 -0
- data/ext/ruby_whisper_transcribe.cpp +1 -1
- data/ext/ruby_whisper_vad_context.c +75 -0
- data/ext/ruby_whisper_vad_context_detect.cpp +50 -0
- data/ext/ruby_whisper_vad_segment.c +139 -0
- data/ext/ruby_whisper_vad_segments.c +106 -0
- data/ext/sources/CMakeLists.txt +4 -1
- data/ext/sources/bindings/javascript/package.json +1 -1
- data/ext/sources/cmake/arm64-apple-clang.cmake +16 -0
- data/ext/sources/cmake/arm64-windows-llvm.cmake +16 -0
- data/ext/sources/cmake/riscv64-spacemit-linux-gnu-gcc.cmake +29 -0
- data/ext/sources/cmake/x64-windows-llvm.cmake +5 -0
- data/ext/sources/examples/CMakeLists.txt +1 -0
- data/ext/sources/examples/addon.node/addon.cpp +19 -19
- data/ext/sources/examples/addon.node/index.js +7 -5
- data/ext/sources/examples/addon.node/vad-example.js +2 -2
- data/ext/sources/examples/bench/bench.cpp +26 -16
- data/ext/sources/examples/bench.wasm/index-tmpl.html +10 -9
- data/ext/sources/examples/cli/cli.cpp +122 -111
- data/ext/sources/examples/command/command.cpp +26 -24
- data/ext/sources/examples/command.wasm/index-tmpl.html +5 -4
- data/ext/sources/examples/common-ggml.cpp +2 -0
- data/ext/sources/examples/lsp/CMakeLists.txt +2 -1
- data/ext/sources/examples/lsp/lsp.cpp +19 -17
- data/ext/sources/examples/quantize/CMakeLists.txt +2 -1
- data/ext/sources/examples/server/server.cpp +34 -24
- data/ext/sources/examples/server.py +6 -1
- data/ext/sources/examples/stream/stream.cpp +4 -2
- data/ext/sources/examples/stream.wasm/emscripten.cpp +6 -6
- data/ext/sources/examples/stream.wasm/index-tmpl.html +82 -5
- data/ext/sources/examples/talk-llama/CMakeLists.txt +7 -3
- data/ext/sources/examples/talk-llama/llama-adapter.cpp +113 -7
- data/ext/sources/examples/talk-llama/llama-adapter.h +13 -1
- data/ext/sources/examples/talk-llama/llama-arch.cpp +2136 -1491
- data/ext/sources/examples/talk-llama/llama-arch.h +125 -3
- data/ext/sources/examples/talk-llama/llama-batch.cpp +174 -100
- data/ext/sources/examples/talk-llama/llama-batch.h +46 -20
- data/ext/sources/examples/talk-llama/llama-chat.cpp +199 -8
- data/ext/sources/examples/talk-llama/llama-chat.h +11 -0
- data/ext/sources/examples/talk-llama/llama-context.cpp +1213 -413
- data/ext/sources/examples/talk-llama/llama-context.h +99 -36
- data/ext/sources/examples/talk-llama/llama-cparams.h +5 -4
- data/ext/sources/examples/talk-llama/llama-grammar.cpp +288 -53
- data/ext/sources/examples/talk-llama/llama-grammar.h +22 -1
- data/ext/sources/examples/talk-llama/llama-graph.cpp +883 -294
- data/ext/sources/examples/talk-llama/llama-graph.h +361 -161
- data/ext/sources/examples/talk-llama/llama-hparams.cpp +144 -6
- data/ext/sources/examples/talk-llama/llama-hparams.h +100 -23
- data/ext/sources/examples/talk-llama/llama-impl.cpp +7 -3
- data/ext/sources/examples/talk-llama/llama-impl.h +3 -1
- data/ext/sources/examples/talk-llama/llama-kv-cache-iswa.cpp +328 -0
- data/ext/sources/examples/talk-llama/{llama-kv-cache-unified-iswa.h → llama-kv-cache-iswa.h} +38 -29
- data/ext/sources/examples/talk-llama/llama-kv-cache.cpp +2100 -0
- data/ext/sources/examples/talk-llama/llama-kv-cache.h +373 -27
- data/ext/sources/examples/talk-llama/llama-kv-cells.h +124 -30
- data/ext/sources/examples/talk-llama/llama-memory-hybrid.cpp +63 -41
- data/ext/sources/examples/talk-llama/llama-memory-hybrid.h +30 -29
- data/ext/sources/examples/talk-llama/llama-memory-recurrent.cpp +77 -35
- data/ext/sources/examples/talk-llama/llama-memory-recurrent.h +15 -16
- data/ext/sources/examples/talk-llama/llama-memory.h +16 -10
- data/ext/sources/examples/talk-llama/llama-mmap.cpp +172 -37
- data/ext/sources/examples/talk-llama/llama-mmap.h +8 -3
- data/ext/sources/examples/talk-llama/llama-model-loader.cpp +93 -9
- data/ext/sources/examples/talk-llama/llama-model-loader.h +9 -2
- data/ext/sources/examples/talk-llama/llama-model-saver.cpp +3 -0
- data/ext/sources/examples/talk-llama/llama-model.cpp +3369 -10145
- data/ext/sources/examples/talk-llama/llama-model.h +104 -12
- data/ext/sources/examples/talk-llama/llama-quant.cpp +53 -30
- data/ext/sources/examples/talk-llama/llama-sampling.cpp +1520 -324
- data/ext/sources/examples/talk-llama/llama-sampling.h +19 -7
- data/ext/sources/examples/talk-llama/llama-vocab.cpp +562 -39
- data/ext/sources/examples/talk-llama/llama-vocab.h +50 -0
- data/ext/sources/examples/talk-llama/llama.cpp +794 -12
- data/ext/sources/examples/talk-llama/llama.h +246 -190
- data/ext/sources/examples/talk-llama/models/afmoe.cpp +191 -0
- data/ext/sources/examples/talk-llama/models/apertus.cpp +125 -0
- data/ext/sources/examples/talk-llama/models/arcee.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/arctic.cpp +138 -0
- data/ext/sources/examples/talk-llama/models/arwkv7.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/baichuan.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/bailingmoe.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/bailingmoe2.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/bert.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/bitnet.cpp +160 -0
- data/ext/sources/examples/talk-llama/models/bloom.cpp +101 -0
- data/ext/sources/examples/talk-llama/models/chameleon.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/chatglm.cpp +132 -0
- data/ext/sources/examples/talk-llama/models/codeshell.cpp +111 -0
- data/ext/sources/examples/talk-llama/models/cogvlm.cpp +102 -0
- data/ext/sources/examples/talk-llama/models/cohere2-iswa.cpp +134 -0
- data/ext/sources/examples/talk-llama/models/command-r.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/dbrx.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/deci.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/deepseek.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/deepseek2.cpp +259 -0
- data/ext/sources/examples/talk-llama/models/dots1.cpp +134 -0
- data/ext/sources/examples/talk-llama/models/dream.cpp +105 -0
- data/ext/sources/examples/talk-llama/models/ernie4-5-moe.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/ernie4-5.cpp +110 -0
- data/ext/sources/examples/talk-llama/models/exaone.cpp +114 -0
- data/ext/sources/examples/talk-llama/models/exaone4.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/falcon-h1.cpp +113 -0
- data/ext/sources/examples/talk-llama/models/falcon.cpp +120 -0
- data/ext/sources/examples/talk-llama/models/gemma-embedding.cpp +116 -0
- data/ext/sources/examples/talk-llama/models/gemma.cpp +112 -0
- data/ext/sources/examples/talk-llama/models/gemma2-iswa.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/gemma3.cpp +155 -0
- data/ext/sources/examples/talk-llama/models/gemma3n-iswa.cpp +384 -0
- data/ext/sources/examples/talk-llama/models/glm4-moe.cpp +170 -0
- data/ext/sources/examples/talk-llama/models/glm4.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/gpt2.cpp +105 -0
- data/ext/sources/examples/talk-llama/models/gptneox.cpp +144 -0
- data/ext/sources/examples/talk-llama/models/granite-hybrid.cpp +196 -0
- data/ext/sources/examples/talk-llama/models/granite.cpp +211 -0
- data/ext/sources/examples/talk-llama/models/graph-context-mamba.cpp +283 -0
- data/ext/sources/examples/talk-llama/models/grok.cpp +159 -0
- data/ext/sources/examples/talk-llama/models/grovemoe.cpp +141 -0
- data/ext/sources/examples/talk-llama/models/hunyuan-dense.cpp +132 -0
- data/ext/sources/examples/talk-llama/models/hunyuan-moe.cpp +154 -0
- data/ext/sources/examples/talk-llama/models/internlm2.cpp +120 -0
- data/ext/sources/examples/talk-llama/models/jais.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/jamba.cpp +106 -0
- data/ext/sources/examples/talk-llama/models/lfm2.cpp +175 -0
- data/ext/sources/examples/talk-llama/models/llada-moe.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/llada.cpp +99 -0
- data/ext/sources/examples/talk-llama/models/llama-iswa.cpp +178 -0
- data/ext/sources/examples/talk-llama/models/llama.cpp +168 -0
- data/ext/sources/examples/talk-llama/models/maincoder.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/mamba.cpp +55 -0
- data/ext/sources/examples/talk-llama/models/mimo2-iswa.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/minicpm3.cpp +199 -0
- data/ext/sources/examples/talk-llama/models/minimax-m2.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/mistral3.cpp +160 -0
- data/ext/sources/examples/talk-llama/models/models.h +569 -0
- data/ext/sources/examples/talk-llama/models/modern-bert.cpp +116 -0
- data/ext/sources/examples/talk-llama/models/mpt.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/nemotron-h.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/nemotron.cpp +122 -0
- data/ext/sources/examples/talk-llama/models/neo-bert.cpp +104 -0
- data/ext/sources/examples/talk-llama/models/olmo.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/olmo2.cpp +150 -0
- data/ext/sources/examples/talk-llama/models/olmoe.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/openai-moe-iswa.cpp +127 -0
- data/ext/sources/examples/talk-llama/models/openelm.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/orion.cpp +123 -0
- data/ext/sources/examples/talk-llama/models/pangu-embedded.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/phi2.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/phi3.cpp +152 -0
- data/ext/sources/examples/talk-llama/models/plamo.cpp +110 -0
- data/ext/sources/examples/talk-llama/models/plamo2.cpp +316 -0
- data/ext/sources/examples/talk-llama/models/plamo3.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/plm.cpp +168 -0
- data/ext/sources/examples/talk-llama/models/qwen.cpp +108 -0
- data/ext/sources/examples/talk-llama/models/qwen2.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/qwen2moe.cpp +151 -0
- data/ext/sources/examples/talk-llama/models/qwen2vl.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/qwen3.cpp +117 -0
- data/ext/sources/examples/talk-llama/models/qwen3moe.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/qwen3next.cpp +873 -0
- data/ext/sources/examples/talk-llama/models/qwen3vl-moe.cpp +149 -0
- data/ext/sources/examples/talk-llama/models/qwen3vl.cpp +141 -0
- data/ext/sources/examples/talk-llama/models/refact.cpp +94 -0
- data/ext/sources/examples/talk-llama/models/rnd1.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/rwkv6-base.cpp +162 -0
- data/ext/sources/examples/talk-llama/models/rwkv6.cpp +94 -0
- data/ext/sources/examples/talk-llama/models/rwkv6qwen2.cpp +86 -0
- data/ext/sources/examples/talk-llama/models/rwkv7-base.cpp +135 -0
- data/ext/sources/examples/talk-llama/models/rwkv7.cpp +90 -0
- data/ext/sources/examples/talk-llama/models/seed-oss.cpp +124 -0
- data/ext/sources/examples/talk-llama/models/smallthinker.cpp +126 -0
- data/ext/sources/examples/talk-llama/models/smollm3.cpp +128 -0
- data/ext/sources/examples/talk-llama/models/stablelm.cpp +146 -0
- data/ext/sources/examples/talk-llama/models/starcoder.cpp +100 -0
- data/ext/sources/examples/talk-llama/models/starcoder2.cpp +121 -0
- data/ext/sources/examples/talk-llama/models/t5-dec.cpp +166 -0
- data/ext/sources/examples/talk-llama/models/t5-enc.cpp +96 -0
- data/ext/sources/examples/talk-llama/models/wavtokenizer-dec.cpp +149 -0
- data/ext/sources/examples/talk-llama/models/xverse.cpp +108 -0
- data/ext/sources/examples/talk-llama/talk-llama.cpp +9 -6
- data/ext/sources/examples/talk-llama/unicode.cpp +309 -16
- data/ext/sources/examples/talk-llama/unicode.h +45 -0
- data/ext/sources/examples/vad-speech-segments/CMakeLists.txt +1 -1
- data/ext/sources/examples/wchess/wchess.cmd/wchess.cmd.cpp +4 -2
- data/ext/sources/examples/whisper.wasm/index-tmpl.html +18 -17
- data/ext/sources/ggml/CMakeLists.txt +135 -79
- data/ext/sources/ggml/cmake/ggml-config.cmake.in +132 -93
- data/ext/sources/ggml/include/ggml-alloc.h +9 -0
- data/ext/sources/ggml/include/ggml-backend.h +21 -2
- data/ext/sources/ggml/include/ggml-cpu.h +2 -1
- data/ext/sources/ggml/include/ggml-hexagon.h +19 -0
- data/ext/sources/ggml/include/ggml-metal.h +1 -6
- data/ext/sources/ggml/include/ggml-opt.h +25 -6
- data/ext/sources/ggml/include/ggml-rpc.h +8 -11
- data/ext/sources/ggml/include/ggml-webgpu.h +19 -0
- data/ext/sources/ggml/include/ggml-zdnn.h +17 -0
- data/ext/sources/ggml/include/ggml-zendnn.h +22 -0
- data/ext/sources/ggml/include/ggml.h +406 -23
- data/ext/sources/ggml/src/CMakeLists.txt +99 -13
- data/ext/sources/ggml/src/ggml-alloc.c +368 -161
- data/ext/sources/ggml/src/ggml-backend-impl.h +5 -5
- data/ext/sources/ggml/src/ggml-backend-reg.cpp +55 -14
- data/ext/sources/ggml/src/ggml-backend.cpp +290 -57
- data/ext/sources/ggml/src/ggml-blas/CMakeLists.txt +17 -3
- data/ext/sources/ggml/src/ggml-blas/ggml-blas.cpp +10 -13
- data/ext/sources/ggml/src/ggml-cann/CMakeLists.txt +14 -0
- data/ext/sources/ggml/src/ggml-cann/acl_tensor.cpp +59 -45
- data/ext/sources/ggml/src/ggml-cann/acl_tensor.h +138 -47
- data/ext/sources/ggml/src/ggml-cann/aclnn_ops.cpp +2586 -1917
- data/ext/sources/ggml/src/ggml-cann/aclnn_ops.h +348 -309
- data/ext/sources/ggml/src/ggml-cann/common.h +350 -133
- data/ext/sources/ggml/src/ggml-cann/ggml-cann.cpp +894 -625
- data/ext/sources/ggml/src/ggml-common.h +17 -0
- data/ext/sources/ggml/src/ggml-cpu/CMakeLists.txt +167 -75
- data/ext/sources/ggml/src/ggml-cpu/amx/amx.cpp +5 -2
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/cpu-feats.cpp +4 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/quants.c +560 -622
- data/ext/sources/ggml/src/ggml-cpu/arch/arm/repack.cpp +1002 -270
- data/ext/sources/ggml/src/ggml-cpu/arch/loongarch/quants.c +107 -587
- data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/quants.c +162 -589
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/cpu-feats.cpp +38 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/quants.c +373 -486
- data/ext/sources/ggml/src/ggml-cpu/arch/riscv/repack.cpp +3 -58
- data/ext/sources/ggml/src/ggml-cpu/arch/s390/cpu-feats.cpp +50 -0
- data/ext/sources/ggml/src/ggml-cpu/arch/s390/quants.c +521 -353
- data/ext/sources/ggml/src/ggml-cpu/arch/wasm/quants.c +54 -314
- data/ext/sources/ggml/src/ggml-cpu/arch/x86/quants.c +184 -675
- data/ext/sources/ggml/src/ggml-cpu/arch/x86/repack.cpp +4682 -1660
- data/ext/sources/ggml/src/ggml-cpu/arch-fallback.h +82 -4
- data/ext/sources/ggml/src/ggml-cpu/common.h +14 -0
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-impl.h +18 -9
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c +263 -111
- data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.cpp +39 -28
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.cpp +683 -82
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.h +38 -43
- data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +435 -119
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm-ppc.h +333 -0
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp +1234 -1182
- data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.h +6 -0
- data/ext/sources/ggml/src/ggml-cpu/ops.cpp +2167 -1480
- data/ext/sources/ggml/src/ggml-cpu/ops.h +10 -12
- data/ext/sources/ggml/src/ggml-cpu/quants.c +35 -0
- data/ext/sources/ggml/src/ggml-cpu/quants.h +8 -0
- data/ext/sources/ggml/src/ggml-cpu/repack.cpp +1132 -81
- data/ext/sources/ggml/src/ggml-cpu/repack.h +36 -0
- data/ext/sources/ggml/src/ggml-cpu/simd-mappings.h +120 -93
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.cpp +1025 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.h +13 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime1_kernels.cpp +3196 -0
- data/ext/sources/ggml/src/ggml-cpu/spacemit/ime_kernels.h +26 -0
- data/ext/sources/ggml/src/ggml-cpu/traits.cpp +2 -2
- data/ext/sources/ggml/src/ggml-cpu/traits.h +1 -1
- data/ext/sources/ggml/src/ggml-cpu/unary-ops.cpp +151 -0
- data/ext/sources/ggml/src/ggml-cpu/unary-ops.h +7 -0
- data/ext/sources/ggml/src/ggml-cpu/vec.cpp +294 -27
- data/ext/sources/ggml/src/ggml-cpu/vec.h +606 -48
- data/ext/sources/ggml/src/ggml-cuda/CMakeLists.txt +92 -17
- data/ext/sources/ggml/src/ggml-cuda/add-id.cu +58 -0
- data/ext/sources/ggml/src/ggml-cuda/add-id.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/argmax.cu +2 -2
- data/ext/sources/ggml/src/ggml-cuda/argsort.cu +123 -6
- data/ext/sources/ggml/src/ggml-cuda/argsort.cuh +16 -0
- data/ext/sources/ggml/src/ggml-cuda/binbcast.cu +330 -191
- data/ext/sources/ggml/src/ggml-cuda/binbcast.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/common.cuh +588 -128
- data/ext/sources/ggml/src/ggml-cuda/conv-transpose-1d.cu +1 -4
- data/ext/sources/ggml/src/ggml-cuda/conv2d.cu +166 -0
- data/ext/sources/ggml/src/ggml-cuda/conv2d.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/convert.cu +95 -22
- data/ext/sources/ggml/src/ggml-cuda/convert.cuh +25 -0
- data/ext/sources/ggml/src/ggml-cuda/cpy-utils.cuh +217 -0
- data/ext/sources/ggml/src/ggml-cuda/cpy.cu +335 -485
- data/ext/sources/ggml/src/ggml-cuda/cpy.cuh +1 -5
- data/ext/sources/ggml/src/ggml-cuda/cross-entropy-loss.cu +2 -14
- data/ext/sources/ggml/src/ggml-cuda/cumsum.cu +307 -0
- data/ext/sources/ggml/src/ggml-cuda/cumsum.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/dequantize.cuh +14 -40
- data/ext/sources/ggml/src/ggml-cuda/diag.cu +77 -0
- data/ext/sources/ggml/src/ggml-cuda/diag.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-common.cuh +519 -378
- data/ext/sources/ggml/src/ggml-cuda/fattn-mma-f16.cuh +750 -637
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cu +49 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cuh +1244 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec.cuh +586 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cu +98 -61
- data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +48 -0
- data/ext/sources/ggml/src/ggml-cuda/fattn.cu +230 -197
- data/ext/sources/ggml/src/ggml-cuda/fattn.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/fill.cu +37 -0
- data/ext/sources/ggml/src/ggml-cuda/fill.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/getrows.cu +50 -39
- data/ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu +1557 -294
- data/ext/sources/ggml/src/ggml-cuda/im2col.cu +196 -35
- data/ext/sources/ggml/src/ggml-cuda/im2col.cuh +1 -0
- data/ext/sources/ggml/src/ggml-cuda/mean.cu +57 -2
- data/ext/sources/ggml/src/ggml-cuda/mma.cuh +915 -69
- data/ext/sources/ggml/src/ggml-cuda/mmf.cu +171 -0
- data/ext/sources/ggml/src/ggml-cuda/mmf.cuh +835 -0
- data/ext/sources/ggml/src/ggml-cuda/mmid.cu +164 -0
- data/ext/sources/ggml/src/ggml-cuda/mmid.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/mmq.cu +109 -67
- data/ext/sources/ggml/src/ggml-cuda/mmq.cuh +1601 -733
- data/ext/sources/ggml/src/ggml-cuda/mmvf.cu +802 -0
- data/ext/sources/ggml/src/ggml-cuda/mmvf.cuh +12 -0
- data/ext/sources/ggml/src/ggml-cuda/mmvq.cu +286 -149
- data/ext/sources/ggml/src/ggml-cuda/mmvq.cuh +1 -1
- data/ext/sources/ggml/src/ggml-cuda/norm.cu +284 -12
- data/ext/sources/ggml/src/ggml-cuda/norm.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cu +49 -0
- data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/pad.cu +86 -32
- data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cu +91 -0
- data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/quantize.cu +163 -10
- data/ext/sources/ggml/src/ggml-cuda/quantize.cuh +14 -0
- data/ext/sources/ggml/src/ggml-cuda/reduce_rows.cuh +53 -0
- data/ext/sources/ggml/src/ggml-cuda/roll.cu +67 -0
- data/ext/sources/ggml/src/ggml-cuda/roll.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/rope.cu +207 -98
- data/ext/sources/ggml/src/ggml-cuda/rope.cuh +2 -0
- data/ext/sources/ggml/src/ggml-cuda/scale.cu +14 -11
- data/ext/sources/ggml/src/ggml-cuda/set-rows.cu +330 -0
- data/ext/sources/ggml/src/ggml-cuda/set-rows.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/set.cu +39 -0
- data/ext/sources/ggml/src/ggml-cuda/set.cuh +7 -0
- data/ext/sources/ggml/src/ggml-cuda/softcap.cu +34 -0
- data/ext/sources/ggml/src/ggml-cuda/softcap.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/softmax.cu +325 -61
- data/ext/sources/ggml/src/ggml-cuda/solve_tri.cu +275 -0
- data/ext/sources/ggml/src/ggml-cuda/solve_tri.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cu +14 -12
- data/ext/sources/ggml/src/ggml-cuda/ssm-scan.cu +291 -104
- data/ext/sources/ggml/src/ggml-cuda/sum.cu +6 -10
- data/ext/sources/ggml/src/ggml-cuda/sumrows.cu +21 -4
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu +7 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +40 -19
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu +5 -0
- data/ext/sources/ggml/src/ggml-cuda/top-k.cu +96 -0
- data/ext/sources/ggml/src/ggml-cuda/top-k.cuh +3 -0
- data/ext/sources/ggml/src/ggml-cuda/topk-moe.cu +351 -0
- data/ext/sources/ggml/src/ggml-cuda/topk-moe.cuh +21 -0
- data/ext/sources/ggml/src/ggml-cuda/tri.cu +136 -0
- data/ext/sources/ggml/src/ggml-cuda/tri.cuh +5 -0
- data/ext/sources/ggml/src/ggml-cuda/tsembd.cu +3 -3
- data/ext/sources/ggml/src/ggml-cuda/unary.cu +189 -5
- data/ext/sources/ggml/src/ggml-cuda/unary.cuh +44 -0
- data/ext/sources/ggml/src/ggml-cuda/upscale.cu +248 -6
- data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh +110 -22
- data/ext/sources/ggml/src/ggml-cuda/vendors/cuda.h +8 -0
- data/ext/sources/ggml/src/ggml-cuda/vendors/hip.h +70 -37
- data/ext/sources/ggml/src/ggml-cuda/vendors/musa.h +10 -3
- data/ext/sources/ggml/src/ggml-hexagon/CMakeLists.txt +80 -0
- data/ext/sources/ggml/src/ggml-hexagon/ggml-hexagon.cpp +3151 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/CMakeLists.txt +44 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/act-ops.c +682 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/binary-ops.c +360 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/cmake-toolchain.cmake +157 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/flash-attn-ops.c +566 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/get-rows-ops.c +112 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ctx.h +35 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.c +63 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.h +157 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-msg.h +165 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ops.h +92 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/htp_iface.idl +16 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.c +94 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.c +72 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c +49 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.c +1020 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.h +1353 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/main.c +1001 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/matmul-ops.c +2503 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/ops-utils.h +149 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/rope-ops.c +487 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/set-rows-ops.c +168 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/softmax-ops.c +402 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/unary-ops.c +287 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.c +297 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.h +57 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp-utils.c +454 -0
- data/ext/sources/ggml/src/ggml-hexagon/htp-utils.h +221 -0
- data/ext/sources/ggml/src/ggml-hexagon/op-desc.h +153 -0
- data/ext/sources/ggml/src/ggml-hip/CMakeLists.txt +16 -13
- data/ext/sources/ggml/src/ggml-impl.h +186 -15
- data/ext/sources/ggml/src/ggml-metal/CMakeLists.txt +10 -7
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.cpp +446 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.h +52 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.h +33 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.m +609 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.cpp +1743 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.h +273 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.m +1686 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-impl.h +356 -61
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.cpp +4161 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.h +94 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.cpp +724 -0
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.metal +4495 -1876
- data/ext/sources/ggml/src/ggml-musa/CMakeLists.txt +21 -9
- data/ext/sources/ggml/src/ggml-opencl/CMakeLists.txt +29 -0
- data/ext/sources/ggml/src/ggml-opencl/ggml-opencl.cpp +4005 -427
- data/ext/sources/ggml/src/ggml-opencl/kernels/add.cl +107 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/add_id.cl +42 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d.cl +185 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d_f16_f32.cl +176 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl +147 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/div.cl +66 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/expm1.cl +82 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/fill.cl +17 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl +370 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32.cl +371 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl +373 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gelu.cl +27 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl +162 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl +156 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/get_rows.cl +36 -12
- data/ext/sources/ggml/src/ggml-opencl/kernels/glu.cl +177 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/group_norm.cl +49 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f16.cl +1 -1
- data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f32.cl +1 -1
- data/ext/sources/ggml/src/ggml-opencl/kernels/mean.cl +39 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul.cl +73 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mat_f16_f32.cl +130 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl +273 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl +146 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl +147 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_l4_lm.cl +154 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32.cl +189 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32_flat.cl +176 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32.cl +140 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32_flat.cl +222 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32.cl +144 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32_flat.cl +167 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl +125 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl +202 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/norm.cl +80 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/pad.cl +29 -20
- data/ext/sources/ggml/src/ggml-opencl/kernels/rms_norm.cl +94 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/rope.cl +50 -24
- data/ext/sources/ggml/src/ggml-opencl/kernels/scale.cl +3 -2
- data/ext/sources/ggml/src/ggml-opencl/kernels/set_rows.cl +208 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f16.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f32.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f16.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f32.cl +34 -13
- data/ext/sources/ggml/src/ggml-opencl/kernels/softplus.cl +88 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sqr.cl +53 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sqrt.cl +53 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/ssm_conv.cl +77 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/sub.cl +66 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl +33 -0
- data/ext/sources/ggml/src/ggml-opencl/kernels/tsembd.cl +2 -2
- data/ext/sources/ggml/src/ggml-opencl/kernels/upscale.cl +2 -3
- data/ext/sources/ggml/src/ggml-opt.cpp +97 -41
- data/ext/sources/ggml/src/ggml-quants.c +111 -16
- data/ext/sources/ggml/src/ggml-quants.h +6 -0
- data/ext/sources/ggml/src/ggml-rpc/ggml-rpc.cpp +497 -195
- data/ext/sources/ggml/src/ggml-sycl/CMakeLists.txt +48 -3
- data/ext/sources/ggml/src/ggml-sycl/add-id.cpp +77 -0
- data/ext/sources/ggml/src/ggml-sycl/add-id.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/backend.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/binbcast.cpp +6 -5
- data/ext/sources/ggml/src/ggml-sycl/common.hpp +117 -15
- data/ext/sources/ggml/src/ggml-sycl/concat.cpp +50 -30
- data/ext/sources/ggml/src/ggml-sycl/conv.cpp +10 -4
- data/ext/sources/ggml/src/ggml-sycl/convert.cpp +200 -99
- data/ext/sources/ggml/src/ggml-sycl/count-equal.cpp +79 -0
- data/ext/sources/ggml/src/ggml-sycl/count-equal.hpp +9 -0
- data/ext/sources/ggml/src/ggml-sycl/cpy.cpp +72 -309
- data/ext/sources/ggml/src/ggml-sycl/cpy.hpp +213 -1
- data/ext/sources/ggml/src/ggml-sycl/dequantize.hpp +18 -0
- data/ext/sources/ggml/src/ggml-sycl/dmmv.cpp +67 -49
- data/ext/sources/ggml/src/ggml-sycl/dpct/helper.hpp +77 -34
- data/ext/sources/ggml/src/ggml-sycl/element_wise.cpp +397 -314
- data/ext/sources/ggml/src/ggml-sycl/element_wise.hpp +12 -2
- data/ext/sources/ggml/src/ggml-sycl/gemm.hpp +14 -26
- data/ext/sources/ggml/src/ggml-sycl/getrows.cpp +9 -6
- data/ext/sources/ggml/src/ggml-sycl/ggml-sycl.cpp +643 -413
- data/ext/sources/ggml/src/ggml-sycl/gla.cpp +2 -2
- data/ext/sources/ggml/src/ggml-sycl/im2col.cpp +2 -2
- data/ext/sources/ggml/src/ggml-sycl/mmq.cpp +80 -60
- data/ext/sources/ggml/src/ggml-sycl/mmvq.cpp +223 -132
- data/ext/sources/ggml/src/ggml-sycl/norm.cpp +230 -55
- data/ext/sources/ggml/src/ggml-sycl/norm.hpp +2 -0
- data/ext/sources/ggml/src/ggml-sycl/pad.cpp +97 -0
- data/ext/sources/ggml/src/ggml-sycl/pad.hpp +24 -0
- data/ext/sources/ggml/src/ggml-sycl/pad_reflect_1d.cpp +100 -0
- data/ext/sources/ggml/src/ggml-sycl/pad_reflect_1d.hpp +10 -0
- data/ext/sources/ggml/src/ggml-sycl/presets.hpp +2 -0
- data/ext/sources/ggml/src/ggml-sycl/quantize.hpp +133 -0
- data/ext/sources/ggml/src/ggml-sycl/quants.hpp +8 -9
- data/ext/sources/ggml/src/ggml-sycl/repeat_back.cpp +76 -0
- data/ext/sources/ggml/src/ggml-sycl/repeat_back.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/roll.cpp +122 -0
- data/ext/sources/ggml/src/ggml-sycl/roll.hpp +20 -0
- data/ext/sources/ggml/src/ggml-sycl/rope.cpp +65 -59
- data/ext/sources/ggml/src/ggml-sycl/set.cpp +73 -0
- data/ext/sources/ggml/src/ggml-sycl/set.hpp +5 -0
- data/ext/sources/ggml/src/ggml-sycl/set_rows.cpp +234 -0
- data/ext/sources/ggml/src/ggml-sycl/set_rows.hpp +8 -0
- data/ext/sources/ggml/src/ggml-sycl/softmax.cpp +330 -165
- data/ext/sources/ggml/src/ggml-sycl/softmax.hpp +4 -0
- data/ext/sources/ggml/src/ggml-sycl/ssm_conv.cpp +127 -0
- data/ext/sources/ggml/src/ggml-sycl/ssm_conv.hpp +5 -0
- data/ext/sources/ggml/src/ggml-sycl/tsembd.cpp +12 -6
- data/ext/sources/ggml/src/ggml-sycl/vecdotq.hpp +60 -6
- data/ext/sources/ggml/src/ggml-sycl/wkv.cpp +16 -12
- data/ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt +38 -18
- data/ext/sources/ggml/src/ggml-vulkan/ggml-vulkan.cpp +7398 -2635
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/abs.comp +21 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +43 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add1.comp +28 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add_id.comp +42 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/arange.comp +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +15 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +56 -39
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp +114 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ceil.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp +347 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +5 -5
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +67 -13
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_transpose.comp +67 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/count_experts.comp +51 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum.comp +83 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum_multipass1.comp +60 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cumsum_multipass2.comp +66 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs.comp → dequant_funcs.glsl} +158 -16
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs_cm2.comp → dequant_funcs_cm2.glsl} +38 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_head.comp → dequant_head.glsl} +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +3 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +7 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +5 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_mxfp4.comp +32 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +4 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag.comp +29 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp +21 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/fill.comp +19 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +103 -36
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.glsl +220 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +139 -45
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +113 -38
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +75 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/floor.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp +27 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp +11 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_erf.comp +39 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_binary_head.comp → generic_binary_head.glsl} +19 -17
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_head.comp → generic_head.glsl} +2 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{generic_unary_head.comp → generic_unary_head.glsl} +7 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +21 -12
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +28 -18
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{glu_head.comp → glu_head.glsl} +4 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardsigmoid.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardswish.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +33 -17
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col_3d.comp +125 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/log.comp +18 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl +227 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iface.glsl +35 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +71 -21
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +41 -25
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +44 -26
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +20 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +9 -7
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +4 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq.comp +143 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq_funcs.glsl +494 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +144 -556
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +230 -51
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl +566 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_id_funcs.glsl +72 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +90 -223
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl +454 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_shmem_types.glsl +78 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/multi_add.comp +195 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/neg.comp +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_sgd.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +41 -5
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +1 -1
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +59 -9
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +104 -14
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_partials.comp +65 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/roll.comp +46 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_funcs.glsl +234 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.glsl +20 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +6 -52
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +6 -35
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +6 -35
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl +28 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +6 -39
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/round.comp +29 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rte.glsl +5 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +3 -3
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +30 -8
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +6 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large1.comp +62 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large2.comp +79 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large3.comp +65 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large_common.glsl +53 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/softplus.comp +23 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/solve_tri.comp +81 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp +17 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp +44 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_scan.comp +124 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/step.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +16 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.glsl +25 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu_oai.comp +14 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +2 -2
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +5 -4
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_argsort.comp +118 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_moe.comp +213 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/topk_nary_search.comp +246 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/tri.comp +43 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/trunc.comp +22 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{types.comp → types.glsl} +435 -24
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +148 -6
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/utils.glsl +25 -0
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +619 -177
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/xielu.comp +35 -0
- data/ext/sources/ggml/src/ggml-webgpu/CMakeLists.txt +80 -0
- data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp +169 -0
- data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu.cpp +3087 -0
- data/ext/sources/ggml/src/ggml-webgpu/pre_wgsl.hpp +778 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl +188 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +45 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +930 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl +101 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +147 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/flash_attn.wgsl +591 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl +874 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/glu.tmpl.wgsl +323 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/memset.wgsl +40 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl +907 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl +97 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.tmpl.wgsl +247 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.tmpl.wgsl +302 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl +267 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl +123 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl +295 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/scale.tmpl.wgsl +90 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.tmpl.wgsl +112 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.wgsl +81 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl +345 -0
- data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary_op.wgsl +483 -0
- data/ext/sources/ggml/src/ggml-zdnn/CMakeLists.txt +36 -0
- data/ext/sources/ggml/src/ggml-zdnn/common.hpp +59 -0
- data/ext/sources/ggml/src/ggml-zdnn/ggml-zdnn.cpp +628 -0
- data/ext/sources/ggml/src/ggml-zdnn/mmf.cpp +80 -0
- data/ext/sources/ggml/src/ggml-zdnn/mmf.hpp +12 -0
- data/ext/sources/ggml/src/ggml-zdnn/utils.cpp +79 -0
- data/ext/sources/ggml/src/ggml-zdnn/utils.hpp +19 -0
- data/ext/sources/ggml/src/ggml-zendnn/CMakeLists.txt +92 -0
- data/ext/sources/ggml/src/ggml-zendnn/ggml-zendnn.cpp +466 -0
- data/ext/sources/ggml/src/ggml.c +901 -129
- data/ext/sources/ggml/src/gguf.cpp +8 -1
- data/ext/sources/include/whisper.h +1 -0
- data/ext/sources/src/CMakeLists.txt +3 -1
- data/ext/sources/src/whisper.cpp +124 -81
- data/ext/sources/tests/CMakeLists.txt +8 -1
- data/ext/sources/tests/test-vad-full.cpp +7 -5
- data/ext/sources/tests/test-vad.cpp +3 -3
- data/extsources.rb +1 -0
- data/lib/whisper/model/uri.rb +17 -18
- data/sig/whisper.rbs +126 -2
- data/test/test_params.rb +24 -8
- data/test/test_segment.rb +0 -1
- data/test/test_token.rb +70 -0
- data/test/test_vad.rb +1 -1
- data/test/test_vad_context.rb +50 -0
- data/test/test_vad_segment.rb +19 -0
- data/test/test_vad_segments.rb +16 -0
- data/test/test_whisper.rb +8 -1
- data/whispercpp.gemspec +1 -1
- metadata +439 -179
- data/ext/sources/build-xcframework.sh +0 -547
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified-iswa.cpp +0 -279
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified.cpp +0 -1841
- data/ext/sources/examples/talk-llama/llama-kv-cache-unified.h +0 -303
- data/ext/sources/ggml/include/ggml-kompute.h +0 -50
- data/ext/sources/ggml/src/ggml-amx/CMakeLists.txt +0 -107
- data/ext/sources/ggml/src/ggml-amx/common.h +0 -94
- data/ext/sources/ggml/src/ggml-amx/ggml-amx.cpp +0 -446
- data/ext/sources/ggml/src/ggml-amx/mmq.cpp +0 -2510
- data/ext/sources/ggml/src/ggml-amx/mmq.h +0 -17
- data/ext/sources/ggml/src/ggml-cann/Doxyfile +0 -2579
- data/ext/sources/ggml/src/ggml-cann/kernels/CMakeLists.txt +0 -30
- data/ext/sources/ggml/src/ggml-cann/kernels/ascendc_kernels.h +0 -19
- data/ext/sources/ggml/src/ggml-cann/kernels/dup.cpp +0 -234
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f16.cpp +0 -197
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f32.cpp +0 -190
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +0 -204
- data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q8_0.cpp +0 -191
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +0 -218
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +0 -216
- data/ext/sources/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +0 -295
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cu +0 -357
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cuh +0 -3
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cu +0 -365
- data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cuh +0 -3
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f16.cuh +0 -482
- data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f32.cuh +0 -472
- data/ext/sources/ggml/src/ggml-cuda/mmv.cu +0 -506
- data/ext/sources/ggml/src/ggml-cuda/mmv.cuh +0 -11
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
- data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
- data/ext/sources/ggml/src/ggml-kompute/CMakeLists.txt +0 -166
- data/ext/sources/ggml/src/ggml-kompute/ggml-kompute.cpp +0 -2251
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/common.comp +0 -112
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_add.comp +0 -58
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_addrow.comp +0 -25
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_diagmask.comp +0 -30
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_gelu.comp +0 -22
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows.comp +0 -17
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f16.comp +0 -31
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f32.comp +0 -31
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_0.comp +0 -38
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_1.comp +0 -39
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q6_k.comp +0 -44
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp +0 -69
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_mat_f32.comp +0 -51
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_0.comp +0 -33
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_1.comp +0 -35
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp +0 -140
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp +0 -106
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q8_0.comp +0 -73
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp +0 -28
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_norm.comp +0 -84
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_relu.comp +0 -21
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rmsnorm.comp +0 -53
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp +0 -52
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale.comp +0 -19
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale_8.comp +0 -23
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_silu.comp +0 -22
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_softmax.comp +0 -72
- data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/rope_common.comp +0 -71
- data/ext/sources/ggml/src/ggml-metal/ggml-metal.m +0 -6280
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.comp +0 -162
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +0 -118
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +0 -99
- data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +0 -58
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_bfloat16_support.comp → feature-tests/bfloat16.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat_support.comp → feature-tests/coopmat.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat2_support.comp → feature-tests/coopmat2.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{test_integer_dot_support.comp → feature-tests/integer_dot.comp} +0 -0
- /data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/{glu_main.comp → glu_main.glsl} +0 -0
|
@@ -114,6 +114,9 @@ message(STATUS "GGML_SYSTEM_ARCH: ${GGML_SYSTEM_ARCH}")
|
|
|
114
114
|
|
|
115
115
|
if (NOT MSVC)
|
|
116
116
|
if (GGML_STATIC)
|
|
117
|
+
if (UNIX AND NOT APPLE)
|
|
118
|
+
set(CMAKE_FIND_LIBRARY_SUFFIXES ".a;.so")
|
|
119
|
+
endif()
|
|
117
120
|
add_link_options(-static)
|
|
118
121
|
if (MINGW)
|
|
119
122
|
add_link_options(-static-libgcc -static-libstdc++)
|
|
@@ -124,10 +127,6 @@ if (NOT MSVC)
|
|
|
124
127
|
endif()
|
|
125
128
|
endif()
|
|
126
129
|
|
|
127
|
-
if (MINGW)
|
|
128
|
-
add_compile_definitions(_WIN32_WINNT=${GGML_WIN_VER})
|
|
129
|
-
endif()
|
|
130
|
-
|
|
131
130
|
#
|
|
132
131
|
# POSIX conformance
|
|
133
132
|
#
|
|
@@ -142,6 +141,9 @@ endif()
|
|
|
142
141
|
# which was introduced in POSIX.1-2008, forcing us to go higher
|
|
143
142
|
if (CMAKE_SYSTEM_NAME MATCHES "OpenBSD")
|
|
144
143
|
add_compile_definitions(_XOPEN_SOURCE=700)
|
|
144
|
+
elseif (CMAKE_SYSTEM_NAME MATCHES "AIX")
|
|
145
|
+
# Don't define _XOPEN_SOURCE. We need _ALL_SOURCE, which is the default,
|
|
146
|
+
# in order to define _SC_PHYS_PAGES.
|
|
145
147
|
else()
|
|
146
148
|
add_compile_definitions(_XOPEN_SOURCE=600)
|
|
147
149
|
endif()
|
|
@@ -205,15 +207,36 @@ add_library(ggml-base
|
|
|
205
207
|
ggml-quants.h
|
|
206
208
|
gguf.cpp)
|
|
207
209
|
|
|
210
|
+
set_target_properties(ggml-base PROPERTIES
|
|
211
|
+
VERSION ${GGML_VERSION}
|
|
212
|
+
SOVERSION ${GGML_VERSION_MAJOR}
|
|
213
|
+
)
|
|
214
|
+
|
|
208
215
|
target_include_directories(ggml-base PRIVATE .)
|
|
209
216
|
if (GGML_BACKEND_DL)
|
|
210
217
|
target_compile_definitions(ggml-base PUBLIC GGML_BACKEND_DL)
|
|
211
218
|
endif()
|
|
212
219
|
|
|
220
|
+
if (GGML_SCHED_NO_REALLOC)
|
|
221
|
+
target_compile_definitions(ggml-base PUBLIC GGML_SCHED_NO_REALLOC)
|
|
222
|
+
endif()
|
|
223
|
+
|
|
213
224
|
add_library(ggml
|
|
214
225
|
ggml-backend-reg.cpp)
|
|
215
226
|
add_library(ggml::ggml ALIAS ggml)
|
|
216
227
|
|
|
228
|
+
set_target_properties(ggml PROPERTIES
|
|
229
|
+
VERSION ${GGML_VERSION}
|
|
230
|
+
SOVERSION ${GGML_VERSION_MAJOR}
|
|
231
|
+
)
|
|
232
|
+
|
|
233
|
+
if (GGML_BACKEND_DIR)
|
|
234
|
+
if (NOT GGML_BACKEND_DL)
|
|
235
|
+
message(FATAL_ERROR "GGML_BACKEND_DIR requires GGML_BACKEND_DL")
|
|
236
|
+
endif()
|
|
237
|
+
target_compile_definitions(ggml PUBLIC GGML_BACKEND_DIR="${GGML_BACKEND_DIR}")
|
|
238
|
+
endif()
|
|
239
|
+
|
|
217
240
|
target_link_libraries(ggml PUBLIC ggml-base)
|
|
218
241
|
|
|
219
242
|
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
|
|
@@ -227,7 +250,11 @@ function(ggml_add_backend_library backend)
|
|
|
227
250
|
set_target_properties(${backend} PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_RUNTIME_OUTPUT_DIRECTORY})
|
|
228
251
|
target_compile_definitions(${backend} PRIVATE GGML_BACKEND_DL)
|
|
229
252
|
add_dependencies(ggml ${backend})
|
|
230
|
-
|
|
253
|
+
if (GGML_BACKEND_DIR)
|
|
254
|
+
install(TARGETS ${backend} LIBRARY DESTINATION ${GGML_BACKEND_DIR})
|
|
255
|
+
else()
|
|
256
|
+
install(TARGETS ${backend} LIBRARY DESTINATION ${CMAKE_INSTALL_BINDIR})
|
|
257
|
+
endif()
|
|
231
258
|
else()
|
|
232
259
|
add_library(${backend} ${ARGN})
|
|
233
260
|
target_link_libraries(ggml PUBLIC ${backend})
|
|
@@ -242,6 +269,15 @@ function(ggml_add_backend_library backend)
|
|
|
242
269
|
target_compile_definitions(${backend} PUBLIC GGML_BACKEND_SHARED)
|
|
243
270
|
endif()
|
|
244
271
|
|
|
272
|
+
# Set versioning properties for all backend libraries
|
|
273
|
+
# Building a MODULE library with a version is not supported on macOS (https://gitlab.kitware.com/cmake/cmake/-/issues/20782)
|
|
274
|
+
if (NOT (APPLE AND GGML_BACKEND_DL))
|
|
275
|
+
set_target_properties(${backend} PROPERTIES
|
|
276
|
+
VERSION ${GGML_VERSION}
|
|
277
|
+
SOVERSION ${GGML_VERSION_MAJOR}
|
|
278
|
+
)
|
|
279
|
+
endif()
|
|
280
|
+
|
|
245
281
|
if(NOT GGML_AVAILABLE_BACKENDS)
|
|
246
282
|
set(GGML_AVAILABLE_BACKENDS "${backend}"
|
|
247
283
|
CACHE INTERNAL "List of backends for cmake package")
|
|
@@ -287,6 +323,22 @@ function(ggml_add_cpu_backend_variant tag_name)
|
|
|
287
323
|
set(GGML_INTERNAL_${feat} ON)
|
|
288
324
|
endforeach()
|
|
289
325
|
elseif (GGML_SYSTEM_ARCH STREQUAL "PowerPC")
|
|
326
|
+
foreach (feat ${ARGN})
|
|
327
|
+
set(GGML_INTERNAL_${feat} ON)
|
|
328
|
+
endforeach()
|
|
329
|
+
elseif (GGML_SYSTEM_ARCH STREQUAL "s390x")
|
|
330
|
+
foreach (feat VXE2 NNPA)
|
|
331
|
+
set(GGML_INTERNAL_${feat} OFF)
|
|
332
|
+
endforeach()
|
|
333
|
+
|
|
334
|
+
foreach (feat ${ARGN})
|
|
335
|
+
set(GGML_INTERNAL_${feat} ON)
|
|
336
|
+
endforeach()
|
|
337
|
+
elseif (GGML_SYSTEM_ARCH STREQUAL "riscv64")
|
|
338
|
+
foreach (feat RVV)
|
|
339
|
+
set(GGML_INTERNAL_${feat} OFF)
|
|
340
|
+
endforeach()
|
|
341
|
+
|
|
290
342
|
foreach (feat ${ARGN})
|
|
291
343
|
set(GGML_INTERNAL_${feat} ON)
|
|
292
344
|
endforeach()
|
|
@@ -305,15 +357,29 @@ if (GGML_CPU_ALL_VARIANTS)
|
|
|
305
357
|
endif()
|
|
306
358
|
if (GGML_SYSTEM_ARCH STREQUAL "x86")
|
|
307
359
|
ggml_add_cpu_backend_variant(x64)
|
|
308
|
-
ggml_add_cpu_backend_variant(sse42
|
|
309
|
-
ggml_add_cpu_backend_variant(sandybridge
|
|
310
|
-
|
|
311
|
-
|
|
312
|
-
|
|
313
|
-
|
|
360
|
+
ggml_add_cpu_backend_variant(sse42 SSE42)
|
|
361
|
+
ggml_add_cpu_backend_variant(sandybridge SSE42 AVX)
|
|
362
|
+
if (NOT MSVC)
|
|
363
|
+
# __FMA__ and __F16C__ are not defined in MSVC, however they are implied with AVX2/AVX512
|
|
364
|
+
ggml_add_cpu_backend_variant(ivybridge SSE42 AVX F16C)
|
|
365
|
+
ggml_add_cpu_backend_variant(piledriver SSE42 AVX F16C FMA)
|
|
366
|
+
endif()
|
|
367
|
+
ggml_add_cpu_backend_variant(haswell SSE42 AVX F16C FMA AVX2 BMI2)
|
|
368
|
+
ggml_add_cpu_backend_variant(skylakex SSE42 AVX F16C FMA AVX2 BMI2 AVX512)
|
|
369
|
+
ggml_add_cpu_backend_variant(cannonlake SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VBMI)
|
|
370
|
+
ggml_add_cpu_backend_variant(cascadelake SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VNNI)
|
|
371
|
+
ggml_add_cpu_backend_variant(icelake SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VBMI AVX512_VNNI)
|
|
372
|
+
if (NOT MSVC)
|
|
373
|
+
# MSVC 2022 doesn't support BF16 intrinsics without `/arch:AVX10.1` ?!
|
|
374
|
+
# https://learn.microsoft.com/en-us/cpp/intrinsics/x64-amd64-intrinsics-list?view=msvc-170
|
|
375
|
+
# https://learn.microsoft.com/en-us/cpp/build/reference/arch-x64?view=msvc-170
|
|
376
|
+
ggml_add_cpu_backend_variant(cooperlake SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VNNI AVX512_BF16)
|
|
377
|
+
ggml_add_cpu_backend_variant(zen4 SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16)
|
|
378
|
+
endif()
|
|
379
|
+
ggml_add_cpu_backend_variant(alderlake SSE42 AVX F16C FMA AVX2 BMI2 AVX_VNNI)
|
|
314
380
|
if (NOT MSVC)
|
|
315
381
|
# MSVC doesn't support AMX
|
|
316
|
-
ggml_add_cpu_backend_variant(sapphirerapids SSE42 AVX F16C AVX2 BMI2
|
|
382
|
+
ggml_add_cpu_backend_variant(sapphirerapids SSE42 AVX F16C FMA AVX2 BMI2 AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8)
|
|
317
383
|
endif()
|
|
318
384
|
elseif(GGML_SYSTEM_ARCH STREQUAL "ARM")
|
|
319
385
|
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
|
|
@@ -334,6 +400,9 @@ if (GGML_CPU_ALL_VARIANTS)
|
|
|
334
400
|
ggml_add_cpu_backend_variant(android_armv8.2_1 DOTPROD)
|
|
335
401
|
ggml_add_cpu_backend_variant(android_armv8.2_2 DOTPROD FP16_VECTOR_ARITHMETIC)
|
|
336
402
|
ggml_add_cpu_backend_variant(android_armv8.6_1 DOTPROD FP16_VECTOR_ARITHMETIC MATMUL_INT8)
|
|
403
|
+
ggml_add_cpu_backend_variant(android_armv9.0_1 DOTPROD MATMUL_INT8 FP16_VECTOR_ARITHMETIC SVE2)
|
|
404
|
+
ggml_add_cpu_backend_variant(android_armv9.2_1 DOTPROD MATMUL_INT8 FP16_VECTOR_ARITHMETIC SVE SME)
|
|
405
|
+
ggml_add_cpu_backend_variant(android_armv9.2_2 DOTPROD MATMUL_INT8 FP16_VECTOR_ARITHMETIC SVE SVE2 SME)
|
|
337
406
|
elseif (APPLE)
|
|
338
407
|
ggml_add_cpu_backend_variant(apple_m1 DOTPROD)
|
|
339
408
|
ggml_add_cpu_backend_variant(apple_m2_m3 DOTPROD MATMUL_INT8)
|
|
@@ -354,6 +423,20 @@ if (GGML_CPU_ALL_VARIANTS)
|
|
|
354
423
|
else()
|
|
355
424
|
message(FATAL_ERROR "Unsupported PowerPC target OS: ${CMAKE_SYSTEM_NAME}")
|
|
356
425
|
endif()
|
|
426
|
+
elseif (GGML_SYSTEM_ARCH STREQUAL "s390x")
|
|
427
|
+
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
|
|
428
|
+
ggml_add_cpu_backend_variant(z15 Z15 VXE2)
|
|
429
|
+
ggml_add_cpu_backend_variant(z16 Z16 VXE2 NNPA)
|
|
430
|
+
else()
|
|
431
|
+
message(FATAL_ERROR "Unsupported s390x target OS: ${CMAKE_SYSTEM_NAME}")
|
|
432
|
+
endif()
|
|
433
|
+
elseif (GGML_SYSTEM_ARCH STREQUAL "riscv64")
|
|
434
|
+
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
|
|
435
|
+
ggml_add_cpu_backend_variant(riscv64_0)
|
|
436
|
+
ggml_add_cpu_backend_variant(riscv64_v RVV)
|
|
437
|
+
else()
|
|
438
|
+
message(FATAL_ERROR "Unsupported RISC-V target OS: ${CMAKE_SYSTEM_NAME}")
|
|
439
|
+
endif()
|
|
357
440
|
else()
|
|
358
441
|
message(FATAL_ERROR "GGML_CPU_ALL_VARIANTS not yet supported with ${GGML_SYSTEM_ARCH} on ${CMAKE_SYSTEM_NAME}")
|
|
359
442
|
endif()
|
|
@@ -365,13 +448,16 @@ ggml_add_backend(BLAS)
|
|
|
365
448
|
ggml_add_backend(CANN)
|
|
366
449
|
ggml_add_backend(CUDA)
|
|
367
450
|
ggml_add_backend(HIP)
|
|
368
|
-
ggml_add_backend(Kompute)
|
|
369
451
|
ggml_add_backend(METAL)
|
|
370
452
|
ggml_add_backend(MUSA)
|
|
371
453
|
ggml_add_backend(RPC)
|
|
372
454
|
ggml_add_backend(SYCL)
|
|
373
455
|
ggml_add_backend(Vulkan)
|
|
456
|
+
ggml_add_backend(WebGPU)
|
|
457
|
+
ggml_add_backend(zDNN)
|
|
374
458
|
ggml_add_backend(OpenCL)
|
|
459
|
+
ggml_add_backend(Hexagon)
|
|
460
|
+
ggml_add_backend(ZenDNN)
|
|
375
461
|
|
|
376
462
|
foreach (target ggml-base ggml)
|
|
377
463
|
target_include_directories(${target} PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/../include> $<INSTALL_INTERFACE:include>)
|