ggml: Add initial WebGPU backend (llama/14521)

* Minimal setup of webgpu backend with dawn. Just prints out the adapter and segfaults

* Initialize webgpu device

* Making progress on setting up the backend

* Finish more boilerplate/utility functions

* Organize file and work on alloc buffer

* Add webgpu_context to prepare for actually running some shaders

* Work on memset and add shader loading

* Work on memset polyfill

* Implement set_tensor as webgpu WriteBuffer, remove host_buffer stubs since webgpu doesn't support it

* Implement get_tensor and buffer_clear

* Finish rest of setup

* Start work on compute graph

* Basic mat mul working

* Work on emscripten build

* Basic WebGPU backend instructions

* Use EMSCRIPTEN flag

* Work on passing ci, implement 4d tensor multiplication

* Pass thread safety test

* Implement permuting for mul_mat and cpy

* minor cleanups

* Address feedback

* Remove division by type size in cpy op

* Fix formatting and add github action workflows for vulkan and metal (m-series) webgpu backends

* Fix name

* Fix macos dawn prefix path

diff --git a/CMakeLists.txt b/CMakeLists.txt
index eaba9c70469efcae2b1d83995bcc3187a5360b1a..de6d789c98a033ada8ba1bd48c6bb525f72f8cbb 100644 (file)
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -181,6 +181,8 @@ option(GGML_VULKAN_MEMORY_DEBUG             "ggml: enable Vulkan memory debug ou
 option(GGML_VULKAN_SHADER_DEBUG_INFO        "ggml: enable Vulkan shader debug info"           OFF)
 option(GGML_VULKAN_VALIDATE                 "ggml: enable Vulkan validation"                  OFF)
 option(GGML_VULKAN_RUN_TESTS                "ggml: run Vulkan tests"                          OFF)
+option(GGML_WEBGPU                          "ggml: use WebGPU"                                OFF)
+option(GGML_WEBGPU_DEBUG                    "ggml: enable WebGPU debug output"                OFF)
 option(GGML_METAL                           "ggml: use Metal"                                 ${GGML_METAL_DEFAULT})
 option(GGML_METAL_USE_BF16                  "ggml: use bfloat if available"                   OFF)
 option(GGML_METAL_NDEBUG                    "ggml: disable Metal debugging"                   OFF)
@@ -270,6 +272,7 @@ set(GGML_PUBLIC_HEADERS
     include/ggml-rpc.h
     include/ggml-sycl.h
     include/ggml-vulkan.h
+    include/ggml-webgpu.h
     include/gguf.h)
 
 set_target_properties(ggml PROPERTIES PUBLIC_HEADER "${GGML_PUBLIC_HEADERS}")

diff --git a/include/ggml-webgpu.h b/include/ggml-webgpu.h
new file mode 100644 (file)
index 0000000..65b8ed9
--- /dev/null
+++ b/include/ggml-webgpu.h
@@ -0,0 +1,19 @@
+#pragma once
+
+#include "ggml.h"
+#include "ggml-backend.h"
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+#define GGML_WEBGPU_NAME "WebGPU"
+
+// Needed for examples in ggml
+GGML_BACKEND_API ggml_backend_t ggml_backend_webgpu_init(void);
+
+GGML_BACKEND_API ggml_backend_reg_t ggml_backend_webgpu_reg(void);
+
+#ifdef  __cplusplus
+}
+#endif

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 8760c2d35eca4ea5db48423bf70de80992cfdc04..0425fd60a9412a46fdc18762325dd24c400e99a7 100644 (file)
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -370,6 +370,7 @@ ggml_add_backend(MUSA)
 ggml_add_backend(RPC)
 ggml_add_backend(SYCL)
 ggml_add_backend(Vulkan)
+ggml_add_backend(WebGPU)
 ggml_add_backend(OpenCL)
 
 foreach (target ggml-base ggml)

diff --git a/src/ggml-backend-reg.cpp b/src/ggml-backend-reg.cpp
index 042ea77aca7212e8985cdfb5ca7b5e23bd3e32d7..f0cdac31eae9afbcb540f2f176f380eb3ab0fa3a 100644 (file)
--- a/src/ggml-backend-reg.cpp
+++ b/src/ggml-backend-reg.cpp
@@ -45,6 +45,10 @@
 #include "ggml-vulkan.h"
 #endif
 
+#ifdef GGML_USE_WEBGPU
+#include "ggml-webgpu.h"
+#endif
+
 #ifdef GGML_USE_OPENCL
 #include "ggml-opencl.h"
 #endif
@@ -173,6 +177,9 @@ struct ggml_backend_registry {
 #ifdef GGML_USE_VULKAN
         register_backend(ggml_backend_vk_reg());
 #endif
+#ifdef GGML_USE_WEBGPU
+        register_backend(ggml_backend_webgpu_reg());
+#endif
 #ifdef GGML_USE_OPENCL
         register_backend(ggml_backend_opencl_reg());
 #endif