ggml-webgpu: remove legacy constants (ggml-org#23672)

Author: reeselevine

ggml/src/ggml-webgpu/ggml-webgpu.cpp

@@ -94,14 +94,6 @@ static inline uint32_t ggml_webgpu_u32_from_f32(float value) {
 #define WEBGPU_SET_ROWS_ERROR_BUF_SIZE_BYTES     4
 #define WEBGPU_STORAGE_BUF_BINDING_MULT          4    // a storage buffer binding size must be a multiple of 4
 
-// For operations which process a row in parallel, this seems like a reasonable
-// default
-#define WEBGPU_ROW_SPLIT_WG_SIZE 64
-
-// Track https://github.com/gpuweb/gpuweb/issues/5315 for fixes to
-// implementations so this can be removed, necessary only for get_rows right now
-#define WEBGPU_MAX_WG_SIZE 288
-
 /* End Constants */
 
 // This is a "fake" base pointer, since WebGPU buffers do not have pointers to
@@ -631,7 +623,7 @@ static void ggml_backend_webgpu_buffer_memset(webgpu_global_context & ctx,
                                               size_t                  size) {
     std::vector<uint32_t>             params       = { (uint32_t) offset, (uint32_t) size, value };
     std::vector<wgpu::BindGroupEntry> entries      = { ggml_webgpu_make_bind_group_entry(0, buf, 0, buf.GetSize()) };
-    size_t                            bytes_per_wg = WEBGPU_MAX_WG_SIZE * ctx->capabilities.memset_bytes_per_thread;
+    size_t                            bytes_per_wg = ctx->capabilities.limits.maxComputeInvocationsPerWorkgroup * ctx->capabilities.memset_bytes_per_thread;
     uint32_t                          wg_x         = CEIL_DIV(size + 3, bytes_per_wg);
 
     ctx->queue.WriteBuffer(ctx->memset_params_buf, 0, params.data(), params.size() * sizeof(uint32_t));
@@ -1366,7 +1358,7 @@ static webgpu_encoded_op ggml_webgpu_get_rows(webgpu_context & ctx,
     shader_lib_ctx.src0                           = src;
     shader_lib_ctx.src1                           = nullptr;
     shader_lib_ctx.dst                            = dst;
-    shader_lib_ctx.max_wg_size                    = WEBGPU_MAX_WG_SIZE;
+    shader_lib_ctx.max_wg_size                    = ctx->global_ctx->capabilities.limits.maxComputeInvocationsPerWorkgroup;
 
     webgpu_pipeline pipeline  = ctx->shader_lib->get_get_rows_pipeline(shader_lib_ctx);
     auto *          decisions = static_cast<ggml_webgpu_generic_shader_decisions *>(pipeline.context.get());
@@ -3716,13 +3708,13 @@ static ggml_guid_t ggml_backend_webgpu_guid(void) {
 
 static void ggml_webgpu_init_memset_pipeline(webgpu_global_context & ctx) {
     // we use the maximum workgroup size for the memset pipeline
-    size_t max_threads = WEBGPU_MAX_WG_SIZE * ctx->capabilities.limits.maxComputeWorkgroupsPerDimension;
+    size_t max_threads = ctx->capabilities.limits.maxComputeInvocationsPerWorkgroup * ctx->capabilities.limits.maxComputeWorkgroupsPerDimension;
     // Size the bytes_per_thread so that the largest buffer size can be handled
     ctx->capabilities.memset_bytes_per_thread =
         CEIL_DIV(ctx->capabilities.limits.maxStorageBufferBindingSize, max_threads);
     std::vector<wgpu::ConstantEntry> constants(2);
     constants[0].key     = "wg_size";
-    constants[0].value   = WEBGPU_MAX_WG_SIZE;
+    constants[0].value   = ctx->capabilities.limits.maxComputeInvocationsPerWorkgroup;
     constants[1].key     = "bytes_per_thread";
     constants[1].value   = ctx->capabilities.memset_bytes_per_thread;
     ctx->memset_pipeline = ggml_webgpu_create_pipeline(ctx->device, wgsl_memset, "memset", constants);