sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-05-15 13:34:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

state

Browse Source
This commit is contained in:
Ruben Ortlam
2026-04-09 11:24:21 +02:00
parent ddaafa3dc1
commit 700270239d
1 changed files with 12 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@@ -13393,9 +13393,13 @@ static void ggml_vk_graph_cleanup(ggml_backend_vk_context * ctx) {
ggml_vk_command_pool_cleanup(ctx->device, ctx->transfer_cmd_pool);
}
GGML_LOG_DEBUG("ggml_vulkan: graph_cleanup(%s) binary_semaphore_idx=%zu\n",
ctx->name.c_str(), ctx->binary_semaphore_idx);
ctx->binary_semaphore_idx = 0;
for (auto& [peer, staging] : ctx->device->peer_staging) {
GGML_LOG_DEBUG("ggml_vulkan: graph_cleanup reset buf_idx %zu->0, tl_sem_value=%lu\n",
staging.buf_idx, (unsigned long)staging.tl_sem_value);
staging.buf_idx = 0;
}
@@ -14031,6 +14035,11 @@ static bool ggml_backend_vk_cpy_tensor_async(ggml_backend_t backend_src, ggml_ba
nbytes, sync_fd, (void *)(VkSemaphore)staging.src_sem,
(void *)dst_compute_ctx.get());
if (sync_fd == -1) {
GGML_LOG_WARN("ggml_vulkan: WARNING: getSemaphoreFdKHR returned -1 "
"(signal already completed before export)\n");
}
// Per-copy dest semaphore (temporary import is consumed per wait)
vk_semaphore * dst_sem = ggml_vk_create_binary_semaphore(ctx);
@@ -14115,6 +14124,9 @@ static bool ggml_backend_vk_cpy_tensor_async(ggml_backend_t backend_src, ggml_ba
static void ggml_vk_synchronize(ggml_backend_vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_synchronize()");
GGML_LOG_DEBUG("ggml_vulkan: synchronize(%s) submit_pending=%d compute_ctx_expired=%d\n",
ctx->name.c_str(), (int)ctx->submit_pending,
(int)ctx->compute_ctx.expired());
bool do_transfer = !ctx->compute_ctx.expired();