From cfdc33ec83d1cab0ec5fae85feed4924ef370e16 Mon Sep 17 00:00:00 2001
From: Lumpiasty <arek.dzski@gmail.com>
Date: Fri, 17 Jul 2026 03:04:16 +0200
Subject: [PATCH] ggml-vulkan: pre-stage host weights when import is
 unavailable

VK_EXT_external_memory_host import fails on RADV for file-backed mmap pages,
so register_host_buffer no-op'd and every MoE-expert upload paid a slow
single-threaded pageable memcpy into the staging buffer each eval (~3 GB/s
effective on the 35B, ~25% of PCIe bandwidth).

When the import fails, fall back to a one-time copy of the region into a
host-visible Vulkan buffer registered in device->pinned_memory under the
mmap's address range. The existing pinned fast path then DMAs straight from
that buffer each eval at full PCIe bandwidth - no per-eval CPU memcpy and no
blocking sync. Costs one-time host-visible memory equal to the CPU-resident
weight region.

Assisted-by: opencode
---
 ggml/src/ggml-vulkan/ggml-vulkan.cpp | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/ggml/src/ggml-vulkan/ggml-vulkan.cpp b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
index 766158b04..f7bcb5afc 100644
--- a/ggml/src/ggml-vulkan/ggml-vulkan.cpp
+++ b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
@@ -18385,11 +18385,27 @@ static bool ggml_backend_vk_register_host_buffer(void * buffer, size_t size) {
         uint8_t * p = static_cast<uint8_t *>(buffer);
         size_t remaining = size;
         bool dev_success = false;
+        bool import_ok = true;   // flips to false once VK_EXT_external_memory_host import fails
         while (remaining > 0) {
             size_t cur = std::min(remaining, chunk);
-            vk_buffer buf = ggml_vk_buffer_from_host_ptr(device, p, cur);
+            vk_buffer buf;
+            if (import_ok) {
+                buf = ggml_vk_buffer_from_host_ptr(device, p, cur);
+            }
             if (!buf || !buf->buffer) {
-                break;
+                // Fallback for drivers that can't import file-backed mmap pages (e.g. RADV):
+                // make a one-time copy of the region into a host-visible Vulkan buffer. The GPU
+                // then DMAs straight from it every eval at full PCIe bandwidth, instead of paying
+                // a slow single-threaded pageable memcpy into the staging buffer each time.
+                import_ok = false;
+                buf = ggml_vk_create_buffer_check(device, cur,
+                    vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
+                    vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
+                if (buf && buf->buffer && buf->ptr) {
+                    memcpy(buf->ptr, p, cur);
+                } else {
+                    break;
+                }
             }
             {
                 std::lock_guard<std::shared_mutex> guard(device->pinned_memory_mutex);