From: Neo Zhang <redacted>
Date: Fri, 23 Jan 2026 12:54:10 +0000 (+0800)
Subject: use malloc to support both iGPU and dGPU in same time (llama/18992)
X-Git-Tag: v0.9.6~38
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=a2bd034959543b67051d4ab9d50e15f55eabe011;p=pkg%2Fggml%2Fsources%2Fggml

use malloc to support both iGPU and dGPU in same time (llama/18992)

* use malloc to support both iGPU and dGPU in same time

* support windows

---------

Co-authored-by: Neo Zhang Jianyu <redacted>
---

diff --git a/src/ggml-sycl/ggml-sycl.cpp b/src/ggml-sycl/ggml-sycl.cpp
index bb8acc92..ce2f0d41 100644
--- a/src/ggml-sycl/ggml-sycl.cpp
+++ b/src/ggml-sycl/ggml-sycl.cpp
@@ -1157,13 +1157,28 @@ static const char * ggml_backend_sycl_host_buffer_type_name(ggml_backend_buffer_
     GGML_UNUSED(buft);
 }
 
+inline void * aligned_malloc_host(size_t alignment, size_t size) {
+#ifdef _WIN32
+    return _aligned_malloc(size, alignment);
+#else
+    return aligned_alloc(alignment, size);
+#endif
+}
+
+inline void free_aligned_mem_host(void * memblock) {
+#ifdef _WIN32
+    _aligned_free(memblock);
+#else
+    free(memblock);
+#endif
+}
+
 static void ggml_backend_sycl_host_buffer_free_buffer(ggml_backend_buffer_t buffer) {
-    ggml_sycl_host_free(buffer->context);
+    free_aligned_mem_host((void *)buffer->context);
 }
 
 static ggml_backend_buffer_t ggml_backend_sycl_host_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
-    void * ptr = ggml_sycl_host_malloc(size);
-
+    void * ptr = aligned_malloc_host(TENSOR_ALIGNMENT, size);
     if (ptr == nullptr) {
         // fallback to cpu buffer
         return ggml_backend_buft_alloc_buffer(ggml_backend_cpu_buffer_type(), size);