From 913905028badc32a8283750c5342340113269865 Mon Sep 17 00:00:00 2001
From: Jeffrey Morgan <jmorganca@gmail.com>
Date: Mon, 5 May 2025 08:02:39 -0700
Subject: [PATCH] all: fix cgo compiler warnings on windows (#10563)

---
 discover/gpu_info_cudart.c       | 8 ++++----
 discover/gpu_info_nvcuda.c       | 4 ++--
 llama/llama.go                   | 1 +
 ml/backend/ggml/ggml/src/ggml.go | 1 +
 4 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/discover/gpu_info_cudart.c b/discover/gpu_info_cudart.c
index 03f15a2c3..79f19f27f 100644
--- a/discover/gpu_info_cudart.c
+++ b/discover/gpu_info_cudart.c
@@ -58,7 +58,7 @@ void cudart_init(char *cudart_lib_path, cudart_init_resp_t *resp) {
     LOG(resp->ch.verbose, "cudaSetDevice err: %d\n", ret);
     UNLOAD_LIBRARY(resp->ch.handle);
     resp->ch.handle = NULL;
-    if (ret == CUDA_ERROR_INSUFFICIENT_DRIVER) {
+    if (ret == CUDART_ERROR_INSUFFICIENT_DRIVER) {
       resp->err = strdup("your nvidia driver is too old or missing.  If you have a CUDA GPU please upgrade to run ollama");
       return;
     }
@@ -168,9 +168,9 @@ void cudart_bootstrap(cudart_handle_t h, int i, mem_info_t *resp) {
   resp->free = memInfo.free;
   resp->used = memInfo.used;
 
-  LOG(h.verbose, "[%s] CUDA totalMem %lu\n", resp->gpu_id, resp->total);
-  LOG(h.verbose, "[%s] CUDA freeMem %lu\n", resp->gpu_id, resp->free);
-  LOG(h.verbose, "[%s] CUDA usedMem %lu\n", resp->gpu_id, resp->used);
+  LOG(h.verbose, "[%s] CUDA totalMem %llu\n", resp->gpu_id, resp->total);
+  LOG(h.verbose, "[%s] CUDA freeMem %llu\n", resp->gpu_id, resp->free);
+  LOG(h.verbose, "[%s] CUDA usedMem %llu\n", resp->gpu_id, resp->used);
   LOG(h.verbose, "[%s] Compute Capability %d.%d\n", resp->gpu_id, resp->major, resp->minor);
 }
 
diff --git a/discover/gpu_info_nvcuda.c b/discover/gpu_info_nvcuda.c
index 466e1ac24..636cba91c 100644
--- a/discover/gpu_info_nvcuda.c
+++ b/discover/gpu_info_nvcuda.c
@@ -193,8 +193,8 @@ void nvcuda_bootstrap(nvcuda_handle_t h, int i, mem_info_t *resp) {
   resp->total = memInfo.total;
   resp->free = memInfo.free;
 
-  LOG(h.verbose, "[%s] CUDA totalMem %lu mb\n", resp->gpu_id, resp->total / 1024 / 1024);
-  LOG(h.verbose, "[%s] CUDA freeMem %lu mb\n", resp->gpu_id, resp->free / 1024 / 1024);
+  LOG(h.verbose, "[%s] CUDA totalMem %llu mb\n", resp->gpu_id, resp->total / 1024 / 1024);
+  LOG(h.verbose, "[%s] CUDA freeMem %llu mb\n", resp->gpu_id, resp->free / 1024 / 1024);
   LOG(h.verbose, "[%s] Compute Capability %d.%d\n", resp->gpu_id, resp->major, resp->minor);
 
   
diff --git a/llama/llama.go b/llama/llama.go
index 5fce0a622..7f9f9549a 100644
--- a/llama/llama.go
+++ b/llama/llama.go
@@ -2,6 +2,7 @@ package llama
 
 /*
 #cgo CFLAGS: -std=c11
+#cgo windows CFLAGS: -Wno-dll-attribute-on-redeclaration
 #cgo CXXFLAGS: -std=c++17
 #cgo CPPFLAGS: -I${SRCDIR}/llama.cpp/include
 #cgo CPPFLAGS: -I${SRCDIR}/llama.cpp/common
diff --git a/ml/backend/ggml/ggml/src/ggml.go b/ml/backend/ggml/ggml/src/ggml.go
index 678f55ca6..91f1f1ada 100644
--- a/ml/backend/ggml/ggml/src/ggml.go
+++ b/ml/backend/ggml/ggml/src/ggml.go
@@ -3,6 +3,7 @@ package ggml
 // #cgo CXXFLAGS: -std=c++17
 // #cgo CPPFLAGS: -DNDEBUG -DGGML_USE_CPU
 // #cgo CPPFLAGS: -I${SRCDIR}/../include -I${SRCDIR}/ggml-cpu
+// #cgo windows CFLAGS: -Wno-dll-attribute-on-redeclaration
 // #cgo windows LDFLAGS: -lmsvcrt -static -static-libgcc -static-libstdc++
 // #include <stdlib.h>
 // #include "ggml-backend.h"