diff --git a/ggml.c b/ggml.c
index 2effeca49..a683c2dc4 100644
--- a/ggml.c
+++ b/ggml.c
@@ -11880,8 +11880,6 @@ static void ggml_compute_forward_mul_mat(
 
     const enum ggml_type type = src0->type;
 
-    const bool src1_cont = ggml_is_contiguous(src1);
-
     enum ggml_type    const vec_dot_type          = type_traits[type].vec_dot_type;
     ggml_from_float_t const from_float_to_vec_dot = type_traits[vec_dot_type].from_float;
     int64_t           const vec_dot_num_rows      = type_traits[type].nrows;
@@ -11901,10 +11899,6 @@ static void ggml_compute_forward_mul_mat(
     GGML_ASSERT(nb1 <= nb2);
     GGML_ASSERT(nb2 <= nb3);
 
-    // broadcast factors
-    const int64_t r2 = ne12/ne02;
-    const int64_t r3 = ne13/ne03;
-
     // nb01 >= nb00 - src0 is not transposed
     //   compute by src0 rows
 
@@ -11985,6 +11979,12 @@ static void ggml_compute_forward_mul_mat(
 #endif
 
 #if GGML_USE_LLAMAFILE
+    const bool src1_cont = ggml_is_contiguous(src1);
+
+    // broadcast factors
+    const int64_t r2 = ne12 / ne02;
+    const int64_t r3 = ne13 / ne03;
+
     if (src1_cont) {
         for (int64_t i13 = 0; i13 < ne13; i13++)
             for (int64_t i12 = 0; i12 < ne12; i12++)
@@ -12036,11 +12036,11 @@ UseGgmlGemm1:;
         return;
     }
 
-    const void * wdata    = (src1->type == vec_dot_type) ? src1->data : params->wdata;
-    const size_t row_size = ggml_row_size(vec_dot_type, ne10);
-
 #if GGML_USE_LLAMAFILE
     if (src1->type != vec_dot_type) {
+        const void* wdata = (src1->type == vec_dot_type) ? src1->data : params->wdata;
+        const size_t row_size = ggml_row_size(vec_dot_type, ne10);
+
         for (int64_t i13 = 0; i13 < ne13; i13++)
             for (int64_t i12 = 0; i12 < ne12; i12++)
                 if (!llamafile_sgemm(ne01, ne11, ne00/ggml_blck_size(src0->type),