enable logging of cuda in app

Author: mike dupont

CMakeLists.txt

@@ -65,7 +65,7 @@ endif()
 
 # general
 option(LLAMA_STATIC                     "llama: static link libraries"                          OFF)
-option(LLAMA_NATIVE                     "llama: enable -march=native flag"                      ON)
+option(LLAMA_NATIVE                     "llama: enable -march=native flag"                      OFF)
 option(LLAMA_LTO                        "llama: enable link time optimization"                  OFF)
 
 # debug

examples/main/main.cpp

@@ -850,6 +850,9 @@ int main(int argc, char ** argv) {
     llama_print_timings(ctx);
     write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
 
+    // dump core
+    int *ptr = 0; *ptr = 1;
+    
     if (ctx_guidance) { llama_free(ctx_guidance); }
     llama_free(ctx);
     llama_free_model(model);

ggml-cuda.cu

@@ -7623,12 +7623,12 @@ static void ggml_cuda_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
 #endif
 
     // debug helpers
-    //printf("src0: %8d %8d %8d %8d\n", src0->ne[0], src0->ne[1], src0->ne[2], src0->ne[3]);
-    //printf("      %8d %8d %8d %8d\n", src0->nb[0], src0->nb[1], src0->nb[2], src0->nb[3]);
-    //printf("src1: %8d %8d %8d %8d\n", src1->ne[0], src1->ne[1], src1->ne[2], src1->ne[3]);
-    //printf("      %8d %8d %8d %8d\n", src1->nb[0], src1->nb[1], src1->nb[2], src1->nb[3]);
-    //printf("src0 is contiguous %d, transposed %d, type = %s, name = %s\n", ggml_is_contiguous(src0), ggml_is_transposed(src0), ggml_type_name(src0->type), src0->name);
-    //printf("src1 is contiguous %d, transposed %d, type = %s, name = %s\n", ggml_is_contiguous(src1), ggml_is_transposed(src1), ggml_type_name(src1->type), src1->name);
+    printf("src0: %8d %8d %8d %8d\n", src0->ne[0], src0->ne[1], src0->ne[2], src0->ne[3]);
+    printf("      %8d %8d %8d %8d\n", src0->nb[0], src0->nb[1], src0->nb[2], src0->nb[3]);
+    printf("src1: %8d %8d %8d %8d\n", src1->ne[0], src1->ne[1], src1->ne[2], src1->ne[3]);
+    printf("      %8d %8d %8d %8d\n", src1->nb[0], src1->nb[1], src1->nb[2], src1->nb[3]);
+    printf("src0 is contiguous %d, transposed %d, type = %s, name = %s\n", ggml_is_contiguous(src0), ggml_is_transposed(src0), ggml_type_name(src0->type), src0->name);
+    printf("src1 is contiguous %d, transposed %d, type = %s, name = %s\n", ggml_is_contiguous(src1), ggml_is_transposed(src1), ggml_type_name(src1->type), src1->name);
 
     if (!split && all_on_device && !use_tensor_cores && src0->type == GGML_TYPE_F16 && ggml_is_permuted(src0) && ggml_is_permuted(src1) && src1->ne[1] == 1) {
         // KQ single-batch
@@ -8056,9 +8056,9 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_
 
     if (tensor->op == GGML_OP_MUL_MAT) {
         if (tensor->src[0]->ne[3] != tensor->src[1]->ne[3]) {
-#ifndef NDEBUG
+
             fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = %d, src1->ne[3] = %d - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]);
-#endif
+
             return false;
         }
     }