Fix: Address various compiler warnings

- ggml/src/ggml-cpu/ggml-cpu.c:
  - Fix cast discarding 'const' qualifier.
  - Remove unused variables 'bs' and 'nbw0'.
- src/llama-quant.cpp:
  - Fix comparison of integer expressions of different signedness.
  - Remove unused variables 'thread_src', 'thread_dst_char', and 'total_size_written'.

Author: google-labs-jules[bot]

ggml/src/ggml-cpu/ggml-cpu.c

@@ -8718,7 +8718,7 @@ static void ggml_compute_forward_mul_mat_one_chunk(
                         // If src1 is already quantized, it must match vec_dot_type_for_src1
                         // This path is less common as per todo.txt and might need more robust handling
                         GGML_ASSERT(src1->type == vec_dot_type_for_src1);
-                        src1_segment_prepared_data = (void *)src1_segment_f32_ptr; // This cast is placeholder, actual pointer would be from src1 data
+                        src1_segment_prepared_data = (const void *)src1_segment_f32_ptr; // This cast is placeholder, actual pointer would be from src1 data
                     }
 
                     float segment_result = 0.0f;
@@ -8863,7 +8863,7 @@ UseGgmlGemm1:;
     } else if (src1->type != vec_dot_type) {
         char * wdata = params->wdata;
 
-        const size_t nbw0 = ggml_type_size(vec_dot_type);
+        // const size_t nbw0 = ggml_type_size(vec_dot_type); // Unused
         const size_t nbw1 = ggml_row_size(vec_dot_type, ne10);
         const size_t nbw2 = nbw1*ne11;
         const size_t nbw3 = nbw2*ne12;
@@ -8875,7 +8875,7 @@ UseGgmlGemm1:;
         for (int64_t i13 = 0; i13 < ne13; ++i13) {
             for (int64_t i12 = 0; i12 < ne12; ++i12) {
                 for (int64_t i11 = 0; i11 < ne11; ++i11) { // Changed loop to cover all of src1
-                    size_t bs = ggml_blck_size(vec_dot_type);
+                    // size_t bs = ggml_blck_size(vec_dot_type); // Unused
                     // Parallelize quantization of src1 rows if multiple threads are available for this part
                     // For simplicity in this change, assuming ith=0, nth=1 for src1 quantization here
                     // or that from_float is thread-safe and handles partitioning if params->ith/nth are used.

src/llama-quant.cpp

@@ -99,7 +99,7 @@ SmarterQuantConfig load_smarter_quant_config(const std::string & fname) {
                     tensor_info.n_cols_for_permutation = 0; // Reset before goto
                     goto next_tensor_label;
                 }
-                for (size_t i = 0; i < tensor_info.n_cols_for_permutation; ++i) {
+                for (size_t i = 0; i < (size_t)tensor_info.n_cols_for_permutation; ++i) {
                     if (!permutation_json[i].is_number_integer()) {
                         LLAMA_LOG_WARN("%s: Invalid type for 'column_permutation[%zu]' for tensor '%s'. Expected integer. Skipping tensor processing.\n", __func__, i, tensor_name.c_str());
                         delete[] tensor_info.column_permutation;
@@ -287,16 +287,16 @@ static size_t llama_tensor_quantize_impl(
         loc_workers.resize(nthread -1); // nthread-1 worker threads, 1 main thread
 
         int64_t rows_per_thread = (nrows + nthread - 1) / nthread;
-        size_t total_size_written = 0;
+        // size_t total_size_written = 0; // Unused
         std::mutex size_mutex;
 
         for (int t = 0; t < nthread; ++t) {
             const int64_t r_start = t * rows_per_thread;
             const int64_t r_end   = std::min(r_start + rows_per_thread, nrows);
             if (r_start >= r_end) continue;
 
-            const float * thread_src = src + r_start * k;
-            char * thread_dst_char = static_cast<char *>(dst);
+            // const float * thread_src = src + r_start * k; // Unused
+            // char * thread_dst_char = static_cast<char *>(dst); // Unused
             // Calculate offset into dst for this thread's rows
             // This requires knowing the size of previously quantized rows by other threads if types vary,
             // or assuming fixed output size per row if type is const for this call.