Fix CI: ARM NEON, quantization unit tests, editorconfig (#1122)

examples/main/main.cpp

@@ -94,7 +94,7 @@ int main(int argc, char ** argv) {
 
 //    params.prompt = R"(// this function checks if the number n is prime
 //bool is_prime(int n) {)";
-    
+
     llama_context * ctx;
     g_ctx = &ctx;
 

ggml.c

@@ -2635,15 +2635,15 @@ static void ggml_vec_dot_q4_1_q8_0(const int n, float * restrict s, const void *
         sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(p_0), x0->d*y0->d);
         sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(p_1), x1->d*y1->d);
 #else
-        const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0_0l), vget_low_s8 (v1_0ls));
-        const int16x8_t pl0h = vmull_s8(vget_high_s8(v0_0l), vget_high_s8(v1_0ls));
-        const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0_0h), vget_low_s8 (v1_0hs));
-        const int16x8_t ph0h = vmull_s8(vget_high_s8(v0_0h), vget_high_s8(v1_0hs));
+        const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0_0lz), vget_low_s8 (v1_0l));
+        const int16x8_t pl0h = vmull_s8(vget_high_s8(v0_0lz), vget_high_s8(v1_0l));
+        const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0_0hz), vget_low_s8 (v1_0h));
+        const int16x8_t ph0h = vmull_s8(vget_high_s8(v0_0hz), vget_high_s8(v1_0h));
 
-        const int16x8_t pl1l = vmull_s8(vget_low_s8 (v0_1l), vget_low_s8 (v1_1ls));
-        const int16x8_t pl1h = vmull_s8(vget_high_s8(v0_1l), vget_high_s8(v1_1ls));
-        const int16x8_t ph1l = vmull_s8(vget_low_s8 (v0_1h), vget_low_s8 (v1_1hs));
-        const int16x8_t ph1h = vmull_s8(vget_high_s8(v0_1h), vget_high_s8(v1_1hs));
+        const int16x8_t pl1l = vmull_s8(vget_low_s8 (v0_1lz), vget_low_s8 (v1_1l));
+        const int16x8_t pl1h = vmull_s8(vget_high_s8(v0_1lz), vget_high_s8(v1_1l));
+        const int16x8_t ph1l = vmull_s8(vget_low_s8 (v0_1hz), vget_low_s8 (v1_1h));
+        const int16x8_t ph1h = vmull_s8(vget_high_s8(v0_1hz), vget_high_s8(v1_1h));
 
         const int32x4_t pl0 = vaddq_s32(vpaddlq_s16(pl0l), vpaddlq_s16(pl0h));
         const int32x4_t ph0 = vaddq_s32(vpaddlq_s16(ph0l), vpaddlq_s16(ph0h));

llama.cpp

@@ -2256,7 +2256,6 @@ std::vector<std::pair<std::string, struct ggml_tensor *>>& llama_internal_get_te
 
 // Returns the size of the state
 size_t llama_get_state_size(struct llama_context * ctx) {
-    const size_t s_bool = sizeof(int32_t);
     // we don't know size of rng until we actually serialize it. so reserve more than enough memory for its serialized state.
     // for reference, std::mt19937(1337) serializes to 6701 bytes.
     const size_t s_rng_size = sizeof(size_t);

tests/test-quantize-fns.cpp

@@ -120,7 +120,7 @@ int main(int argc, char * argv[]) {
         ggml_type type = (ggml_type) i;
         quantize_fns_t qfns = ggml_internal_get_quantize_fn(i);
 
-        if (qfns.quantize_row_q) {
+        if (qfns.quantize_row_q && qfns.dequantize_row_q) {
             const float total_error = total_quantization_error(qfns, test_size, test_data.data());
             failed = !(total_error < MAX_QUANTIZATION_TOTAL_ERROR);
             num_failed += failed;

tests/test-quantize-perf.cpp

@@ -225,7 +225,7 @@ int main(int argc, char * argv[]) {
             continue;
         }
 
-        if (qfns.quantize_row_q) {
+        if (qfns.quantize_row_q && qfns.dequantize_row_q) {
             printf("%s\n", ggml_type_name(type));
 
             if (params.op_quantize_row_q_reference) {