update bench

bin/benchmark/main.c

@@ -97,7 +97,7 @@ int main(void) {
 
             // get buffer sizes large enough that all the simd code paths get hit hard, but
             // also measure the smaller buffer paths since they often can't be optimized as thoroughly.
-            size_t buffer_sizes[] = {8, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536};
+            size_t buffer_sizes[] = {8, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 524288};
             size_t buffer_sizes_len = AWS_ARRAY_SIZE(buffer_sizes);
 
             // warm it up to factor out the cpuid checks:
@@ -115,9 +115,10 @@ int main(void) {
                 aws_high_res_clock_get_ticks(&end_time);
                 fprintf(
                     stdout,
-                    "buffer size %zu (bytes), latency: %" PRIu64 " ns\n",
+                    "buffer size %zu (bytes), latency: %" PRIu64 " ns throughput: %d  GB/s\n",
                     buffer_sizes[k],
-                    end_time - start_time);
+                    end_time - start_time,
+                    (buffer_sizes[k] * 1000000000.0 / 1024.0 / 1024.0 / 1024.0) / (end_time - start_time));
                 aws_byte_buf_clean_up(&x_bytes);
             }
             fprintf(stdout, "\n");