examples : add idle tool for investigating GPU idle overhead

Makefile

@@ -18,6 +18,7 @@ BUILD_TARGETS = \
 	llama-gguf-hash \
 	llama-gguf-split \
 	llama-gritlm \
+	llama-idle \
 	llama-imatrix \
 	llama-infill \
 	llama-llava-cli \

examples/CMakeLists.txt

@@ -30,6 +30,7 @@ else()
     add_subdirectory(gguf-split)
     add_subdirectory(gguf)
     add_subdirectory(gritlm)
+    add_subdirectory(idle)
     add_subdirectory(imatrix)
     add_subdirectory(infill)
     add_subdirectory(llama-bench)

examples/idle/CMakeLists.txt

@@ -0,0 +1,5 @@
+set(TARGET llama-idle)
+add_executable(${TARGET} idle.cpp)
+install(TARGETS ${TARGET} RUNTIME)
+target_link_libraries(${TARGET} PRIVATE llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)

examples/idle/README.md

@@ -0,0 +1,3 @@
+# llama.cpp/example/idle
+
+

examples/idle/idle.cpp

@@ -0,0 +1,133 @@
+#include "llama.h"
+
+#include <cmath>
+#include <cstdio>
+#include <cstring>
+#include <string>
+#include <thread>
+#include <vector>
+
+static void print_usage(int /*argc*/, char ** argv) {
+    printf("\nexample usage:\n");
+    printf("\n    %s -m model.gguf [-ngl n_gpu_layers]\n", argv[0]);
+    printf("\n");
+}
+
+int main(int argc, char ** argv) {
+    // path to the model gguf file
+    std::string model_path;
+
+    // number of layers to offload to the GPU
+    int ngl = 99;
+
+    // parse command line arguments
+
+    {
+        int i = 1;
+        for (; i < argc; i++) {
+            if (strcmp(argv[i], "-m") == 0) {
+                if (i + 1 < argc) {
+                    model_path = argv[++i];
+                } else {
+                    print_usage(argc, argv);
+                    return 1;
+                }
+            } else if (strcmp(argv[i], "-ngl") == 0) {
+                if (i + 1 < argc) {
+                    try {
+                        ngl = std::stoi(argv[++i]);
+                    } catch (...) {
+                        print_usage(argc, argv);
+                        return 1;
+                    }
+                } else {
+                    print_usage(argc, argv);
+                    return 1;
+                }
+            } else {
+                // prompt starts here
+                break;
+            }
+        }
+        if (model_path.empty()) {
+            print_usage(argc, argv);
+            return 1;
+        }
+    }
+
+    llama_model_params model_params = llama_model_default_params();
+    model_params.n_gpu_layers = ngl;
+
+    llama_model * model = llama_load_model_from_file(model_path.c_str(), model_params);
+    if (model == NULL) {
+        fprintf(stderr , "%s: error: unable to load model\n" , __func__);
+        return 1;
+    }
+
+    // we need just a dummy token to evaluate
+    std::vector<llama_token> prompt_tokens(1, llama_token_bos(model));
+
+    llama_context_params ctx_params = llama_context_default_params();
+    ctx_params.n_ctx   = 512;
+    ctx_params.n_batch = 512;
+    ctx_params.no_perf = false;
+
+    llama_context * ctx = llama_new_context_with_model(model, ctx_params);
+    if (ctx == NULL) {
+        fprintf(stderr , "%s: error: failed to create the llama_context\n" , __func__);
+        return 1;
+    }
+
+    llama_batch batch = llama_batch_get_one(prompt_tokens.data(), prompt_tokens.size());
+
+    const int n_iters = 10;
+
+    // warm-up
+    llama_decode(ctx, batch);
+    llama_kv_cache_clear (ctx);
+    llama_kv_cache_update(ctx);
+    llama_synchronize    (ctx);
+
+    for (int64_t t_pause_ms = 0; t_pause_ms <= 2200; t_pause_ms += 200) {
+        double t_sum_us  = 0.0;
+        double t_sum2_us = 0.0;
+
+        for (int i = 0; i < n_iters; i++) {
+            // this pause is important - it simulates "idle GPU"
+            std::this_thread::sleep_for(std::chrono::milliseconds(t_pause_ms));
+
+            const int64_t t_start_us = llama_time_us();
+
+            // this should take constant time
+            llama_decode(ctx, batch);
+            llama_synchronize(ctx);
+
+            const int64_t t_end_us = llama_time_us();
+
+            const double t_cur_us = t_end_us - t_start_us;
+
+#if 0
+            // print individual decode times
+            printf("  - decode time: %8.2f ms\n", t_cur_us / 1000);
+#endif
+
+            t_sum_us  += t_cur_us;
+            t_sum2_us += t_cur_us * t_cur_us;
+
+            llama_kv_cache_clear (ctx);
+            llama_kv_cache_update(ctx);
+            llama_synchronize    (ctx); // just in case
+        }
+
+        const double t_avg_us = t_sum_us / n_iters;
+        const double t_dev_us = sqrt((t_sum2_us / (n_iters - 1)) - (t_avg_us * t_avg_us * n_iters) / (n_iters - 1));
+
+        printf("iters: %4d, pause: %5d ms, avg decode time: %8.2f +/- %4.2f ms\n", n_iters, (int) t_pause_ms, t_avg_us / 1000, t_dev_us / 1000);
+        fflush(stdout);
+    }
+
+    llama_free(ctx);
+    llama_free_model(model);
+
+    return 0;
+}