finished timing

Part1-Timing/main.c

@@ -8,10 +8,10 @@
 
 #define LINE_SIZE 64
 // [1.2] TODO: Uncomment the following lines and fill in the correct size
-//#define L1_SIZE TODO
+#define L1_SIZE 768*1024
-//#define L2_SIZE TODO
+#define L2_SIZE 16*1024*1024
-//#define L3_SIZE TODO
+#define L3_SIZE 64*1024*1024
-//#define BUFF_SIZE TODO
+#define BUFF_SIZE 16*1024*1024
  
 int main (int ac, char **av) {
 
@@ -37,7 +37,7 @@ int main (int ac, char **av) {
 
     // [1.2] TODO: Uncomment the following line to allocate a buffer of a size
     // of your chosing. This will help you measure the latencies at L2 and L3.
-    //volatile uint8_t *eviction_buffer = (uint8_t *)malloc(BUFF_SIZE);
+    volatile uint8_t *eviction_buffer = (uint8_t *)malloc(BUFF_SIZE);
 
     // Example: Measure L1 access latency, store results in l1_latency array
     for (int i=0; i<SAMPLES; i++){
@@ -53,17 +53,45 @@ int main (int ac, char **av) {
     // [1.2] TODO: Measure DRAM Latency, store results in dram_latency array
     // ======
     //
+    for (int i=0; i<SAMPLES; i++){
+        // Step 1: bring the target cache line into DRAM
+        clflush((void *)target_buffer);
+
+        // Step 2: measure the access latency
+        dram_latency[i] = measure_one_block_access_time((uint64_t)target_buffer);
+    }
 
     // ======
     // [1.2] TODO: Measure L2 Latency, store results in l2_latency array
     // ======
     //
+    for (int i=0; i<SAMPLES; i++){
+        // Step 1: bring the target cache line into L1 by simply accessing
+        //         the line
+        tmp = target_buffer[0];
+        for(int j=0; j<L1_SIZE/LINE_SIZE; j++){
+            tmp = eviction_buffer[j*LINE_SIZE];
+        }
+
+        // Step 2: measure the access latency
+        l2_latency[i] = measure_one_block_access_time((uint64_t)target_buffer);
+    }
 
     // ======
     // [1.2] TODO: Measure L3 Latency, store results in l3_latency array
     // ======
     //
+    for (int i=0; i<SAMPLES; i++){
+        // Step 1: bring the target cache line into L1 by simply accessing
+        //         the line
+        tmp = target_buffer[0];
+        for(int j=0; j<L2_SIZE/LINE_SIZE; j++){
+            tmp = eviction_buffer[j*LINE_SIZE];
+        }
 
+        // Step 2: measure the access latency
+        l3_latency[i] = measure_one_block_access_time((uint64_t)target_buffer);
+    }
 
     // Print the results to the screen
     // When compile to main and used by `make run`,

report.md

@@ -4,9 +4,9 @@
 
 | Cache | Cache Line Size | Total Size | Associativity | Number of Sets | Raw Latency |
 | ----- | --------------- | ---------- | ------------- | -------------- | ----------- |
-| L1-D  | 64              |            |               |                |             |
+| L1-D  | 64              | 768K       | 12            | 1K             |             |
-| L2    |                 |            |               |                |             |
+| L2    | 64              | 16M        | 16            | 16K            |             |
-| L3    |                 |            |               |                |             |
+| L3    | 64              | 64M        | 16            | 64K            |             |
 
 ## 1-3