Added backend switch logic as well as sgemm naive matmul

main.cpp

@@ -16,32 +16,36 @@ int main() {
     cout << "=========================================== \n";
     cout << "=============TESTING DAG VERIFICATION============== \n";
 
-    float val_a[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
-    float val_b[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
-    float val_c[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
-    float val_d[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
+    float val_a[900]{};
+    float val_b[900]{};
+    float val_c[900]{};
+    float val_d[900]{};
 
     // TACTICAL FIX: Make 'e' the exact same size to avoid a GPU segfault 
     // before we implement a dedicated GPU broadcasting kernel!
-    float val_e[] = {7.0f, 8.0f, 7.0f, 8.0f, 7.0f, 8.0f}; 
+    float val_e[900]{};
 
-    uint32_t dims_a[] = {3, 2};
-    uint32_t dims_b[] = {3, 2};
-    uint32_t dims_c[] = {3, 2};
-    uint32_t dims_d[] = {3, 2};
-    uint32_t dims_e[] = {3, 2}; // Matched dimensions!
+    uint32_t dims_a[] = {30, 30};
+    uint32_t dims_b[] = {30, 30};
+    uint32_t dims_c[] = {30, 30};
+    uint32_t dims_d[] = {30, 30};
+    uint32_t dims_e[] = {30, 30}; // Matched dimensions!
 
     tensor_t *a = tensor_create(pool, tensor_dtype_t::FLOAT32_T, 2, dims_a, val_a);
     tensor_t *b = tensor_create(pool, tensor_dtype_t::FLOAT32_T, 2, dims_b, val_b);
     tensor_t *c = tensor_create(pool, tensor_dtype_t::FLOAT32_T, 2, dims_c, val_c);
     tensor_t *d = tensor_create(pool, tensor_dtype_t::FLOAT32_T, 2, dims_d, val_d);
     tensor_t *e = tensor_create(pool, tensor_dtype_t::FLOAT32_T, 2, dims_e, val_e);
-
-    tensor_t *f = tensor_add(pool, a, b);
-    tensor_t *g = tensor_add(pool, b, c);
-    tensor_t *h = tensor_add(pool, e, f);
-    tensor_t *i = tensor_add(pool, h, a);
-    tensor_t *j = tensor_add(pool, i, g);
+    tensor_fill_random_normal(a, 10.1, 5.7);
+    tensor_fill_random_normal(b, 10.1, 5.7);
+    tensor_fill_random_normal(c, 10.1, 5.7);
+    tensor_fill_random_normal(d, 10.1, 5.7);
+    tensor_fill_random_normal(e, 10.1, 5.7);
+    tensor_t *f = tensor_mul(pool, a, b);
+    tensor_t *g = tensor_mul(pool, b, c);
+    tensor_t *h = tensor_mul(pool, e, f);
+    tensor_t *i = tensor_mul(pool, h, a);
+    tensor_t *j = tensor_mul(pool, i, g);
 
     cout << "=========================================== \n";
     cout << "==============LAZY EVAL DONE=============== \n";

scripts/Experiments/config/data.json

@@ -0,0 +1,4 @@
+{
+  "hello":"world",
+  "Yes":"No"
+}

scripts/Experiments/config/json.cpp

@@ -0,0 +1,17 @@
+#include <iostream>
+#include "json_utils.h"
+
+int main() {
+    try {
+        auto data = readJsonToMap("data.json");
+
+        for (const auto& [key, value] : data) {
+            std::cout << key << " : " << value << std::endl;
+        }
+
+    } catch (const std::exception& e) {
+        std::cout << "Error: " << e.what() << std::endl;
+    }
+
+    return 0;
+}