Added jit benchmark image, fixed some format issues

edwinsolisf · syurkevi · commit 6959b30c9e10 · 2025-09-04T18:23:12.000-04:00
diff --git a/docs/TODO b/docs/TODO
@@ -0,0 +1,7 @@
+TODO
+=====
+
+- Add docstring to function image processing
+- Change to where the Restructured Files for the undocumented functions point to (currently some point to docs in the arrayfire-binary-python-wrapper)
+- Add more detail to the constants and classes
+- Add more information in the python installation section
diff --git a/docs/afjit.py b/docs/afjit.py
@@ -6,21 +6,45 @@
 # following function.
 
 import arrayfire as af
+import time
+
+samples = int(9e8)
+x = af.randu((samples))
+y = af.randu((samples))
 
 def pi_no_jit(x, y, samples):
     temp = x * x
-    temp.eval()
+    af.eval(temp)
     temp += y * y
-    temp.eval()
+    af.eval(temp)
     temp = af.sqrt(temp)
-    temp.eval()
+    af.eval(temp)
     temp = temp < 1
-    temp.eval()
+    af.eval(temp)
     return 4.0 * af.sum(temp) / samples
 
 def pi_jit(x, y, samples):
     temp = af.sqrt(x * x + y * y) < 1
-    temp.eval()
+    af.eval(temp)
     return 4.0 * af.sum(temp) / samples
 
+# Print device info
+af.info()
+
+# Time JIT code
+start = time.perf_counter()
+res = pi_jit(x, y, samples)
+af.sync()
+end = time.perf_counter()
+
+print("jit:", end - start, res)
+af.device_gc()
+
+# Time no JIT code
+start = time.perf_counter()
+res = pi_no_jit(x, y, samples)
+af.sync()
+end = time.perf_counter()
+print("no jit:", end - start, res)
+
 # [jit-endsnippet]
diff --git a/docs/arrayfirejitcodegeneration.rst b/docs/arrayfirejitcodegeneration.rst
@@ -34,12 +34,15 @@ Specifically, as JIT is an integral feature of the ArrayFire library, it cannot
 
 Timing these two implementations results in the following performance benchmark:
 
-**Add Picture**
+.. image:: images/afjit_benchmark.png
+   :alt: Benchmark Results
+   :align: center
+   :class: responsive-img
 
-The above figure depicts the execution time (abscissa) as a function of the number of samples (ordinate) for the two implementations discussed above.
+The above figure depicts the execution time as a function of the number of samples for the two implementations discussed above.
 
-When the number of samples is small, the execution time of pi_no_jit is dominated by the launch of multiple kernels and the execution time pi_jit is dominated by on-the-fly compilation of the JIT code required to launch a single kernel. Even with this JIT compilation time, pi_jit outperforms pi_no_jit by 1.4-2.0X for smaller sample sizes.
+When the number of samples is small (less than 100k samples), the execution time of pi_no_jit is dominated by the launch of multiple kernels and the execution time pi_jit is dominated by on-the-fly compilation of the JIT code required to launch a single kernel. Even with this JIT compilation time, pi_jit outperforms pi_no_jit by 1.4-2.0X for smaller sample sizes.
 
-When the number of samples is large, both the kernel launch overhead and the JIT code creation are no longer the limiting factors – the kernel’s computational load dominates the execution time. Here, the pi_jit outperforms pi_no_jit by 2.0-2.7X.
+When the number of samples is large (more than 1 million samples), both the kernel launch overhead and the JIT code creation are no longer the limiting factors – the kernel’s computational load dominates the execution time. Here, the pi_jit outperforms pi_no_jit by 3-25X.
 
 The number of applications that benefit from the JIT code generation is significant. The actual performance benefits are also application-dependent.
diff --git a/docs/configuringarrayfireenvironment.rst b/docs/configuringarrayfireenvironment.rst
@@ -150,7 +150,7 @@ deallocations, and garbage collection.
 All trace statements printed to the console have a suffix with the following
 pattern.
 
-**[category][Seconds since Epoch][Thread Id][source file relative path] \<Message\>**
+:literal:`[category][Seconds since Epoch][Thread Id][source file relative path] \<Message\>`
 
 `AF_OPENCL_SHOW_BUILD_INFO`
 ***************************
diff --git a/docs/images/afjit_benchmark.png b/docs/images/afjit_benchmark.png
diff --git a/docs/index.rst b/docs/index.rst
@@ -11,9 +11,9 @@ Welcome to ArrayFire's documentation!
    :caption: The Basics
 
    overview
+   tutorial
    installation
    gettingstarted
-   tutorial
 
 .. toctree::
    :maxdepth: 1
diff --git a/docs/installation.rst b/docs/installation.rst
@@ -20,8 +20,6 @@ Once the ArrayFire has been downloaded, run the installer.
 
 The installer offers the option to automatically add ArrayFire to the path for all users. If the installer did not do this, simply append :literal:`%AF_PATH%/lib` to the PATH variable so that the loader can find ArrayFire DLLs.
 
-For more information on using ArrayFire on Windows, visit the following page.
-
 
 .. _Linux:
 
@@ -131,5 +129,6 @@ Getting help
 * Google Groups: https://groups.google.com/forum/#!forum/arrayfire-users
 * ArrayFire Services: `Consulting <https://arrayfire.com/consulting/>`_ | `Training <https://arrayfire.com/training/>`_
 * ArrayFire Blogs: http://arrayfire.com/blog/
+* ArrayFire `Contact Us <https://arrayfire.com/company/#contact-us>`_
 * Email: support@arrayfire.com
 
diff --git a/docs/mathematical_operations_functions.rst b/docs/mathematical_operations_functions.rst
@@ -109,8 +109,6 @@ Functions for performing fundamental arithmetic and mathematical operations on a
         - Elementwise minimum between two arrays
     * - :doc:`af.mod() <functions/mod>`
         - Calculate the modulus.
-    * - :doc:`af.moddims() <functions/moddims>`
-        - Modify the dimensions of an array without changing the order of its elements.
     * - :doc:`af.mul() <functions/mul>`
         - Elementwise multiply.
     * - :doc:`af.neg() <functions/neg>`
@@ -141,6 +139,7 @@ Functions for performing fundamental arithmetic and mathematical operations on a
         - Evaluate the square root.
     * - :doc:`af.sub() <functions/sub>`
         - Elementwise subtraction.
+    * - :doc:`af.tan() <functions/tan>`
         - Evaluate the tangent function.
     * - :doc:`af.tanh() <functions/tanh>`
         - Evaluate the hyperbolic tangent function.
diff --git a/docs/special_functions.rst b/docs/special_functions.rst
@@ -18,7 +18,7 @@ Functions for convolutions, creating and applying specific types of filters, com
     * - :doc:`af.canny() <functions/canny>`
         - Canny Edge Detector.
     * - :doc:`af.convolve1() <functions/convolve1>`
-            - Convolution Integral for one dimensional data.
+        - Convolution Integral for one dimensional data.
     * - :doc:`af.convolve2() <functions/convolve2>`
         - Convolution Integral for two dimensional data.
     * - :doc:`af.convolve2_gradient_nn() <functions/convolve2_gradient_nn>`
