Update README

Author: andrewdalpino

README.md

@@ -1,8 +1,15 @@
-# Tensor: A PHP Extension for Scientific Computing
+# Tensor: Scientific Computing for [PHP](https://php.net)
 
 [![PHP from Packagist](https://img.shields.io/packagist/php-v/rubix/tensor_ext.svg?style=flat&colorB=8892BF)](https://www.php.net/) [![Latest Stable Version](https://img.shields.io/packagist/v/rubix/tensor_ext.svg?style=flat&colorB=orange)](https://packagist.org/packages/rubix/tensor_ext) [![Code Checks](https://github.com/RubixML/Tensor-Ext/actions/workflows/ci.yml/badge.svg)](https://github.com/RubixML/Tensor-Ext/actions/workflows/ci.yml) [![Downloads from Packagist](https://img.shields.io/packagist/dt/rubix/tensor_ext.svg?style=flat&colorB=red)](https://packagist.org/packages/rubix/tensor_ext) [![GitHub](https://img.shields.io/github/license/RubixML/Tensor-Ext)](https://github.com/RubixML/Tensor-Ext/blob/master/LICENSE.md)
 
-Scientific Computing for the [PHP](https://php.net) language.
+Tensor Ext is a PHP extension that brings fast Matrix and Vector computing to the PHP language. The object-oriented API is written in [Zephir](https://zephir-lang.com) and backed by native C kernels.
+
+## Features
+
+- **Object-oriented API** — Matrix and Vector APIs are fluent and expressive.
+- **Native BLAS and LAPACK acceleration** — Highly-optimized multi-threaded linear algebra kernels.
+- **Universal element-wise functions** — Elementwise operations are broadcast and vectorized.
+- **Dynamic kernel dispatching** — Operations are dispatched at runtime based on CPU architecture.
 
 ## Installation
 
@@ -15,9 +22,9 @@ $ pie install rubix/tensor_ext
 ## Requirements
 
 - [PHP](https://php.net) 8.1 or above
+- The PHP development package (source code and tooling)
 - A C compiler such as [GCC](https://gcc.gnu.org/), [Clang](https://clang.llvm.org/), or [Visual C++](https://support.microsoft.com/en-us/help/2977003/the-latest-supported-visual-c-downloads)
 - A Fortran compiler such as [GFortran](https://gcc.gnu.org/wiki/GFortran)
-- The PHP development package (source code and tooling)
 - [OpenBLAS](https://www.openblas.net/) development package
 - [LAPACKE](https://www.netlib.org/lapack/lapacke.html) C interface to [LAPACK](http://www.netlib.org/lapack/)
 - [re2c](https://re2c.org/) 0.13.6 or later

ext/include/buffer.c

@@ -69,7 +69,7 @@ int tensor_tensorbuffer_create_uninit(zval * ret, zend_long len, zval * buffer)
  * bytes decoding to Inf would give `Inf * 0 = NaN` and contaminate the result.
  * Everywhere else, prefer tensor_tensorbuffer_create_uninit().
  */
-int tensor_tensorbuffer_zeros(zval * ret, zend_long len, zval * buffer)
+int tensor_tensorbuffer_create_zeros(zval * ret, zend_long len, zval * buffer)
 {
 	if (UNEXPECTED(zephir_buffer_create(buffer, len, ZEPHIR_BUFFER_DOUBLE) == FAILURE)) {
 		ZVAL_UNDEF(buffer);

ext/include/buffer.h

@@ -30,7 +30,7 @@ int tensor_tensorbuffer_create_uninit(zval * ret, zend_long len, zval * buffer);
  * Zero-fills because some callers hand the buffer to BLAS with `beta = 0.0`,
  * which must not read uninitialized bytes. Prefer the _uninit variant
  * everywhere the caller fully overwrites the buffer. */
-int tensor_tensorbuffer_zeros(zval * ret, zend_long len, zval * buffer);
+int tensor_tensorbuffer_create_zeros(zval * ret, zend_long len, zval * buffer);
 
 /* Unwrap the double buffer hidden inside a `Tensor\TensorBuffer` object,
  * returning a raw pointer into it. Sets `*success` to 1 on success and 0 on

ext/include/linear_algebra.c

@@ -46,7 +46,7 @@ void tensor_matmul(zval * return_value, zval * a, zval * b, zval * m, zval * p,
 
     zval c;
 
-    if (UNEXPECTED(tensor_tensorbuffer_zeros(return_value, ma * nb, &c) == FAILURE)) {
+    if (UNEXPECTED(tensor_tensorbuffer_create_zeros(return_value, ma * nb, &c) == FAILURE)) {
         return;
     }
 
@@ -88,7 +88,7 @@ void tensor_matrix_dot(zval * return_value, zval * a, zval * b, zval * m, zval *
 
     zval c;
 
-    if (UNEXPECTED(tensor_tensorbuffer_zeros(return_value, ma, &c) == FAILURE)) {
+    if (UNEXPECTED(tensor_tensorbuffer_create_zeros(return_value, ma, &c) == FAILURE)) {
         return;
     }