Stdlib: sort() for packed string and integer arrays (#209)

docs/capabilities.md

@@ -44,7 +44,7 @@ Auto-generated by `script/capability-matrix.php`. Do not edit by hand.
 | `header` | yes | yes | yes | standard | AOT PHPT |
 | `hexdec` | yes | yes | yes | standard |  |
 | `htmlspecialchars` | yes | yes | yes | standard | AOT PHPT |
-| `implode` | yes | yes | yes | standard | doc: VM only; AOT PHPT |
+| `implode` | yes | yes | yes | standard | doc: VM only; JIT PHPT; AOT PHPT |
 | `in_array` | yes | yes | yes | standard | doc: VM only; JIT PHPT |
 | `intdiv` | yes | yes | yes | standard |  |
 | `intval` | yes | yes | yes | standard | JIT PHPT; AOT PHPT |
@@ -86,6 +86,7 @@ Auto-generated by `script/capability-matrix.php`. Do not edit by hand.
 | `scandir` | yes | yes | yes | standard |  |
 | `sin` | yes | yes | yes | standard |  |
 | `sizeof` | yes | yes | yes | standard |  |
+| `sort` | yes | yes | yes | standard | JIT PHPT; AOT PHPT |
 | `sqrt` | yes | yes | yes | standard |  |
 | `str_contains` | yes | yes | yes | standard |  |
 | `str_ends_with` | yes | yes | yes | standard | AOT PHPT |

ext/standard/Module.php

@@ -79,6 +79,7 @@ public function getFunctions(): array
             new array_push(),
             new array_pop(),
             new array_shift(),
+            new sort_(),
             new array_values(),
             new array_keys(),
             new array_merge(),

ext/standard/sort_.php

@@ -0,0 +1,88 @@
+<?php
+
+declare(strict_types=1);
+
+/**
+ * This file is part of PHP-Compiler, a PHP CFG Compiler for PHP code
+ *
+ * @copyright 2015 Anthony Ferrara. All rights reserved
+ * @license MIT See LICENSE at the root of the project for more info
+ */
+
+namespace PHPCompiler\ext\standard;
+
+use PHPCompiler\Frame;
+use PHPCompiler\Func\Internal;
+use PHPCompiler\JIT\ArrayBuiltinHelper;
+use PHPCompiler\JIT\Context;
+use PHPCompiler\JIT\Variable as JITVariable;
+use PHPCompiler\VM\Variable;
+use PHPLLVM\Value;
+
+/**
+ * sort() for homogeneous packed string or integer arrays (subset of PHP).
+ *
+ * VM: full support. JIT/AOT: dynamic hashtable arrays only (not fixed native literals).
+ */
+final class sort_ extends Internal
+{
+    public function __construct()
+    {
+        parent::__construct('sort');
+    }
+
+    public function execute(Frame $frame): void
+    {
+        if (1 !== \count($frame->calledArgs)) {
+            throw new \LogicException('sort() requires exactly one argument');
+        }
+        $array = $frame->calledArgs[0]->resolveIndirect();
+        if (Variable::TYPE_ARRAY !== $array->type) {
+            throw new \LogicException('sort() argument must be an array in this compiler build');
+        }
+        $ht = $array->toArray();
+        if ($ht->getNumElements() < 2) {
+            if (null !== $frame->returnVar) {
+                $frame->returnVar->bool(true);
+            }
+
+            return;
+        }
+        $values = [];
+        foreach ($ht->iterate(true) as $value) {
+            $copy = new Variable();
+            $copy->copyFrom($value);
+            $values[] = $copy;
+        }
+        \usort($values, static function (Variable $a, Variable $b): int {
+            $a = $a->resolveIndirect();
+            $b = $b->resolveIndirect();
+            if (Variable::TYPE_STRING === $a->type && Variable::TYPE_STRING === $b->type) {
+                return VmString::strcmp($a->toString(), $b->toString());
+            }
+            if (Variable::TYPE_INTEGER === $a->type && Variable::TYPE_INTEGER === $b->type) {
+                return $a->toInt() <=> $b->toInt();
+            }
+
+            throw new \LogicException(
+                'sort() only supports homogeneous string or integer arrays in this compiler build'
+            );
+        });
+        $ht->replacePackedValues($values);
+        if (null !== $frame->returnVar) {
+            $frame->returnVar->bool(true);
+        }
+    }
+
+    public Context $context;
+
+    public function call(Context $context, JITVariable ...$args): Value
+    {
+        if (1 !== \count($args)) {
+            throw new \LogicException('sort() requires exactly one argument');
+        }
+        ArrayBuiltinHelper::sortPacked($context, $args[0]);
+
+        return $context->getTypeFromString('int1')->constInt(1, false);
+    }
+}

lib/JIT/ArrayBuiltinHelper.php

@@ -431,6 +431,17 @@ private static function looseEqual(Context $context, Variable $left, Variable $r
         return $context->constantFromBool(false);
     }
 
+    public static function sortPacked(Context $context, Variable $array): void
+    {
+        if (self::isNativeArray($array->type)) {
+            throw new \LogicException(
+                'sort() cannot compile fixed-size literal arrays in JIT/AOT yet; use bin/vm.php or bin/serve.php, or build the list with [] append'
+            );
+        }
+        $ht = self::loadHashTable($context, $array);
+        $context->builder->call($context->lookupFunction('__hashtable__sortPacked'), $ht);
+    }
+
     private static function sameTypeEqual(Context $context, Variable $left, Variable $right): Value
     {
         switch ($left->type) {

lib/JIT/Builtin/Type/HashTable.php

@@ -72,6 +72,7 @@ public function register(): void
         $this->registerFn('__hashtable__setStringKeyLong', 'void', ['__hashtable__*', '__string__*', 'int64']);
         $this->registerFn('__hashtable__offsetIsSetStringKey', 'int1', ['__hashtable__*', '__string__*']);
         $this->registerFn('__hashtable__readStringKeyValue', '__value__*', ['__hashtable__*', '__string__*']);
+        $this->registerFn('__hashtable__sortPacked', 'void', ['__hashtable__*']);
 
         $this->pointer = $this->context->getTypeFromString('__hashtable__*');
     }
@@ -106,6 +107,7 @@ public function implement(): void
         $this->implementSetStringKeyString();
         $this->implementOffsetIsSetStringKey();
         $this->implementReadStringKeyValue();
+        $this->implementSortPacked();
     }
 
     private function ensureLibcStringCompare(): void
@@ -726,6 +728,222 @@ private function lookupStringKeyValue(
         return $this->context->builder->load($resultSlot);
     }
 
+    private function implementSortPacked(): void
+    {
+        $fn = $this->context->lookupFunction('__hashtable__sortPacked');
+        $main = $fn->appendBasicBlock('main');
+        $this->context->builder->positionAtEnd($main);
+        $ht = $fn->getParam(0);
+        $map = $this->context->structFieldMap['__hashtable__'];
+        $sizeT = $this->context->getTypeFromString('size_t');
+        $two = $sizeT->constInt(2, false);
+        $num = $this->context->builder->load($this->context->builder->structGep($ht, $map['nextFreeElement']));
+        $tooSmall = $this->context->builder->icmp(Builder::INT_ULT, $num, $two);
+        $done = $fn->appendBasicBlock('sort_done');
+        $work = $fn->appendBasicBlock('sort_work');
+        $this->context->builder->branchIf($tooSmall, $done, $work);
+
+        $this->context->builder->positionAtEnd($work);
+        $zero = $sizeT->constInt(0, false);
+        $firstEntry = $this->listEntryAt($ht, $map, $zero);
+        $valueMap = $this->context->structFieldMap['__value__'];
+        $firstType = $this->context->builder->load(
+            $this->context->builder->structGep($firstEntry, $valueMap['type'])
+        );
+        $i8 = $this->context->getTypeFromString('int8');
+        $stringTag = $i8->constInt(Variable::TYPE_STRING, false);
+        $longTag = $i8->constInt(Variable::TYPE_NATIVE_LONG, false);
+        $isString = $this->context->builder->icmp(Builder::INT_EQ, $firstType, $stringTag);
+        $sortStrings = $fn->appendBasicBlock('sort_strings');
+        $sortLongs = $fn->appendBasicBlock('sort_longs');
+        $this->context->builder->branchIf($isString, $sortStrings, $sortLongs);
+
+        $this->context->builder->positionAtEnd($sortStrings);
+        $this->emitBubbleSortStrings($fn, $ht, $map, $num);
+        $this->context->builder->branch($done);
+
+        $this->context->builder->positionAtEnd($sortLongs);
+        $this->emitBubbleSortLongs($fn, $ht, $map, $num);
+        $this->context->builder->branch($done);
+
+        $this->context->builder->positionAtEnd($done);
+        $this->context->builder->returnVoid();
+    }
+
+    private function emitBubbleSortStrings(
+        PHPLLVM\LLVMAbstract\Value\Function_ $fn,
+        PHPLLVM\Value $ht,
+        array $map,
+        PHPLLVM\Value $num
+    ): void {
+        $sizeT = $this->context->getTypeFromString('size_t');
+        $one = $sizeT->constInt(1, false);
+        $zero = $sizeT->constInt(0, false);
+        $outerSlot = $this->context->builder->alloca($sizeT, 1, 'sort_outer');
+        $this->context->builder->store($zero, $outerSlot);
+        $outerHead = $fn->appendBasicBlock('sort_str_outer_head');
+        $outerBody = $fn->appendBasicBlock('sort_str_outer_body');
+        $outerDone = $fn->appendBasicBlock('sort_str_outer_done');
+        $this->context->builder->branch($outerHead);
+
+        $this->context->builder->positionAtEnd($outerHead);
+        $outer = $this->context->builder->load($outerSlot);
+        $outerEnd = $this->context->builder->sub($num, $one);
+        $outerAtEnd = $this->context->builder->icmp(Builder::INT_SGE, $outer, $outerEnd);
+        $this->context->builder->branchIf($outerAtEnd, $outerDone, $outerBody);
+
+        $this->context->builder->positionAtEnd($outerBody);
+        $innerSlot = $this->context->builder->alloca($sizeT, 1, 'sort_inner');
+        $this->context->builder->store($zero, $innerSlot);
+        $limit = $this->context->builder->sub($num, $outer);
+        $limit = $this->context->builder->sub($limit, $one);
+        $innerHead = $fn->appendBasicBlock('sort_str_inner_head');
+        $innerBody = $fn->appendBasicBlock('sort_str_inner_body');
+        $innerDone = $fn->appendBasicBlock('sort_str_inner_done');
+        $this->context->builder->branch($innerHead);
+
+        $this->context->builder->positionAtEnd($innerHead);
+        $inner = $this->context->builder->load($innerSlot);
+        $innerAtEnd = $this->context->builder->icmp(Builder::INT_SGE, $inner, $limit);
+        $this->context->builder->branchIf($innerAtEnd, $innerDone, $innerBody);
+
+        $this->context->builder->positionAtEnd($innerBody);
+        $nextInner = $this->context->builder->addNoSignedWrap($inner, $one);
+        $strA = $this->context->builder->call(
+            $this->context->lookupFunction('__hashtable__readStringAt'),
+            $ht,
+            $inner
+        );
+        $strB = $this->context->builder->call(
+            $this->context->lookupFunction('__hashtable__readStringAt'),
+            $ht,
+            $nextInner
+        );
+        $cmp = $this->context->builder->call(
+            $this->context->lookupFunction('strcmp'),
+            $this->stringDataPtr($strA),
+            $this->stringDataPtr($strB)
+        );
+        $i32 = $this->context->getTypeFromString('int32');
+        $needsSwap = $this->context->builder->icmp(Builder::INT_SGT, $cmp, $i32->constInt(0, false));
+        $swapBlock = $fn->appendBasicBlock('sort_str_swap');
+        $noSwap = $fn->appendBasicBlock('sort_str_no_swap');
+        $afterSwap = $fn->appendBasicBlock('sort_str_after_swap');
+        $this->context->builder->branchIf($needsSwap, $swapBlock, $noSwap);
+
+        $this->context->builder->positionAtEnd($swapBlock);
+        $entryA = $this->listEntryAt($ht, $map, $inner);
+        $entryB = $this->listEntryAt($ht, $map, $nextInner);
+        $this->context->builder->call($this->context->lookupFunction('__value__writeString'), $entryA, $strB);
+        $this->context->builder->call($this->context->lookupFunction('__value__writeString'), $entryB, $strA);
+        $this->context->builder->branch($afterSwap);
+
+        $this->context->builder->positionAtEnd($noSwap);
+        $this->context->builder->branch($afterSwap);
+
+        $this->context->builder->positionAtEnd($afterSwap);
+        $this->context->builder->store($nextInner, $innerSlot);
+        $this->context->builder->branch($innerHead);
+
+        $this->context->builder->positionAtEnd($innerDone);
+        $this->context->builder->store(
+            $this->context->builder->addNoSignedWrap($outer, $one),
+            $outerSlot
+        );
+        $this->context->builder->branch($outerHead);
+
+        $this->context->builder->positionAtEnd($outerDone);
+    }
+
+    private function emitBubbleSortLongs(
+        PHPLLVM\LLVMAbstract\Value\Function_ $fn,
+        PHPLLVM\Value $ht,
+        array $map,
+        PHPLLVM\Value $num
+    ): void {
+        $sizeT = $this->context->getTypeFromString('size_t');
+        $one = $sizeT->constInt(1, false);
+        $zero = $sizeT->constInt(0, false);
+        $outerSlot = $this->context->builder->alloca($sizeT, 1, 'sort_long_outer');
+        $this->context->builder->store($zero, $outerSlot);
+        $outerHead = $fn->appendBasicBlock('sort_long_outer_head');
+        $outerBody = $fn->appendBasicBlock('sort_long_outer_body');
+        $outerDone = $fn->appendBasicBlock('sort_long_outer_done');
+        $this->context->builder->branch($outerHead);
+
+        $this->context->builder->positionAtEnd($outerHead);
+        $outer = $this->context->builder->load($outerSlot);
+        $outerEnd = $this->context->builder->sub($num, $one);
+        $outerAtEnd = $this->context->builder->icmp(Builder::INT_SGE, $outer, $outerEnd);
+        $this->context->builder->branchIf($outerAtEnd, $outerDone, $outerBody);
+
+        $this->context->builder->positionAtEnd($outerBody);
+        $innerSlot = $this->context->builder->alloca($sizeT, 1, 'sort_long_inner');
+        $this->context->builder->store($zero, $innerSlot);
+        $limit = $this->context->builder->sub($num, $outer);
+        $limit = $this->context->builder->sub($limit, $one);
+        $innerHead = $fn->appendBasicBlock('sort_long_inner_head');
+        $innerBody = $fn->appendBasicBlock('sort_long_inner_body');
+        $innerDone = $fn->appendBasicBlock('sort_long_inner_done');
+        $this->context->builder->branch($innerHead);
+
+        $this->context->builder->positionAtEnd($innerHead);
+        $inner = $this->context->builder->load($innerSlot);
+        $innerAtEnd = $this->context->builder->icmp(Builder::INT_SGE, $inner, $limit);
+        $this->context->builder->branchIf($innerAtEnd, $innerDone, $innerBody);
+
+        $this->context->builder->positionAtEnd($innerBody);
+        $nextInner = $this->context->builder->addNoSignedWrap($inner, $one);
+        $longA = $this->context->builder->call(
+            $this->context->lookupFunction('__hashtable__readLongAt'),
+            $ht,
+            $inner
+        );
+        $longB = $this->context->builder->call(
+            $this->context->lookupFunction('__hashtable__readLongAt'),
+            $ht,
+            $nextInner
+        );
+        $needsSwap = $this->context->builder->icmp(Builder::INT_SGT, $longA, $longB);
+        $swapBlock = $fn->appendBasicBlock('sort_long_swap');
+        $noSwap = $fn->appendBasicBlock('sort_long_no_swap');
+        $afterSwap = $fn->appendBasicBlock('sort_long_after_swap');
+        $this->context->builder->branchIf($needsSwap, $swapBlock, $noSwap);
+
+        $this->context->builder->positionAtEnd($swapBlock);
+        $entryA = $this->listEntryAt($ht, $map, $inner);
+        $entryB = $this->listEntryAt($ht, $map, $nextInner);
+        $this->context->builder->call($this->context->lookupFunction('__value__writeLong'), $entryA, $longB);
+        $this->context->builder->call($this->context->lookupFunction('__value__writeLong'), $entryB, $longA);
+        $this->context->builder->branch($afterSwap);
+
+        $this->context->builder->positionAtEnd($noSwap);
+        $this->context->builder->branch($afterSwap);
+
+        $this->context->builder->positionAtEnd($afterSwap);
+        $this->context->builder->store($nextInner, $innerSlot);
+        $this->context->builder->branch($innerHead);
+
+        $this->context->builder->positionAtEnd($innerDone);
+        $this->context->builder->store(
+            $this->context->builder->addNoSignedWrap($outer, $one),
+            $outerSlot
+        );
+        $this->context->builder->branch($outerHead);
+
+        $this->context->builder->positionAtEnd($outerDone);
+    }
+
+    /**
+     * @param array<string, int> $map
+     */
+    private function listEntryAt(PHPLLVM\Value $ht, array $map, PHPLLVM\Value $index): PHPLLVM\Value
+    {
+        $values = $this->context->builder->load($this->context->builder->structGep($ht, $map['values']));
+
+        return $this->context->builder->inBoundsGep($values, $index);
+    }
+
     private function stringDataPtr(PHPLLVM\Value $str): PHPLLVM\Value
     {
         $map = $this->context->structFieldMap['__string__'];