sort: use a different codegen strategy

The existing codegen strategy in sort.go relied on parsing the sort.go source
with go/ast and a combination of an AST rewrite + code text rewrite with regexes
to generate zfuncversion -- the same sort functionality with a different variant
of data.

In preparation for implementing #47619, we need a more robust codegen
strategy. To generate variants required for the generic sort functions
in the slices package, we'd need significanly more complicated AST
rewrites, which would make genzfunc.go much heavier.

Instead, redo the codegen strategy to use text/template instead of AST rewrites.
gen_sort_variants.go now contains the code for the underlying sort functions,
and generates multiple versions of them based on Variant configuration structs.
With this approach, adding new variants to generate generic sort functions for
the slices package becomes trivial.

See the discussion in #47619 for more details on the design decisions.

Change-Id: I8af784c41b1dc8ef92aaf6321359e8faa5fe106c
Reviewed-on: https://go-review.googlesource.com/c/go/+/353069
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Trust: Than McIntosh <thanm@google.com>

6 files changed, 1211 insertions, 728 deletions

diff --git a/src/sort/gen_sort_variants.go b/src/sort/gen_sort_variants.go
new file mode 100644
index 0000000000..5f817221e1
--- /dev/null
+++ b/src/sort/gen_sort_variants.go
@@ -0,0 +1,526 @@
+// Copyright 2022 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build ignore
+// +build ignore
+
+// This program is run via "go generate" (via a directive in sort.go)
+// to generate implementation variants of the underlying sorting algorithm.
+// When passed the -generic flag it generates generic variants of sorting;
+// otherwise it generates the non-generic variants used by the sort package.
+
+package main
+
+import (
+	"bytes"
+	"flag"
+	"fmt"
+	"go/format"
+	"log"
+	"os"
+	"text/template"
+)
+
+type Variant struct {
+	// Name is the variant name: should be unique among variants.
+	Name string
+
+	// Path is the file path into which the generator will emit the code for this
+	// variant.
+	Path string
+
+	// Package is the package this code will be emitted into.
+	Package string
+
+	// Imports is the imports needed for this package.
+	Imports string
+
+	// FuncSuffix is appended to all function names in this variant's code. All
+	// suffixes should be unique within a package.
+	FuncSuffix string
+
+	// DataType is the type of the data parameter of functions in this variant's
+	// code.
+	DataType string
+
+	// TypeParam is the optional type parameter for the function.
+	TypeParam string
+
+	// ExtraParam is an extra parameter to pass to the function. Should begin with
+	// ", " to separate from other params.
+	ExtraParam string
+
+	// ExtraArg is an extra argument to pass to calls between functions; typically
+	// it invokes ExtraParam. Should begin with ", " to separate from other args.
+	ExtraArg string
+
+	// Funcs is a map of functions used from within the template. The following
+	// functions are expected to exist:
+	//
+	//    Less (name, i, j):
+	//      emits a comparison expression that checks if the value `name` at
+	//      index `i` is smaller than at index `j`.
+	//
+	//    Swap (name, i, j):
+	//      emits a statement that performs a data swap between elements `i` and
+	//      `j` of the value `name`.
+	Funcs template.FuncMap
+}
+
+func main() {
+	genGeneric := flag.Bool("generic", false, "generate generic versions")
+	flag.Parse()
+
+	if *genGeneric {
+		generate(&Variant{
+			Name:       "generic_ordered",
+			Path:       "zsortordered.go",
+			Package:    "slices",
+			Imports:    "import \"constraints\"\n",
+			FuncSuffix: "Ordered",
+			TypeParam:  "[Elem constraints.Ordered]",
+			ExtraParam: "",
+			ExtraArg:   "",
+			DataType:   "[]Elem",
+			Funcs: template.FuncMap{
+				"Less": func(name, i, j string) string {
+					return fmt.Sprintf("(%s[%s] < %s[%s])", name, i, name, j)
+				},
+				"Swap": func(name, i, j string) string {
+					return fmt.Sprintf("%s[%s], %s[%s] = %s[%s], %s[%s]", name, i, name, j, name, j, name, i)
+				},
+			},
+		})
+
+		generate(&Variant{
+			Name:       "generic_func",
+			Path:       "zsortanyfunc.go",
+			Package:    "slices",
+			FuncSuffix: "LessFunc",
+			TypeParam:  "[Elem any]",
+			ExtraParam: ", less func(a, b Elem) bool",
+			ExtraArg:   ", less",
+			DataType:   "[]Elem",
+			Funcs: template.FuncMap{
+				"Less": func(name, i, j string) string {
+					return fmt.Sprintf("less(%s[%s], %s[%s])", name, i, name, j)
+				},
+				"Swap": func(name, i, j string) string {
+					return fmt.Sprintf("%s[%s], %s[%s] = %s[%s], %s[%s]", name, i, name, j, name, j, name, i)
+				},
+			},
+		})
+	} else {
+		generate(&Variant{
+			Name:       "interface",
+			Path:       "zsortinterface.go",
+			Package:    "sort",
+			Imports:    "",
+			FuncSuffix: "",
+			TypeParam:  "",
+			ExtraParam: "",
+			ExtraArg:   "",
+			DataType:   "Interface",
+			Funcs: template.FuncMap{
+				"Less": func(name, i, j string) string {
+					return fmt.Sprintf("%s.Less(%s, %s)", name, i, j)
+				},
+				"Swap": func(name, i, j string) string {
+					return fmt.Sprintf("%s.Swap(%s, %s)", name, i, j)
+				},
+			},
+		})
+
+		generate(&Variant{
+			Name:       "func",
+			Path:       "zsortfunc.go",
+			Package:    "sort",
+			Imports:    "",
+			FuncSuffix: "_func",
+			TypeParam:  "",
+			ExtraParam: "",
+			ExtraArg:   "",
+			DataType:   "lessSwap",
+			Funcs: template.FuncMap{
+				"Less": func(name, i, j string) string {
+					return fmt.Sprintf("%s.Less(%s, %s)", name, i, j)
+				},
+				"Swap": func(name, i, j string) string {
+					return fmt.Sprintf("%s.Swap(%s, %s)", name, i, j)
+				},
+			},
+		})
+	}
+}
+
+// generate generates the code for variant `v` into a file named by `v.Path`.
+func generate(v *Variant) {
+	// Parse templateCode anew for each variant because Parse requires Funcs to be
+	// registered, and it helps type-check the funcs.
+	tmpl, err := template.New("gen").Funcs(v.Funcs).Parse(templateCode)
+	if err != nil {
+		log.Fatal("template Parse:", err)
+	}
+
+	var out bytes.Buffer
+	err = tmpl.Execute(&out, v)
+	if err != nil {
+		log.Fatal("template Execute:", err)
+	}
+
+	formatted, err := format.Source(out.Bytes())
+	if err != nil {
+		log.Fatal("format:", err)
+	}
+
+	if err := os.WriteFile(v.Path, formatted, 0644); err != nil {
+		log.Fatal("WriteFile:", err)
+	}
+}
+
+var templateCode = `// Code generated by gen_sort_variants.go; DO NOT EDIT.
+
+// Copyright 2022 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package {{.Package}}
+
+{{.Imports}}
+
+// insertionSort{{.FuncSuffix}} sorts data[a:b] using insertion sort.
+func insertionSort{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, b int {{.ExtraParam}}) {
+	for i := a + 1; i < b; i++ {
+		for j := i; j > a && {{Less "data" "j" "j-1"}}; j-- {
+			{{Swap "data" "j" "j-1"}}
+		}
+	}
+}
+
+// siftDown{{.FuncSuffix}} implements the heap property on data[lo:hi].
+// first is an offset into the array where the root of the heap lies.
+func siftDown{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, lo, hi, first int {{.ExtraParam}}) {
+	root := lo
+	for {
+		child := 2*root + 1
+		if child >= hi {
+			break
+		}
+		if child+1 < hi && {{Less "data" "first+child" "first+child+1"}} {
+			child++
+		}
+		if !{{Less "data" "first+root" "first+child"}} {
+			return
+		}
+		{{Swap "data" "first+root" "first+child"}}
+		root = child
+	}
+}
+
+func heapSort{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, b int {{.ExtraParam}}) {
+	first := a
+	lo := 0
+	hi := b - a
+
+	// Build heap with greatest element at top.
+	for i := (hi - 1) / 2; i >= 0; i-- {
+		siftDown{{.FuncSuffix}}(data, i, hi, first {{.ExtraArg}})
+	}
+
+	// Pop elements, largest first, into end of data.
+	for i := hi - 1; i >= 0; i-- {
+		{{Swap "data" "first" "first+i"}}
+		siftDown{{.FuncSuffix}}(data, lo, i, first {{.ExtraArg}})
+	}
+}
+
+// Quicksort, loosely following Bentley and McIlroy,
+// "Engineering a Sort Function" SP&E November 1993.
+
+// medianOfThree{{.FuncSuffix}} moves the median of the three values data[m0], data[m1], data[m2] into data[m1].
+func medianOfThree{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, m1, m0, m2 int {{.ExtraParam}}) {
+	// sort 3 elements
+	if {{Less "data" "m1" "m0"}} {
+		{{Swap "data" "m1" "m0"}}
+	}
+	// data[m0] <= data[m1]
+	if {{Less "data" "m2" "m1"}} {
+		{{Swap "data" "m2" "m1"}}
+		// data[m0] <= data[m2] && data[m1] < data[m2]
+		if {{Less "data" "m1" "m0"}} {
+			{{Swap "data" "m1" "m0"}}
+		}
+	}
+	// now data[m0] <= data[m1] <= data[m2]
+}
+
+func swapRange{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, b, n int {{.ExtraParam}}) {
+	for i := 0; i < n; i++ {
+		{{Swap "data" "a+i" "b+i"}}
+	}
+}
+
+func doPivot{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, lo, hi int {{.ExtraParam}}) (midlo, midhi int) {
+	m := int(uint(lo+hi) >> 1) // Written like this to avoid integer overflow.
+	if hi-lo > 40 {
+		// Tukey's "Ninther" median of three medians of three.
+		s := (hi - lo) / 8
+		medianOfThree{{.FuncSuffix}}(data, lo, lo+s, lo+2*s {{.ExtraArg}})
+		medianOfThree{{.FuncSuffix}}(data, m, m-s, m+s {{.ExtraArg}})
+		medianOfThree{{.FuncSuffix}}(data, hi-1, hi-1-s, hi-1-2*s {{.ExtraArg}})
+	}
+	medianOfThree{{.FuncSuffix}}(data, lo, m, hi-1 {{.ExtraArg}})
+
+	// Invariants are:
+	//	data[lo] = pivot (set up by ChoosePivot)
+	//	data[lo < i < a] < pivot
+	//	data[a <= i < b] <= pivot
+	//	data[b <= i < c] unexamined
+	//	data[c <= i < hi-1] > pivot
+	//	data[hi-1] >= pivot
+	pivot := lo
+	a, c := lo+1, hi-1
+
+	for ; a < c && {{Less "data" "a" "pivot"}}; a++ {
+	}
+	b := a
+	for {
+		for ; b < c && !{{Less "data" "pivot" "b"}}; b++ { // data[b] <= pivot
+		}
+		for ; b < c && {{Less "data" "pivot" "c-1"}}; c-- { // data[c-1] > pivot
+		}
+		if b >= c {
+			break
+		}
+		// data[b] > pivot; data[c-1] <= pivot
+		{{Swap "data" "b" "c-1"}}
+		b++
+		c--
+	}
+	// If hi-c<3 then there are duplicates (by property of median of nine).
+	// Let's be a bit more conservative, and set border to 5.
+	protect := hi-c < 5
+	if !protect && hi-c < (hi-lo)/4 {
+		// Lets test some points for equality to pivot
+		dups := 0
+		if !{{Less "data" "pivot" "hi-1"}} { // data[hi-1] = pivot
+			{{Swap "data" "c" "hi-1"}}
+			c++
+			dups++
+		}
+		if !{{Less "data" "b-1" "pivot"}} { // data[b-1] = pivot
+			b--
+			dups++
+		}
+		// m-lo = (hi-lo)/2 > 6
+		// b-lo > (hi-lo)*3/4-1 > 8
+		// ==> m < b ==> data[m] <= pivot
+		if !{{Less "data" "m" "pivot"}} { // data[m] = pivot
+			{{Swap "data" "m" "b-1"}}
+			b--
+			dups++
+		}
+		// if at least 2 points are equal to pivot, assume skewed distribution
+		protect = dups > 1
+	}
+	if protect {
+		// Protect against a lot of duplicates
+		// Add invariant:
+		//	data[a <= i < b] unexamined
+		//	data[b <= i < c] = pivot
+		for {
+			for ; a < b && !{{Less "data" "b-1" "pivot"}}; b-- { // data[b] == pivot
+			}
+			for ; a < b && {{Less "data" "a" "pivot"}}; a++ { // data[a] < pivot
+			}
+			if a >= b {
+				break
+			}
+			// data[a] == pivot; data[b-1] < pivot
+			{{Swap "data" "a" "b-1"}}
+			a++
+			b--
+		}
+	}
+	// Swap pivot into middle
+	{{Swap "data" "pivot" "b-1"}}
+	return b - 1, c
+}
+
+func quickSort{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, b, maxDepth int {{.ExtraParam}}) {
+	for b-a > 12 { // Use ShellSort for slices <= 12 elements
+		if maxDepth == 0 {
+			heapSort{{.FuncSuffix}}(data, a, b {{.ExtraArg}})
+			return
+		}
+		maxDepth--
+		mlo, mhi := doPivot{{.FuncSuffix}}(data, a, b {{.ExtraArg}})
+		// Avoiding recursion on the larger subproblem guarantees
+		// a stack depth of at most lg(b-a).
+		if mlo-a < b-mhi {
+			quickSort{{.FuncSuffix}}(data, a, mlo, maxDepth {{.ExtraArg}})
+			a = mhi // i.e., quickSort{{.FuncSuffix}}(data, mhi, b)
+		} else {
+			quickSort{{.FuncSuffix}}(data, mhi, b, maxDepth {{.ExtraArg}})
+			b = mlo // i.e., quickSort{{.FuncSuffix}}(data, a, mlo)
+		}
+	}
+	if b-a > 1 {
+		// Do ShellSort pass with gap 6
+		// It could be written in this simplified form cause b-a <= 12
+		for i := a + 6; i < b; i++ {
+			if {{Less "data" "i" "i-6"}} {
+				{{Swap "data" "i" "i-6"}}
+			}
+		}
+		insertionSort{{.FuncSuffix}}(data, a, b {{.ExtraArg}})
+	}
+}
+
+func stable{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, n int {{.ExtraParam}}) {
+	blockSize := 20 // must be > 0
+	a, b := 0, blockSize
+	for b <= n {
+		insertionSort{{.FuncSuffix}}(data, a, b {{.ExtraArg}})
+		a = b
+		b += blockSize
+	}
+	insertionSort{{.FuncSuffix}}(data, a, n {{.ExtraArg}})
+
+	for blockSize < n {
+		a, b = 0, 2*blockSize
+		for b <= n {
+			symMerge{{.FuncSuffix}}(data, a, a+blockSize, b {{.ExtraArg}})
+			a = b
+			b += 2 * blockSize
+		}
+		if m := a + blockSize; m < n {
+			symMerge{{.FuncSuffix}}(data, a, m, n {{.ExtraArg}})
+		}
+		blockSize *= 2
+	}
+}
+
+// symMerge{{.FuncSuffix}} merges the two sorted subsequences data[a:m] and data[m:b] using
+// the SymMerge algorithm from Pok-Son Kim and Arne Kutzner, "Stable Minimum
+// Storage Merging by Symmetric Comparisons", in Susanne Albers and Tomasz
+// Radzik, editors, Algorithms - ESA 2004, volume 3221 of Lecture Notes in
+// Computer Science, pages 714-723. Springer, 2004.
+//
+// Let M = m-a and N = b-n. Wolog M < N.
+// The recursion depth is bound by ceil(log(N+M)).
+// The algorithm needs O(M*log(N/M + 1)) calls to data.Less.
+// The algorithm needs O((M+N)*log(M)) calls to data.Swap.
+//
+// The paper gives O((M+N)*log(M)) as the number of assignments assuming a
+// rotation algorithm which uses O(M+N+gcd(M+N)) assignments. The argumentation
+// in the paper carries through for Swap operations, especially as the block
+// swapping rotate uses only O(M+N) Swaps.
+//
+// symMerge assumes non-degenerate arguments: a < m && m < b.
+// Having the caller check this condition eliminates many leaf recursion calls,
+// which improves performance.
+func symMerge{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, m, b int {{.ExtraParam}}) {
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[a] into data[m:b]
+	// if data[a:m] only contains one element.
+	if m-a == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] >= data[a] for m <= i < b.
+		// Exit the search loop with i == b in case no such index exists.
+		i := m
+		j := b
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if {{Less "data" "h" "a"}} {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[a] reaches the position before i.
+		for k := a; k < i-1; k++ {
+			{{Swap "data" "k" "k+1"}}
+		}
+		return
+	}
+
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[m] into data[a:m]
+	// if data[m:b] only contains one element.
+	if b-m == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] > data[m] for a <= i < m.
+		// Exit the search loop with i == m in case no such index exists.
+		i := a
+		j := m
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if !{{Less "data" "m" "h"}} {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[m] reaches the position i.
+		for k := m; k > i; k-- {
+			{{Swap "data" "k" "k-1"}}
+		}
+		return
+	}
+
+	mid := int(uint(a+b) >> 1)
+	n := mid + m
+	var start, r int
+	if m > mid {
+		start = n - b
+		r = mid
+	} else {
+		start = a
+		r = m
+	}
+	p := n - 1
+
+	for start < r {
+		c := int(uint(start+r) >> 1)
+		if !{{Less "data" "p-c" "c"}} {
+			start = c + 1
+		} else {
+			r = c
+		}
+	}
+
+	end := n - start
+	if start < m && m < end {
+		rotate{{.FuncSuffix}}(data, start, m, end {{.ExtraArg}})
+	}
+	if a < start && start < mid {
+		symMerge{{.FuncSuffix}}(data, a, start, mid {{.ExtraArg}})
+	}
+	if mid < end && end < b {
+		symMerge{{.FuncSuffix}}(data, mid, end, b {{.ExtraArg}})
+	}
+}
+
+// rotate{{.FuncSuffix}} rotates two consecutive blocks u = data[a:m] and v = data[m:b] in data:
+// Data of the form 'x u v y' is changed to 'x v u y'.
+// rotate performs at most b-a many calls to data.Swap,
+// and it assumes non-degenerate arguments: a < m && m < b.
+func rotate{{.FuncSuffix}}{{.TypeParam}}(data {{.DataType}}, a, m, b int {{.ExtraParam}}) {
+	i := m - a
+	j := b - m
+
+	for i != j {
+		if i > j {
+			swapRange{{.FuncSuffix}}(data, m-i, m, j {{.ExtraArg}})
+			i -= j
+		} else {
+			swapRange{{.FuncSuffix}}(data, m-i, m+j-i, i {{.ExtraArg}})
+			j -= i
+		}
+	}
+	// i == j
+	swapRange{{.FuncSuffix}}(data, m-i, m, i {{.ExtraArg}})
+}
+`
diff --git a/src/sort/genzfunc.go b/src/sort/genzfunc.go
deleted file mode 100644
index ed04e33568..0000000000
--- a/src/sort/genzfunc.go
+++ /dev/null
@@ -1,127 +0,0 @@
-// Copyright 2016 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-//go:build ignore
-// +build ignore
-
-// This program is run via "go generate" (via a directive in sort.go)
-// to generate zfuncversion.go.
-//
-// It copies sort.go to zfuncversion.go, only retaining funcs which
-// take a "data Interface" parameter, and renaming each to have a
-// "_func" suffix and taking a "data lessSwap" instead. It then rewrites
-// each internal function call to the appropriate _func variants.
-
-package main
-
-import (
-	"bytes"
-	"go/ast"
-	"go/format"
-	"go/parser"
-	"go/token"
-	"log"
-	"os"
-	"regexp"
-)
-
-var fset = token.NewFileSet()
-
-func main() {
-	af, err := parser.ParseFile(fset, "sort.go", nil, 0)
-	if err != nil {
-		log.Fatal(err)
-	}
-	af.Doc = nil
-	af.Imports = nil
-	af.Comments = nil
-
-	var newDecl []ast.Decl
-	for _, d := range af.Decls {
-		fd, ok := d.(*ast.FuncDecl)
-		if !ok {
-			continue
-		}
-		if fd.Recv != nil || fd.Name.IsExported() {
-			continue
-		}
-		typ := fd.Type
-		if len(typ.Params.List) < 1 {
-			continue
-		}
-		arg0 := typ.Params.List[0]
-		arg0Name := arg0.Names[0].Name
-		arg0Type := arg0.Type.(*ast.Ident)
-		if arg0Name != "data" || arg0Type.Name != "Interface" {
-			continue
-		}
-		arg0Type.Name = "lessSwap"
-
-		newDecl = append(newDecl, fd)
-	}
-	af.Decls = newDecl
-	ast.Walk(visitFunc(rewriteCalls), af)
-
-	var out bytes.Buffer
-	if err := format.Node(&out, fset, af); err != nil {
-		log.Fatalf("format.Node: %v", err)
-	}
-
-	// Get rid of blank lines after removal of comments.
-	src := regexp.MustCompile(`\n{2,}`).ReplaceAll(out.Bytes(), []byte("\n"))
-
-	// Add comments to each func, for the lost reader.
-	// This is so much easier than adding comments via the AST
-	// and trying to get position info correct.
-	src = regexp.MustCompile(`(?m)^func (\w+)`).ReplaceAll(src, []byte("\n// Auto-generated variant of sort.go:$1\nfunc ${1}_func"))
-
-	// Final gofmt.
-	src, err = format.Source(src)
-	if err != nil {
-		log.Fatalf("format.Source: %v on\n%s", err, src)
-	}
-
-	out.Reset()
-	out.WriteString(`// Code generated from sort.go using genzfunc.go; DO NOT EDIT.
-
-// Copyright 2016 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-`)
-	out.Write(src)
-
-	const target = "zfuncversion.go"
-	if err := os.WriteFile(target, out.Bytes(), 0644); err != nil {
-		log.Fatal(err)
-	}
-}
-
-type visitFunc func(ast.Node) ast.Visitor
-
-func (f visitFunc) Visit(n ast.Node) ast.Visitor { return f(n) }
-
-func rewriteCalls(n ast.Node) ast.Visitor {
-	ce, ok := n.(*ast.CallExpr)
-	if ok {
-		rewriteCall(ce)
-	}
-	return visitFunc(rewriteCalls)
-}
-
-func rewriteCall(ce *ast.CallExpr) {
-	ident, ok := ce.Fun.(*ast.Ident)
-	if !ok {
-		// e.g. skip SelectorExpr (data.Less(..) calls)
-		return
-	}
-	// skip casts
-	if ident.Name == "int" || ident.Name == "uint" {
-		return
-	}
-	if len(ce.Args) < 1 {
-		return
-	}
-	ident.Name += "_func"
-}
diff --git a/src/sort/sort.go b/src/sort/sort.go
index 749310764a..2c197afc03 100644
--- a/src/sort/sort.go
+++ b/src/sort/sort.go
@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:generate go run genzfunc.go
+//go:generate go run gen_sort_variants.go
 
 // Package sort provides primitives for sorting slices and user-defined collections.
 package sort
@@ -34,195 +34,6 @@ type Interface interface {
 	Swap(i, j int)
 }
 
-// insertionSort sorts data[a:b] using insertion sort.
-func insertionSort(data Interface, a, b int) {
-	for i := a + 1; i < b; i++ {
-		for j := i; j > a && data.Less(j, j-1); j-- {
-			data.Swap(j, j-1)
-		}
-	}
-}
-
-// siftDown implements the heap property on data[lo:hi].
-// first is an offset into the array where the root of the heap lies.
-func siftDown(data Interface, lo, hi, first int) {
-	root := lo
-	for {
-		child := 2*root + 1
-		if child >= hi {
-			break
-		}
-		if child+1 < hi && data.Less(first+child, first+child+1) {
-			child++
-		}
-		if !data.Less(first+root, first+child) {
-			return
-		}
-		data.Swap(first+root, first+child)
-		root = child
-	}
-}
-
-func heapSort(data Interface, a, b int) {
-	first := a
-	lo := 0
-	hi := b - a
-
-	// Build heap with greatest element at top.
-	for i := (hi - 1) / 2; i >= 0; i-- {
-		siftDown(data, i, hi, first)
-	}
-
-	// Pop elements, largest first, into end of data.
-	for i := hi - 1; i >= 0; i-- {
-		data.Swap(first, first+i)
-		siftDown(data, lo, i, first)
-	}
-}
-
-// Quicksort, loosely following Bentley and McIlroy,
-// ``Engineering a Sort Function,'' SP&E November 1993.
-
-// medianOfThree moves the median of the three values data[m0], data[m1], data[m2] into data[m1].
-func medianOfThree(data Interface, m1, m0, m2 int) {
-	// sort 3 elements
-	if data.Less(m1, m0) {
-		data.Swap(m1, m0)
-	}
-	// data[m0] <= data[m1]
-	if data.Less(m2, m1) {
-		data.Swap(m2, m1)
-		// data[m0] <= data[m2] && data[m1] < data[m2]
-		if data.Less(m1, m0) {
-			data.Swap(m1, m0)
-		}
-	}
-	// now data[m0] <= data[m1] <= data[m2]
-}
-
-func swapRange(data Interface, a, b, n int) {
-	for i := 0; i < n; i++ {
-		data.Swap(a+i, b+i)
-	}
-}
-
-func doPivot(data Interface, lo, hi int) (midlo, midhi int) {
-	m := int(uint(lo+hi) >> 1) // Written like this to avoid integer overflow.
-	if hi-lo > 40 {
-		// Tukey's ``Ninther,'' median of three medians of three.
-		s := (hi - lo) / 8
-		medianOfThree(data, lo, lo+s, lo+2*s)
-		medianOfThree(data, m, m-s, m+s)
-		medianOfThree(data, hi-1, hi-1-s, hi-1-2*s)
-	}
-	medianOfThree(data, lo, m, hi-1)
-
-	// Invariants are:
-	//	data[lo] = pivot (set up by ChoosePivot)
-	//	data[lo < i < a] < pivot
-	//	data[a <= i < b] <= pivot
-	//	data[b <= i < c] unexamined
-	//	data[c <= i < hi-1] > pivot
-	//	data[hi-1] >= pivot
-	pivot := lo
-	a, c := lo+1, hi-1
-
-	for ; a < c && data.Less(a, pivot); a++ {
-	}
-	b := a
-	for {
-		for ; b < c && !data.Less(pivot, b); b++ { // data[b] <= pivot
-		}
-		for ; b < c && data.Less(pivot, c-1); c-- { // data[c-1] > pivot
-		}
-		if b >= c {
-			break
-		}
-		// data[b] > pivot; data[c-1] <= pivot
-		data.Swap(b, c-1)
-		b++
-		c--
-	}
-	// If hi-c<3 then there are duplicates (by property of median of nine).
-	// Let's be a bit more conservative, and set border to 5.
-	protect := hi-c < 5
-	if !protect && hi-c < (hi-lo)/4 {
-		// Lets test some points for equality to pivot
-		dups := 0
-		if !data.Less(pivot, hi-1) { // data[hi-1] = pivot
-			data.Swap(c, hi-1)
-			c++
-			dups++
-		}
-		if !data.Less(b-1, pivot) { // data[b-1] = pivot
-			b--
-			dups++
-		}
-		// m-lo = (hi-lo)/2 > 6
-		// b-lo > (hi-lo)*3/4-1 > 8
-		// ==> m < b ==> data[m] <= pivot
-		if !data.Less(m, pivot) { // data[m] = pivot
-			data.Swap(m, b-1)
-			b--
-			dups++
-		}
-		// if at least 2 points are equal to pivot, assume skewed distribution
-		protect = dups > 1
-	}
-	if protect {
-		// Protect against a lot of duplicates
-		// Add invariant:
-		//	data[a <= i < b] unexamined
-		//	data[b <= i < c] = pivot
-		for {
-			for ; a < b && !data.Less(b-1, pivot); b-- { // data[b] == pivot
-			}
-			for ; a < b && data.Less(a, pivot); a++ { // data[a] < pivot
-			}
-			if a >= b {
-				break
-			}
-			// data[a] == pivot; data[b-1] < pivot
-			data.Swap(a, b-1)
-			a++
-			b--
-		}
-	}
-	// Swap pivot into middle
-	data.Swap(pivot, b-1)
-	return b - 1, c
-}
-
-func quickSort(data Interface, a, b, maxDepth int) {
-	for b-a > 12 { // Use ShellSort for slices <= 12 elements
-		if maxDepth == 0 {
-			heapSort(data, a, b)
-			return
-		}
-		maxDepth--
-		mlo, mhi := doPivot(data, a, b)
-		// Avoiding recursion on the larger subproblem guarantees
-		// a stack depth of at most lg(b-a).
-		if mlo-a < b-mhi {
-			quickSort(data, a, mlo, maxDepth)
-			a = mhi // i.e., quickSort(data, mhi, b)
-		} else {
-			quickSort(data, mhi, b, maxDepth)
-			b = mlo // i.e., quickSort(data, a, mlo)
-		}
-	}
-	if b-a > 1 {
-		// Do ShellSort pass with gap 6
-		// It could be written in this simplified form cause b-a <= 12
-		for i := a + 6; i < b; i++ {
-			if data.Less(i, i-6) {
-				data.Swap(i, i-6)
-			}
-		}
-		insertionSort(data, a, b)
-	}
-}
-
 // Sort sorts data in ascending order as determined by the Less method.
 // It makes one call to data.Len to determine n and O(n*log(n)) calls to
 // data.Less and data.Swap. The sort is not guaranteed to be stable.
@@ -379,152 +190,6 @@ func Stable(data Interface) {
 	stable(data, data.Len())
 }
 
-func stable(data Interface, n int) {
-	blockSize := 20 // must be > 0
-	a, b := 0, blockSize
-	for b <= n {
-		insertionSort(data, a, b)
-		a = b
-		b += blockSize
-	}
-	insertionSort(data, a, n)
-
-	for blockSize < n {
-		a, b = 0, 2*blockSize
-		for b <= n {
-			symMerge(data, a, a+blockSize, b)
-			a = b
-			b += 2 * blockSize
-		}
-		if m := a + blockSize; m < n {
-			symMerge(data, a, m, n)
-		}
-		blockSize *= 2
-	}
-}
-
-// symMerge merges the two sorted subsequences data[a:m] and data[m:b] using
-// the SymMerge algorithm from Pok-Son Kim and Arne Kutzner, "Stable Minimum
-// Storage Merging by Symmetric Comparisons", in Susanne Albers and Tomasz
-// Radzik, editors, Algorithms - ESA 2004, volume 3221 of Lecture Notes in
-// Computer Science, pages 714-723. Springer, 2004.
-//
-// Let M = m-a and N = b-n. Wolog M < N.
-// The recursion depth is bound by ceil(log(N+M)).
-// The algorithm needs O(M*log(N/M + 1)) calls to data.Less.
-// The algorithm needs O((M+N)*log(M)) calls to data.Swap.
-//
-// The paper gives O((M+N)*log(M)) as the number of assignments assuming a
-// rotation algorithm which uses O(M+N+gcd(M+N)) assignments. The argumentation
-// in the paper carries through for Swap operations, especially as the block
-// swapping rotate uses only O(M+N) Swaps.
-//
-// symMerge assumes non-degenerate arguments: a < m && m < b.
-// Having the caller check this condition eliminates many leaf recursion calls,
-// which improves performance.
-func symMerge(data Interface, a, m, b int) {
-	// Avoid unnecessary recursions of symMerge
-	// by direct insertion of data[a] into data[m:b]
-	// if data[a:m] only contains one element.
-	if m-a == 1 {
-		// Use binary search to find the lowest index i
-		// such that data[i] >= data[a] for m <= i < b.
-		// Exit the search loop with i == b in case no such index exists.
-		i := m
-		j := b
-		for i < j {
-			h := int(uint(i+j) >> 1)
-			if data.Less(h, a) {
-				i = h + 1
-			} else {
-				j = h
-			}
-		}
-		// Swap values until data[a] reaches the position before i.
-		for k := a; k < i-1; k++ {
-			data.Swap(k, k+1)
-		}
-		return
-	}
-
-	// Avoid unnecessary recursions of symMerge
-	// by direct insertion of data[m] into data[a:m]
-	// if data[m:b] only contains one element.
-	if b-m == 1 {
-		// Use binary search to find the lowest index i
-		// such that data[i] > data[m] for a <= i < m.
-		// Exit the search loop with i == m in case no such index exists.
-		i := a
-		j := m
-		for i < j {
-			h := int(uint(i+j) >> 1)
-			if !data.Less(m, h) {
-				i = h + 1
-			} else {
-				j = h
-			}
-		}
-		// Swap values until data[m] reaches the position i.
-		for k := m; k > i; k-- {
-			data.Swap(k, k-1)
-		}
-		return
-	}
-
-	mid := int(uint(a+b) >> 1)
-	n := mid + m
-	var start, r int
-	if m > mid {
-		start = n - b
-		r = mid
-	} else {
-		start = a
-		r = m
-	}
-	p := n - 1
-
-	for start < r {
-		c := int(uint(start+r) >> 1)
-		if !data.Less(p-c, c) {
-			start = c + 1
-		} else {
-			r = c
-		}
-	}
-
-	end := n - start
-	if start < m && m < end {
-		rotate(data, start, m, end)
-	}
-	if a < start && start < mid {
-		symMerge(data, a, start, mid)
-	}
-	if mid < end && end < b {
-		symMerge(data, mid, end, b)
-	}
-}
-
-// rotate rotates two consecutive blocks u = data[a:m] and v = data[m:b] in data:
-// Data of the form 'x u v y' is changed to 'x v u y'.
-// rotate performs at most b-a many calls to data.Swap,
-// and it assumes non-degenerate arguments: a < m && m < b.
-func rotate(data Interface, a, m, b int) {
-	i := m - a
-	j := b - m
-
-	for i != j {
-		if i > j {
-			swapRange(data, m-i, m, j)
-			i -= j
-		} else {
-			swapRange(data, m-i, m+j-i, i)
-			j -= i
-		}
-	}
-	// i == j
-	swapRange(data, m-i, m, i)
-}
-
 /*
 Complexity of Stable Sorting
 
diff --git a/src/sort/zfuncversion.go b/src/sort/zfuncversion.go
deleted file mode 100644
index 30067cbe07..0000000000
--- a/src/sort/zfuncversion.go
+++ /dev/null
@@ -1,265 +0,0 @@
-// Code generated from sort.go using genzfunc.go; DO NOT EDIT.
-
-// Copyright 2016 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-package sort
-
-// Auto-generated variant of sort.go:insertionSort
-func insertionSort_func(data lessSwap, a, b int) {
-	for i := a + 1; i < b; i++ {
-		for j := i; j > a && data.Less(j, j-1); j-- {
-			data.Swap(j, j-1)
-		}
-	}
-}
-
-// Auto-generated variant of sort.go:siftDown
-func siftDown_func(data lessSwap, lo, hi, first int) {
-	root := lo
-	for {
-		child := 2*root + 1
-		if child >= hi {
-			break
-		}
-		if child+1 < hi && data.Less(first+child, first+child+1) {
-			child++
-		}
-		if !data.Less(first+root, first+child) {
-			return
-		}
-		data.Swap(first+root, first+child)
-		root = child
-	}
-}
-
-// Auto-generated variant of sort.go:heapSort
-func heapSort_func(data lessSwap, a, b int) {
-	first := a
-	lo := 0
-	hi := b - a
-	for i := (hi - 1) / 2; i >= 0; i-- {
-		siftDown_func(data, i, hi, first)
-	}
-	for i := hi - 1; i >= 0; i-- {
-		data.Swap(first, first+i)
-		siftDown_func(data, lo, i, first)
-	}
-}
-
-// Auto-generated variant of sort.go:medianOfThree
-func medianOfThree_func(data lessSwap, m1, m0, m2 int) {
-	if data.Less(m1, m0) {
-		data.Swap(m1, m0)
-	}
-	if data.Less(m2, m1) {
-		data.Swap(m2, m1)
-		if data.Less(m1, m0) {
-			data.Swap(m1, m0)
-		}
-	}
-}
-
-// Auto-generated variant of sort.go:swapRange
-func swapRange_func(data lessSwap, a, b, n int) {
-	for i := 0; i < n; i++ {
-		data.Swap(a+i, b+i)
-	}
-}
-
-// Auto-generated variant of sort.go:doPivot
-func doPivot_func(data lessSwap, lo, hi int) (midlo, midhi int) {
-	m := int(uint(lo+hi) >> 1)
-	if hi-lo > 40 {
-		s := (hi - lo) / 8
-		medianOfThree_func(data, lo, lo+s, lo+2*s)
-		medianOfThree_func(data, m, m-s, m+s)
-		medianOfThree_func(data, hi-1, hi-1-s, hi-1-2*s)
-	}
-	medianOfThree_func(data, lo, m, hi-1)
-	pivot := lo
-	a, c := lo+1, hi-1
-	for ; a < c && data.Less(a, pivot); a++ {
-	}
-	b := a
-	for {
-		for ; b < c && !data.Less(pivot, b); b++ {
-		}
-		for ; b < c && data.Less(pivot, c-1); c-- {
-		}
-		if b >= c {
-			break
-		}
-		data.Swap(b, c-1)
-		b++
-		c--
-	}
-	protect := hi-c < 5
-	if !protect && hi-c < (hi-lo)/4 {
-		dups := 0
-		if !data.Less(pivot, hi-1) {
-			data.Swap(c, hi-1)
-			c++
-			dups++
-		}
-		if !data.Less(b-1, pivot) {
-			b--
-			dups++
-		}
-		if !data.Less(m, pivot) {
-			data.Swap(m, b-1)
-			b--
-			dups++
-		}
-		protect = dups > 1
-	}
-	if protect {
-		for {
-			for ; a < b && !data.Less(b-1, pivot); b-- {
-			}
-			for ; a < b && data.Less(a, pivot); a++ {
-			}
-			if a >= b {
-				break
-			}
-			data.Swap(a, b-1)
-			a++
-			b--
-		}
-	}
-	data.Swap(pivot, b-1)
-	return b - 1, c
-}
-
-// Auto-generated variant of sort.go:quickSort
-func quickSort_func(data lessSwap, a, b, maxDepth int) {
-	for b-a > 12 {
-		if maxDepth == 0 {
-			heapSort_func(data, a, b)
-			return
-		}
-		maxDepth--
-		mlo, mhi := doPivot_func(data, a, b)
-		if mlo-a < b-mhi {
-			quickSort_func(data, a, mlo, maxDepth)
-			a = mhi
-		} else {
-			quickSort_func(data, mhi, b, maxDepth)
-			b = mlo
-		}
-	}
-	if b-a > 1 {
-		for i := a + 6; i < b; i++ {
-			if data.Less(i, i-6) {
-				data.Swap(i, i-6)
-			}
-		}
-		insertionSort_func(data, a, b)
-	}
-}
-
-// Auto-generated variant of sort.go:stable
-func stable_func(data lessSwap, n int) {
-	blockSize := 20
-	a, b := 0, blockSize
-	for b <= n {
-		insertionSort_func(data, a, b)
-		a = b
-		b += blockSize
-	}
-	insertionSort_func(data, a, n)
-	for blockSize < n {
-		a, b = 0, 2*blockSize
-		for b <= n {
-			symMerge_func(data, a, a+blockSize, b)
-			a = b
-			b += 2 * blockSize
-		}
-		if m := a + blockSize; m < n {
-			symMerge_func(data, a, m, n)
-		}
-		blockSize *= 2
-	}
-}
-
-// Auto-generated variant of sort.go:symMerge
-func symMerge_func(data lessSwap, a, m, b int) {
-	if m-a == 1 {
-		i := m
-		j := b
-		for i < j {
-			h := int(uint(i+j) >> 1)
-			if data.Less(h, a) {
-				i = h + 1
-			} else {
-				j = h
-			}
-		}
-		for k := a; k < i-1; k++ {
-			data.Swap(k, k+1)
-		}
-		return
-	}
-	if b-m == 1 {
-		i := a
-		j := m
-		for i < j {
-			h := int(uint(i+j) >> 1)
-			if !data.Less(m, h) {
-				i = h + 1
-			} else {
-				j = h
-			}
-		}
-		for k := m; k > i; k-- {
-			data.Swap(k, k-1)
-		}
-		return
-	}
-	mid := int(uint(a+b) >> 1)
-	n := mid + m
-	var start, r int
-	if m > mid {
-		start = n - b
-		r = mid
-	} else {
-		start = a
-		r = m
-	}
-	p := n - 1
-	for start < r {
-		c := int(uint(start+r) >> 1)
-		if !data.Less(p-c, c) {
-			start = c + 1
-		} else {
-			r = c
-		}
-	}
-	end := n - start
-	if start < m && m < end {
-		rotate_func(data, start, m, end)
-	}
-	if a < start && start < mid {
-		symMerge_func(data, a, start, mid)
-	}
-	if mid < end && end < b {
-		symMerge_func(data, mid, end, b)
-	}
-}
-
-// Auto-generated variant of sort.go:rotate
-func rotate_func(data lessSwap, a, m, b int) {
-	i := m - a
-	j := b - m
-	for i != j {
-		if i > j {
-			swapRange_func(data, m-i, m, j)
-			i -= j
-		} else {
-			swapRange_func(data, m-i, m+j-i, i)
-			j -= i
-		}
-	}
-	swapRange_func(data, m-i, m, i)
-}
diff --git a/src/sort/zsortfunc.go b/src/sort/zsortfunc.go
new file mode 100644
index 0000000000..80c8a77995
--- /dev/null
+++ b/src/sort/zsortfunc.go
@@ -0,0 +1,342 @@
+// Code generated by gen_sort_variants.go; DO NOT EDIT.
+
+// Copyright 2022 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sort
+
+// insertionSort_func sorts data[a:b] using insertion sort.
+func insertionSort_func(data lessSwap, a, b int) {
+	for i := a + 1; i < b; i++ {
+		for j := i; j > a && data.Less(j, j-1); j-- {
+			data.Swap(j, j-1)
+		}
+	}
+}
+
+// siftDown_func implements the heap property on data[lo:hi].
+// first is an offset into the array where the root of the heap lies.
+func siftDown_func(data lessSwap, lo, hi, first int) {
+	root := lo
+	for {
+		child := 2*root + 1
+		if child >= hi {
+			break
+		}
+		if child+1 < hi && data.Less(first+child, first+child+1) {
+			child++
+		}
+		if !data.Less(first+root, first+child) {
+			return
+		}
+		data.Swap(first+root, first+child)
+		root = child
+	}
+}
+
+func heapSort_func(data lessSwap, a, b int) {
+	first := a
+	lo := 0
+	hi := b - a
+
+	// Build heap with greatest element at top.
+	for i := (hi - 1) / 2; i >= 0; i-- {
+		siftDown_func(data, i, hi, first)
+	}
+
+	// Pop elements, largest first, into end of data.
+	for i := hi - 1; i >= 0; i-- {
+		data.Swap(first, first+i)
+		siftDown_func(data, lo, i, first)
+	}
+}
+
+// Quicksort, loosely following Bentley and McIlroy,
+// "Engineering a Sort Function" SP&E November 1993.
+
+// medianOfThree_func moves the median of the three values data[m0], data[m1], data[m2] into data[m1].
+func medianOfThree_func(data lessSwap, m1, m0, m2 int) {
+	// sort 3 elements
+	if data.Less(m1, m0) {
+		data.Swap(m1, m0)
+	}
+	// data[m0] <= data[m1]
+	if data.Less(m2, m1) {
+		data.Swap(m2, m1)
+		// data[m0] <= data[m2] && data[m1] < data[m2]
+		if data.Less(m1, m0) {
+			data.Swap(m1, m0)
+		}
+	}
+	// now data[m0] <= data[m1] <= data[m2]
+}
+
+func swapRange_func(data lessSwap, a, b, n int) {
+	for i := 0; i < n; i++ {
+		data.Swap(a+i, b+i)
+	}
+}
+
+func doPivot_func(data lessSwap, lo, hi int) (midlo, midhi int) {
+	m := int(uint(lo+hi) >> 1) // Written like this to avoid integer overflow.
+	if hi-lo > 40 {
+		// Tukey's "Ninther" median of three medians of three.
+		s := (hi - lo) / 8
+		medianOfThree_func(data, lo, lo+s, lo+2*s)
+		medianOfThree_func(data, m, m-s, m+s)
+		medianOfThree_func(data, hi-1, hi-1-s, hi-1-2*s)
+	}
+	medianOfThree_func(data, lo, m, hi-1)
+
+	// Invariants are:
+	//	data[lo] = pivot (set up by ChoosePivot)
+	//	data[lo < i < a] < pivot
+	//	data[a <= i < b] <= pivot
+	//	data[b <= i < c] unexamined
+	//	data[c <= i < hi-1] > pivot
+	//	data[hi-1] >= pivot
+	pivot := lo
+	a, c := lo+1, hi-1
+
+	for ; a < c && data.Less(a, pivot); a++ {
+	}
+	b := a
+	for {
+		for ; b < c && !data.Less(pivot, b); b++ { // data[b] <= pivot
+		}
+		for ; b < c && data.Less(pivot, c-1); c-- { // data[c-1] > pivot
+		}
+		if b >= c {
+			break
+		}
+		// data[b] > pivot; data[c-1] <= pivot
+		data.Swap(b, c-1)
+		b++
+		c--
+	}
+	// If hi-c<3 then there are duplicates (by property of median of nine).
+	// Let's be a bit more conservative, and set border to 5.
+	protect := hi-c < 5
+	if !protect && hi-c < (hi-lo)/4 {
+		// Lets test some points for equality to pivot
+		dups := 0
+		if !data.Less(pivot, hi-1) { // data[hi-1] = pivot
+			data.Swap(c, hi-1)
+			c++
+			dups++
+		}
+		if !data.Less(b-1, pivot) { // data[b-1] = pivot
+			b--
+			dups++
+		}
+		// m-lo = (hi-lo)/2 > 6
+		// b-lo > (hi-lo)*3/4-1 > 8
+		// ==> m < b ==> data[m] <= pivot
+		if !data.Less(m, pivot) { // data[m] = pivot
+			data.Swap(m, b-1)
+			b--
+			dups++
+		}
+		// if at least 2 points are equal to pivot, assume skewed distribution
+		protect = dups > 1
+	}
+	if protect {
+		// Protect against a lot of duplicates
+		// Add invariant:
+		//	data[a <= i < b] unexamined
+		//	data[b <= i < c] = pivot
+		for {
+			for ; a < b && !data.Less(b-1, pivot); b-- { // data[b] == pivot
+			}
+			for ; a < b && data.Less(a, pivot); a++ { // data[a] < pivot
+			}
+			if a >= b {
+				break
+			}
+			// data[a] == pivot; data[b-1] < pivot
+			data.Swap(a, b-1)
+			a++
+			b--
+		}
+	}
+	// Swap pivot into middle
+	data.Swap(pivot, b-1)
+	return b - 1, c
+}
+
+func quickSort_func(data lessSwap, a, b, maxDepth int) {
+	for b-a > 12 { // Use ShellSort for slices <= 12 elements
+		if maxDepth == 0 {
+			heapSort_func(data, a, b)
+			return
+		}
+		maxDepth--
+		mlo, mhi := doPivot_func(data, a, b)
+		// Avoiding recursion on the larger subproblem guarantees
+		// a stack depth of at most lg(b-a).
+		if mlo-a < b-mhi {
+			quickSort_func(data, a, mlo, maxDepth)
+			a = mhi // i.e., quickSort_func(data, mhi, b)
+		} else {
+			quickSort_func(data, mhi, b, maxDepth)
+			b = mlo // i.e., quickSort_func(data, a, mlo)
+		}
+	}
+	if b-a > 1 {
+		// Do ShellSort pass with gap 6
+		// It could be written in this simplified form cause b-a <= 12
+		for i := a + 6; i < b; i++ {
+			if data.Less(i, i-6) {
+				data.Swap(i, i-6)
+			}
+		}
+		insertionSort_func(data, a, b)
+	}
+}
+
+func stable_func(data lessSwap, n int) {
+	blockSize := 20 // must be > 0
+	a, b := 0, blockSize
+	for b <= n {
+		insertionSort_func(data, a, b)
+		a = b
+		b += blockSize
+	}
+	insertionSort_func(data, a, n)
+
+	for blockSize < n {
+		a, b = 0, 2*blockSize
+		for b <= n {
+			symMerge_func(data, a, a+blockSize, b)
+			a = b
+			b += 2 * blockSize
+		}
+		if m := a + blockSize; m < n {
+			symMerge_func(data, a, m, n)
+		}
+		blockSize *= 2
+	}
+}
+
+// symMerge_func merges the two sorted subsequences data[a:m] and data[m:b] using
+// the SymMerge algorithm from Pok-Son Kim and Arne Kutzner, "Stable Minimum
+// Storage Merging by Symmetric Comparisons", in Susanne Albers and Tomasz
+// Radzik, editors, Algorithms - ESA 2004, volume 3221 of Lecture Notes in
+// Computer Science, pages 714-723. Springer, 2004.
+//
+// Let M = m-a and N = b-n. Wolog M < N.
+// The recursion depth is bound by ceil(log(N+M)).
+// The algorithm needs O(M*log(N/M + 1)) calls to data.Less.
+// The algorithm needs O((M+N)*log(M)) calls to data.Swap.
+//
+// The paper gives O((M+N)*log(M)) as the number of assignments assuming a
+// rotation algorithm which uses O(M+N+gcd(M+N)) assignments. The argumentation
+// in the paper carries through for Swap operations, especially as the block
+// swapping rotate uses only O(M+N) Swaps.
+//
+// symMerge assumes non-degenerate arguments: a < m && m < b.
+// Having the caller check this condition eliminates many leaf recursion calls,
+// which improves performance.
+func symMerge_func(data lessSwap, a, m, b int) {
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[a] into data[m:b]
+	// if data[a:m] only contains one element.
+	if m-a == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] >= data[a] for m <= i < b.
+		// Exit the search loop with i == b in case no such index exists.
+		i := m
+		j := b
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if data.Less(h, a) {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[a] reaches the position before i.
+		for k := a; k < i-1; k++ {
+			data.Swap(k, k+1)
+		}
+		return
+	}
+
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[m] into data[a:m]
+	// if data[m:b] only contains one element.
+	if b-m == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] > data[m] for a <= i < m.
+		// Exit the search loop with i == m in case no such index exists.
+		i := a
+		j := m
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if !data.Less(m, h) {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[m] reaches the position i.
+		for k := m; k > i; k-- {
+			data.Swap(k, k-1)
+		}
+		return
+	}
+
+	mid := int(uint(a+b) >> 1)
+	n := mid + m
+	var start, r int
+	if m > mid {
+		start = n - b
+		r = mid
+	} else {
+		start = a
+		r = m
+	}
+	p := n - 1
+
+	for start < r {
+		c := int(uint(start+r) >> 1)
+		if !data.Less(p-c, c) {
+			start = c + 1
+		} else {
+			r = c
+		}
+	}
+
+	end := n - start
+	if start < m && m < end {
+		rotate_func(data, start, m, end)
+	}
+	if a < start && start < mid {
+		symMerge_func(data, a, start, mid)
+	}
+	if mid < end && end < b {
+		symMerge_func(data, mid, end, b)
+	}
+}
+
+// rotate_func rotates two consecutive blocks u = data[a:m] and v = data[m:b] in data:
+// Data of the form 'x u v y' is changed to 'x v u y'.
+// rotate performs at most b-a many calls to data.Swap,
+// and it assumes non-degenerate arguments: a < m && m < b.
+func rotate_func(data lessSwap, a, m, b int) {
+	i := m - a
+	j := b - m
+
+	for i != j {
+		if i > j {
+			swapRange_func(data, m-i, m, j)
+			i -= j
+		} else {
+			swapRange_func(data, m-i, m+j-i, i)
+			j -= i
+		}
+	}
+	// i == j
+	swapRange_func(data, m-i, m, i)
+}
diff --git a/src/sort/zsortinterface.go b/src/sort/zsortinterface.go
new file mode 100644
index 0000000000..e0d7093678
--- /dev/null
+++ b/src/sort/zsortinterface.go
@@ -0,0 +1,342 @@
+// Code generated by gen_sort_variants.go; DO NOT EDIT.
+
+// Copyright 2022 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sort
+
+// insertionSort sorts data[a:b] using insertion sort.
+func insertionSort(data Interface, a, b int) {
+	for i := a + 1; i < b; i++ {
+		for j := i; j > a && data.Less(j, j-1); j-- {
+			data.Swap(j, j-1)
+		}
+	}
+}
+
+// siftDown implements the heap property on data[lo:hi].
+// first is an offset into the array where the root of the heap lies.
+func siftDown(data Interface, lo, hi, first int) {
+	root := lo
+	for {
+		child := 2*root + 1
+		if child >= hi {
+			break
+		}
+		if child+1 < hi && data.Less(first+child, first+child+1) {
+			child++
+		}
+		if !data.Less(first+root, first+child) {
+			return
+		}
+		data.Swap(first+root, first+child)
+		root = child
+	}
+}
+
+func heapSort(data Interface, a, b int) {
+	first := a
+	lo := 0
+	hi := b - a
+
+	// Build heap with greatest element at top.
+	for i := (hi - 1) / 2; i >= 0; i-- {
+		siftDown(data, i, hi, first)
+	}
+
+	// Pop elements, largest first, into end of data.
+	for i := hi - 1; i >= 0; i-- {
+		data.Swap(first, first+i)
+		siftDown(data, lo, i, first)
+	}
+}
+
+// Quicksort, loosely following Bentley and McIlroy,
+// "Engineering a Sort Function" SP&E November 1993.
+
+// medianOfThree moves the median of the three values data[m0], data[m1], data[m2] into data[m1].
+func medianOfThree(data Interface, m1, m0, m2 int) {
+	// sort 3 elements
+	if data.Less(m1, m0) {
+		data.Swap(m1, m0)
+	}
+	// data[m0] <= data[m1]
+	if data.Less(m2, m1) {
+		data.Swap(m2, m1)
+		// data[m0] <= data[m2] && data[m1] < data[m2]
+		if data.Less(m1, m0) {
+			data.Swap(m1, m0)
+		}
+	}
+	// now data[m0] <= data[m1] <= data[m2]
+}
+
+func swapRange(data Interface, a, b, n int) {
+	for i := 0; i < n; i++ {
+		data.Swap(a+i, b+i)
+	}
+}
+
+func doPivot(data Interface, lo, hi int) (midlo, midhi int) {
+	m := int(uint(lo+hi) >> 1) // Written like this to avoid integer overflow.
+	if hi-lo > 40 {
+		// Tukey's "Ninther" median of three medians of three.
+		s := (hi - lo) / 8
+		medianOfThree(data, lo, lo+s, lo+2*s)
+		medianOfThree(data, m, m-s, m+s)
+		medianOfThree(data, hi-1, hi-1-s, hi-1-2*s)
+	}
+	medianOfThree(data, lo, m, hi-1)
+
+	// Invariants are:
+	//	data[lo] = pivot (set up by ChoosePivot)
+	//	data[lo < i < a] < pivot
+	//	data[a <= i < b] <= pivot
+	//	data[b <= i < c] unexamined
+	//	data[c <= i < hi-1] > pivot
+	//	data[hi-1] >= pivot
+	pivot := lo
+	a, c := lo+1, hi-1
+
+	for ; a < c && data.Less(a, pivot); a++ {
+	}
+	b := a
+	for {
+		for ; b < c && !data.Less(pivot, b); b++ { // data[b] <= pivot
+		}
+		for ; b < c && data.Less(pivot, c-1); c-- { // data[c-1] > pivot
+		}
+		if b >= c {
+			break
+		}
+		// data[b] > pivot; data[c-1] <= pivot
+		data.Swap(b, c-1)
+		b++
+		c--
+	}
+	// If hi-c<3 then there are duplicates (by property of median of nine).
+	// Let's be a bit more conservative, and set border to 5.
+	protect := hi-c < 5
+	if !protect && hi-c < (hi-lo)/4 {
+		// Lets test some points for equality to pivot
+		dups := 0
+		if !data.Less(pivot, hi-1) { // data[hi-1] = pivot
+			data.Swap(c, hi-1)
+			c++
+			dups++
+		}
+		if !data.Less(b-1, pivot) { // data[b-1] = pivot
+			b--
+			dups++
+		}
+		// m-lo = (hi-lo)/2 > 6
+		// b-lo > (hi-lo)*3/4-1 > 8
+		// ==> m < b ==> data[m] <= pivot
+		if !data.Less(m, pivot) { // data[m] = pivot
+			data.Swap(m, b-1)
+			b--
+			dups++
+		}
+		// if at least 2 points are equal to pivot, assume skewed distribution
+		protect = dups > 1
+	}
+	if protect {
+		// Protect against a lot of duplicates
+		// Add invariant:
+		//	data[a <= i < b] unexamined
+		//	data[b <= i < c] = pivot
+		for {
+			for ; a < b && !data.Less(b-1, pivot); b-- { // data[b] == pivot
+			}
+			for ; a < b && data.Less(a, pivot); a++ { // data[a] < pivot
+			}
+			if a >= b {
+				break
+			}
+			// data[a] == pivot; data[b-1] < pivot
+			data.Swap(a, b-1)
+			a++
+			b--
+		}
+	}
+	// Swap pivot into middle
+	data.Swap(pivot, b-1)
+	return b - 1, c
+}
+
+func quickSort(data Interface, a, b, maxDepth int) {
+	for b-a > 12 { // Use ShellSort for slices <= 12 elements
+		if maxDepth == 0 {
+			heapSort(data, a, b)
+			return
+		}
+		maxDepth--
+		mlo, mhi := doPivot(data, a, b)
+		// Avoiding recursion on the larger subproblem guarantees
+		// a stack depth of at most lg(b-a).
+		if mlo-a < b-mhi {
+			quickSort(data, a, mlo, maxDepth)
+			a = mhi // i.e., quickSort(data, mhi, b)
+		} else {
+			quickSort(data, mhi, b, maxDepth)
+			b = mlo // i.e., quickSort(data, a, mlo)
+		}
+	}
+	if b-a > 1 {
+		// Do ShellSort pass with gap 6
+		// It could be written in this simplified form cause b-a <= 12
+		for i := a + 6; i < b; i++ {
+			if data.Less(i, i-6) {
+				data.Swap(i, i-6)
+			}
+		}
+		insertionSort(data, a, b)
+	}
+}
+
+func stable(data Interface, n int) {
+	blockSize := 20 // must be > 0
+	a, b := 0, blockSize
+	for b <= n {
+		insertionSort(data, a, b)
+		a = b
+		b += blockSize
+	}
+	insertionSort(data, a, n)
+
+	for blockSize < n {
+		a, b = 0, 2*blockSize
+		for b <= n {
+			symMerge(data, a, a+blockSize, b)
+			a = b
+			b += 2 * blockSize
+		}
+		if m := a + blockSize; m < n {
+			symMerge(data, a, m, n)
+		}
+		blockSize *= 2
+	}
+}
+
+// symMerge merges the two sorted subsequences data[a:m] and data[m:b] using
+// the SymMerge algorithm from Pok-Son Kim and Arne Kutzner, "Stable Minimum
+// Storage Merging by Symmetric Comparisons", in Susanne Albers and Tomasz
+// Radzik, editors, Algorithms - ESA 2004, volume 3221 of Lecture Notes in
+// Computer Science, pages 714-723. Springer, 2004.
+//
+// Let M = m-a and N = b-n. Wolog M < N.
+// The recursion depth is bound by ceil(log(N+M)).
+// The algorithm needs O(M*log(N/M + 1)) calls to data.Less.
+// The algorithm needs O((M+N)*log(M)) calls to data.Swap.
+//
+// The paper gives O((M+N)*log(M)) as the number of assignments assuming a
+// rotation algorithm which uses O(M+N+gcd(M+N)) assignments. The argumentation
+// in the paper carries through for Swap operations, especially as the block
+// swapping rotate uses only O(M+N) Swaps.
+//
+// symMerge assumes non-degenerate arguments: a < m && m < b.
+// Having the caller check this condition eliminates many leaf recursion calls,
+// which improves performance.
+func symMerge(data Interface, a, m, b int) {
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[a] into data[m:b]
+	// if data[a:m] only contains one element.
+	if m-a == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] >= data[a] for m <= i < b.
+		// Exit the search loop with i == b in case no such index exists.
+		i := m
+		j := b
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if data.Less(h, a) {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[a] reaches the position before i.
+		for k := a; k < i-1; k++ {
+			data.Swap(k, k+1)
+		}
+		return
+	}
+
+	// Avoid unnecessary recursions of symMerge
+	// by direct insertion of data[m] into data[a:m]
+	// if data[m:b] only contains one element.
+	if b-m == 1 {
+		// Use binary search to find the lowest index i
+		// such that data[i] > data[m] for a <= i < m.
+		// Exit the search loop with i == m in case no such index exists.
+		i := a
+		j := m
+		for i < j {
+			h := int(uint(i+j) >> 1)
+			if !data.Less(m, h) {
+				i = h + 1
+			} else {
+				j = h
+			}
+		}
+		// Swap values until data[m] reaches the position i.
+		for k := m; k > i; k-- {
+			data.Swap(k, k-1)
+		}
+		return
+	}
+
+	mid := int(uint(a+b) >> 1)
+	n := mid + m
+	var start, r int
+	if m > mid {
+		start = n - b
+		r = mid
+	} else {
+		start = a
+		r = m
+	}
+	p := n - 1
+
+	for start < r {
+		c := int(uint(start+r) >> 1)
+		if !data.Less(p-c, c) {
+			start = c + 1
+		} else {
+			r = c
+		}
+	}
+
+	end := n - start
+	if start < m && m < end {
+		rotate(data, start, m, end)
+	}
+	if a < start && start < mid {
+		symMerge(data, a, start, mid)
+	}
+	if mid < end && end < b {
+		symMerge(data, mid, end, b)
+	}
+}
+
+// rotate rotates two consecutive blocks u = data[a:m] and v = data[m:b] in data:
+// Data of the form 'x u v y' is changed to 'x v u y'.
+// rotate performs at most b-a many calls to data.Swap,
+// and it assumes non-degenerate arguments: a < m && m < b.
+func rotate(data Interface, a, m, b int) {
+	i := m - a
+	j := b - m
+
+	for i != j {
+		if i > j {
+			swapRange(data, m-i, m, j)
+			i -= j
+		} else {
+			swapRange(data, m-i, m+j-i, i)
+			j -= i
+		}
+	}
+	// i == j
+	swapRange(data, m-i, m, i)
+}