From aaedf1d5857868150abb0d36f9795a606b8121a8 Mon Sep 17 00:00:00 2001
From: Avril <flanchan@cumallover.me>
Date: Fri, 27 Nov 2020 16:41:13 +0000
Subject: [PATCH] update TODO

---
 .gitignore                      |   1 -
 TODO.org                        |  27 ++
 profiling/debug-flame-old.svg   | 419 +++++++++++++++++++++++
 profiling/debug-flame.svg       | 419 +++++++++++++++++++++++
 profiling/debug-flameu.svg      | 419 +++++++++++++++++++++++
 profiling/massif.out.520307     | 567 ++++++++++++++++++++++++++++++++
 profiling/mem-small.png         | Bin 0 -> 4224 bytes
 profiling/mem-smallu.png        | Bin 0 -> 4560 bytes
 profiling/mem.png               | Bin 0 -> 3243 bytes
 profiling/old-mem-small.png     | Bin 0 -> 3655 bytes
 profiling/old-mem-smallu.png    | Bin 0 -> 4240 bytes
 profiling/old-mem.png           | Bin 0 -> 3048 bytes
 profiling/release-flame-old.svg | 419 +++++++++++++++++++++++
 profiling/release-flame.svg     | 419 +++++++++++++++++++++++
 profiling/release-flameu.svg    | 419 +++++++++++++++++++++++
 15 files changed, 3108 insertions(+), 1 deletion(-)
 create mode 100644 profiling/debug-flame-old.svg
 create mode 100644 profiling/debug-flame.svg
 create mode 100644 profiling/debug-flameu.svg
 create mode 100644 profiling/massif.out.520307
 create mode 100644 profiling/mem-small.png
 create mode 100644 profiling/mem-smallu.png
 create mode 100644 profiling/mem.png
 create mode 100644 profiling/old-mem-small.png
 create mode 100644 profiling/old-mem-smallu.png
 create mode 100644 profiling/old-mem.png
 create mode 100644 profiling/release-flame-old.svg
 create mode 100644 profiling/release-flame.svg
 create mode 100644 profiling/release-flameu.svg

diff --git a/.gitignore b/.gitignore
index 3fa48f6..28eeec4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,4 +3,3 @@ obj
 build/
 test/
 shuffle3-*
-profiling/
diff --git a/TODO.org b/TODO.org
index ad35f41..5d9dd3e 100644
--- a/TODO.org
+++ b/TODO.org
@@ -26,6 +26,33 @@
   - Can properly handle large files without core dumping
   - Doesn't dump huge amounts of trash onto each stack frame
 
+** Performance
+   ~[[https://github.com/sharkdp/hyperfine][hyperfine]]~ reports a *700-800%* speedup over =v1=.
+
+   It's easy to see why.
+*** V1 flamegraph
+    V1 uses a pesudo-array adaptor to perform filesystem reads, seeks, and writes. This causes a massive syscall overhead.
+    [[./profiling/release-flame-old.svg]]
+*** V2 flamegraph
+    Whereas V2 uses a single ~mmap()~.
+    [[./profiling/release-flame.svg]]
+    
+** Memory usage
+   The ~[[https://www.systutorials.com/docs/linux/man/1-memusage/][memusage]]~ graph for =v1= shows extremely inefficient stack usage.
+   [[./profiling/old-mem.png]]
+   ( the green is supposed to be a line, not a bar)
+   This is due to how the unshuffler buffers RNG results.
+
+   =v1= naively used VLAs to store this buffer, which can baloon to 8 times the size of the file being unshuffled.
+   It dumps this massive buffer onto the stack frame of a function that is called multiple times, causing massive and inefficient stack usage.
+
+   This can cause a segfault when attempting to unshuffle a large file, while shuffling a file of the same size might succeed.
+
+*** V2 improvement
+    The ~memusage~ graph for =v2= is a lot more sane.
+    [[./profiling/mem.png]]
+    
+    ~v2~ instead allocates this buffer on the heap. Note the stable stack and heap usage.
 * Todo
   - [X] impl rng
   - [X] impl shuffling
diff --git a/profiling/debug-flame-old.svg b/profiling/debug-flame-old.svg
new file mode 100644
index 0000000..f76d18b
--- /dev/null
+++ b/profiling/debug-flame-old.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="358" onload="init(evt)" viewBox="0 0 1200 358" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="358" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="341.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="341.00"> </text><svg id="frames" x="10" width="1180"><g><title>__GI__IO_free_backup_area (28 samples, 0.30%)</title><rect x="0.1703%" y="261" width="0.2980%" height="15" fill="rgb(227,0,7)"/><text x="0.4203%" y="271.50"></text></g><g><title>__GI__IO_sgetn (30 samples, 0.32%)</title><rect x="0.4683%" y="261" width="0.3193%" height="15" fill="rgb(217,0,24)"/><text x="0.7183%" y="271.50"></text></g><g><title>__GI__IO_switch_to_get_mode (23 samples, 0.24%)</title><rect x="0.7876%" y="261" width="0.2448%" height="15" fill="rgb(221,193,54)"/><text x="1.0376%" y="271.50"></text></g><g><title>__GI___libc_free (14 samples, 0.15%)</title><rect x="1.0324%" y="261" width="0.1490%" height="15" fill="rgb(248,212,6)"/><text x="1.2824%" y="271.50"></text></g><g><title>__lseek64 (85 samples, 0.90%)</title><rect x="1.1814%" y="261" width="0.9046%" height="15" fill="rgb(208,68,35)"/><text x="1.4314%" y="271.50"></text></g><g><title>new_do_write (15 samples, 0.16%)</title><rect x="2.1818%" y="261" width="0.1596%" height="15" fill="rgb(232,128,0)"/><text x="2.4318%" y="271.50"></text></g><g><title>[[heap]] (224 samples, 2.38%)</title><rect x="0.0000%" y="277" width="2.3840%" height="15" fill="rgb(207,160,47)"/><text x="0.2500%" y="287.50">[[..</text></g><g><title>__GI__IO_file_overflow (12 samples, 0.13%)</title><rect x="2.4798%" y="261" width="0.1277%" height="15" fill="rgb(228,23,34)"/><text x="2.7298%" y="271.50"></text></g><g><title>__GI__IO_file_seekoff (30 samples, 0.32%)</title><rect x="2.6075%" y="261" width="0.3193%" height="15" fill="rgb(218,30,26)"/><text x="2.8575%" y="271.50"></text></g><g><title>_ar_file_set (23 samples, 0.24%)</title><rect x="2.9906%" y="261" width="0.2448%" height="15" fill="rgb(220,122,19)"/><text x="3.2406%" y="271.50"></text></g><g><title>[[stack]] (116 samples, 1.23%)</title><rect x="2.3840%" y="277" width="1.2346%" height="15" fill="rgb(250,228,42)"/><text x="2.6340%" y="287.50"></text></g><g><title>_IO_file_xsgetn (31 samples, 0.33%)</title><rect x="3.6186%" y="261" width="0.3299%" height="15" fill="rgb(240,193,28)"/><text x="3.8686%" y="271.50"></text></g><g><title>_IO_seekoff_unlocked (17 samples, 0.18%)</title><rect x="3.9485%" y="261" width="0.1809%" height="15" fill="rgb(216,20,37)"/><text x="4.1985%" y="271.50"></text></g><g><title>[anon] (53 samples, 0.56%)</title><rect x="3.6186%" y="277" width="0.5641%" height="15" fill="rgb(206,188,39)"/><text x="3.8686%" y="287.50"></text></g><g><title>_IO_new_file_xsputn (68 samples, 0.72%)</title><rect x="4.1826%" y="261" width="0.7237%" height="15" fill="rgb(217,207,13)"/><text x="4.4326%" y="271.50"></text></g><g><title>[libc-2.32.so] (72 samples, 0.77%)</title><rect x="4.1826%" y="277" width="0.7663%" height="15" fill="rgb(231,73,38)"/><text x="4.4326%" y="287.50"></text></g><g><title>_IO_file_xsgetn (17 samples, 0.18%)</title><rect x="5.0128%" y="261" width="0.1809%" height="15" fill="rgb(225,20,46)"/><text x="5.2628%" y="271.50"></text></g><g><title>_IO_new_file_write (40 samples, 0.43%)</title><rect x="5.2469%" y="261" width="0.4257%" height="15" fill="rgb(210,31,41)"/><text x="5.4969%" y="271.50"></text></g><g><title>_IO_new_file_xsputn (19 samples, 0.20%)</title><rect x="5.6726%" y="261" width="0.2022%" height="15" fill="rgb(221,200,47)"/><text x="5.9226%" y="271.50"></text></g><g><title>_IO_seekoff_unlocked (16 samples, 0.17%)</title><rect x="5.8748%" y="261" width="0.1703%" height="15" fill="rgb(226,26,5)"/><text x="6.1248%" y="271.50"></text></g><g><title>__GI__IO_default_xsputn (15 samples, 0.16%)</title><rect x="6.0451%" y="261" width="0.1596%" height="15" fill="rgb(249,33,26)"/><text x="6.2951%" y="271.50"></text></g><g><title>__GI__IO_file_overflow (27 samples, 0.29%)</title><rect x="6.2048%" y="261" width="0.2874%" height="15" fill="rgb(235,183,28)"/><text x="6.4548%" y="271.50"></text></g><g><title>__GI__IO_file_seekoff (151 samples, 1.61%)</title><rect x="6.4921%" y="261" width="1.6071%" height="15" fill="rgb(221,5,38)"/><text x="6.7421%" y="271.50"></text></g><g><title>__GI__IO_fread (43 samples, 0.46%)</title><rect x="8.0992%" y="261" width="0.4576%" height="15" fill="rgb(247,18,42)"/><text x="8.3492%" y="271.50"></text></g><g><title>__GI__IO_fwrite (275 samples, 2.93%)</title><rect x="8.5568%" y="261" width="2.9268%" height="15" fill="rgb(241,131,45)"/><text x="8.8068%" y="271.50">__..</text></g><g><title>__GI___libc_malloc (11 samples, 0.12%)</title><rect x="11.6220%" y="261" width="0.1171%" height="15" fill="rgb(249,31,29)"/><text x="11.8720%" y="271.50"></text></g><g><title>__GI___libc_write (48 samples, 0.51%)</title><rect x="11.7390%" y="261" width="0.5109%" height="15" fill="rgb(225,111,53)"/><text x="11.9890%" y="271.50"></text></g><g><title>__GI___read (109 samples, 1.16%)</title><rect x="12.2499%" y="261" width="1.1601%" height="15" fill="rgb(238,160,17)"/><text x="12.4999%" y="271.50"></text></g><g><title>__GI_fseek (40 samples, 0.43%)</title><rect x="13.4100%" y="261" width="0.4257%" height="15" fill="rgb(214,148,48)"/><text x="13.6600%" y="271.50"></text></g><g><title>__lseek64 (119 samples, 1.27%)</title><rect x="13.8357%" y="261" width="1.2665%" height="15" fill="rgb(232,36,49)"/><text x="14.0857%" y="271.50"></text></g><g><title>_int_free (26 samples, 0.28%)</title><rect x="15.1980%" y="261" width="0.2767%" height="15" fill="rgb(209,103,24)"/><text x="15.4480%" y="271.50"></text></g><g><title>ar_set (18 samples, 0.19%)</title><rect x="15.5172%" y="261" width="0.1916%" height="15" fill="rgb(229,88,8)"/><text x="15.7672%" y="271.50"></text></g><g><title>new_do_write (130 samples, 1.38%)</title><rect x="15.7088%" y="261" width="1.3836%" height="15" fill="rgb(213,181,19)"/><text x="15.9588%" y="271.50"></text></g><g><title>[unknown] (1,142 samples, 12.15%)</title><rect x="4.9915%" y="277" width="12.1541%" height="15" fill="rgb(254,191,54)"/><text x="5.2415%" y="287.50">[unknown]</text></g><g><title>_IO_file_xsgetn (11 samples, 0.12%)</title><rect x="17.1456%" y="277" width="0.1171%" height="15" fill="rgb(241,83,37)"/><text x="17.3956%" y="287.50"></text></g><g><title>__GI__IO_default_xsputn (13 samples, 0.14%)</title><rect x="17.2946%" y="277" width="0.1384%" height="15" fill="rgb(233,36,39)"/><text x="17.5446%" y="287.50"></text></g><g><title>_IO_file_xsgetn (12 samples, 0.13%)</title><rect x="18.3802%" y="149" width="0.1277%" height="15" fill="rgb(226,3,54)"/><text x="18.6302%" y="159.50"></text></g><g><title>__GI__IO_fread (26 samples, 0.28%)</title><rect x="18.2418%" y="165" width="0.2767%" height="15" fill="rgb(245,192,40)"/><text x="18.4918%" y="175.50"></text></g><g><title>__GI_fseek (111 samples, 1.18%)</title><rect x="18.6888%" y="165" width="1.1814%" height="15" fill="rgb(238,167,29)"/><text x="18.9388%" y="175.50"></text></g><g><title>__GI__IO_file_seekoff (71 samples, 0.76%)</title><rect x="19.1145%" y="149" width="0.7556%" height="15" fill="rgb(232,182,51)"/><text x="19.3645%" y="159.50"></text></g><g><title>__lseek64 (46 samples, 0.49%)</title><rect x="19.3806%" y="133" width="0.4896%" height="15" fill="rgb(231,60,39)"/><text x="19.6306%" y="143.50"></text></g><g><title>_int_free (13 samples, 0.14%)</title><rect x="19.8702%" y="165" width="0.1384%" height="15" fill="rgb(208,69,12)"/><text x="20.1202%" y="175.50"></text></g><g><title>ar_get (197 samples, 2.10%)</title><rect x="17.9332%" y="197" width="2.0966%" height="15" fill="rgb(235,93,37)"/><text x="18.1832%" y="207.50">a..</text></g><g><title>_ar_file_get (190 samples, 2.02%)</title><rect x="18.0077%" y="181" width="2.0221%" height="15" fill="rgb(213,116,39)"/><text x="18.2577%" y="191.50">_..</text></g><g><title>minmax_floats (214 samples, 2.28%)</title><rect x="17.7948%" y="213" width="2.2776%" height="15" fill="rgb(222,207,29)"/><text x="18.0448%" y="223.50">m..</text></g><g><title>__GI__IO_fread (13 samples, 0.14%)</title><rect x="20.1469%" y="165" width="0.1384%" height="15" fill="rgb(206,96,30)"/><text x="20.3969%" y="175.50"></text></g><g><title>__GI_fseek (29 samples, 0.31%)</title><rect x="20.3172%" y="165" width="0.3086%" height="15" fill="rgb(218,138,4)"/><text x="20.5672%" y="175.50"></text></g><g><title>__GI__IO_file_seekoff (17 samples, 0.18%)</title><rect x="20.4449%" y="149" width="0.1809%" height="15" fill="rgb(250,191,14)"/><text x="20.6949%" y="159.50"></text></g><g><title>__lseek64 (11 samples, 0.12%)</title><rect x="20.5087%" y="133" width="0.1171%" height="15" fill="rgb(239,60,40)"/><text x="20.7587%" y="143.50"></text></g><g><title>minmax_int64_ts (61 samples, 0.65%)</title><rect x="20.0724%" y="213" width="0.6492%" height="15" fill="rgb(206,27,48)"/><text x="20.3224%" y="223.50"></text></g><g><title>ar_get (61 samples, 0.65%)</title><rect x="20.0724%" y="197" width="0.6492%" height="15" fill="rgb(225,35,8)"/><text x="20.3224%" y="207.50"></text></g><g><title>_ar_file_get (56 samples, 0.60%)</title><rect x="20.1256%" y="181" width="0.5960%" height="15" fill="rgb(250,213,24)"/><text x="20.3756%" y="191.50"></text></g><g><title>_IO_file_xsgetn (26 samples, 0.28%)</title><rect x="21.8391%" y="149" width="0.2767%" height="15" fill="rgb(247,123,22)"/><text x="22.0891%" y="159.50"></text></g><g><title>__GI__IO_fread (59 samples, 0.63%)</title><rect x="21.5198%" y="165" width="0.6279%" height="15" fill="rgb(231,138,38)"/><text x="21.7698%" y="175.50"></text></g><g><title>__GI___libc_free (17 samples, 0.18%)</title><rect x="22.1477%" y="165" width="0.1809%" height="15" fill="rgb(231,145,46)"/><text x="22.3977%" y="175.50"></text></g><g><title>__GI___libc_malloc (25 samples, 0.27%)</title><rect x="22.3287%" y="165" width="0.2661%" height="15" fill="rgb(251,118,11)"/><text x="22.5787%" y="175.50"></text></g><g><title>_IO_seekoff_unlocked (20 samples, 0.21%)</title><rect x="23.7654%" y="149" width="0.2129%" height="15" fill="rgb(217,147,25)"/><text x="24.0154%" y="159.50"></text></g><g><title>__GI__IO_free_backup_area (15 samples, 0.16%)</title><rect x="24.7765%" y="133" width="0.1596%" height="15" fill="rgb(247,81,37)"/><text x="25.0265%" y="143.50"></text></g><g><title>__GI_fseek (314 samples, 3.34%)</title><rect x="22.5947%" y="165" width="3.3418%" height="15" fill="rgb(209,12,38)"/><text x="22.8447%" y="175.50">__G..</text></g><g><title>__GI__IO_file_seekoff (184 samples, 1.96%)</title><rect x="23.9783%" y="149" width="1.9583%" height="15" fill="rgb(227,1,9)"/><text x="24.2283%" y="159.50">_..</text></g><g><title>__lseek64 (94 samples, 1.00%)</title><rect x="24.9361%" y="133" width="1.0004%" height="15" fill="rgb(248,47,43)"/><text x="25.1861%" y="143.50"></text></g><g><title>_int_free (32 samples, 0.34%)</title><rect x="25.9472%" y="165" width="0.3406%" height="15" fill="rgb(221,10,30)"/><text x="26.1972%" y="175.50"></text></g><g><title>ar_get (519 samples, 5.52%)</title><rect x="20.9132%" y="197" width="5.5236%" height="15" fill="rgb(210,229,1)"/><text x="21.1632%" y="207.50">ar_get</text></g><g><title>_ar_file_get (503 samples, 5.35%)</title><rect x="21.0834%" y="181" width="5.3533%" height="15" fill="rgb(222,148,37)"/><text x="21.3334%" y="191.50">_ar_fil..</text></g><g><title>minmax_sbytes (548 samples, 5.83%)</title><rect x="20.7216%" y="213" width="5.8323%" height="15" fill="rgb(234,67,33)"/><text x="20.9716%" y="223.50">minmax_..</text></g><g><title>ar_size (11 samples, 0.12%)</title><rect x="26.4368%" y="197" width="0.1171%" height="15" fill="rgb(247,98,35)"/><text x="26.6868%" y="207.50"></text></g><g><title>ar_set (36 samples, 0.38%)</title><rect x="27.4691%" y="197" width="0.3831%" height="15" fill="rgb(247,138,52)"/><text x="27.7191%" y="207.50"></text></g><g><title>_IO_file_xsgetn (47 samples, 0.50%)</title><rect x="31.6837%" y="133" width="0.5002%" height="15" fill="rgb(213,79,30)"/><text x="31.9337%" y="143.50"></text></g><g><title>__memmove_avx_unaligned_erms (14 samples, 0.15%)</title><rect x="32.0349%" y="117" width="0.1490%" height="15" fill="rgb(246,177,23)"/><text x="32.2849%" y="127.50"></text></g><g><title>__GI__IO_fread (118 samples, 1.26%)</title><rect x="31.0345%" y="149" width="1.2559%" height="15" fill="rgb(230,62,27)"/><text x="31.2845%" y="159.50"></text></g><g><title>__GI__IO_sgetn (10 samples, 0.11%)</title><rect x="32.1839%" y="133" width="0.1064%" height="15" fill="rgb(216,154,8)"/><text x="32.4339%" y="143.50"></text></g><g><title>__GI___libc_free (58 samples, 0.62%)</title><rect x="32.2903%" y="149" width="0.6173%" height="15" fill="rgb(244,35,45)"/><text x="32.5403%" y="159.50"></text></g><g><title>__GI___libc_malloc (32 samples, 0.34%)</title><rect x="32.9076%" y="149" width="0.3406%" height="15" fill="rgb(251,115,12)"/><text x="33.1576%" y="159.50"></text></g><g><title>_IO_seekoff_unlocked (20 samples, 0.21%)</title><rect x="33.9506%" y="133" width="0.2129%" height="15" fill="rgb(240,54,50)"/><text x="34.2006%" y="143.50"></text></g><g><title>__GI__IO_free_backup_area (17 samples, 0.18%)</title><rect x="36.2920%" y="117" width="0.1809%" height="15" fill="rgb(233,84,52)"/><text x="36.5420%" y="127.50"></text></g><g><title>__GI___read (159 samples, 1.69%)</title><rect x="36.4730%" y="117" width="1.6922%" height="15" fill="rgb(207,117,47)"/><text x="36.7230%" y="127.50"></text></g><g><title>__GI_fseek (622 samples, 6.62%)</title><rect x="33.2482%" y="149" width="6.6198%" height="15" fill="rgb(249,43,39)"/><text x="33.4982%" y="159.50">__GI_fseek</text></g><g><title>__GI__IO_file_seekoff (536 samples, 5.70%)</title><rect x="34.1635%" y="133" width="5.7046%" height="15" fill="rgb(209,38,44)"/><text x="34.4135%" y="143.50">__GI__I..</text></g><g><title>__lseek64 (160 samples, 1.70%)</title><rect x="38.1652%" y="117" width="1.7029%" height="15" fill="rgb(236,212,23)"/><text x="38.4152%" y="127.50"></text></g><g><title>_int_free (62 samples, 0.66%)</title><rect x="39.9425%" y="149" width="0.6599%" height="15" fill="rgb(242,79,21)"/><text x="40.1925%" y="159.50"></text></g><g><title>ar_get (1,054 samples, 11.22%)</title><rect x="29.5658%" y="181" width="11.2175%" height="15" fill="rgb(211,96,35)"/><text x="29.8158%" y="191.50">ar_get</text></g><g><title>_ar_file_get (1,023 samples, 10.89%)</title><rect x="29.8957%" y="165" width="10.8876%" height="15" fill="rgb(253,215,40)"/><text x="30.1457%" y="175.50">_ar_file_get</text></g><g><title>__memmove_avx_unaligned_erms (20 samples, 0.21%)</title><rect x="41.4645%" y="165" width="0.2129%" height="15" fill="rgb(211,81,21)"/><text x="41.7145%" y="175.50"></text></g><g><title>_IO_file_xsgetn (44 samples, 0.47%)</title><rect x="44.4977%" y="117" width="0.4683%" height="15" fill="rgb(208,190,38)"/><text x="44.7477%" y="127.50"></text></g><g><title>__memmove_avx_unaligned_erms (11 samples, 0.12%)</title><rect x="44.8489%" y="101" width="0.1171%" height="15" fill="rgb(235,213,38)"/><text x="45.0989%" y="111.50"></text></g><g><title>__GI__IO_fread (132 samples, 1.40%)</title><rect x="43.7527%" y="133" width="1.4049%" height="15" fill="rgb(237,122,38)"/><text x="44.0027%" y="143.50"></text></g><g><title>__GI__IO_sgetn (18 samples, 0.19%)</title><rect x="44.9659%" y="117" width="0.1916%" height="15" fill="rgb(244,218,35)"/><text x="45.2159%" y="127.50"></text></g><g><title>__GI___libc_free (61 samples, 0.65%)</title><rect x="45.1575%" y="133" width="0.6492%" height="15" fill="rgb(240,68,47)"/><text x="45.4075%" y="143.50"></text></g><g><title>__GI___libc_malloc (28 samples, 0.30%)</title><rect x="45.8067%" y="133" width="0.2980%" height="15" fill="rgb(210,16,53)"/><text x="46.0567%" y="143.50"></text></g><g><title>_IO_seekoff_unlocked (20 samples, 0.21%)</title><rect x="46.6794%" y="117" width="0.2129%" height="15" fill="rgb(235,124,12)"/><text x="46.9294%" y="127.50"></text></g><g><title>[libc-2.32.so] (21 samples, 0.22%)</title><rect x="49.1166%" y="85" width="0.2235%" height="15" fill="rgb(224,169,11)"/><text x="49.3666%" y="95.50"></text></g><g><title>__GI__IO_free_backup_area (88 samples, 0.94%)</title><rect x="49.0528%" y="101" width="0.9366%" height="15" fill="rgb(250,166,2)"/><text x="49.3028%" y="111.50"></text></g><g><title>__GI___libc_free (61 samples, 0.65%)</title><rect x="49.3401%" y="85" width="0.6492%" height="15" fill="rgb(242,216,29)"/><text x="49.5901%" y="95.50"></text></g><g><title>_IO_new_file_write (245 samples, 2.61%)</title><rect x="51.1814%" y="53" width="2.6075%" height="15" fill="rgb(230,116,27)"/><text x="51.4314%" y="63.50">_I..</text></g><g><title>__GI___libc_write (147 samples, 1.56%)</title><rect x="52.2244%" y="37" width="1.5645%" height="15" fill="rgb(228,99,48)"/><text x="52.4744%" y="47.50"></text></g><g><title>_IO_new_do_write (477 samples, 5.08%)</title><rect x="50.3193%" y="85" width="5.0766%" height="15" fill="rgb(253,11,6)"/><text x="50.5693%" y="95.50">_IO_ne..</text></g><g><title>new_do_write (472 samples, 5.02%)</title><rect x="50.3725%" y="69" width="5.0234%" height="15" fill="rgb(247,143,39)"/><text x="50.6225%" y="79.50">new_do..</text></g><g><title>__lseek64 (149 samples, 1.59%)</title><rect x="53.8101%" y="53" width="1.5858%" height="15" fill="rgb(236,97,10)"/><text x="54.0601%" y="63.50"></text></g><g><title>__GI__IO_switch_to_get_mode (519 samples, 5.52%)</title><rect x="49.9894%" y="101" width="5.5236%" height="15" fill="rgb(233,208,19)"/><text x="50.2394%" y="111.50">__GI__I..</text></g><g><title>__GI__IO_file_overflow (11 samples, 0.12%)</title><rect x="55.3959%" y="85" width="0.1171%" height="15" fill="rgb(216,164,2)"/><text x="55.6459%" y="95.50"></text></g><g><title>__GI___read (153 samples, 1.63%)</title><rect x="55.5130%" y="101" width="1.6284%" height="15" fill="rgb(220,129,5)"/><text x="55.7630%" y="111.50"></text></g><g><title>__GI_fseek (1,166 samples, 12.41%)</title><rect x="46.1047%" y="133" width="12.4095%" height="15" fill="rgb(242,17,10)"/><text x="46.3547%" y="143.50">__GI_fseek</text></g><g><title>__GI__IO_file_seekoff (1,092 samples, 11.62%)</title><rect x="46.8923%" y="117" width="11.6220%" height="15" fill="rgb(242,107,0)"/><text x="47.1423%" y="127.50">__GI__IO_file_see..</text></g><g><title>__lseek64 (129 samples, 1.37%)</title><rect x="57.1413%" y="101" width="1.3729%" height="15" fill="rgb(251,28,31)"/><text x="57.3913%" y="111.50"></text></g><g><title>_init (45 samples, 0.48%)</title><rect x="58.5143%" y="133" width="0.4789%" height="15" fill="rgb(233,223,10)"/><text x="58.7643%" y="143.50"></text></g><g><title>_int_free (74 samples, 0.79%)</title><rect x="58.9932%" y="133" width="0.7876%" height="15" fill="rgb(215,21,27)"/><text x="59.2432%" y="143.50"></text></g><g><title>ar_get (1,715 samples, 18.25%)</title><rect x="41.7092%" y="165" width="18.2524%" height="15" fill="rgb(232,23,21)"/><text x="41.9592%" y="175.50">ar_get</text></g><g><title>_ar_file_get (1,687 samples, 17.95%)</title><rect x="42.0072%" y="149" width="17.9544%" height="15" fill="rgb(244,5,23)"/><text x="42.2572%" y="159.50">_ar_file_get</text></g><g><title>ar_get_into (1,803 samples, 19.19%)</title><rect x="40.7833%" y="181" width="19.1890%" height="15" fill="rgb(226,81,46)"/><text x="41.0333%" y="191.50">ar_get_into</text></g><g><title>_IO_new_do_write (10 samples, 0.11%)</title><rect x="71.0515%" y="117" width="0.1064%" height="15" fill="rgb(247,70,30)"/><text x="71.3015%" y="127.50"></text></g><g><title>__GI__IO_default_xsputn (104 samples, 1.11%)</title><rect x="71.1579%" y="117" width="1.1069%" height="15" fill="rgb(212,68,19)"/><text x="71.4079%" y="127.50"></text></g><g><title>__GI__IO_fwrite (974 samples, 10.37%)</title><rect x="62.7927%" y="149" width="10.3661%" height="15" fill="rgb(240,187,13)"/><text x="63.0427%" y="159.50">__GI__IO_fwrite</text></g><g><title>_IO_new_file_xsputn (566 samples, 6.02%)</title><rect x="67.1350%" y="133" width="6.0238%" height="15" fill="rgb(223,113,26)"/><text x="67.3850%" y="143.50">_IO_new_..</text></g><g><title>__GI__IO_file_overflow (84 samples, 0.89%)</title><rect x="72.2648%" y="117" width="0.8940%" height="15" fill="rgb(206,192,2)"/><text x="72.5148%" y="127.50"></text></g><g><title>_IO_seekoff_unlocked (44 samples, 0.47%)</title><rect x="74.5743%" y="133" width="0.4683%" height="15" fill="rgb(241,108,4)"/><text x="74.8243%" y="143.50"></text></g><g><title>__GI__IO_file_read (12 samples, 0.13%)</title><rect x="79.3316%" y="117" width="0.1277%" height="15" fill="rgb(247,173,49)"/><text x="79.5816%" y="127.50"></text></g><g><title>[libc-2.32.so] (25 samples, 0.27%)</title><rect x="79.6828%" y="101" width="0.2661%" height="15" fill="rgb(224,114,35)"/><text x="79.9328%" y="111.50"></text></g><g><title>__GI__IO_free_backup_area (123 samples, 1.31%)</title><rect x="79.5445%" y="117" width="1.3091%" height="15" fill="rgb(245,159,27)"/><text x="79.7945%" y="127.50"></text></g><g><title>__GI___libc_free (85 samples, 0.90%)</title><rect x="79.9489%" y="101" width="0.9046%" height="15" fill="rgb(245,172,44)"/><text x="80.1989%" y="111.50"></text></g><g><title>_IO_new_file_write (239 samples, 2.54%)</title><rect x="81.9817%" y="69" width="2.5436%" height="15" fill="rgb(236,23,11)"/><text x="82.2317%" y="79.50">_I..</text></g><g><title>__GI___libc_write (146 samples, 1.55%)</title><rect x="82.9715%" y="53" width="1.5539%" height="15" fill="rgb(205,117,38)"/><text x="83.2215%" y="63.50"></text></g><g><title>_IO_new_do_write (492 samples, 5.24%)</title><rect x="81.0451%" y="101" width="5.2363%" height="15" fill="rgb(237,72,25)"/><text x="81.2951%" y="111.50">_IO_ne..</text></g><g><title>new_do_write (482 samples, 5.13%)</title><rect x="81.1516%" y="85" width="5.1298%" height="15" fill="rgb(244,70,9)"/><text x="81.4016%" y="95.50">new_do..</text></g><g><title>__lseek64 (162 samples, 1.72%)</title><rect x="84.5573%" y="69" width="1.7241%" height="15" fill="rgb(217,125,39)"/><text x="84.8073%" y="79.50"></text></g><g><title>__GI__IO_switch_to_get_mode (524 samples, 5.58%)</title><rect x="80.8536%" y="117" width="5.5768%" height="15" fill="rgb(235,36,10)"/><text x="81.1036%" y="127.50">__GI__I..</text></g><g><title>__GI__IO_file_overflow (14 samples, 0.15%)</title><rect x="86.2814%" y="101" width="0.1490%" height="15" fill="rgb(251,123,47)"/><text x="86.5314%" y="111.50"></text></g><g><title>__GI___read (294 samples, 3.13%)</title><rect x="86.4304%" y="117" width="3.1290%" height="15" fill="rgb(221,13,13)"/><text x="86.6804%" y="127.50">__G..</text></g><g><title>__GI_fseek (1,837 samples, 19.55%)</title><rect x="73.1588%" y="149" width="19.5509%" height="15" fill="rgb(238,131,9)"/><text x="73.4088%" y="159.50">__GI_fseek</text></g><g><title>__GI__IO_file_seekoff (1,660 samples, 17.67%)</title><rect x="75.0426%" y="133" width="17.6671%" height="15" fill="rgb(211,50,8)"/><text x="75.2926%" y="143.50">__GI__IO_file_seekoff</text></g><g><title>__lseek64 (296 samples, 3.15%)</title><rect x="89.5594%" y="117" width="3.1503%" height="15" fill="rgb(245,182,24)"/><text x="89.8094%" y="127.50">__l..</text></g><g><title>ar_swap (6,142 samples, 65.37%)</title><rect x="27.8523%" y="197" width="65.3682%" height="15" fill="rgb(242,14,37)"/><text x="28.1023%" y="207.50">ar_swap</text></g><g><title>ar_set (3,124 samples, 33.25%)</title><rect x="59.9723%" y="181" width="33.2482%" height="15" fill="rgb(246,228,12)"/><text x="60.2223%" y="191.50">ar_set</text></g><g><title>_ar_file_set (3,031 samples, 32.26%)</title><rect x="60.9621%" y="165" width="32.2584%" height="15" fill="rgb(213,55,15)"/><text x="61.2121%" y="175.50">_ar_file_set</text></g><g><title>fwrite@plt (41 samples, 0.44%)</title><rect x="92.7842%" y="149" width="0.4364%" height="15" fill="rgb(209,9,3)"/><text x="93.0342%" y="159.50"></text></g><g><title>[libm-2.32.so] (18 samples, 0.19%)</title><rect x="94.3806%" y="181" width="0.1916%" height="15" fill="rgb(230,59,30)"/><text x="94.6306%" y="191.50"></text></g><g><title>rand (89 samples, 0.95%)</title><rect x="95.3491%" y="165" width="0.9472%" height="15" fill="rgb(209,121,21)"/><text x="95.5991%" y="175.50"></text></g><g><title>__random (89 samples, 0.95%)</title><rect x="95.3491%" y="149" width="0.9472%" height="15" fill="rgb(220,109,13)"/><text x="95.5991%" y="159.50"></text></g><g><title>__random_r (60 samples, 0.64%)</title><rect x="95.6577%" y="133" width="0.6386%" height="15" fill="rgb(232,18,1)"/><text x="95.9077%" y="143.50"></text></g><g><title>_next (164 samples, 1.75%)</title><rect x="94.5722%" y="181" width="1.7454%" height="15" fill="rgb(215,41,42)"/><text x="94.8222%" y="191.50"></text></g><g><title>[libm-2.32.so] (87 samples, 0.93%)</title><rect x="97.1584%" y="133" width="0.9259%" height="15" fill="rgb(224,123,36)"/><text x="97.4084%" y="143.50"></text></g><g><title>dot (20 samples, 0.21%)</title><rect x="98.0843%" y="133" width="0.2129%" height="15" fill="rgb(240,125,3)"/><text x="98.3343%" y="143.50"></text></g><g><title>fract (24 samples, 0.26%)</title><rect x="98.2971%" y="133" width="0.2554%" height="15" fill="rgb(205,98,50)"/><text x="98.5471%" y="143.50"></text></g><g><title>next (205 samples, 2.18%)</title><rect x="96.3814%" y="165" width="2.1818%" height="15" fill="rgb(205,185,37)"/><text x="96.6314%" y="175.50">n..</text></g><g><title>sample_double (165 samples, 1.76%)</title><rect x="96.8072%" y="149" width="1.7561%" height="15" fill="rgb(238,207,15)"/><text x="97.0572%" y="159.50"></text></g><g><title>[libm-2.32.so] (64 samples, 0.68%)</title><rect x="98.7974%" y="149" width="0.6811%" height="15" fill="rgb(213,199,42)"/><text x="99.0474%" y="159.50"></text></g><g><title>dot (27 samples, 0.29%)</title><rect x="99.4785%" y="149" width="0.2874%" height="15" fill="rgb(235,201,11)"/><text x="99.7285%" y="159.50"></text></g><g><title>fract (16 samples, 0.17%)</title><rect x="99.7659%" y="149" width="0.1703%" height="15" fill="rgb(207,46,11)"/><text x="100.0159%" y="159.50"></text></g><g><title>_sample (341 samples, 3.63%)</title><rect x="96.3176%" y="181" width="3.6292%" height="15" fill="rgb(241,35,35)"/><text x="96.5676%" y="191.50">_sam..</text></g><g><title>sample_double (130 samples, 1.38%)</title><rect x="98.5632%" y="165" width="1.3836%" height="15" fill="rgb(243,32,47)"/><text x="98.8132%" y="175.50"></text></g><g><title>shuffle (6,898 samples, 73.41%)</title><rect x="26.5751%" y="213" width="73.4142%" height="15" fill="rgb(247,202,23)"/><text x="26.8251%" y="223.50">shuffle</text></g><g><title>rng_next_int (636 samples, 6.77%)</title><rect x="93.2205%" y="197" width="6.7688%" height="15" fill="rgb(219,102,11)"/><text x="93.4705%" y="207.50">rng_next_..</text></g><g><title>all (9,396 samples, 100%)</title><rect x="0.0000%" y="309" width="100.0000%" height="15" fill="rgb(243,110,44)"/><text x="0.2500%" y="319.50"></text></g><g><title>shuffle3-debug (9,396 samples, 100.00%)</title><rect x="0.0000%" y="293" width="100.0000%" height="15" fill="rgb(222,74,54)"/><text x="0.2500%" y="303.50">shuffle3-debug</text></g><g><title>_start (7,735 samples, 82.32%)</title><rect x="17.6777%" y="277" width="82.3223%" height="15" fill="rgb(216,99,12)"/><text x="17.9277%" y="287.50">_start</text></g><g><title>__libc_start_main (7,731 samples, 82.28%)</title><rect x="17.7203%" y="261" width="82.2797%" height="15" fill="rgb(226,22,26)"/><text x="17.9703%" y="271.50">__libc_start_main</text></g><g><title>main (7,731 samples, 82.28%)</title><rect x="17.7203%" y="245" width="82.2797%" height="15" fill="rgb(217,163,10)"/><text x="17.9703%" y="255.50">main</text></g><g><title>shuffle3 (7,731 samples, 82.28%)</title><rect x="17.7203%" y="229" width="82.2797%" height="15" fill="rgb(213,25,53)"/><text x="17.9703%" y="239.50">shuffle3</text></g></svg></svg>
\ No newline at end of file
diff --git a/profiling/debug-flame.svg b/profiling/debug-flame.svg
new file mode 100644
index 0000000..fd3789c
--- /dev/null
+++ b/profiling/debug-flame.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="326" onload="init(evt)" viewBox="0 0 1200 326" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="326" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="309.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="309.00"> </text><svg id="frames" x="10" width="1180"><g><title>[[stack]] (1 samples, 0.75%)</title><rect x="0.0000%" y="245" width="0.7519%" height="15" fill="rgb(227,0,7)"/><text x="0.2500%" y="255.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (1 samples, 0.75%)</title><rect x="0.0000%" y="229" width="0.7519%" height="15" fill="rgb(217,0,24)"/><text x="0.2500%" y="239.50"></text></g><g><title>RNG::next_long (2 samples, 1.50%)</title><rect x="0.7519%" y="229" width="1.5038%" height="15" fill="rgb(221,193,54)"/><text x="1.0019%" y="239.50"></text></g><g><title>[unknown] (3 samples, 2.26%)</title><rect x="0.7519%" y="245" width="2.2556%" height="15" fill="rgb(248,212,6)"/><text x="1.0019%" y="255.50">[..</text></g><g><title>rng::drng::sample (1 samples, 0.75%)</title><rect x="2.2556%" y="229" width="0.7519%" height="15" fill="rgb(208,68,35)"/><text x="2.5056%" y="239.50"></text></g><g><title>minmax_t&lt;long&gt; (1 samples, 0.75%)</title><rect x="6.0150%" y="165" width="0.7519%" height="15" fill="rgb(232,128,0)"/><text x="6.2650%" y="175.50"></text></g><g><title>minmax_t&lt;long, minmax_t&lt;long&gt;(span&lt;long&gt; const&amp;)::{lambda(long)#1}&gt; (1 samples, 0.75%)</title><rect x="6.0150%" y="149" width="0.7519%" height="15" fill="rgb(207,160,47)"/><text x="6.2650%" y="159.50"></text></g><g><title>span&lt;long&gt;::operator[] (1 samples, 0.75%)</title><rect x="6.0150%" y="133" width="0.7519%" height="15" fill="rgb(228,23,34)"/><text x="6.2650%" y="143.50"></text></g><g><title>minmax_t&lt;signed char&gt; (1 samples, 0.75%)</title><rect x="11.2782%" y="133" width="0.7519%" height="15" fill="rgb(218,30,26)"/><text x="11.5282%" y="143.50"></text></g><g><title>minmax_t&lt;signed char&gt; (9 samples, 6.77%)</title><rect x="6.7669%" y="165" width="6.7669%" height="15" fill="rgb(220,122,19)"/><text x="7.0169%" y="175.50">minmax_t&lt;..</text></g><g><title>minmax_t&lt;signed char, minmax_t&lt;signed char&gt;(span&lt;signed char&gt; const&amp;)::{lambda(signed char)#1}&gt; (9 samples, 6.77%)</title><rect x="6.7669%" y="149" width="6.7669%" height="15" fill="rgb(250,228,42)"/><text x="7.0169%" y="159.50">minmax_t&lt;..</text></g><g><title>span&lt;signed char&gt;::operator[] (2 samples, 1.50%)</title><rect x="12.0301%" y="133" width="1.5038%" height="15" fill="rgb(240,193,28)"/><text x="12.2801%" y="143.50"></text></g><g><title>[libm-2.32.so] (1 samples, 0.75%)</title><rect x="15.0376%" y="117" width="0.7519%" height="15" fill="rgb(216,20,37)"/><text x="15.2876%" y="127.50"></text></g><g><title>[libm-2.32.so] (9 samples, 6.77%)</title><rect x="15.7895%" y="85" width="6.7669%" height="15" fill="rgb(206,188,39)"/><text x="16.0395%" y="95.50">[libm-2.3..</text></g><g><title>rng::frng::sample_double (14 samples, 10.53%)</title><rect x="15.7895%" y="101" width="10.5263%" height="15" fill="rgb(217,207,13)"/><text x="16.0395%" y="111.50">rng::frng::samp..</text></g><g><title>rng::frng::dot&lt;2ul&gt; (5 samples, 3.76%)</title><rect x="22.5564%" y="85" width="3.7594%" height="15" fill="rgb(231,73,38)"/><text x="22.8064%" y="95.50">rng:..</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (2 samples, 1.50%)</title><rect x="24.8120%" y="69" width="1.5038%" height="15" fill="rgb(225,20,46)"/><text x="25.0620%" y="79.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (2 samples, 1.50%)</title><rect x="24.8120%" y="53" width="1.5038%" height="15" fill="rgb(210,31,41)"/><text x="25.0620%" y="63.50"></text></g><g><title>[libm-2.32.so] (16 samples, 12.03%)</title><rect x="30.8271%" y="69" width="12.0301%" height="15" fill="rgb(221,200,47)"/><text x="31.0771%" y="79.50">[libm-2.32.so]</text></g><g><title>rng::frng::dot&lt;2ul&gt; (14 samples, 10.53%)</title><rect x="42.8571%" y="69" width="10.5263%" height="15" fill="rgb(226,26,5)"/><text x="43.1071%" y="79.50">rng::frng::dot&lt;..</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (8 samples, 6.02%)</title><rect x="47.3684%" y="53" width="6.0150%" height="15" fill="rgb(249,33,26)"/><text x="47.6184%" y="63.50">std::arr..</text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (4 samples, 3.01%)</title><rect x="50.3759%" y="37" width="3.0075%" height="15" fill="rgb(235,183,28)"/><text x="50.6259%" y="47.50">std..</text></g><g><title>rng::frng::sample_double (36 samples, 27.07%)</title><rect x="28.5714%" y="85" width="27.0677%" height="15" fill="rgb(221,5,38)"/><text x="28.8214%" y="95.50">rng::frng::sample_double</text></g><g><title>rng::frng::fract (3 samples, 2.26%)</title><rect x="53.3835%" y="69" width="2.2556%" height="15" fill="rgb(247,18,42)"/><text x="53.6335%" y="79.50">r..</text></g><g><title>[libm-2.32.so] (2 samples, 1.50%)</title><rect x="54.1353%" y="53" width="1.5038%" height="15" fill="rgb(241,131,45)"/><text x="54.3853%" y="63.50"></text></g><g><title>RNG::next_long (58 samples, 43.61%)</title><rect x="13.5338%" y="149" width="43.6090%" height="15" fill="rgb(249,31,29)"/><text x="13.7838%" y="159.50">RNG::next_long</text></g><g><title>RNG::next_long (57 samples, 42.86%)</title><rect x="14.2857%" y="133" width="42.8571%" height="15" fill="rgb(225,111,53)"/><text x="14.5357%" y="143.50">RNG::next_long</text></g><g><title>rng::frng::sample (55 samples, 41.35%)</title><rect x="15.7895%" y="117" width="41.3534%" height="15" fill="rgb(238,160,17)"/><text x="16.0395%" y="127.50">rng::frng::sample</text></g><g><title>rng::frng::update_state (41 samples, 30.83%)</title><rect x="26.3158%" y="101" width="30.8271%" height="15" fill="rgb(214,148,48)"/><text x="26.5658%" y="111.50">rng::frng::update_state</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (2 samples, 1.50%)</title><rect x="55.6391%" y="85" width="1.5038%" height="15" fill="rgb(232,36,49)"/><text x="55.8891%" y="95.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (2 samples, 1.50%)</title><rect x="55.6391%" y="69" width="1.5038%" height="15" fill="rgb(209,103,24)"/><text x="55.8891%" y="79.50"></text></g><g><title>rng::shuffle&lt;float, rng::frng&gt; (65 samples, 48.87%)</title><rect x="13.5338%" y="165" width="48.8722%" height="15" fill="rgb(229,88,8)"/><text x="13.7838%" y="175.50">rng::shuffle&lt;float, rng::frng&gt;</text></g><g><title>std::swap&lt;float&gt; (7 samples, 5.26%)</title><rect x="57.1429%" y="149" width="5.2632%" height="15" fill="rgb(213,181,19)"/><text x="57.3929%" y="159.50">std::s..</text></g><g><title>RNG::next_long (10 samples, 7.52%)</title><rect x="65.4135%" y="133" width="7.5188%" height="15" fill="rgb(254,191,54)"/><text x="65.6635%" y="143.50">RNG::next_..</text></g><g><title>rng::drng::sample (5 samples, 3.76%)</title><rect x="69.1729%" y="117" width="3.7594%" height="15" fill="rgb(241,83,37)"/><text x="69.4229%" y="127.50">rng:..</text></g><g><title>rand_r (2 samples, 1.50%)</title><rect x="71.4286%" y="101" width="1.5038%" height="15" fill="rgb(233,36,39)"/><text x="71.6786%" y="111.50"></text></g><g><title>RNG::next_long (11 samples, 8.27%)</title><rect x="65.4135%" y="149" width="8.2707%" height="15" fill="rgb(226,3,54)"/><text x="65.6635%" y="159.50">RNG::next_lo..</text></g><g><title>rng::drng::sample (1 samples, 0.75%)</title><rect x="72.9323%" y="133" width="0.7519%" height="15" fill="rgb(245,192,40)"/><text x="73.1823%" y="143.50"></text></g><g><title>span&lt;unsigned char&gt;::operator[] (3 samples, 2.26%)</title><rect x="73.6842%" y="149" width="2.2556%" height="15" fill="rgb(238,167,29)"/><text x="73.9342%" y="159.50">s..</text></g><g><title>rng::shuffle&lt;unsigned char, rng::drng&gt; (47 samples, 35.34%)</title><rect x="62.4060%" y="165" width="35.3383%" height="15" fill="rgb(232,182,51)"/><text x="62.6560%" y="175.50">rng::shuffle&lt;unsigned char, rng::drng&gt;</text></g><g><title>std::swap&lt;unsigned char&gt; (29 samples, 21.80%)</title><rect x="75.9398%" y="149" width="21.8045%" height="15" fill="rgb(231,60,39)"/><text x="76.1898%" y="159.50">std::swap&lt;unsigned char&gt;</text></g><g><title>__libc_start_main (123 samples, 92.48%)</title><rect x="6.0150%" y="229" width="92.4812%" height="15" fill="rgb(208,69,12)"/><text x="6.2650%" y="239.50">__libc_start_main</text></g><g><title>main (123 samples, 92.48%)</title><rect x="6.0150%" y="213" width="92.4812%" height="15" fill="rgb(235,93,37)"/><text x="6.2650%" y="223.50">main</text></g><g><title>do_work (123 samples, 92.48%)</title><rect x="6.0150%" y="197" width="92.4812%" height="15" fill="rgb(213,116,39)"/><text x="6.2650%" y="207.50">do_work</text></g><g><title>work::xshuffle_ip&lt;false&gt; (123 samples, 92.48%)</title><rect x="6.0150%" y="181" width="92.4812%" height="15" fill="rgb(222,207,29)"/><text x="6.2650%" y="191.50">work::xshuffle_ip&lt;false&gt;</text></g><g><title>std::swap&lt;unsigned char&gt; (1 samples, 0.75%)</title><rect x="97.7444%" y="165" width="0.7519%" height="15" fill="rgb(206,96,30)"/><text x="97.9944%" y="175.50"></text></g><g><title>_start (128 samples, 96.24%)</title><rect x="3.0075%" y="245" width="96.2406%" height="15" fill="rgb(218,138,4)"/><text x="3.2575%" y="255.50">_start</text></g><g><title>_dl_start (1 samples, 0.75%)</title><rect x="98.4962%" y="229" width="0.7519%" height="15" fill="rgb(250,191,14)"/><text x="98.7462%" y="239.50"></text></g><g><title>_dl_sysdep_start (1 samples, 0.75%)</title><rect x="98.4962%" y="213" width="0.7519%" height="15" fill="rgb(239,60,40)"/><text x="98.7462%" y="223.50"></text></g><g><title>dl_main (1 samples, 0.75%)</title><rect x="98.4962%" y="197" width="0.7519%" height="15" fill="rgb(206,27,48)"/><text x="98.7462%" y="207.50"></text></g><g><title>all (133 samples, 100%)</title><rect x="0.0000%" y="277" width="100.0000%" height="15" fill="rgb(225,35,8)"/><text x="0.2500%" y="287.50"></text></g><g><title>shuffle3-debug (133 samples, 100.00%)</title><rect x="0.0000%" y="261" width="100.0000%" height="15" fill="rgb(250,213,24)"/><text x="0.2500%" y="271.50">shuffle3-debug</text></g><g><title>rng::frng::fract (1 samples, 0.75%)</title><rect x="99.2481%" y="245" width="0.7519%" height="15" fill="rgb(247,123,22)"/><text x="99.4981%" y="255.50"></text></g></svg></svg>
\ No newline at end of file
diff --git a/profiling/debug-flameu.svg b/profiling/debug-flameu.svg
new file mode 100644
index 0000000..6db6ca2
--- /dev/null
+++ b/profiling/debug-flameu.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="342" onload="init(evt)" viewBox="0 0 1200 342" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="342" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="325.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="325.00"> </text><svg id="frames" x="10" width="1180"><g><title>[libm-2.32.so] (2 samples, 0.82%)</title><rect x="0.0000%" y="245" width="0.8163%" height="15" fill="rgb(227,0,7)"/><text x="0.2500%" y="255.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::__normal_iterator (1 samples, 0.41%)</title><rect x="0.8163%" y="245" width="0.4082%" height="15" fill="rgb(217,0,24)"/><text x="1.0663%" y="255.50"></text></g><g><title>operator new (1 samples, 0.41%)</title><rect x="1.2245%" y="245" width="0.4082%" height="15" fill="rgb(221,193,54)"/><text x="1.4745%" y="255.50"></text></g><g><title>rng::frng::sample_double (1 samples, 0.41%)</title><rect x="1.6327%" y="245" width="0.4082%" height="15" fill="rgb(248,212,6)"/><text x="1.8827%" y="255.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (1 samples, 0.41%)</title><rect x="2.0408%" y="245" width="0.4082%" height="15" fill="rgb(208,68,35)"/><text x="2.2908%" y="255.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::end (2 samples, 0.82%)</title><rect x="2.4490%" y="245" width="0.8163%" height="15" fill="rgb(232,128,0)"/><text x="2.6990%" y="255.50"></text></g><g><title>[[stack]] (9 samples, 3.67%)</title><rect x="0.0000%" y="261" width="3.6735%" height="15" fill="rgb(207,160,47)"/><text x="0.2500%" y="271.50">[[st..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::push_back (1 samples, 0.41%)</title><rect x="3.2653%" y="245" width="0.4082%" height="15" fill="rgb(228,23,34)"/><text x="3.5153%" y="255.50"></text></g><g><title>[anon] (2 samples, 0.82%)</title><rect x="3.6735%" y="261" width="0.8163%" height="15" fill="rgb(218,30,26)"/><text x="3.9235%" y="271.50"></text></g><g><title>RNG::next_long (2 samples, 0.82%)</title><rect x="3.6735%" y="245" width="0.8163%" height="15" fill="rgb(220,122,19)"/><text x="3.9235%" y="255.50"></text></g><g><title>[unknown] (1 samples, 0.41%)</title><rect x="4.4898%" y="261" width="0.4082%" height="15" fill="rgb(250,228,42)"/><text x="4.7398%" y="271.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (1 samples, 0.41%)</title><rect x="4.4898%" y="245" width="0.4082%" height="15" fill="rgb(240,193,28)"/><text x="4.7398%" y="255.50"></text></g><g><title>minmax_t&lt;float, work::xshuffle_ip&lt;true&gt;(char const*)::{lambda(float)#1}&gt; (4 samples, 1.63%)</title><rect x="6.5306%" y="181" width="1.6327%" height="15" fill="rgb(216,20,37)"/><text x="6.7806%" y="191.50"></text></g><g><title>work::xshuffle_ip&lt;true&gt; (1 samples, 0.41%)</title><rect x="7.7551%" y="165" width="0.4082%" height="15" fill="rgb(206,188,39)"/><text x="8.0051%" y="175.50"></text></g><g><title>minmax_t&lt;long&gt; (1 samples, 0.41%)</title><rect x="8.1633%" y="181" width="0.4082%" height="15" fill="rgb(217,207,13)"/><text x="8.4133%" y="191.50"></text></g><g><title>minmax_t&lt;long, minmax_t&lt;long&gt;(span&lt;long&gt; const&amp;)::{lambda(long)#1}&gt; (1 samples, 0.41%)</title><rect x="8.1633%" y="165" width="0.4082%" height="15" fill="rgb(231,73,38)"/><text x="8.4133%" y="175.50"></text></g><g><title>span&lt;long&gt;::operator[] (1 samples, 0.41%)</title><rect x="8.1633%" y="149" width="0.4082%" height="15" fill="rgb(225,20,46)"/><text x="8.4133%" y="159.50"></text></g><g><title>minmax_t&lt;signed char&gt; (1 samples, 0.41%)</title><rect x="8.5714%" y="181" width="0.4082%" height="15" fill="rgb(210,31,41)"/><text x="8.8214%" y="191.50"></text></g><g><title>minmax_t&lt;signed char, minmax_t&lt;signed char&gt;(span&lt;signed char&gt; const&amp;)::{lambda(signed char)#1}&gt; (1 samples, 0.41%)</title><rect x="8.5714%" y="165" width="0.4082%" height="15" fill="rgb(221,200,47)"/><text x="8.8214%" y="175.50"></text></g><g><title>[libm-2.32.so] (9 samples, 3.67%)</title><rect x="10.6122%" y="101" width="3.6735%" height="15" fill="rgb(226,26,5)"/><text x="10.8622%" y="111.50">[lib..</text></g><g><title>rng::frng::dot&lt;2ul&gt; (4 samples, 1.63%)</title><rect x="14.2857%" y="101" width="1.6327%" height="15" fill="rgb(249,33,26)"/><text x="14.5357%" y="111.50"></text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (1 samples, 0.41%)</title><rect x="15.5102%" y="85" width="0.4082%" height="15" fill="rgb(235,183,28)"/><text x="15.7602%" y="95.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (1 samples, 0.41%)</title><rect x="15.5102%" y="69" width="0.4082%" height="15" fill="rgb(221,5,38)"/><text x="15.7602%" y="79.50"></text></g><g><title>rng::frng::fract (4 samples, 1.63%)</title><rect x="15.9184%" y="101" width="1.6327%" height="15" fill="rgb(247,18,42)"/><text x="16.1684%" y="111.50"></text></g><g><title>[libm-2.32.so] (1 samples, 0.41%)</title><rect x="17.1429%" y="85" width="0.4082%" height="15" fill="rgb(241,131,45)"/><text x="17.3929%" y="95.50"></text></g><g><title>rng::frng::sample_double (19 samples, 7.76%)</title><rect x="10.2041%" y="117" width="7.7551%" height="15" fill="rgb(249,31,29)"/><text x="10.4541%" y="127.50">rng::frng::..</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (1 samples, 0.41%)</title><rect x="17.5510%" y="101" width="0.4082%" height="15" fill="rgb(225,111,53)"/><text x="17.8010%" y="111.50"></text></g><g><title>[libm-2.32.so] (15 samples, 6.12%)</title><rect x="19.1837%" y="85" width="6.1224%" height="15" fill="rgb(238,160,17)"/><text x="19.4337%" y="95.50">[libm-2...</text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (2 samples, 0.82%)</title><rect x="27.7551%" y="69" width="0.8163%" height="15" fill="rgb(214,148,48)"/><text x="28.0051%" y="79.50"></text></g><g><title>rng::frng::dot&lt;2ul&gt; (14 samples, 5.71%)</title><rect x="25.3061%" y="85" width="5.7143%" height="15" fill="rgb(232,36,49)"/><text x="25.5561%" y="95.50">rng::fr..</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (6 samples, 2.45%)</title><rect x="28.5714%" y="69" width="2.4490%" height="15" fill="rgb(209,103,24)"/><text x="28.8214%" y="79.50">st..</text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (4 samples, 1.63%)</title><rect x="29.3878%" y="53" width="1.6327%" height="15" fill="rgb(229,88,8)"/><text x="29.6378%" y="63.50"></text></g><g><title>rng::frng::sample_double (36 samples, 14.69%)</title><rect x="17.9592%" y="101" width="14.6939%" height="15" fill="rgb(213,181,19)"/><text x="18.2092%" y="111.50">rng::frng::sample_doub..</text></g><g><title>rng::frng::fract (4 samples, 1.63%)</title><rect x="31.0204%" y="85" width="1.6327%" height="15" fill="rgb(254,191,54)"/><text x="31.2704%" y="95.50"></text></g><g><title>[libm-2.32.so] (1 samples, 0.41%)</title><rect x="32.2449%" y="69" width="0.4082%" height="15" fill="rgb(241,83,37)"/><text x="32.4949%" y="79.50"></text></g><g><title>rng::frng::update_state (39 samples, 15.92%)</title><rect x="17.9592%" y="117" width="15.9184%" height="15" fill="rgb(233,36,39)"/><text x="18.2092%" y="127.50">rng::frng::update_state</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (3 samples, 1.22%)</title><rect x="32.6531%" y="101" width="1.2245%" height="15" fill="rgb(226,3,54)"/><text x="32.9031%" y="111.50"></text></g><g><title>std::__array_traits&lt;double, 2ul&gt;::_S_ref (2 samples, 0.82%)</title><rect x="33.0612%" y="85" width="0.8163%" height="15" fill="rgb(245,192,40)"/><text x="33.3112%" y="95.50"></text></g><g><title>RNG::next_long (59 samples, 24.08%)</title><rect x="10.2041%" y="165" width="24.0816%" height="15" fill="rgb(238,167,29)"/><text x="10.4541%" y="175.50">RNG::next_long</text></g><g><title>RNG::next_long (59 samples, 24.08%)</title><rect x="10.2041%" y="149" width="24.0816%" height="15" fill="rgb(232,182,51)"/><text x="10.4541%" y="159.50">RNG::next_long</text></g><g><title>rng::frng::sample (59 samples, 24.08%)</title><rect x="10.2041%" y="133" width="24.0816%" height="15" fill="rgb(231,60,39)"/><text x="10.4541%" y="143.50">rng::frng::sample</text></g><g><title>std::array&lt;double, 2ul&gt;::operator[] (1 samples, 0.41%)</title><rect x="33.8776%" y="117" width="0.4082%" height="15" fill="rgb(208,69,12)"/><text x="34.1276%" y="127.50"></text></g><g><title>std::swap&lt;float&gt; (10 samples, 4.08%)</title><rect x="34.2857%" y="165" width="4.0816%" height="15" fill="rgb(235,93,37)"/><text x="34.5357%" y="175.50">std:..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (2 samples, 0.82%)</title><rect x="38.3673%" y="165" width="0.8163%" height="15" fill="rgb(213,116,39)"/><text x="38.6173%" y="175.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::pop_back (1 samples, 0.41%)</title><rect x="39.1837%" y="165" width="0.4082%" height="15" fill="rgb(222,207,29)"/><text x="39.4337%" y="175.50"></text></g><g><title>std::allocator_traits&lt;std::allocator&lt;unsigned long&gt; &gt;::construct&lt;unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.0000%" y="133" width="0.4082%" height="15" fill="rgb(206,96,30)"/><text x="40.2500%" y="143.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::push_back (3 samples, 1.22%)</title><rect x="39.5918%" y="165" width="1.2245%" height="15" fill="rgb(218,138,4)"/><text x="39.8418%" y="175.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (3 samples, 1.22%)</title><rect x="39.5918%" y="149" width="1.2245%" height="15" fill="rgb(250,191,14)"/><text x="39.8418%" y="159.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (1 samples, 0.41%)</title><rect x="40.4082%" y="133" width="0.4082%" height="15" fill="rgb(239,60,40)"/><text x="40.6582%" y="143.50"></text></g><g><title>rng::unshuffle&lt;float, rng::frng&gt; (79 samples, 32.24%)</title><rect x="8.9796%" y="181" width="32.2449%" height="15" fill="rgb(206,27,48)"/><text x="9.2296%" y="191.50">rng::unshuffle&lt;float, rng::frng&gt;</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::vector (1 samples, 0.41%)</title><rect x="40.8163%" y="165" width="0.4082%" height="15" fill="rgb(225,35,8)"/><text x="41.0663%" y="175.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::_M_default_initialize (1 samples, 0.41%)</title><rect x="40.8163%" y="149" width="0.4082%" height="15" fill="rgb(250,213,24)"/><text x="41.0663%" y="159.50"></text></g><g><title>std::__uninitialized_default_n_a&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="133" width="0.4082%" height="15" fill="rgb(247,123,22)"/><text x="41.0663%" y="143.50"></text></g><g><title>std::__uninitialized_default_n&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="117" width="0.4082%" height="15" fill="rgb(231,138,38)"/><text x="41.0663%" y="127.50"></text></g><g><title>std::__uninitialized_default_n_1&lt;true&gt;::__uninit_default_n&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="101" width="0.4082%" height="15" fill="rgb(231,145,46)"/><text x="41.0663%" y="111.50"></text></g><g><title>std::fill_n&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="85" width="0.4082%" height="15" fill="rgb(251,118,11)"/><text x="41.0663%" y="95.50"></text></g><g><title>std::__fill_n_a&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="69" width="0.4082%" height="15" fill="rgb(217,147,25)"/><text x="41.0663%" y="79.50"></text></g><g><title>std::__fill_a&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="53" width="0.4082%" height="15" fill="rgb(247,81,37)"/><text x="41.0663%" y="63.50"></text></g><g><title>std::__fill_a1&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="40.8163%" y="37" width="0.4082%" height="15" fill="rgb(209,12,38)"/><text x="41.0663%" y="47.50"></text></g><g><title>[libm-2.32.so] (2 samples, 0.82%)</title><rect x="42.0408%" y="133" width="0.8163%" height="15" fill="rgb(227,1,9)"/><text x="42.2908%" y="143.50"></text></g><g><title>RNG::next_long (9 samples, 3.67%)</title><rect x="42.0408%" y="165" width="3.6735%" height="15" fill="rgb(248,47,43)"/><text x="42.2908%" y="175.50">RNG:..</text></g><g><title>RNG::next_long (9 samples, 3.67%)</title><rect x="42.0408%" y="149" width="3.6735%" height="15" fill="rgb(221,10,30)"/><text x="42.2908%" y="159.50">RNG:..</text></g><g><title>rng::xoroshiro128plus::sample (7 samples, 2.86%)</title><rect x="42.8571%" y="133" width="2.8571%" height="15" fill="rgb(210,229,1)"/><text x="43.1071%" y="143.50">rn..</text></g><g><title>rng::next (5 samples, 2.04%)</title><rect x="43.6735%" y="117" width="2.0408%" height="15" fill="rgb(222,148,37)"/><text x="43.9235%" y="127.50">r..</text></g><g><title>std::array&lt;unsigned long, 2ul&gt;::operator[] (1 samples, 0.41%)</title><rect x="45.3061%" y="101" width="0.4082%" height="15" fill="rgb(234,67,33)"/><text x="45.5561%" y="111.50"></text></g><g><title>std::swap&lt;long&gt; (4 samples, 1.63%)</title><rect x="45.7143%" y="165" width="1.6327%" height="15" fill="rgb(247,98,35)"/><text x="45.9643%" y="175.50"></text></g><g><title>std::move&lt;long&amp;&gt; (1 samples, 0.41%)</title><rect x="46.9388%" y="149" width="0.4082%" height="15" fill="rgb(247,138,52)"/><text x="47.1888%" y="159.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (2 samples, 0.82%)</title><rect x="47.3469%" y="165" width="0.8163%" height="15" fill="rgb(213,79,30)"/><text x="47.5969%" y="175.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator* (2 samples, 0.82%)</title><rect x="47.3469%" y="149" width="0.8163%" height="15" fill="rgb(246,177,23)"/><text x="47.5969%" y="159.50"></text></g><g><title>std::allocator_traits&lt;std::allocator&lt;unsigned long&gt; &gt;::construct&lt;unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.1633%" y="133" width="0.4082%" height="15" fill="rgb(230,62,27)"/><text x="48.4133%" y="143.50"></text></g><g><title>std::construct_at&lt;unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.1633%" y="117" width="0.4082%" height="15" fill="rgb(216,154,8)"/><text x="48.4133%" y="127.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::push_back (2 samples, 0.82%)</title><rect x="48.1633%" y="165" width="0.8163%" height="15" fill="rgb(244,35,45)"/><text x="48.4133%" y="175.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (2 samples, 0.82%)</title><rect x="48.1633%" y="149" width="0.8163%" height="15" fill="rgb(251,115,12)"/><text x="48.4133%" y="159.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (1 samples, 0.41%)</title><rect x="48.5714%" y="133" width="0.4082%" height="15" fill="rgb(240,54,50)"/><text x="48.8214%" y="143.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator- (1 samples, 0.41%)</title><rect x="48.5714%" y="117" width="0.4082%" height="15" fill="rgb(233,84,52)"/><text x="48.8214%" y="127.50"></text></g><g><title>rng::unshuffle&lt;long, rng::xoroshiro128plus&gt; (20 samples, 8.16%)</title><rect x="41.2245%" y="181" width="8.1633%" height="15" fill="rgb(207,117,47)"/><text x="41.4745%" y="191.50">rng::unshuf..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::vector (1 samples, 0.41%)</title><rect x="48.9796%" y="165" width="0.4082%" height="15" fill="rgb(249,43,39)"/><text x="49.2296%" y="175.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::_M_default_initialize (1 samples, 0.41%)</title><rect x="48.9796%" y="149" width="0.4082%" height="15" fill="rgb(209,38,44)"/><text x="49.2296%" y="159.50"></text></g><g><title>std::__uninitialized_default_n_a&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="133" width="0.4082%" height="15" fill="rgb(236,212,23)"/><text x="49.2296%" y="143.50"></text></g><g><title>std::__uninitialized_default_n&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="117" width="0.4082%" height="15" fill="rgb(242,79,21)"/><text x="49.2296%" y="127.50"></text></g><g><title>std::__uninitialized_default_n_1&lt;true&gt;::__uninit_default_n&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="101" width="0.4082%" height="15" fill="rgb(211,96,35)"/><text x="49.2296%" y="111.50"></text></g><g><title>std::fill_n&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="85" width="0.4082%" height="15" fill="rgb(253,215,40)"/><text x="49.2296%" y="95.50"></text></g><g><title>std::__fill_n_a&lt;unsigned long*, unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="69" width="0.4082%" height="15" fill="rgb(211,81,21)"/><text x="49.2296%" y="79.50"></text></g><g><title>std::__fill_a&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="53" width="0.4082%" height="15" fill="rgb(208,190,38)"/><text x="49.2296%" y="63.50"></text></g><g><title>std::__fill_a1&lt;unsigned long*, unsigned long&gt; (1 samples, 0.41%)</title><rect x="48.9796%" y="37" width="0.4082%" height="15" fill="rgb(235,213,38)"/><text x="49.2296%" y="47.50"></text></g><g><title>[libm-2.32.so] (3 samples, 1.22%)</title><rect x="55.9184%" y="133" width="1.2245%" height="15" fill="rgb(237,122,38)"/><text x="56.1684%" y="143.50"></text></g><g><title>rand_r (7 samples, 2.86%)</title><rect x="60.8163%" y="117" width="2.8571%" height="15" fill="rgb(244,218,35)"/><text x="61.0663%" y="127.50">ra..</text></g><g><title>RNG::next_long (29 samples, 11.84%)</title><rect x="52.6531%" y="165" width="11.8367%" height="15" fill="rgb(240,68,47)"/><text x="52.9031%" y="175.50">RNG::next_long</text></g><g><title>RNG::next_long (29 samples, 11.84%)</title><rect x="52.6531%" y="149" width="11.8367%" height="15" fill="rgb(210,16,53)"/><text x="52.9031%" y="159.50">RNG::next_long</text></g><g><title>rng::drng::sample (18 samples, 7.35%)</title><rect x="57.1429%" y="133" width="7.3469%" height="15" fill="rgb(235,124,12)"/><text x="57.3929%" y="143.50">rng::drng:..</text></g><g><title>rand_r@plt (2 samples, 0.82%)</title><rect x="63.6735%" y="117" width="0.8163%" height="15" fill="rgb(224,169,11)"/><text x="63.9235%" y="127.50"></text></g><g><title>span&lt;unsigned char&gt;::operator[] (3 samples, 1.22%)</title><rect x="64.4898%" y="165" width="1.2245%" height="15" fill="rgb(250,166,2)"/><text x="64.7398%" y="175.50"></text></g><g><title>span&lt;unsigned char&gt;::size (1 samples, 0.41%)</title><rect x="65.7143%" y="165" width="0.4082%" height="15" fill="rgb(242,216,29)"/><text x="65.9643%" y="175.50"></text></g><g><title>std::swap&lt;unsigned char&gt; (42 samples, 17.14%)</title><rect x="66.1224%" y="165" width="17.1429%" height="15" fill="rgb(230,116,27)"/><text x="66.3724%" y="175.50">std::swap&lt;unsigned char&gt;</text></g><g><title>std::move&lt;unsigned char&amp;&gt; (3 samples, 1.22%)</title><rect x="82.0408%" y="149" width="1.2245%" height="15" fill="rgb(228,99,48)"/><text x="82.2908%" y="159.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator* (3 samples, 1.22%)</title><rect x="84.0816%" y="149" width="1.2245%" height="15" fill="rgb(253,11,6)"/><text x="84.3316%" y="159.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator- (4 samples, 1.63%)</title><rect x="85.3061%" y="149" width="1.6327%" height="15" fill="rgb(247,143,39)"/><text x="85.5561%" y="159.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (10 samples, 4.08%)</title><rect x="83.2653%" y="165" width="4.0816%" height="15" fill="rgb(236,97,10)"/><text x="83.5153%" y="175.50">std:..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::end (1 samples, 0.41%)</title><rect x="86.9388%" y="149" width="0.4082%" height="15" fill="rgb(233,208,19)"/><text x="87.1888%" y="159.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::__normal_iterator (1 samples, 0.41%)</title><rect x="86.9388%" y="133" width="0.4082%" height="15" fill="rgb(216,164,2)"/><text x="87.1888%" y="143.50"></text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::pop_back (3 samples, 1.22%)</title><rect x="87.3469%" y="165" width="1.2245%" height="15" fill="rgb(220,129,5)"/><text x="87.5969%" y="175.50"></text></g><g><title>std::allocator_traits&lt;std::allocator&lt;unsigned long&gt; &gt;::destroy&lt;unsigned long&gt; (2 samples, 0.82%)</title><rect x="87.7551%" y="149" width="0.8163%" height="15" fill="rgb(242,17,10)"/><text x="88.0051%" y="159.50"></text></g><g><title>std::destroy_at&lt;unsigned long&gt; (2 samples, 0.82%)</title><rect x="87.7551%" y="133" width="0.8163%" height="15" fill="rgb(242,107,0)"/><text x="88.0051%" y="143.50"></text></g><g><title>std::allocator_traits&lt;std::allocator&lt;unsigned long&gt; &gt;::construct&lt;unsigned long, unsigned long&gt; (1 samples, 0.41%)</title><rect x="88.5714%" y="149" width="0.4082%" height="15" fill="rgb(251,28,31)"/><text x="88.8214%" y="159.50"></text></g><g><title>std::construct_at&lt;unsigned long, unsigned long&gt; (3 samples, 1.22%)</title><rect x="91.0204%" y="117" width="1.2245%" height="15" fill="rgb(233,223,10)"/><text x="91.2704%" y="127.50"></text></g><g><title>std::allocator_traits&lt;std::allocator&lt;unsigned long&gt; &gt;::construct&lt;unsigned long, unsigned long&gt; (5 samples, 2.04%)</title><rect x="90.6122%" y="133" width="2.0408%" height="15" fill="rgb(215,21,27)"/><text x="90.8622%" y="143.50">s..</text></g><g><title>std::forward&lt;unsigned long&gt; (1 samples, 0.41%)</title><rect x="92.2449%" y="117" width="0.4082%" height="15" fill="rgb(232,23,21)"/><text x="92.4949%" y="127.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator* (2 samples, 0.82%)</title><rect x="94.6939%" y="117" width="0.8163%" height="15" fill="rgb(244,5,23)"/><text x="94.9439%" y="127.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::operator- (3 samples, 1.22%)</title><rect x="95.5102%" y="117" width="1.2245%" height="15" fill="rgb(226,81,46)"/><text x="95.7602%" y="127.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::__normal_iterator (2 samples, 0.82%)</title><rect x="95.9184%" y="101" width="0.8163%" height="15" fill="rgb(247,70,30)"/><text x="96.1684%" y="111.50"></text></g><g><title>rng::unshuffle&lt;unsigned char, rng::drng&gt; (120 samples, 48.98%)</title><rect x="49.3878%" y="181" width="48.9796%" height="15" fill="rgb(212,68,19)"/><text x="49.6378%" y="191.50">rng::unshuffle&lt;unsigned char, rng::drng&gt;</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::push_back (24 samples, 9.80%)</title><rect x="88.5714%" y="165" width="9.7959%" height="15" fill="rgb(240,187,13)"/><text x="88.8214%" y="175.50">std::vector&lt;un..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (23 samples, 9.39%)</title><rect x="88.9796%" y="149" width="9.3878%" height="15" fill="rgb(223,113,26)"/><text x="89.2296%" y="159.50">std::vector&lt;u..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::back (14 samples, 5.71%)</title><rect x="92.6531%" y="133" width="5.7143%" height="15" fill="rgb(206,192,2)"/><text x="92.9031%" y="143.50">std::ve..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::end (4 samples, 1.63%)</title><rect x="96.7347%" y="117" width="1.6327%" height="15" fill="rgb(241,108,4)"/><text x="96.9847%" y="127.50"></text></g><g><title>__gnu_cxx::__normal_iterator&lt;unsigned long*, std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt; &gt;::__normal_iterator (4 samples, 1.63%)</title><rect x="96.7347%" y="101" width="1.6327%" height="15" fill="rgb(247,173,49)"/><text x="96.9847%" y="111.50"></text></g><g><title>span&lt;unsigned char&gt;::operator[] (1 samples, 0.41%)</title><rect x="98.3673%" y="181" width="0.4082%" height="15" fill="rgb(224,114,35)"/><text x="98.6173%" y="191.50"></text></g><g><title>__libc_start_main (228 samples, 93.06%)</title><rect x="6.5306%" y="245" width="93.0612%" height="15" fill="rgb(245,159,27)"/><text x="6.7806%" y="255.50">__libc_start_main</text></g><g><title>main (228 samples, 93.06%)</title><rect x="6.5306%" y="229" width="93.0612%" height="15" fill="rgb(245,172,44)"/><text x="6.7806%" y="239.50">main</text></g><g><title>do_work (228 samples, 93.06%)</title><rect x="6.5306%" y="213" width="93.0612%" height="15" fill="rgb(236,23,11)"/><text x="6.7806%" y="223.50">do_work</text></g><g><title>work::xshuffle_ip&lt;true&gt; (228 samples, 93.06%)</title><rect x="6.5306%" y="197" width="93.0612%" height="15" fill="rgb(205,117,38)"/><text x="6.7806%" y="207.50">work::xshuffle_ip&lt;true&gt;</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::push_back (2 samples, 0.82%)</title><rect x="98.7755%" y="181" width="0.8163%" height="15" fill="rgb(237,72,25)"/><text x="99.0255%" y="191.50"></text></g><g><title>all (245 samples, 100%)</title><rect x="0.0000%" y="293" width="100.0000%" height="15" fill="rgb(244,70,9)"/><text x="0.2500%" y="303.50"></text></g><g><title>shuffle3-debug (245 samples, 100.00%)</title><rect x="0.0000%" y="277" width="100.0000%" height="15" fill="rgb(217,125,39)"/><text x="0.2500%" y="287.50">shuffle3-debug</text></g><g><title>_start (233 samples, 95.10%)</title><rect x="4.8980%" y="261" width="95.1020%" height="15" fill="rgb(235,36,10)"/><text x="5.1480%" y="271.50">_start</text></g><g><title>_dl_start (1 samples, 0.41%)</title><rect x="99.5918%" y="245" width="0.4082%" height="15" fill="rgb(251,123,47)"/><text x="99.8418%" y="255.50"></text></g><g><title>_dl_sysdep_start (1 samples, 0.41%)</title><rect x="99.5918%" y="229" width="0.4082%" height="15" fill="rgb(221,13,13)"/><text x="99.8418%" y="239.50"></text></g><g><title>dl_main (1 samples, 0.41%)</title><rect x="99.5918%" y="213" width="0.4082%" height="15" fill="rgb(238,131,9)"/><text x="99.8418%" y="223.50"></text></g><g><title>_dl_lookup_symbol_x (1 samples, 0.41%)</title><rect x="99.5918%" y="197" width="0.4082%" height="15" fill="rgb(211,50,8)"/><text x="99.8418%" y="207.50"></text></g></svg></svg>
\ No newline at end of file
diff --git a/profiling/massif.out.520307 b/profiling/massif.out.520307
new file mode 100644
index 0000000..e4c07fd
--- /dev/null
+++ b/profiling/massif.out.520307
@@ -0,0 +1,567 @@
+desc: (none)
+cmd: ./shuffle3-release small-u
+time_unit: i
+#-----------
+snapshot=0
+#-----------
+time=0
+mem_heap_B=0
+mem_heap_extra_B=0
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=1
+#-----------
+time=4048133
+mem_heap_B=32816
+mem_heap_extra_B=8
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=2
+#-----------
+time=4050820
+mem_heap_B=8
+mem_heap_extra_B=16
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=3
+#-----------
+time=4057134
+mem_heap_B=72712
+mem_heap_extra_B=24
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=4
+#-----------
+time=4282565
+mem_heap_B=73736
+mem_heap_extra_B=32
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=5
+#-----------
+time=4289663
+mem_heap_B=74248
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=6
+#-----------
+time=4295729
+mem_heap_B=75784
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=7
+#-----------
+time=4303612
+mem_heap_B=77832
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=8
+#-----------
+time=4318766
+mem_heap_B=86024
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=9
+#-----------
+time=4322930
+mem_heap_B=81928
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=10
+#-----------
+time=4353188
+mem_heap_B=98312
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=11
+#-----------
+time=4361448
+mem_heap_B=90120
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=12
+#-----------
+time=4421914
+mem_heap_B=122888
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=13
+#-----------
+time=4438366
+mem_heap_B=122888
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=detailed
+n3: 122888 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n1: 49152 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 49152 0x109385: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 49152 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n0: 1032 in 3 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=14
+#-----------
+time=4438366
+mem_heap_B=106504
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=15
+#-----------
+time=4559248
+mem_heap_B=172040
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=16
+#-----------
+time=4592084
+mem_heap_B=172040
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=detailed
+n3: 172040 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 98304 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 98304 0x109385: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 98304 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n0: 1032 in 3 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=17
+#-----------
+time=4592084
+mem_heap_B=139272
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=18
+#-----------
+time=4833798
+mem_heap_B=270344
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=19
+#-----------
+time=4899402
+mem_heap_B=270344
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=detailed
+n3: 270344 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 196608 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 196608 0x109385: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 196608 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n0: 1032 in 3 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=20
+#-----------
+time=4899402
+mem_heap_B=204808
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=21
+#-----------
+time=5382780
+mem_heap_B=466952
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=22
+#-----------
+time=5513920
+mem_heap_B=466952
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=peak
+n3: 466952 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 393216 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 393216 0x109385: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 393216 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n0: 1032 in 3 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=23
+#-----------
+time=5513920
+mem_heap_B=335880
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=24
+#-----------
+time=5892087
+mem_heap_B=73736
+mem_heap_extra_B=32
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=25
+#-----------
+time=5989421
+mem_heap_B=73744
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=26
+#-----------
+time=5996279
+mem_heap_B=73992
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=27
+#-----------
+time=6002698
+mem_heap_B=74248
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=28
+#-----------
+time=6015289
+mem_heap_B=75272
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=29
+#-----------
+time=6015414
+mem_heap_B=74760
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=30
+#-----------
+time=6040651
+mem_heap_B=76808
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=31
+#-----------
+time=6040828
+mem_heap_B=75784
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=32
+#-----------
+time=6091310
+mem_heap_B=79880
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=detailed
+n4: 79880 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n2: 6144 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 6144 0x1095C2: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 6144 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+  n0: 0 in 1 place, below massif's threshold (1.00%)
+ n1: 1024 0x4CB6E03: _IO_file_doallocate (in /usr/lib/libc-2.32.so)
+  n1: 1024 0x4CC570F: _IO_doallocbuf (in /usr/lib/libc-2.32.so)
+   n1: 1024 0x4CC48A7: _IO_file_overflow@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+    n1: 1024 0x4CC3955: _IO_file_xsputn@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+     n1: 1024 0x4CB8330: fwrite (in /usr/lib/libc-2.32.so)
+      n1: 1024 0x4877FD7: fmt::v7::vprint(_IO_FILE*, fmt::v7::basic_string_view<char>, fmt::v7::format_args) (in /usr/lib/libfmt.so.7.1.2)
+       n1: 1024 0x10A258: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+        n1: 1024 0x10933E: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+         n0: 1024 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n0: 8 in 2 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=33
+#-----------
+time=6091591
+mem_heap_B=77832
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=34
+#-----------
+time=6192490
+mem_heap_B=86024
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=35
+#-----------
+time=6196654
+mem_heap_B=81928
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=36
+#-----------
+time=6398233
+mem_heap_B=98312
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=37
+#-----------
+time=6406493
+mem_heap_B=90120
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=38
+#-----------
+time=6809904
+mem_heap_B=122888
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=39
+#-----------
+time=6826356
+mem_heap_B=106504
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=40
+#-----------
+time=7633039
+mem_heap_B=172040
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=41
+#-----------
+time=7665875
+mem_heap_B=139272
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=42
+#-----------
+time=7951930
+mem_heap_B=73736
+mem_heap_extra_B=32
+mem_stacks_B=0
+heap_tree=detailed
+n3: 73736 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n1: 1024 0x4CB6E03: _IO_file_doallocate (in /usr/lib/libc-2.32.so)
+  n1: 1024 0x4CC570F: _IO_doallocbuf (in /usr/lib/libc-2.32.so)
+   n1: 1024 0x4CC48A7: _IO_file_overflow@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+    n1: 1024 0x4CC3955: _IO_file_xsputn@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+     n1: 1024 0x4CB8330: fwrite (in /usr/lib/libc-2.32.so)
+      n1: 1024 0x4877FD7: fmt::v7::vprint(_IO_FILE*, fmt::v7::basic_string_view<char>, fmt::v7::format_args) (in /usr/lib/libfmt.so.7.1.2)
+       n1: 1024 0x10A258: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+        n1: 1024 0x10933E: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+         n0: 1024 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n0: 8 in 3 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=43
+#-----------
+time=7971885
+mem_heap_B=73744
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=44
+#-----------
+time=7979276
+mem_heap_B=75784
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=45
+#-----------
+time=7985879
+mem_heap_B=77832
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=46
+#-----------
+time=7998473
+mem_heap_B=86024
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=detailed
+n4: 86024 (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
+ n1: 72704 0x493F45A: pool (eh_alloc.cc:123)
+  n1: 72704 0x493F45A: __static_initialization_and_destruction_0 (eh_alloc.cc:262)
+   n1: 72704 0x493F45A: _GLOBAL__sub_I_eh_alloc.cc (eh_alloc.cc:338)
+    n1: 72704 0x40112DD: call_init.part.0 (in /usr/lib/ld-2.32.so)
+     n1: 72704 0x40113C7: _dl_init (in /usr/lib/ld-2.32.so)
+      n1: 72704 0x40020C9: ??? (in /usr/lib/ld-2.32.so)
+       n1: 72704 0x1: ???
+        n1: 72704 0x1FFF000AC6: ???
+         n0: 72704 0x1FFF000AD9: ???
+ n2: 12288 0x10A384: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+  n1: 12288 0x10973E: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+   n0: 12288 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+  n0: 0 in 2 places, all below massif's threshold (1.00%)
+ n1: 1024 0x4CB6E03: _IO_file_doallocate (in /usr/lib/libc-2.32.so)
+  n1: 1024 0x4CC570F: _IO_doallocbuf (in /usr/lib/libc-2.32.so)
+   n1: 1024 0x4CC48A7: _IO_file_overflow@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+    n1: 1024 0x4CC3955: _IO_file_xsputn@@GLIBC_2.2.5 (in /usr/lib/libc-2.32.so)
+     n1: 1024 0x4CB8330: fwrite (in /usr/lib/libc-2.32.so)
+      n1: 1024 0x4877FD7: fmt::v7::vprint(_IO_FILE*, fmt::v7::basic_string_view<char>, fmt::v7::format_args) (in /usr/lib/libfmt.so.7.1.2)
+       n1: 1024 0x10A258: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+        n1: 1024 0x10933E: ??? (in /home/avril/work/shuffle3/lean/shuffle3-release)
+         n0: 1024 0x4C6A151: (below main) (in /usr/lib/libc-2.32.so)
+ n0: 8 in 2 places, all below massif's threshold (1.00%)
+#-----------
+snapshot=47
+#-----------
+time=8002637
+mem_heap_B=81928
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=48
+#-----------
+time=8027775
+mem_heap_B=98312
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=49
+#-----------
+time=8036035
+mem_heap_B=90120
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=50
+#-----------
+time=8086261
+mem_heap_B=122888
+mem_heap_extra_B=48
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=51
+#-----------
+time=8102713
+mem_heap_B=106504
+mem_heap_extra_B=40
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=52
+#-----------
+time=8147526
+mem_heap_B=73736
+mem_heap_extra_B=32
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=53
+#-----------
+time=8156172
+mem_heap_B=1032
+mem_heap_extra_B=24
+mem_stacks_B=0
+heap_tree=empty
+#-----------
+snapshot=54
+#-----------
+time=8157131
+mem_heap_B=8
+mem_heap_extra_B=16
+mem_stacks_B=0
+heap_tree=empty
diff --git a/profiling/mem-small.png b/profiling/mem-small.png
new file mode 100644
index 0000000000000000000000000000000000000000..982bfa1334312754b363f482c6f6400bb4644398
GIT binary patch
literal 4224
zcmZu!3piA37oIL^Tw;$5A+jfw+#+(6j@oXc48I`^sYq^xT*@Vv(jF$Mc5=_<q={s5
z85JdTN+RKe%ye^?rjat5BH{l`=l`Gof1WeXv!8dbZ+&}xYprj6@A~HKzwB*RNhnF+
zaJW^rWJ@Xzw@eL(Tk5ua85r?*E3^P0`!73ZtAT+5@S59hh{G-562XJ>7L#cPLrWs4
z_D;B+@43D}5I;tyh2wD2O^eTx3~gy;91dD(Yq`rgD(_RiN7#^^to2!3kdMp2n)ztg
zrl>yQqxl-c`?wX#>U;P4j&2CGHwiwJRc;fl<@Gvw(S`s1c1pF6hSf$;z6Qt2mCSca
zQMUY~$HaC@@~k&_-l{*{4Z)vgKY!nJAGT~*I60r3%pW$-DmX!=ZGpR_60Y<;%bK~4
zTirhTEIY{>{`C9<Y@LMP#W3hp4bE$jCo!*hGc6c|iXCrr(C0l^u$x8UZHU^On7qBD
zqWk-IWO(~&Y;A8$ZMb_}0U>QCJ9)b<BGvl+*h<(XJ-YS`q@6l(Sgi?v?M+;Xa7+=d
zR<o2s8K)rwR^T=ybthvDp0Esd&wd$!a!)}95NQR{w7HCjuUZ0oFV6J$a_uH~FGZR}
zGar{1uWwqu__;2tPry-yyc4}BHVm;v8{TVrBT%=$ZQQP8@$d$?Up5$CG++bw%f^dG
zw~N;^*NR+#NJ#+{8?vd$1-E&rzm~wGTwq1+%%Uerj=(hWE=3+m*#+Eu@hD0u5u>JF
z(LvJ_Mfypj<?!`2CN#X3*y2Kx{M-u|B8mVw+iR=>(|<(}Hw9j^IT2_&2piZHwnw6J
z-hDnWQ<Mqa@e&xX^{Am0ArnM_a2+flDjIj0j*%9v-9DC`q{dvHtSJd=h>|zBdDSkm
zxHNEV_fCeqs;5380TxsRvIrODFaB(ZDvDTM56FVh05+5+Bg%kYH-wH1clp5T>qLo+
zEr#)}%_U&<t)kVZ?JS_Q^A2G3qfimuiE|Fy2Yhw{vIzepi?E393eXPdMOcIyW$ii8
zm^*s95De69^odd%(f301-d*L>Bk<A9Wqi*ISaHn@bLsk=q^EPCRWyg$u_Nb-ZD18G
zzG6*z<cyg~oaZ{ya`<AcF-@D=%RTI9Q6RqBF#W<V@)>Ny5XzMKiGO>4KRPnlD<rAQ
zp#Fm#fRZR<kOKk|;eTWi`VVpet_1pjM*qQAK)(#={|Qw%Am_YaIe!i2{a5CcD=dyn
z_5*z@(DwuRA5d}(jF$raKjGOZP&u~CMJ4(f%=<q%0FQJX3juoujsw`+39x6(K~OR3
zt3-wS7F3Mwqeif2vUtMdM!wZA{2y6@n9GVBn+a2lubxXez0T?n%wlT089`Q?Ia_}m
zb>_to(h#~PKknImj6Dx&sV59v5I>LgE8CDW0$OI3x*RRyovs0UOYwyD&Unw>1A%rH
zAK0=tUih@<`H@CUt|Hl`3!v<KvXh>_jo)y1u4wmV<JGYC%6RE2M6hxq(+zI_k}^2l
zmV_@gT$Phk8kt<eSw)~;Z*{+aToArY3<ic>TBZ}p2M{RinmwhY&7D$-RppszuZWit
z(#<q#CVOqa#|-#UJ;<|WvT$RmQ+Gng6fb&YBbqMU83Z>X0kKc8oFB_F=&|2fO4C)&
z_huTt<o+0z<QM8Osg%#1Q^s5E_0PpjZT?bSx?$vIQQWI<q!DW?__750+-dCn>`Ax>
zJK!A1ZcZ1zQw?6{w=&*{NPke`r5aY^i{2h?>7Rmv^UWLslni^Z6D;`CFqh|9+k<Vt
zlgqM!FTUs@z@3(oIRpE9^K+5%xt9U)I*5)Nizn&XYtixJ7!0NrdTdB2Dvgd;CIXEH
zbH`kMse`9HGT}_@nkRM<U<c<iUL@S#uOFPhS>~SO-j+|_W``ci6yL-+CpWkE<RsfU
z69qqrj|WTob4vSfxOA6Hy;Tja@9k8T66Ps|U1MVkHTKr9Wy|>}m{S;Ek_6@(8b2tZ
zCr-UUKE90TyHF|dH`7_y19sYr7w45EwO^+h9OHbz2Q-@~lbU>uHMPnDiX%_Ov#}K8
zxJfe;&v4!ftwR$-j$fzyQw+}i$!+sba<n9yeo8SoLUq|<iAPdsPKb+~aPqSsXYm`#
zn>QS6zZ8;UKuG)jShAYsxm>`0+9JvZ#M}e8n6~6^`Yj5i*t8gmm`2Sq^(L6tCOTk%
z(mw*~)e%uEsQMco-2^s%@gsQ`jpeYlb2(2)Uvv0W;J7FG(D74ufN>YXyYjWJ9^{%k
zd3*1*A}zAp6}q?@qr1WSuZ`>PViT2B&5>hB?ricc&=yYUF;7m{t_lP^tWy~Rw|ARg
zIoOZfKd=ytL;W4=i(QFbKW8jeQASk={O(+56X5)pM`9>&dz-_ZS9h?9_OWM+hRv5n
z+|Xi1&st{v0KQj`yOT|(?1(7Can|pYG?8vp?O;{f@AGoCV3xa;K*AFaJw@i3H*`l>
zr@)zhkN0Qa6rWN4soX-_E-x7RYbyk%tUstk-P2NL(cMW~0!aKlw*t%J*_!*omgg7&
zWDU4N-&mMOeBtwwAb8)>V*w~*+_k0dt$IN<3G`53?k%k5s(wK=RAfcwcv!W}YE!oP
z(R;osA->Xf+bkhcb1?n+sK#g>F%&eSEvONVh(5P&2<t!?Yfcgk6ZYLV)%H~y()Ssx
zjj&KJ28s0ew#i3pL+g7TltI|CDpY|TmEZcFm95$4*&^^NO#-{ZEy0N5wnlzp6bndj
z;SbF=kCxn8&T|@ww-9kr$?)xv$W8i9k{0Ql2Fv#Nz#}Ska%@9It@o@Ofa*fP=XbpH
zv#X18b<hW1S%}NJ^n)EKA7eb}Ec1=^5J<hFXGIByS~9a1!*b*BwbVyRPHi+VO6E0;
zd59`WK$+$5a&}1Awjti8+AO|R%aKR9I_B`%UHX~TP$H%sj8J84H~BrGSiD8m*QGf#
znv7}q=vHYu(@*Vs9vx$8{A657iV|<FZAuSw*$3KN*0=+?gTNbMJuK(gl6wg}6psiL
z!r;o*Fxj-MIG-@kZ!^QxD)~Y}S~R_Ng!N9mb%$C8Bt9NP5Tw+Lx7zO@nQF6vE0QNR
zsGAQNhq?HG?x-+MA^$n)Kf-D$=U9+iC&|F<(M;p@Z2obGntnn(Eb}%tFXegP_sPSj
z8Uja1_O(#2U~Erg0woP&^Y^Iv*-G*}Q;}_`dbh9c9lgWn*U7Z|nJHZp-lnzbdR}ru
zBlu1(JPp-#q^q8vQ}BwY5#Cm_0%ggu8<ONU3{3yN$DJI0FEuD+$`tXwqBC{YYi5*~
zU;Ry^X~4h8Lj5k^it*;#n+H!)>24!e2J(n+grugRsh~dZyNKPR9l!ZKIZ=<Fd0dWi
z*V_4aNKC{8(buTh`7_j-f{d$_L$+jXA$n&VA%)8$YdNd$ecBpsx(D_x)im^jo5HOW
zmq*qYBjbjYgB`CgL67;{^el_&_M^WA(RMKo7l^mfUVzI<eARlmI6za!&ngnE^yC7{
z_074FM|>xIEpE_Id<Ic}(26iReZ^3>vFvQHbAfyhn<&l^%_g%aO(U!?nr(KBIWZMk
z62A0abBc0Mb(WFr5Oxp=bD0MBGrc#d3PYC;ECrhvSP_1Dx|~C1?1>R?q16$F!VNew
zm!?4r*Yh|!rC>}LO5gW!j5Q4k)M4C;;Tl|;U%edMHO$xwaB%7M6CJW2VkqdeXG;b+
zf_t3hNq-65@u7A)86fsdZZVdY8w@vXsc>$mNl-FRL!7eOK!%!+BDk}BZQo=Zo{4bI
zLCHaPG$a)yL+!zUJmPW+>p9(LdfmH&nw=yk8QPf0r!y=s)1g`2ClZsL8(#vid#g<0
zpch?qWRrS&py{3>a6Dz|AnN+Uo@ca^5Owt9Eh0%4To1BCG@Z%$N2z|=c-ktYSNg4+
zVQF>7-J_Ev*n^*N^C(l-V=dgP9MC4OPoM@pb2)6G1ai-+A!srCj@S_IqL>`S_C5F{
ztIi}srdoGj!_F4@1mtOgWYC}hZ_tN$)W$Yol=<|rlTccdFPh%_bt!zj|J_KvaR$hs
z>s@`O;L4Q;CN*(zM*l@#{f-Pkcd4%ZndNG(WZ^<`Q}h|Utr$GD?1&%h`~#=5+Tt;Z
zS6>nu$L@!rj9X@Q<hj(*ItuDjDFz2kDhW{Zr^!Hpat64z8de>L35spU%aj$^lLw&D
zn)ia-3sASQjriGCw+nBq$g2409h=9$7u^qXD=A*@<1>sD8D`EWTz!{WL9!sboth3(
zf^!%PN-NIf7OYp-I}{zH48Kwv`TX!<>gb)&{41aFiiDdyMTE2^v%A+sSKVl5-qwp)
zY&zYo=W7I9rPy|Ob~m79X=Mf3yeejbV-y4XfScqlyAC3h;&9zr>W<{7W&u|Wl{<|+
zJw2(|7Qc=Oi21IrrP`rapPMr28<#U?_Ny`FhU>ALyZ}VqFmomti0IoDfj&C?xt(N#
z;@PWvyDy^*ADv;9SCvo(uIk-tbj>Nu>kqqTc00`-v>IjkqRUlw;k~|eyci<@oWG@Y
z`Ghi_IFfFd>DBT24=mU$27P{L!^rK*@?yn{^XAsdMvrv-c;W$XW{wM&bQ#YQNiN+o
z2jWiU`Es`6dEb8&hA5w?D_AI4nAsYYiuhVE6~cH6Gv6DJR=Rh74EEurID2%eG4qu}
z&qJuQd#5)lWmUh6TrUKWuiSM$0iqt+QSziY5;P>f!M2OOYgEi}<EzWoiiemMuC^CU
zeyCZ$`BFrC!rA1O$9-`sjRi6jKc{$;UxsRDCvE80wijkSeR~NqTwQz-Q7w=vR9RQh
zxc|>AYis*$PZ6Z!>Ep@Rx#{B#6SRfauV1H)W3M(2l~k;?7N4u;@n+1LAO99M-D~74
z7iM?w_k3<HR{Ak6Zq-=q4{qVBfX<F#no5CcWAW<Pg<f@^m~GctJf5(-P-fDm_D<37
zpDaU)Ou}D=)j20!jw@&ogeO*%cW^ozgN;Y#Zw+dX@u#N^(@xl;pC<;j5NUJdf-C&J
zF%ovr;-AN6r%i;{NfNeO&YwihBotGcBc44KggVf=BV{gQ!!KV>ZW_=V35XrI8g#h~
zo9t?gGg~{Mb;P+RA#b<v5W27NxN2bgTwQYMjsE74&p(_Gmn-x=wOUAMwkB({m4{;&
P|I^!A*;|$)2NVAdd)|j1

literal 0
HcmV?d00001

diff --git a/profiling/mem-smallu.png b/profiling/mem-smallu.png
new file mode 100644
index 0000000000000000000000000000000000000000..b0b14862fe480d4a9d2f70e6a2e23d620993617e
GIT binary patch
literal 4560
zcmb7I2Ut_t@{b~r1dz~0q+F^JiWEUmKrSsPf`Wo53mTdzEus`@3O7h|xnQA_6(cB}
zP(%@MX$Cd0NGK98G+`;i5|AqJ;_mag|M&Lo`~LU)?sw0bbLKa5=FH5QbK_1}TkxTG
zqY(%MpC!)J7J=Z9K_IrD*}=i4Ts%WEVSfZqSe-H(8X96ZGiaI!#1<lo{UY$3{H<)#
zwyU<*c8DY6eID!%${*(#fI#eQ{`zjahutZGKnOTkni`)9&ia&nE~QXE!#wV&c2c~Z
zkbk930~YUDht1_*ARU$S6-`ShOI|7K?PtJ!Q7#u9V-sGQZDT|J>+wgopE`X(IUvxm
zKAy7q^ajD1h{dGbVY&k)CBTQcCy$OO9(6I9nX#ziw%avFzRy~p*nHYLy|se{&$@3Y
zUVad-Ez%UM^m#>RlvPvOB4t=x{ow`X&{n!%=Q)52cve$PhjRzrR=0Kn=@%iDN+4no
zyVVdF1zQcgW-{+dGG&Gr&LJ&dTIpFRnc}u%WVyiRBeA>0gl)5hbcB!s8~|R%loyJ<
ziP88r_MJF?9}{}sJXeD>=)?bS^J;#AJ7z!L_|7$lEMIz=Xc`T-KuAl(??S!<{U>@N
z7QOnb^UlAU@_)7RL+O2-)IvI11^)>1Z*?AoVhLt|^x&Iq|E3swaeoJzcnrNdmN4~q
zQ~o43*aYV&GZy^KGO5C>wz@qS+17I)I@HZx)|hGwLgb+<U<Q{YfNbOZpQ~m=6OIGY
z1AnD;A*ei&^s}<3hmfp1=VkB<KRvxa#u#VVqwARGU%wn_0OAzwi7QjXPtVrJ(=*kz
zBJSl;x|TKT3hs<}1Q(O?E>FdH*`73Em6f-#T+Dm5Wsz-7_79y`eo*O$!@TE~dhRu6
z8snr6?ic>H>3^Y1Yba<Pyv7TPe$@ggFnR@a+W*D$n*si}**;+K0x5{!{Yg!EETFbQ
z?!S81e>%mVEpJ0bC2@dMJb3Q+4QTC#(mH?TF#h}9{c7Kn04en0S`=L1fqkHgxpMf}
zfo3ZQaUkT3;7DGk7c|Cp7!_ZNvW{c^OXvH$5)QaMZ)F<_FEMh_#QG0sTB0<cS3j2M
z4$xt{A>0pa(DJag?BhNARC-|=RDFaxUzV!S$|B&py=&|R`(;ZpTD<Y<!XsHOABUa;
zAghcpHAdLZ352kHG2!;}1{qsjws*WGE>Wx)1yy0$aGsuSA<E|_Jc@c9vN)77PLIp9
z>AS!K?ddqt9_Un)3|=t|AAbYg9i{j{MZT2R7vMN`A<g$|+77%9N%RUHf{Vh~Q$^)r
z1ej8G{kXx4GBsIp$1mJGlT+q;jK`TW!#g1P;hw^>irPA48?D%ZunjQOa>1IG^R0>k
zA4^r&hA%a0AT1lG+RPqa+rv@?k4|sx&3^@)%eTg)>|n<9<?)s41lQK_r0U2U-Gs~h
zv#;~5_9=0zAiV?*rG8cgpKlaa?{hqMkPEOp!+e9Tc4S-7e`;)UiX_f7SCDCDtZ)I<
zVTQJT%M?rs$>O2&+ZlI2{!ksvWTx9G;a>qvS#si(Q51Yw#YLm2x!t{ryMQU<wi*8*
z0B<qHlp{~GbY{8fl@qo;R9#fM-nDng2Qg!wmbl*aY#n*s;}ZCUd>uEn;WQyAZ4Sf2
zgOL26@_;>&{)$M8)Jny$)f2*_yS6NFYV0to$Lt2!Y4VKNyFI&mGgZ30>)-zRIY~No
zd~)MdX@R)I`;qL$v!DEX!uEyp5q4)5Eu~QRe@-sy!SsGeJ|D)%j-4Kf&m})Tt^{1v
z$pv5X7;{UzA-54*?ii%wv9xosSPUhDEi%{+*P7oU<0c1yDeMGUqR=ekrpIpkYXH&Q
zkc>P(JFE&o!zJ@;-_vf*eK>cwO@z!p11#|k(kn9nQemfT!SAs*iU$u~qhzr~{GQ}S
zuK_x{JP^P+l+XURFDo55z&yXc^J&{4Z-vUA_j4fMAH5iTfeS;q_X}4OuZ$b^0bh*x
z5vn~X!K-xxE{4AIBEV|F{_|-!pCBMBX4(IPRqO52qWD*FGz`EUjnPoUj@kL0viPeQ
zt%-;8Sr*hqJ_fJAmHmI2NBnqJizh$qQ*xEi&1u&;q|?<>4TdesiApCkENl8&W>pfO
z5~0Zc`I;FP<v`BcCGE{p>C8zy4YNvX@%y+$dUZW!SXQj|>3$3=8GP+~P>S&&V00>2
zpV8H~BCFBWl3_8lEZT67F;$*x{0Ck%R2j**;1u{QNz%rHlz6+M{)<uqKlGxyug}RL
zY2!YjW?A%*k+nRzwEtvCpOb>F;JXe{K;Cy=;@JTYd)yd)S+w;e_&PoHb|{UeAED=v
zM7vsk6J|?EKIbZRc?~I6+m}r|xkq8FTiul2+xP7HdBp^2R7gc&(RhEF;cL5^AU3|5
zE$V&w<bbn+=XSuP=Qw=p=~sTc8e^)Q(^nDZgf!S8Ncs8vlwrJ<)>nS7^5kbr4fJKC
z7CLBA%3kX*j#^kdsKL$Cepo$1c8ul<S2=|O;UeiVp|0f>u0}_w<|U=d<gCO`_48Mk
z{AY-FeV?x+!w#Yn>C@w{%s#Drs!MR84$iFX1yEiQ$-(bb3q@?(^Nd8hvZ}l`t)$v-
zJ$npmmcsSP)`|Q-sW%JfqY23L8*m6F3Ks1;ByORbPWdH5JOek`<Oq33!TWf?sHSML
zj2jvp*R#Wg?$ykFyO-?O-E|8qKtX|dDm-oFy%NYa%&Z{jq@hGE3xU#R-52(JJ_C8*
zfR$dL0}9}hXCRIT96u>!`&>ouSK{3e`<fT=@I)p*hNXR35u@4h1dL+7lq_3?(!7V!
z3Gg)xt2q}|(F;blRWz=Fc}5HaI3IE%aYI*C(5p(Nux1B3RZKuLpu@);p!te|XEBr;
zIhahjFDAJA{;2|*(G8ud2GOvSr($lCjJ!q7F`hQT51dUZPQiEXDI5m#aLBe5vP1}w
z@G-|MR{Zvr8}MliYX>OHJ}CK+j+}svjJzfQzWWzB;vSqhVNEBhMCw-DP0md0ie@ut
zasuTjffTL_fQfWfh;3>zPc&KVokdLvOs~yFrsFdr=H5sv?PM;}-YAjh(0f&yvz40{
zb!CNz4-#tdCb5&gM|HGNTF6Pink=O%c-FT|P2DS+b2+vc{Ul?H3z+Q4Hbp%tOfI!-
zU!^Od95*kGWAIXTadCw}UijEg3ul!b(ti8g;kLGtx-fcaenI(Km{9ZnI9G!n&ugOx
zkpv}giU&zvck{49iJx)6j_Jk_*-N<7tGlnn!0E3vZyVag@u2r5z`m#DmMPX3O)xB|
zRN`|UjYosMne7il&mY`6lQ%zsBKWMQE1j9|d(<WJPKql_fo#C_T5*g(eWdC(F0RBe
zNvyO0#iI<QcHM-#9~nVp!=NkT0QOOi3DtxH6eSJ)a>qEf4NDfMq8-D|zdUHHynEX!
zR5|?Rh!fQ-Q$>vTRiwhsTWt5!bgy_3fOO2U7PLM~^+3ssuquD61mDUb!4c6MGwQ&-
z_hy(B!#&`LSkCZcZXi^2^D|h@BjY7t5H3lX=>pu610Z3oa4}J1T#SDPsH5+gl%nvG
zFM3aB&e~Nh+N%`uRy-ygL2uI!smwPVi~@D8t1QVY=Ms`y=$(qgro^Hkc-m62c}zre
zGAUolt#I82R^uF<HG1;UU|9#GLY%;p_G~A1)?Fs$XoHG_ky2y<)&K)%M?E_clXDep
zZ0F;cgmw(O5sQ8}HfOolZ8qly1e&MLeuTW8pGzto&X|0}Xybm;)P{}&ECp26v>(C#
zXRUEx;`8_Bek?T+1hjg8b{54mX>YePRmWDVlh_NE%eo6wD_(GLi*<f+{BWvM)JB;H
zMTl@&)&^Q@y4|H-*geHqTgIYhr{hP?0}Dzb(H;CPv|B<#c)WEw?;dRP>ch?~aJ)%R
z5}@t76Dzma$$P8w?(r%L)s?6c(<LAn4jKAChFjzobJzqD1{R<+1PdV@hfe_5&|YQG
z`3zHz=vEIkP=Zut<SBMdLfokLeS?xIl0rTY_*>-Ou7XwzbduQ7<y}DeFu$O!NcUR%
zz$zFx_XAO@{&pN+WFoBi%T)V79N3Q!4mV>8+xlc-^>mzOUgzv3hv>g=<*2m}DTJ)n
zUzAB6SSa@I#D=o#J(kU`sd#+xlJESxqC8f|<LP+C4Z=k+-&e2AZ-t!0b$4#m1uc07
z-h(G|pB`_;GNXeUA0Zdy$w}gGeza;S$F8l1BGMb{l$*ck(#*z7C%Sc-`=YhBtA0PM
z<_j>M{O!ofzqv8xV(!fBEb!z0ueXv!G2%f;e*1qP?}gH?c6{0h{C@6#E7#ct^Aa-T
zaH9aw7JNL`1gFftiAVuSE@V@jrn5-OY;SGDo`04Ut!h9MJIAGzo8W#knL%3qaqCk@
zX-r1ctxvyKAV*#`0^gOJuu-_a30jA));c0BxxwaGu@QF}c|D|nD+w++ZGqEw797E(
z{862c5Vh4XDH+*d|DXy){-&CM>R4nZ0NPgA;X3Co1M&jw8b<_(s3pRtGNy@0prrf;
ze0X3g_Xl++*WElxZzhqJZN+7Qb^g=n5nk{Lp%9CEcD?B*F<ZvUdQdZoX|&uB-5ahM
z{;4;dxkxga?uwjC^M#5c1}_7r4c15v=3hKhraYCeGZ)$C*_w-?)5coRZ#f;erI+8Q
zkNN6vkQ#LIqRnhPBej+VDw6Uffa)_33ISw@D!XgM<5&#TpVUQ$9QZm&)V}>()!v_G
z1iiSw9{XMid`**Na((N_rWZ9fxK4zUB5Bafv8htmG&OyBX8iV-1XDl8(CLWTL-DJg
z{U?4N9*Adj{8%ecS%`ycJ&GF#K1fGy3TwGE>W7CJ%ahec8C6ixsjd^a{li;psJsa=
z|E@k+=F(jX46i3K`>UH6b#ELuEDo&(EcnrxD@)nA$wLjHw6Tp3^jnJVo!(Woe9Ohh
z!W4~`3sKLQnUrOb^`4#6DLw&#en-1)HV68*OBr4wR^IufSeQMt8OGc0(j*xIU!*gv
z=!%!LvhPPIgk4}Vf}}VHo;?qWWUlGBt^BGJrmB#UOftx9U#K`#0JNA;ZIs+rjF93!
z6j6wGQmj8Uk$$<iWY~SzCEVVzi5?!WjHC;9x+Bh3f)V--M%g!vse3o#3ImVp12l&Q
zY0;LlwoAg@P1R3!UdAnV4n955lyAm(ygb?AM>ir?g%5YEyxTIuXH|Jvs>E-8{X=4D
LW^Gymo{#zmAkfvS

literal 0
HcmV?d00001

diff --git a/profiling/mem.png b/profiling/mem.png
new file mode 100644
index 0000000000000000000000000000000000000000..fb01a2b5d06ce6bff8e0bd691b2ec53262554eaf
GIT binary patch
literal 3243
zcmbtX2{e>@8=p2y)`V=CnVTCTCXpk<^rkWq$`wf(vXrYC*VvNnG-|ZbG?h^)-6&<N
zlw}xntTUkzS!X1axyBd`rlHxssjqWBeWm-I`<?Ut&-=d5|Nm^i=Xrn6dC%>2cUD=c
zw-N*bsq9AX@C1RDK|r9T2bM3BNsb-JagyCM_U`f8`S$Hw+2f6ND-cKu!pRos$O^S)
znP^G0r~7Z9ZDW0*vKPfakiJnMkZR-Swd9(asy+y$Y_)p_;B_K<a&R#AE_La)z#fIE
zSK%*}9%ncR%DfqZLk2_A{i-V-te9T2`o-WCar-eZm+j2y+}&GuLU$95{(aRr$JA)|
z-ajcgl4wTSTjEwyFjrS<dI4*L?6L{NZp=8ZV`T1u{>?g4%g$Irwy54wCv25gAKK+(
zyv&f4OlKxp{dtoy%AO+6)7v+$Ma)~1X1OHY@uT#6NtA|x4W3QX!dNedjAtfH`p5u@
z8C$KF`BFPyTTt!M>_C@>VRE5Ys_7fCJZtFFt(d3Yj*(hmNhO}6?75yLu>335=AxVt
ziAz6AL+9qZoY2%+zlVjet@xmW5L=Iz^|Riw%SDiwMD60lk^3m;8`jhT$2X81FA{GC
z-!$$9pGMmr3?r0n%mf@CT3C{>^rMmj$Z&FNK2GyeytoyYnD~}mNF*ehv{Yo$mT4mL
zNcCNV4JYbh6)lyp%~CS?+9jSx&U&)&UFLkGolf4|XlhX{Qkj-UhR_0!AuZ12S4XTc
zRlz>3QMwL4IUdB<fmM0APaSq&nxQXPefx8$f9Oj^VTN(2K0ivJ@e#sEg1n-@&rvrx
zFe;R=DK)WYjl3f5%=KP%nrjEFR&#xv+Nk1qc=|BBt!`KR4aQ^#hT8?VY>Te1pKPHS
z7{%A8y}Ab@CzzDu%qhpLsYs-S&O0qR-LEI+2OuFQO)nZp`!t9;W{(e-!*0RsrSl1y
zGpolk6Q)i#Q1pVi`p1ifcc!t5w3-MXaD^LKL>n|-&5d$xZf{S)U>fd`J!UAdE=462
zE3&@x?jAQBr#8oNtlP#YQGD0LO5q|9pu99lT_^Rq&hlmmOzzaCZ$x8FkcheYtcd#W
zw|;Q?ry@tgJYn2b6|Cl0SC(>$^5*A%S;X;AO)U~}RxfT&$t&&K4E~9sg=n;e1NEye
z+@(Tb`ar=%naPa;7}CNI7@<LQ(Irr~SI+>w)ZkJw-V2@n(d$`A0XG`$5m?C?Z}n*R
zeqw<fU2jHe&n+tOHy+VDLV%as`VQ{P6cP@V_Yh*AYNvY1(Hhs_zk~gM37VqWh`O^i
zV?q7KMrE-W(Ek5YErRbiuxuwx*hoTz{io#YUU%wgNJW#PZK(#on<Bf!L8Pc<Dp-@$
z=jNEw60@5{e=w&L6D-$;oN?0L`D#SZ$`?>3(ncKfeEZW*k7lwMn?JW4`~{9@q7-U(
zHk|55nRRh6sDKaM1_CmNlH;!MwOhlfoig(;a?P{e_d<Iy4SaY9B<*1GJww}rG=Yi=
zv8CD1_>%2jWnvl-pdH8N>Hjim(FnO7e|5#vPeI$dnKI=KidLU|iQM8Vp_n^>YD`LY
z!(8AN>P;q?g%ve;`>Wgf&O2=J4UA3@taU;?WW^<9TNeAv5-IAQdYC_Sz!qu9M-Q4D
zjbrb(`rz0EkuQiat>*+j0<eV<rpon?H5-$tHyho0e$}j)=|1Hk!-C)5^r9Eq-(+Ld
zsPuj|=4`|*%1&n5X$)9nJ2pz{SIKpncE#n{^P6qnxoP8O6D`f^g!_6PoB8VZZKnR&
zbE%Yw&IG0(fTxa};a&Y*6fSF)bzG+xEFZRbpKK$6^ZLT|T=Uf6I3+B(jS@35K@bru
zhHT|ODq|mX=OhbiibQNo_<ndw3rsRw{pj_~gduA$Goy@KB_<0AqN<kovEnNhXTy=n
zt^htKISx$pt58#H>T(Rm1kfRA)4i+y!wsQkNMYyWD(^D@_90_*@LI|mGCR4y12X@*
zDMSn|C@GNGZc`yT`)3wDQY^e*wPllwko|<?-u~fL7c_q^y&j9!3Jdb$p*4F6@Tw%_
z0o{%j50#0oHMMxJBy|C_f7aPIF!E$VAA{>O0IPuT&W6WVIu^=UzI)k_@cD2qkij4J
zVawTNF6j4#U&)M%x>S>U>Tfz1tk5ZC_%3c`aU%Tct4$tfJ&4+7MSb!&p8lIi7sk4n
z7m9|FLAkY+Li!>(e$nHodGe*rIY36p_oO{D3`qx6f3UHb$FH2<zp`^z<(4O>f$6@L
zSoCIDpB-I?+|U2QY!O$q=!}^61JU_I4bC3&**DrDkt$)#;P<69_(sSz0NcN&F$uZF
z&z>LKyGkv=EH!vU$!Q@|ESxZ*DTMnM0tu<QKbkdPAc}{cO}ROT{3r1JaDaV_d)6Qz
zF|9&W1pO?8@8$<srBf0?Shsl4-)fAVfecsPy9^v#yhS*wowG;f?6Q7V=pUrRQRq}o
zse24V%Ej!%h&luI>l)=&DFiwot8vz{N<3s}<sC}87(^&<rnj~~$|0dx{`n~TW)aiw
zk-<z&c3p5Si{4hDR9bcZ8Bj8rCY3nwa%Nq?w&Y$E!|6PCEV>pkQk<`*+RwZGrW}UA
z-<;sVd+^b~_9sUUI4Ki1QQ3MLaqMfr;e%P=hIX)G=j3arCwIpsn~^1u_hY9bS|DT`
zXI4;<Kb(SAi@*K;k`JCUw<p3ci!=ruSK_=W9}kIPuDkWJy7{&{BWTtxe+!jdTslFn
zi|S(-x%#wNn)?sDGmYC2+QW|(o%_(;9V4Ykqi{t<_rnP9Tqw~lC3N~7{PD|-UlHlI
z5fedbhi|O?1Yu$989BWb`zbXYD2Hwy{-f^DkZ1I&fpBs64A~D{bve=#abKx4ijkD(
zNCiK`XVV9RP%SMK1Y%80T`-HgFWtu@MhY4-<^!J!#k~yFS_{XQZzNhhVxP)zJ9T@0
zGIT&=607Em@whe@kR$XnRQu4uh94@45E6uWCtBlm)cVq2i$BH;;V3DQ&f!R&0#*;l
zl-9}XqD!@L%oa-y%IZAo4Y%R&X`vob?wozP1UeN97PxsM%Kl2t#(U>OpF2>GiS+B9
zA-dG_64*m8FO1aE*cxRlamiMHC;V<!<s-7k#!t6Cp+z~fZNYx_+LAc<xKz!`e|Y%@
U*W4wtUkcFfo$fnIwg=$;1r7Inb^rhX

literal 0
HcmV?d00001

diff --git a/profiling/old-mem-small.png b/profiling/old-mem-small.png
new file mode 100644
index 0000000000000000000000000000000000000000..8a3b47a4b22487b5cb4cdfeb631cb27d70e2039d
GIT binary patch
literal 3655
zcmeHKdo)|w9zG|Y^$b<5_fgbpnHoX8iPEW36m3)$(~6pSloF#f2<;G6qz}vxO;<%L
zD6O}8bQL8g6|Jfn?}ju+2^A_bh?|<;wPx<UYaVOOx@+CF?jPrz@7w#g_x^sr^_{c#
z$vEbISXN3^3IG7vBhF4<03ZSZ0PLiwh_L0WlVnHXL-E+r<A>h7c_V}cy$Ar<1X6_;
za8^vdOV|Wrz1+P4B>Ts6Lb5H&**69N<T^iIV6K6jns8pm5hsV^7fGD4#2bYz{awt;
zT`P+NmTFNNA)x<`wdAt#ML`31cS<;TS9ZH}-0k)GD^GP^sy;Qt+iGP<YF~jL{Pj)i
z`_|dbdh+dOfw|=eW{V!wE*e=HW|@jRL))M+&sW-tNNimVXYb>vtLqcN1>}5smB8b9
zm21NC2>-*yslE2EE(hjM&m}P$s_(b{v|*WO+IC6Vb4dx)LB4|snUmg5x$5xP=x1SM
zjtGTI8k;PgX?Z#yaWAJUu8&y|oTaqihbaT<I+fTFPMKe4W@{b;ws}(J`mdzVI)3Pj
zU*Mtg8e);5DW|J?qvwJ@_9gcW??N6t4lKfvBr$jh*y#&gP~L)k9c00T$ji6Qm_@O>
zTo}3^lVfsZ8wF$q9|wkPPxh7q)t4R?c!Fr&muAH<+WgbdFT{vrmtS9#f|@1DRdyi1
zmPeWF!_R3}ock{fIDPr6Q)XE!D#x+i#z$YhID_fhllueY)?^vHq*R|9T2JxauNph9
zz4So8m45q~8gkI(1fYNfd_Ow@IsqWM?zk8n6eYg)+JO{>BT=sZ()l(}!pjDaKbimx
zoi2kB02nQlDh0tLq4JV_cu5Ffkyi%n#sk%-{#57pT4McxaBw`Z*X&)3Bim1eTXgF$
z>G?OY`p~nWC$g09B4ufD2m>);4i#QV0PNiNA0U4gIXCRP7s3N)b74XmsJ{h<i#PG$
z{swpp5YwTBT*r&vsFaui7Ga8}!v_MIhm)FQ!y?++mcM>=uQhhqh?-b!`c3^x*y)M1
z5|xJLVAr`3tMM4m;)O#Gl-Z?2T(kX)L6nl7!JH=ZAw^ygVh~MY8&@^Dg+~t0e{;C;
zT<jSG2GOgSVO>VnembwofFoV9tQJa>z4Mu6VubC)HQUr${3;MHz7+(-pDPJ<<y0ua
zk-#Gt5EO-%|F1&U4%&vMAhXW~e@p0RPmWCddSLn9Ux+M4O%OQQqD{7f@au0OPLXS#
z0rAk5$%@BS5Y`R=f7zBYrTNjo&eQ=N$-Zi?iY4GHx>closZnW-u=ov^4Jk?J9IC8-
zfZ?K%I=WyK`3_G@ftIW#A+q?V*s_#brga*>K^FYz;No$;a%tEQ?-SeYjLZA54wHgP
zz9@<VxX)c)=MDQ(4-;2quM3o6gsbwzUyNToUQ+=PpPZDS43sWq$E6-+t3Sc^yQy-i
zB3>F!_iJFf<d^!$f|*-7lKF=GH+o@K)Kv|~04lv3S(*6UQp78CAimo=^4-l&6Wz@&
ztF7f)K(vi{!AwP0VM3u9%UgmnsnUOK#Dgv4Y*!rQfK^u|_H^;DspF&MdqS}9^=4I@
ztJrd2v^p)}R6sh8C6n<A!>)2%?eFQH*H%p*#GyJOnbGpsJyKaQ^A?U=l|)UnWFY-U
z&ZuMBv`<!(BVJvi`u6(HUoxNiEQDr39;z6gsqzH3)cMV8_kNDGn`ZhdLtF*8e=3%w
z&A(;`M7JupVcTcn{88J^Q4NrVB%U2+=2}%b+5ujI>@%w~al$l17{tbn<!Sun2I&$3
zC$1~^z!lDbUl2hEmaTcjj}p$uqyv^*nIXk3_XDi&a?3uetNG`PPwa&j-=0}5F#1#(
zin0DB5Bp<M&Q#B7jYy#B)Xj(yra9FLnGM6Vt5mlcMXaWx`gRAZCadmlxhkXVws~iB
zL?%azGZ9cZ#O*a+j<aiPJ>9LBbgkBOC}C<;f6A2J*5=lsN4GlEPOhK25&M&E=a0D{
zMT~Wfm>(j^BL+DP^7L?h51*vlY!=<S;EF6uS&rMjma(vFOYp@Gm1^TH>w+{BgA;zC
zAtiVKMmQn#AO7V3;h(KTvM9mbM_4##3<D!3EZr?O3H8?~PL4HR2P0?^QKyiD!q3VD
zC1(kO8ZoNpoyT=QeNOAm5!=c-{^NI?dv2FzYXiO3&={!*&JT@kvaTlQ<LG%Ss8T<*
zg{IP%yt^prn1OJ<YW}Kme?NM*d$ce-W6fA44B@`?t<UJ5dh-ZgzYa-TBj3<I#2LVV
zi(>7zoeIvl;St!R1R6123lnZ#E-~%~!p)0f`nS+>wpEt0%b)}M_0p~TQE*yj(+N2!
zv)@W0Aq?=nHE8+5PlR^2mXR`E5P3P=LRFlw@d));9%kH4EfOalqFTChfNDcCA>%+x
zVPl!|JL8SZd8f0VsEj4=n%IfVd%2Omm<XbOLx&Q*JunLtmp0kzy6ia@X+jk+X^6Lw
zh8QqH&F{l%4YZ)>K|#&WnKc@$5BBd}$BCQ?ZwVH0YZ$#DS%xsM?N7|}dkbZqAB$c}
zW5sbffrp`3VdVQ+$v;;1do%uVZWrn7DH)nEQ|8n0@>eI#-@*fUtbzzJC_3-nW>854
zX!&`FXPv{(+bLn7<-TcwDCT)G6M<h=wNLg>3gEou-X%vI2;c7T+?yiNl)1h-k$jIo
ztu&q{tLe5`6uJBUv#)&VITiax&RQ}m^OCP>g$2@caHHXcJ>3QKv>>}po9^#qZe}e7
zA>8ezk_sj7X8H!J2u~C2OIBmod8yxqI5$!o@P;ec$2tol8RoJ6M-5yr4v&C{iB~h(
zEee!)ey6z{jOGcAAT(^`ZpyJHUl_7i5@;3iv5gV&iw?Ynx#0oSfz6PT#E3QC^Kan@
zI=kxNjRiI~Bb~3LDU<vlD?iSj%O<^5pP$Kp|K(~}B<{z~wwvK!U#Q{^og-uP_mXBe
zRtm|Z1HDZ?7FjViBR+aNp)O3pW%R<8!}sSlxPZ~h#zRtCjc$@u(cm8IE(Mxi06M*4
znr>0E6pZ63d3+F<2uPLf9p~j10`@QB2!2;72C<@#jE2%#BxG6<BBxQBNY~R`i)m@^
zYRJb@TRUwf7jAbI#bafmj%G}V4lC$Mw>f#d<4p7g{k|%9xR&{cGm5QF_A}S@63eb!
zPCU}kj>R7?pZ53Cgs5fq#$6oyWVItEH)?~78U)yC6Q|qIeFYYOR5#`5eqRG9hZBhV
zLX()pYuA+XVKE;q*)WLex}&CPj+&8LIj(~*3`A&{TB|HTp4hXZHCktHqT{?i;1Q`;
z^p1H)J~G`eHYx3ze|}U|{POov-X0^?jnqxov$i`0ak&0Xd&TBGJ55&TAAcEvBZu6b
JDi3~{`VYmn_#OZN

literal 0
HcmV?d00001

diff --git a/profiling/old-mem-smallu.png b/profiling/old-mem-smallu.png
new file mode 100644
index 0000000000000000000000000000000000000000..7043e14d88f16fcb1b23df6186e5e61c98711f1d
GIT binary patch
literal 4240
zcmeHKX;hO}8cxSx!xn;2m4e|zLF)jbs8D2yA4^F}4Hy?>uM*tYcLD;(Pz@a|q^v4S
zh7w8ZLJ&n3fs7M0(FhSjLDmqG5W*5l2x|oT>1==I&vd3|ezf<Td+xpOdCq;e=Y8+@
zeeHwy)YbY}3kHMf;;`-n7;FOogZ<$Yd;=soe~RV?Et`D4IPNhrG6HRHwcEmA>o77j
zVP`b-o<pJ!5(xNXu)||-&O!&x7_47B42EobH$Nz}L>j|js4X~m@OVn;bXjK9=&6rA
z?&q}R_PpBodarf^;n@0)BVp?SMwk-e(?G#YEH1H3;BjbQ$ftH64jd-jseWh0KSzoI
zO+Cy1Nl~{{HlF-Mc}~<{VO)hl(M`jxG<M@Yr>4iGXQXw{bSF1uO7~X1DGrik4$N>v
z?AJq{W$G2*ommN&pDb^Zr`lO?8}h9b+ufF#$4Q;+)7GgCs<k?8o|-M4@DI9_o_Nk`
zNOkSatGm567kQ#PQ)a@Q*~nr+P(N)8dS+rRu4XbyvHrxM*gh>$iYpauC7#y1$sf0P
z8IYwq2)0h4gihH_aa@084xfa<P!#Qq`L(L2KW_2XXnb;5@e;7W`r;>e{D)@|d8Z#H
zQ2`@Fm%fCAP<})v0$5L~%#7=c$h!<|1IsBuh#B^3h&fkL_DSOpEG#wo<wG?Bbo%@y
zV)Q8CGgNM#Ejgv2K)<b$ejPwtgu!P&*}Ws%6KC{0<TntPc+iD|QdjK2Y8eneS=Tqu
zb|&f$W$4uI!e$>~4J58ck*;}tbNWV35PM(`MmRp4jLB<F$5luvUpkzfrZ`jX8H&;0
zwMNX6>k`yfX<Yt;enn96knzw&5>9xk(&or!bQat2N<21amOfm(Q@bm6@djB3t{JOy
z#{MQ?WVdS4=C7G|T2T1#q(lA<kW3tDvHG2m|5+9Gxl;LjrT3*^SE`Oz8n-wzybv%d
zB2UROi<jb_8k-@&JwAlTDcltaHPIbs#IJ=)fAb=B00w_ilY;DvM>>4>ugVwu6+xx{
z9#fQ6Loqo@cy5{a_l#JxRy)ilv-4`D&TZCMS1J=_&Q?xR`SQOc`X8q1@R;3Rm#@)`
zH<@RVCM9n-YJs*WsHsHE(t|Sq+j3WgGKf3~*{GMPJFa|o2(%4^H}+4+@8_@nNVZox
zZtScp2F?`}l#x}v#?Bfd>cI;7)Fp0H4<55mQn>O}UTr9Kh|7jJlCrdao@aW`z??nS
z5Uon~o?t(NE(p+YpuZksw7}&`aLf?y=Z@<mpgPQ;8w=ffs<+B=d62nm`z;>^Qw$-Q
zGjY>GJ<}~^QUW;?hwMy1cHT~23uq{H+gVT$1aBOia8BbIS%hhWXH#s#U2%PKzaG)?
zS_Ek8lu?v;^fE+VS@MIZX7+qY<#z1VIHa7_N}$og^obC5nLa#3D)OL`-y5yy2-ww(
zyy-(onq5$bBJxC{Xx3bC59CpM!lQ}SwYHkK3y#e0T?GY^IiCfsqhBYu$V_Z2=n-0R
zviteF5oufqdK;h!Gh`-REn*YG^_JO9<{L7%J^gnrJCG`uo+j~&0i!@FUu;tVr9;zJ
zXHH1(H!%MSL)4|I_jxTg(uVs}`BItjkXqr{B||TT=^0&pSIs3+jEwq#!BGe9VX9$C
zYeZx;IR^RlzEwoMpex{-^xRdk1M!~nm>{~lMOK?u86bGV+$U7Y78u7CztnrNciOB)
z?c9|E?9Kt-xD_2G*p+pMWjc9K4W$=m4BI02gSJ0KF@Jwse!DGBrAk(XxcH?9z-Pzh
zGYwi@rqSXQ67)Hey5fk}MJTmYssmu~pUTU3a=khQ0yt@RYU(Y(#>3m&KyrIreu3O<
z=<IxiTKpY@5n?Qf$0Wc>7guhBSEB%KzJRG$yNkPDU<0vS8fxD>t(^Ntk-jMYfwZmH
zS9}{2ST(TpVC?bR4Xi1%G^F3%KBHIlOh=+F{KCO$yDRb8p4N<XYU~+?XDGzh!_?0N
zF5IY6`RZl^dix_?+s8M^pH6xwuKw+P``B9m(vTD@^;V+Zh5apKCWHl)r3JP`UiY5(
zJZQaF2t-z`;?kW$ix@3YROm1nvK0+rOD>SRD~km1A0U02gj|2T0Ucdex)bYs(cuU+
zwL9(>F#o;u061}!_iF*yuLEpZG@Gd}S?qJ$NKW~m=)?>blmp>DJo&&8d9LZ0HhA*S
zMS^y6^E~G~hSuce54k*L_TP9GZ7&@&bzcEd{|H&@P(ev7($CW8R}0ktF-N~Z^*Qn&
zi=B`x-(MdS_;#8+==W&!4q3B{nK5F}X5}Ay;;9C1#>^|(8btaUv{jzCb~5k;fOYn-
zJOAv221LnUDmcGcQb7UkR{s%2N8k5{bh>5%YI#gunq`J{2si%j9gcku25p+98k@D(
zIffEnw!03}IpFhcL=4K@9n&+erYt6D+A6jOWY_El(TeS%RDQKuj9G6I3G20yUVB)J
zV%~jgtL4yV*6XB|BFvt%crU(iUM4THi5lhSPhJ#p<eR;#^Pfk`NmmpRy{uO&fkNeS
zQWdZgXA`t{#tgpPbXE4z#cdK1nt-3!04gFNp#R33_=%M``yik41%;+EQI~JoeE=Ix
zTf&EbLA41_39hIDduBP)k=C`!jWpIWW(H3y8a8fL4=rvE9v*E0AG#*TOoW%SIIt13
zyTU79v3Q)MrV`A{UD7$GY=9@C^tx)>4ONv2>ZSR!{G}ISQFrR9^`4%n{yV(7mNzxI
zijG#74rWH&oWEUYb}su+6nbWZa632aM6*uZtjF?{cyv;2U$eZ^Q<6)vxruf)73|_&
zsssk)fMe!hgkJdLK6LBb8w(4w`WPE|PKiTi;}BlWKG2uQ$(lEmh(-F_|2npn*#CV9
zz)f--S`6b1zcQ_#>r8U_sUsvKVocXD=-`JpSx%8tNgXK-dPr?yQ!Ga;fAp;&sy6+R
zt8kR=!fZQrCa~A7NbGiduy3GU6p_-LrFTE)+>7?u5^vu^ivjjqwf7Fr!CSq2Ctj+W
z&4_tGNss}i6o7$Y0KSxaR{#d-;pzv^+bsh=HT{<1-Rj72me_JRkX?t;OObvQN}q{D
z{CYx+8ZneyT>GYEAv_LY>#WK(eIl+7R;ZT6m+rI<x*>fdd5vQUk@mFi+&SS0eN9i)
g886Cwp0&1q%6Q+IvR4DX?|$%c9(ea^*DuLG1IWahV*mgE

literal 0
HcmV?d00001

diff --git a/profiling/old-mem.png b/profiling/old-mem.png
new file mode 100644
index 0000000000000000000000000000000000000000..4fd2389b27003b64584db9aa20110f8aa7908dad
GIT binary patch
literal 3048
zcmds3dsNcd7N#t;L@9I3M<En7)|i#~$QpGJE$mu8voOp@fElJ<ll0DjPSRjbPX40n
zqfD8rDOo-W-{PYN71GiymC-`-m4cyRf&!w0_c4Fob?2_Rf81H?oOSlzXP@t^z0Wz{
zK50HDjvA~tUk?I-3_LOJejtzz90Xc@dX0|Oa{e?EtsRVfyaGJly?dt>@4T=DftEoj
z+71d|3vJh$R>k|B@CSYM{%4|ASr?1JUIKv(TUYw3%r6WrKp_1XPj^&6;$5lGAUYQV
z`YJ4d>-<2m`n*u*^aaCjt7bJ7Y)GgD)K#bCKx3IGJNnK+!>x<9kR@@A71EROk9LoS
z%T=5r&shz1@QQFt;U@!xdkOv~<HrC1xh68s;HZaQ9Ck6hulW2=#>)A{F#)eqG!g95
zFMT^mYU5;`3O3H}+vWFNt<R1JrWYfe#HFqGFGowOM}j*a8uCi3HPF<D`_K@^Q6a26
zkceq1NVNh#E41M@G*fh{CK?5#lU`Z!#NN&`Ihm$?U1d#yjikTyNi9XNs#jb)3{2i1
zW{!l?yjnR~*{Gdb&`$U!S@}2tNJ|vV+Vb7C-e&l1L)DEC+l@0pXTi6xCG6Sz`?#y3
zfW*Vu;p5fr{%O8#xr3ErY;ATvdGFUjZ{i~XCt0x~>gz0&kx}9Ij@1#bD<KTP^kid+
zb8bEXxY~Mu<hXQT&YFfq9{DWhLU(b584|JoL8AXR4$htMb^pQxdt%ae5B26B@3zIb
zg$MjbNX>UB!>!{M`}NXMp^(+*Hki_$n67Du?*m`?gPBU;f?}&X9q<8UO2F_7xY-UU
z_GC{L%AAI0K#VQ~WFA0f?{q@B0qyYmdqj+XfZ*W)By+65WlJ@L;klbHA6#1kVWcHM
z!v2r4WRilb7-lZmVUqrI0J-?77VPxI4=7$vNHbvMqC!oe*i%q!>PKsR;-Bg6n_A(w
zt;-QVkobtdN7&5{{jxI!Yd8XNx;o#hjk<F2BPX>KZQTv<dvvIq^9fI_MbP@ox;8X?
z4aM7eOVyEX3z+lEgqw+zV=_?3iVye-7T3$&v`uOh4m@8$qnh~;fDcDd9V~lf?|BGA
z=*%yo9^lKSwU%|SbY(!MXcUm!p;Qr&CRTokb3@a0`=ijg;8GWsOH#OI@=Eh{m4jo)
zfJc)phA7|DMo?4<;PrjA#^A8i>qpGvfLOjtFST?@!!FcS*#uhRN*gU1!8r(Fuhpm;
z;SwKr?GYW<U_WN-f|n=j3+$0CzJRP?=pfpVy<yD78^RdSSd&g4Guf(*<8e}!{UPFt
z<m(D+`<{rAFADD8fK77rNEoLk))13mhHT*#oPB;*D{#&L3a+?dS_3Q|fwoi}{`$w6
zaCzU1OIBuxP}tixXU6>~Nth{_5WE1U0r}#gAETWU44zf!+kSfKxw+MhA@T!;`)0}!
z^<)yC&_MUvqTO~l<mk7*KEj*;V0t=?07OX?lmr!C#g!{kUabTv89BEHb5fps>%o4^
zPBON9==0K_bb>CL(C|k2R@Q=IOrk!ta0RZX4AHp>uQvjDRy<^f#-Z8bZng;vli2kr
z7(KXqCCk0zJf13T=AAap`ZFcmKPC_SkBQA91(&NOnR~PV4+21O07pU?&uY+t<B(@X
z19oV0+8A33Xq9&RHN?@#7y{tccuw!XX5CL>cyj-gTX~Cor*eG}s#qaPuQ&&osbks&
z`}ye_$I&TMeEU+aj0~|oxBOz^5cu}vWg;a{in0kgxL!y-gdz83FT?|7B3AHj+N1yA
zymO?ek%*e+xPX`_wSS#>`Dw-3C;zM!6`d)+Ue@1GM2Z`un%=+vLgGzP_06cvS5D`-
z0^VHXsQ>#HBD>s}2-?^%<lmIGN(AP0+dd48>{}B<de8wk>uVcl!|vt5j}ri#<Z{Yi
zA1<eMZn#$_o0?Y&c&g-yRzXu2eJUOm$ii5R1sRK$lpB=;%MK0icd5E!YNxzfrg)>d
zoEh_A1(lsA&SKgq9css2>yU%dbtfo2-Ig*BSI*6hy=rCrRQghB4`JR_^Ef0J==4hE
zB+*Url%7+%*gQm-Zv|vFzULHJY$EUUy2gnee6uo_4c>L@^JXD44A|xmT%0bvPJRDP
zvFOW&59(R4tD5&=mP+KjvALt{s^5xh@_I#zN;39i25x3uN)!z_M<rKJ;nk4~p<R&;
z<&w6+wM|q~jMxVCW(OsxJMUF|2+{jy&!i-_*cK*L%7#mFZcj<&j?A}|qS(>Y1nHdK
zc4^1ej0hfGk8OLLhdRQ%nsK7Ad&1ENeS=SS{rM%5@^UNN?Ur(CHT|uw(>$*~scs?+
zFV)qsSV;+zVMf9A4VP!%THfR+V_}Nfc-W0W?yx$PoP1c6blAZ1QUs>uNguiHg?fHs
ziB~dtqgJ~vEMq!ELYreMv=sk(nA@`_cCw4ib^3w+dRlsHLu<hhv|!K1_Q8mH&Hbwu
zY8>>n=CTg7jNT@zak<fKY~utK=vr?VVpK)jA#-!4^B*!Z9Pc37w(}$P;+Ew(pGMLB
zZh%_8xj#FdUMZZO9MBa*OH!dxJyd!+?XKhu5Us4UWKX*Wv222zx3DD>aBqQYtYC>r
zhYK^7q*ULxP;794g1UUkJ|Dr&?g}fbPp+QaEPo?kbzKz_F2ya(Ec|U5`qYP%=kS@Q
RxAs2{^7J_2UhWo}@(aKQBq;y@

literal 0
HcmV?d00001

diff --git a/profiling/release-flame-old.svg b/profiling/release-flame-old.svg
new file mode 100644
index 0000000..e6586b8
--- /dev/null
+++ b/profiling/release-flame-old.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="310" onload="init(evt)" viewBox="0 0 1200 310" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="310" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="293.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="293.00"> </text><svg id="frames" x="10" width="1180"><g><title>_IO_file_xsgetn (20 samples, 0.22%)</title><rect x="0.0000%" y="213" width="0.2179%" height="15" fill="rgb(227,0,7)"/><text x="0.2500%" y="223.50"></text></g><g><title>_IO_new_file_xsputn (12 samples, 0.13%)</title><rect x="0.3595%" y="213" width="0.1307%" height="15" fill="rgb(217,0,24)"/><text x="0.6095%" y="223.50"></text></g><g><title>__GI__IO_file_seekoff (156 samples, 1.70%)</title><rect x="0.5229%" y="213" width="1.6995%" height="15" fill="rgb(221,193,54)"/><text x="0.7729%" y="223.50"></text></g><g><title>__GI__IO_fread (37 samples, 0.40%)</title><rect x="2.2225%" y="213" width="0.4031%" height="15" fill="rgb(248,212,6)"/><text x="2.4725%" y="223.50"></text></g><g><title>__GI__IO_free_backup_area (25 samples, 0.27%)</title><rect x="2.6256%" y="213" width="0.2724%" height="15" fill="rgb(208,68,35)"/><text x="2.8756%" y="223.50"></text></g><g><title>__GI__IO_fwrite (66 samples, 0.72%)</title><rect x="2.8979%" y="213" width="0.7190%" height="15" fill="rgb(232,128,0)"/><text x="3.1479%" y="223.50"></text></g><g><title>__GI__IO_sgetn (30 samples, 0.33%)</title><rect x="3.6170%" y="213" width="0.3268%" height="15" fill="rgb(207,160,47)"/><text x="3.8670%" y="223.50"></text></g><g><title>__GI__IO_switch_to_get_mode (21 samples, 0.23%)</title><rect x="3.9438%" y="213" width="0.2288%" height="15" fill="rgb(228,23,34)"/><text x="4.1938%" y="223.50"></text></g><g><title>__GI___libc_free (24 samples, 0.26%)</title><rect x="4.1726%" y="213" width="0.2615%" height="15" fill="rgb(218,30,26)"/><text x="4.4226%" y="223.50"></text></g><g><title>__GI___libc_malloc (11 samples, 0.12%)</title><rect x="4.4340%" y="213" width="0.1198%" height="15" fill="rgb(220,122,19)"/><text x="4.6840%" y="223.50"></text></g><g><title>__GI_fseek (34 samples, 0.37%)</title><rect x="4.5539%" y="213" width="0.3704%" height="15" fill="rgb(250,228,42)"/><text x="4.8039%" y="223.50"></text></g><g><title>__lseek64 (74 samples, 0.81%)</title><rect x="4.9243%" y="213" width="0.8062%" height="15" fill="rgb(240,193,28)"/><text x="5.1743%" y="223.50"></text></g><g><title>_ar_file_get (16 samples, 0.17%)</title><rect x="5.7849%" y="213" width="0.1743%" height="15" fill="rgb(216,20,37)"/><text x="6.0349%" y="223.50"></text></g><g><title>_ar_file_set (21 samples, 0.23%)</title><rect x="5.9593%" y="213" width="0.2288%" height="15" fill="rgb(206,188,39)"/><text x="6.2093%" y="223.50"></text></g><g><title>_int_free (46 samples, 0.50%)</title><rect x="6.1880%" y="213" width="0.5011%" height="15" fill="rgb(217,207,13)"/><text x="6.4380%" y="223.50"></text></g><g><title>new_do_write (29 samples, 0.32%)</title><rect x="6.7545%" y="213" width="0.3159%" height="15" fill="rgb(231,73,38)"/><text x="7.0045%" y="223.50"></text></g><g><title>[[heap]] (651 samples, 7.09%)</title><rect x="0.0000%" y="229" width="7.0923%" height="15" fill="rgb(225,20,46)"/><text x="0.2500%" y="239.50">[[heap]]</text></g><g><title>_ar_file_get (12 samples, 0.13%)</title><rect x="7.2230%" y="213" width="0.1307%" height="15" fill="rgb(210,31,41)"/><text x="7.4730%" y="223.50"></text></g><g><title>_ar_file_set (19 samples, 0.21%)</title><rect x="7.3537%" y="213" width="0.2070%" height="15" fill="rgb(221,200,47)"/><text x="7.6037%" y="223.50"></text></g><g><title>[[stack]] (70 samples, 0.76%)</title><rect x="7.0923%" y="229" width="0.7626%" height="15" fill="rgb(226,26,5)"/><text x="7.3423%" y="239.50"></text></g><g><title>shuffle3 (27 samples, 0.29%)</title><rect x="7.5607%" y="213" width="0.2941%" height="15" fill="rgb(249,33,26)"/><text x="7.8107%" y="223.50"></text></g><g><title>_IO_file_xsgetn (42 samples, 0.46%)</title><rect x="7.8549%" y="213" width="0.4576%" height="15" fill="rgb(235,183,28)"/><text x="8.1049%" y="223.50"></text></g><g><title>_IO_seekoff_unlocked (20 samples, 0.22%)</title><rect x="8.3125%" y="213" width="0.2179%" height="15" fill="rgb(221,5,38)"/><text x="8.5625%" y="223.50"></text></g><g><title>[anon] (70 samples, 0.76%)</title><rect x="7.8549%" y="229" width="0.7626%" height="15" fill="rgb(247,18,42)"/><text x="8.1049%" y="239.50"></text></g><g><title>_IO_new_file_xsputn (45 samples, 0.49%)</title><rect x="8.6175%" y="213" width="0.4902%" height="15" fill="rgb(241,131,45)"/><text x="8.8675%" y="223.50"></text></g><g><title>[libc-2.32.so] (91 samples, 0.99%)</title><rect x="8.6175%" y="229" width="0.9914%" height="15" fill="rgb(249,31,29)"/><text x="8.8675%" y="239.50"></text></g><g><title>__GI__IO_file_seekoff (38 samples, 0.41%)</title><rect x="9.1949%" y="213" width="0.4140%" height="15" fill="rgb(225,111,53)"/><text x="9.4449%" y="223.50"></text></g><g><title>_IO_new_file_write (33 samples, 0.36%)</title><rect x="9.6852%" y="213" width="0.3595%" height="15" fill="rgb(238,160,17)"/><text x="9.9352%" y="223.50"></text></g><g><title>_IO_seekoff_unlocked (12 samples, 0.13%)</title><rect x="10.0447%" y="213" width="0.1307%" height="15" fill="rgb(214,148,48)"/><text x="10.2947%" y="223.50"></text></g><g><title>__GI__IO_default_xsputn (28 samples, 0.31%)</title><rect x="10.1754%" y="213" width="0.3050%" height="15" fill="rgb(232,36,49)"/><text x="10.4254%" y="223.50"></text></g><g><title>__GI__IO_file_overflow (33 samples, 0.36%)</title><rect x="10.4804%" y="213" width="0.3595%" height="15" fill="rgb(209,103,24)"/><text x="10.7304%" y="223.50"></text></g><g><title>__GI__IO_free_backup_area (12 samples, 0.13%)</title><rect x="10.9271%" y="213" width="0.1307%" height="15" fill="rgb(229,88,8)"/><text x="11.1771%" y="223.50"></text></g><g><title>__GI__IO_switch_to_get_mode (16 samples, 0.17%)</title><rect x="11.0796%" y="213" width="0.1743%" height="15" fill="rgb(213,181,19)"/><text x="11.3296%" y="223.50"></text></g><g><title>__GI___libc_malloc (14 samples, 0.15%)</title><rect x="11.2757%" y="213" width="0.1525%" height="15" fill="rgb(254,191,54)"/><text x="11.5257%" y="223.50"></text></g><g><title>__GI___libc_write (72 samples, 0.78%)</title><rect x="11.4283%" y="213" width="0.7844%" height="15" fill="rgb(241,83,37)"/><text x="11.6783%" y="223.50"></text></g><g><title>__GI___read (149 samples, 1.62%)</title><rect x="12.2127%" y="213" width="1.6233%" height="15" fill="rgb(233,36,39)"/><text x="12.4627%" y="223.50"></text></g><g><title>__GI_fseek (64 samples, 0.70%)</title><rect x="13.8359%" y="213" width="0.6972%" height="15" fill="rgb(226,3,54)"/><text x="14.0859%" y="223.50"></text></g><g><title>__lseek64 (151 samples, 1.65%)</title><rect x="14.5332%" y="213" width="1.6451%" height="15" fill="rgb(245,192,40)"/><text x="14.7832%" y="223.50"></text></g><g><title>__random (10 samples, 0.11%)</title><rect x="16.1782%" y="213" width="0.1089%" height="15" fill="rgb(238,167,29)"/><text x="16.4282%" y="223.50"></text></g><g><title>_ar_file_get (25 samples, 0.27%)</title><rect x="16.2872%" y="213" width="0.2724%" height="15" fill="rgb(232,182,51)"/><text x="16.5372%" y="223.50"></text></g><g><title>_ar_file_set (12 samples, 0.13%)</title><rect x="16.5595%" y="213" width="0.1307%" height="15" fill="rgb(231,60,39)"/><text x="16.8095%" y="223.50"></text></g><g><title>new_do_write (151 samples, 1.65%)</title><rect x="16.7883%" y="213" width="1.6451%" height="15" fill="rgb(208,69,12)"/><text x="17.0383%" y="223.50"></text></g><g><title>[unknown] (811 samples, 8.84%)</title><rect x="9.6089%" y="229" width="8.8354%" height="15" fill="rgb(235,93,37)"/><text x="9.8589%" y="239.50">[unknown]</text></g><g><title>_IO_seekoff_unlocked (12 samples, 0.13%)</title><rect x="18.4770%" y="229" width="0.1307%" height="15" fill="rgb(213,116,39)"/><text x="18.7270%" y="239.50"></text></g><g><title>__GI__IO_file_seekoff (19 samples, 0.21%)</title><rect x="18.6077%" y="229" width="0.2070%" height="15" fill="rgb(222,207,29)"/><text x="18.8577%" y="239.50"></text></g><g><title>rand (86 samples, 0.94%)</title><rect x="20.5796%" y="149" width="0.9369%" height="15" fill="rgb(206,96,30)"/><text x="20.8296%" y="159.50"></text></g><g><title>__random (85 samples, 0.93%)</title><rect x="20.5905%" y="133" width="0.9260%" height="15" fill="rgb(218,138,4)"/><text x="20.8405%" y="143.50"></text></g><g><title>__random_r (62 samples, 0.68%)</title><rect x="20.8411%" y="117" width="0.6755%" height="15" fill="rgb(250,191,14)"/><text x="21.0911%" y="127.50"></text></g><g><title>_next (160 samples, 1.74%)</title><rect x="19.8170%" y="165" width="1.7431%" height="15" fill="rgb(239,60,40)"/><text x="20.0670%" y="175.50"></text></g><g><title>_sample (20 samples, 0.22%)</title><rect x="21.5601%" y="165" width="0.2179%" height="15" fill="rgb(206,27,48)"/><text x="21.8101%" y="175.50"></text></g><g><title>[libm-2.32.so] (202 samples, 2.20%)</title><rect x="22.3445%" y="149" width="2.2007%" height="15" fill="rgb(225,35,8)"/><text x="22.5945%" y="159.50">[..</text></g><g><title>_sample.lto_priv.0 (259 samples, 2.82%)</title><rect x="21.7780%" y="165" width="2.8217%" height="15" fill="rgb(250,213,24)"/><text x="22.0280%" y="175.50">_s..</text></g><g><title>__GI__IO_default_xsputn (125 samples, 1.36%)</title><rect x="31.2779%" y="117" width="1.3618%" height="15" fill="rgb(247,123,22)"/><text x="31.5279%" y="127.50"></text></g><g><title>__GI__IO_fwrite (752 samples, 8.19%)</title><rect x="25.2860%" y="149" width="8.1926%" height="15" fill="rgb(231,138,38)"/><text x="25.5360%" y="159.50">__GI__IO_fw..</text></g><g><title>_IO_new_file_xsputn (572 samples, 6.23%)</title><rect x="27.2470%" y="133" width="6.2316%" height="15" fill="rgb(231,145,46)"/><text x="27.4970%" y="143.50">_IO_new_..</text></g><g><title>__GI__IO_file_overflow (77 samples, 0.84%)</title><rect x="32.6397%" y="117" width="0.8389%" height="15" fill="rgb(251,118,11)"/><text x="32.8897%" y="127.50"></text></g><g><title>__memmove_avx_unaligned_erms (20 samples, 0.22%)</title><rect x="33.4786%" y="149" width="0.2179%" height="15" fill="rgb(217,147,25)"/><text x="33.7286%" y="159.50"></text></g><g><title>_IO_file_xsgetn (91 samples, 0.99%)</title><rect x="36.7905%" y="117" width="0.9914%" height="15" fill="rgb(247,81,37)"/><text x="37.0405%" y="127.50"></text></g><g><title>__memmove_avx_unaligned_erms (29 samples, 0.32%)</title><rect x="37.4660%" y="101" width="0.3159%" height="15" fill="rgb(209,12,38)"/><text x="37.7160%" y="111.50"></text></g><g><title>__GI__IO_fread (270 samples, 2.94%)</title><rect x="35.2435%" y="133" width="2.9415%" height="15" fill="rgb(227,1,9)"/><text x="35.4935%" y="143.50">__..</text></g><g><title>__GI__IO_sgetn (37 samples, 0.40%)</title><rect x="37.7819%" y="117" width="0.4031%" height="15" fill="rgb(248,47,43)"/><text x="38.0319%" y="127.50"></text></g><g><title>__GI___libc_free (151 samples, 1.65%)</title><rect x="38.1850%" y="133" width="1.6451%" height="15" fill="rgb(221,10,30)"/><text x="38.4350%" y="143.50"></text></g><g><title>__GI___libc_malloc (65 samples, 0.71%)</title><rect x="39.8300%" y="133" width="0.7081%" height="15" fill="rgb(210,229,1)"/><text x="40.0800%" y="143.50"></text></g><g><title>_IO_seekoff_unlocked (35 samples, 0.38%)</title><rect x="42.0852%" y="117" width="0.3813%" height="15" fill="rgb(222,148,37)"/><text x="42.3352%" y="127.50"></text></g><g><title>__GI__IO_file_read (20 samples, 0.22%)</title><rect x="47.2165%" y="101" width="0.2179%" height="15" fill="rgb(234,67,33)"/><text x="47.4665%" y="111.50"></text></g><g><title>__GI__IO_file_seek (12 samples, 0.13%)</title><rect x="47.4344%" y="101" width="0.1307%" height="15" fill="rgb(247,98,35)"/><text x="47.6844%" y="111.50"></text></g><g><title>[libc-2.32.so] (30 samples, 0.33%)</title><rect x="47.6849%" y="85" width="0.3268%" height="15" fill="rgb(247,138,52)"/><text x="47.9349%" y="95.50"></text></g><g><title>__GI__IO_free_backup_area (128 samples, 1.39%)</title><rect x="47.5651%" y="101" width="1.3945%" height="15" fill="rgb(213,79,30)"/><text x="47.8151%" y="111.50"></text></g><g><title>__GI___libc_free (87 samples, 0.95%)</title><rect x="48.0118%" y="85" width="0.9478%" height="15" fill="rgb(246,177,23)"/><text x="48.2618%" y="95.50"></text></g><g><title>_IO_new_file_write (285 samples, 3.10%)</title><rect x="50.5066%" y="53" width="3.1049%" height="15" fill="rgb(230,62,27)"/><text x="50.7566%" y="63.50">_IO..</text></g><g><title>__GI___libc_write (179 samples, 1.95%)</title><rect x="51.6614%" y="37" width="1.9501%" height="15" fill="rgb(216,154,8)"/><text x="51.9114%" y="47.50">_..</text></g><g><title>_IO_new_do_write (541 samples, 5.89%)</title><rect x="49.3736%" y="85" width="5.8939%" height="15" fill="rgb(244,35,45)"/><text x="49.6236%" y="95.50">_IO_new..</text></g><g><title>new_do_write (530 samples, 5.77%)</title><rect x="49.4934%" y="69" width="5.7740%" height="15" fill="rgb(251,115,12)"/><text x="49.7434%" y="79.50">new_do_..</text></g><g><title>__lseek64 (149 samples, 1.62%)</title><rect x="53.6442%" y="53" width="1.6233%" height="15" fill="rgb(240,54,50)"/><text x="53.8942%" y="63.50"></text></g><g><title>__GI__IO_switch_to_get_mode (591 samples, 6.44%)</title><rect x="48.9596%" y="101" width="6.4386%" height="15" fill="rgb(233,84,52)"/><text x="49.2096%" y="111.50">__GI__IO..</text></g><g><title>__GI__IO_file_overflow (12 samples, 0.13%)</title><rect x="55.2675%" y="85" width="0.1307%" height="15" fill="rgb(207,117,47)"/><text x="55.5175%" y="95.50"></text></g><g><title>__GI___read (371 samples, 4.04%)</title><rect x="55.3982%" y="101" width="4.0418%" height="15" fill="rgb(249,43,39)"/><text x="55.6482%" y="111.50">__GI..</text></g><g><title>__GI_fseek (2,112 samples, 23.01%)</title><rect x="40.5382%" y="133" width="23.0090%" height="15" fill="rgb(209,38,44)"/><text x="40.7882%" y="143.50">__GI_fseek</text></g><g><title>__GI__IO_file_seekoff (1,935 samples, 21.08%)</title><rect x="42.4665%" y="117" width="21.0807%" height="15" fill="rgb(236,212,23)"/><text x="42.7165%" y="127.50">__GI__IO_file_seekoff</text></g><g><title>__lseek64 (377 samples, 4.11%)</title><rect x="59.4400%" y="101" width="4.1072%" height="15" fill="rgb(242,79,21)"/><text x="59.6900%" y="111.50">__ls..</text></g><g><title>_init (74 samples, 0.81%)</title><rect x="63.5472%" y="133" width="0.8062%" height="15" fill="rgb(211,96,35)"/><text x="63.7972%" y="143.50"></text></g><g><title>_int_free (115 samples, 1.25%)</title><rect x="64.3534%" y="133" width="1.2529%" height="15" fill="rgb(253,215,40)"/><text x="64.6034%" y="143.50"></text></g><g><title>fseek@plt (10 samples, 0.11%)</title><rect x="65.6607%" y="133" width="0.1089%" height="15" fill="rgb(211,81,21)"/><text x="65.9107%" y="143.50"></text></g><g><title>_ar_file_get (2,954 samples, 32.18%)</title><rect x="33.6965%" y="149" width="32.1822%" height="15" fill="rgb(208,190,38)"/><text x="33.9465%" y="159.50">_ar_file_get</text></g><g><title>malloc@plt (10 samples, 0.11%)</title><rect x="65.7697%" y="133" width="0.1089%" height="15" fill="rgb(235,213,38)"/><text x="66.0197%" y="143.50"></text></g><g><title>_IO_seekoff_unlocked (47 samples, 0.51%)</title><rect x="68.7657%" y="117" width="0.5120%" height="15" fill="rgb(237,122,38)"/><text x="69.0157%" y="127.50"></text></g><g><title>__GI__IO_file_read (10 samples, 0.11%)</title><rect x="74.5070%" y="101" width="0.1089%" height="15" fill="rgb(244,218,35)"/><text x="74.7570%" y="111.50"></text></g><g><title>[libc-2.32.so] (21 samples, 0.23%)</title><rect x="74.8121%" y="85" width="0.2288%" height="15" fill="rgb(240,68,47)"/><text x="75.0621%" y="95.50"></text></g><g><title>__GI__IO_free_backup_area (114 samples, 1.24%)</title><rect x="74.6922%" y="101" width="1.2420%" height="15" fill="rgb(210,16,53)"/><text x="74.9422%" y="111.50"></text></g><g><title>__GI___libc_free (82 samples, 0.89%)</title><rect x="75.0409%" y="85" width="0.8933%" height="15" fill="rgb(235,124,12)"/><text x="75.2909%" y="95.50"></text></g><g><title>_IO_new_file_write (287 samples, 3.13%)</title><rect x="77.1980%" y="53" width="3.1267%" height="15" fill="rgb(224,169,11)"/><text x="77.4480%" y="63.50">_IO..</text></g><g><title>__GI___libc_write (174 samples, 1.90%)</title><rect x="78.4290%" y="37" width="1.8956%" height="15" fill="rgb(250,166,2)"/><text x="78.6790%" y="47.50">_..</text></g><g><title>_IO_new_do_write (571 samples, 6.22%)</title><rect x="76.1848%" y="85" width="6.2207%" height="15" fill="rgb(242,216,29)"/><text x="76.4348%" y="95.50">_IO_new_..</text></g><g><title>new_do_write (562 samples, 6.12%)</title><rect x="76.2828%" y="69" width="6.1227%" height="15" fill="rgb(230,116,27)"/><text x="76.5328%" y="79.50">new_do_w..</text></g><g><title>__lseek64 (190 samples, 2.07%)</title><rect x="80.3355%" y="53" width="2.0699%" height="15" fill="rgb(228,99,48)"/><text x="80.5855%" y="63.50">_..</text></g><g><title>__GI__IO_switch_to_get_mode (608 samples, 6.62%)</title><rect x="75.9342%" y="101" width="6.6238%" height="15" fill="rgb(253,11,6)"/><text x="76.1842%" y="111.50">__GI__IO_..</text></g><g><title>__GI__IO_file_overflow (14 samples, 0.15%)</title><rect x="82.4055%" y="85" width="0.1525%" height="15" fill="rgb(247,143,39)"/><text x="82.6555%" y="95.50"></text></g><g><title>__GI___read (377 samples, 4.11%)</title><rect x="82.5580%" y="101" width="4.1072%" height="15" fill="rgb(236,97,10)"/><text x="82.8080%" y="111.50">__GI..</text></g><g><title>__GI_fseek (2,180 samples, 23.75%)</title><rect x="66.9790%" y="133" width="23.7499%" height="15" fill="rgb(233,208,19)"/><text x="67.2290%" y="143.50">__GI_fseek</text></g><g><title>__GI__IO_file_seekoff (1,969 samples, 21.45%)</title><rect x="69.2777%" y="117" width="21.4511%" height="15" fill="rgb(216,164,2)"/><text x="69.5277%" y="127.50">__GI__IO_file_seekoff</text></g><g><title>__lseek64 (373 samples, 4.06%)</title><rect x="86.6652%" y="101" width="4.0636%" height="15" fill="rgb(220,129,5)"/><text x="86.9152%" y="111.50">__ls..</text></g><g><title>_ar_file_set (2,290 samples, 24.95%)</title><rect x="65.8786%" y="149" width="24.9483%" height="15" fill="rgb(242,17,10)"/><text x="66.1286%" y="159.50">_ar_file_set</text></g><g><title>fwrite@plt (43 samples, 0.47%)</title><rect x="90.8269%" y="149" width="0.4685%" height="15" fill="rgb(242,107,0)"/><text x="91.0769%" y="159.50"></text></g><g><title>ar_swap (6,127 samples, 66.75%)</title><rect x="24.5996%" y="165" width="66.7502%" height="15" fill="rgb(251,28,31)"/><text x="24.8496%" y="175.50">ar_swap</text></g><g><title>_IO_file_xsgetn (16 samples, 0.17%)</title><rect x="91.6767%" y="117" width="0.1743%" height="15" fill="rgb(233,223,10)"/><text x="91.9267%" y="127.50"></text></g><g><title>__GI__IO_fread (36 samples, 0.39%)</title><rect x="91.4806%" y="133" width="0.3922%" height="15" fill="rgb(215,21,27)"/><text x="91.7306%" y="143.50"></text></g><g><title>__GI_fseek (120 samples, 1.31%)</title><rect x="92.0035%" y="133" width="1.3073%" height="15" fill="rgb(232,23,21)"/><text x="92.2535%" y="143.50"></text></g><g><title>__GI__IO_file_seekoff (80 samples, 0.87%)</title><rect x="92.4393%" y="117" width="0.8716%" height="15" fill="rgb(244,5,23)"/><text x="92.6893%" y="127.50"></text></g><g><title>__lseek64 (44 samples, 0.48%)</title><rect x="92.8315%" y="101" width="0.4794%" height="15" fill="rgb(226,81,46)"/><text x="93.0815%" y="111.50"></text></g><g><title>_int_free (14 samples, 0.15%)</title><rect x="93.3217%" y="133" width="0.1525%" height="15" fill="rgb(247,70,30)"/><text x="93.5717%" y="143.50"></text></g><g><title>minmax_floats (201 samples, 2.19%)</title><rect x="91.3498%" y="165" width="2.1898%" height="15" fill="rgb(212,68,19)"/><text x="91.5998%" y="175.50">m..</text></g><g><title>_ar_file_get (197 samples, 2.15%)</title><rect x="91.3934%" y="149" width="2.1462%" height="15" fill="rgb(240,187,13)"/><text x="91.6434%" y="159.50">_..</text></g><g><title>__GI__IO_fread (10 samples, 0.11%)</title><rect x="93.5941%" y="133" width="0.1089%" height="15" fill="rgb(223,113,26)"/><text x="93.8441%" y="143.50"></text></g><g><title>__GI_fseek (31 samples, 0.34%)</title><rect x="93.7684%" y="133" width="0.3377%" height="15" fill="rgb(206,192,2)"/><text x="94.0184%" y="143.50"></text></g><g><title>__GI__IO_file_seekoff (19 samples, 0.21%)</title><rect x="93.8991%" y="117" width="0.2070%" height="15" fill="rgb(241,108,4)"/><text x="94.1491%" y="127.50"></text></g><g><title>__lseek64 (12 samples, 0.13%)</title><rect x="93.9754%" y="101" width="0.1307%" height="15" fill="rgb(247,173,49)"/><text x="94.2254%" y="111.50"></text></g><g><title>minmax_int64_ts (58 samples, 0.63%)</title><rect x="93.5396%" y="165" width="0.6319%" height="15" fill="rgb(224,114,35)"/><text x="93.7896%" y="175.50"></text></g><g><title>_ar_file_get (58 samples, 0.63%)</title><rect x="93.5396%" y="149" width="0.6319%" height="15" fill="rgb(245,159,27)"/><text x="93.7896%" y="159.50"></text></g><g><title>_IO_file_xsgetn (33 samples, 0.36%)</title><rect x="94.8905%" y="117" width="0.3595%" height="15" fill="rgb(245,172,44)"/><text x="95.1405%" y="127.50"></text></g><g><title>__GI__IO_fread (86 samples, 0.94%)</title><rect x="94.4329%" y="133" width="0.9369%" height="15" fill="rgb(236,23,11)"/><text x="94.6829%" y="143.50"></text></g><g><title>__GI__IO_sgetn (11 samples, 0.12%)</title><rect x="95.2500%" y="117" width="0.1198%" height="15" fill="rgb(205,117,38)"/><text x="95.5000%" y="127.50"></text></g><g><title>__GI___libc_free (15 samples, 0.16%)</title><rect x="95.3699%" y="133" width="0.1634%" height="15" fill="rgb(237,72,25)"/><text x="95.6199%" y="143.50"></text></g><g><title>__GI___libc_malloc (14 samples, 0.15%)</title><rect x="95.5333%" y="133" width="0.1525%" height="15" fill="rgb(244,70,9)"/><text x="95.7833%" y="143.50"></text></g><g><title>_IO_seekoff_unlocked (17 samples, 0.19%)</title><rect x="96.6772%" y="117" width="0.1852%" height="15" fill="rgb(217,125,39)"/><text x="96.9272%" y="127.50"></text></g><g><title>[libc-2.32.so] (10 samples, 0.11%)</title><rect x="97.7884%" y="85" width="0.1089%" height="15" fill="rgb(235,36,10)"/><text x="98.0384%" y="95.50"></text></g><g><title>__GI__IO_free_backup_area (19 samples, 0.21%)</title><rect x="97.7557%" y="101" width="0.2070%" height="15" fill="rgb(251,123,47)"/><text x="98.0057%" y="111.50"></text></g><g><title>__GI_fseek (338 samples, 3.68%)</title><rect x="95.6858%" y="133" width="3.6823%" height="15" fill="rgb(221,13,13)"/><text x="95.9358%" y="143.50">__GI..</text></g><g><title>__GI__IO_file_seekoff (230 samples, 2.51%)</title><rect x="96.8624%" y="117" width="2.5057%" height="15" fill="rgb(238,131,9)"/><text x="97.1124%" y="127.50">__..</text></g><g><title>__lseek64 (129 samples, 1.41%)</title><rect x="97.9627%" y="101" width="1.4054%" height="15" fill="rgb(211,50,8)"/><text x="98.2127%" y="111.50"></text></g><g><title>_int_free (36 samples, 0.39%)</title><rect x="99.3790%" y="133" width="0.3922%" height="15" fill="rgb(245,182,24)"/><text x="99.6290%" y="143.50"></text></g><g><title>__libc_start_main (7,433 samples, 80.98%)</title><rect x="18.9127%" y="213" width="80.9783%" height="15" fill="rgb(242,14,37)"/><text x="19.1627%" y="223.50">__libc_start_main</text></g><g><title>main (7,433 samples, 80.98%)</title><rect x="18.9127%" y="197" width="80.9783%" height="15" fill="rgb(246,228,12)"/><text x="19.1627%" y="207.50">main</text></g><g><title>shuffle3 (7,433 samples, 80.98%)</title><rect x="18.9127%" y="181" width="80.9783%" height="15" fill="rgb(213,55,15)"/><text x="19.1627%" y="191.50">shuffle3</text></g><g><title>minmax_sbytes (525 samples, 5.72%)</title><rect x="94.1715%" y="165" width="5.7196%" height="15" fill="rgb(209,9,3)"/><text x="94.4215%" y="175.50">minmax_..</text></g><g><title>_ar_file_get (514 samples, 5.60%)</title><rect x="94.2913%" y="149" width="5.5997%" height="15" fill="rgb(230,59,30)"/><text x="94.5413%" y="159.50">_ar_fil..</text></g><g><title>_start (7,438 samples, 81.03%)</title><rect x="18.8692%" y="229" width="81.0328%" height="15" fill="rgb(209,121,21)"/><text x="19.1192%" y="239.50">_start</text></g><g><title>all (9,179 samples, 100%)</title><rect x="0.0000%" y="261" width="100.0000%" height="15" fill="rgb(220,109,13)"/><text x="0.2500%" y="271.50"></text></g><g><title>shuffle3-releas (9,179 samples, 100.00%)</title><rect x="0.0000%" y="245" width="100.0000%" height="15" fill="rgb(232,18,1)"/><text x="0.2500%" y="255.50">shuffle3-releas</text></g></svg></svg>
\ No newline at end of file
diff --git a/profiling/release-flame.svg b/profiling/release-flame.svg
new file mode 100644
index 0000000..cdf9049
--- /dev/null
+++ b/profiling/release-flame.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="182" onload="init(evt)" viewBox="0 0 1200 182" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="182" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="165.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="165.00"> </text><svg id="frames" x="10" width="1180"><g><title>[libm-2.32.so] (9 samples, 26.47%)</title><rect x="58.8235%" y="53" width="26.4706%" height="15" fill="rgb(227,0,7)"/><text x="59.0735%" y="63.50">[libm-2.32.so]</text></g><g><title>__libc_start_main (29 samples, 85.29%)</title><rect x="11.7647%" y="85" width="85.2941%" height="15" fill="rgb(217,0,24)"/><text x="12.0147%" y="95.50">__libc_start_main</text></g><g><title>main (29 samples, 85.29%)</title><rect x="11.7647%" y="69" width="85.2941%" height="15" fill="rgb(221,193,54)"/><text x="12.0147%" y="79.50">main</text></g><g><title>rand_r (4 samples, 11.76%)</title><rect x="85.2941%" y="53" width="11.7647%" height="15" fill="rgb(248,212,6)"/><text x="85.5441%" y="63.50">rand_r</text></g><g><title>all (34 samples, 100%)</title><rect x="0.0000%" y="133" width="100.0000%" height="15" fill="rgb(208,68,35)"/><text x="0.2500%" y="143.50"></text></g><g><title>shuffle3-releas (34 samples, 100.00%)</title><rect x="0.0000%" y="117" width="100.0000%" height="15" fill="rgb(232,128,0)"/><text x="0.2500%" y="127.50">shuffle3-releas</text></g><g><title>_start (34 samples, 100.00%)</title><rect x="0.0000%" y="101" width="100.0000%" height="15" fill="rgb(207,160,47)"/><text x="0.2500%" y="111.50">_start</text></g><g><title>_dl_start (1 samples, 2.94%)</title><rect x="97.0588%" y="85" width="2.9412%" height="15" fill="rgb(228,23,34)"/><text x="97.3088%" y="95.50">_d..</text></g><g><title>_dl_sysdep_start (1 samples, 2.94%)</title><rect x="97.0588%" y="69" width="2.9412%" height="15" fill="rgb(218,30,26)"/><text x="97.3088%" y="79.50">_d..</text></g><g><title>dl_main (1 samples, 2.94%)</title><rect x="97.0588%" y="53" width="2.9412%" height="15" fill="rgb(220,122,19)"/><text x="97.3088%" y="63.50">dl..</text></g><g><title>_dl_lookup_symbol_x (1 samples, 2.94%)</title><rect x="97.0588%" y="37" width="2.9412%" height="15" fill="rgb(250,228,42)"/><text x="97.3088%" y="47.50">_d..</text></g></svg></svg>
\ No newline at end of file
diff --git a/profiling/release-flameu.svg b/profiling/release-flameu.svg
new file mode 100644
index 0000000..39a6ea5
--- /dev/null
+++ b/profiling/release-flameu.svg
@@ -0,0 +1,419 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg version="1.1" width="1200" height="214" onload="init(evt)" viewBox="0 0 1200 214" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><!--Flame graph stack visualization. See https://github.com/brendangregg/FlameGraph for latest version, and http://www.brendangregg.com/flamegraphs.html for examples.--><!--NOTES: --><defs><linearGradient id="background" y1="0" y2="1" x1="0" x2="0"><stop stop-color="#eeeeee" offset="5%"/><stop stop-color="#eeeeb0" offset="95%"/></linearGradient></defs><style type="text/css">
+text { font-family:"Verdana"; font-size:12px; fill:rgb(0,0,0); }
+#title { text-anchor:middle; font-size:17px; }
+#search { opacity:0.1; cursor:pointer; }
+#search:hover, #search.show { opacity:1; }
+#subtitle { text-anchor:middle; font-color:rgb(160,160,160); }
+#unzoom { cursor:pointer; }
+#frames > *:hover { stroke:black; stroke-width:0.5; cursor:pointer; }
+.hide { display:none; }
+.parent { opacity:0.5; }
+</style><script type="text/ecmascript"><![CDATA[var nametype = 'Function:';
+var fontsize = 12;
+var fontwidth = 0.59;
+var xpad = 10;
+var inverted = false;
+var searchcolor = 'rgb(230,0,230)';
+var fluiddrawing = true;
+var truncate_text_right = false;]]><![CDATA["use strict";
+var details, searchbtn, unzoombtn, matchedtxt, svg, searching, frames;
+function init(evt) {
+    details = document.getElementById("details").firstChild;
+    searchbtn = document.getElementById("search");
+    unzoombtn = document.getElementById("unzoom");
+    matchedtxt = document.getElementById("matched");
+    svg = document.getElementsByTagName("svg")[0];
+    frames = document.getElementById("frames");
+    searching = 0;
+
+    // Use GET parameters to restore a flamegraph's state.
+    var restore_state = function() {
+        var params = get_params();
+        if (params.x && params.y)
+            zoom(find_group(document.querySelector('[x="' + params.x + '"][y="' + params.y + '"]')));
+        if (params.s)
+            search(params.s);
+    };
+
+    if (fluiddrawing) {
+        // Make width dynamic so the SVG fits its parent's width.
+        svg.removeAttribute("width");
+        // Edge requires us to have a viewBox that gets updated with size changes.
+        var isEdge = /Edge\/\d./i.test(navigator.userAgent);
+        if (!isEdge) {
+          svg.removeAttribute("viewBox");
+        }
+        var update_for_width_change = function() {
+            if (isEdge) {
+                svg.attributes.viewBox.value = "0 0 " + svg.width.baseVal.value + " " + svg.height.baseVal.value;
+            }
+
+            // Keep consistent padding on left and right of frames container.
+            frames.attributes.width.value = svg.width.baseVal.value - xpad * 2;
+
+            // Text truncation needs to be adjusted for the current width.
+            var el = frames.children;
+            for(var i = 0; i < el.length; i++) {
+                update_text(el[i]);
+            }
+
+            // Keep search elements at a fixed distance from right edge.
+            var svgWidth = svg.width.baseVal.value;
+            searchbtn.attributes.x.value = svgWidth - xpad - 100;
+            matchedtxt.attributes.x.value = svgWidth - xpad - 100;
+        };
+        window.addEventListener('resize', function() {
+            update_for_width_change();
+        });
+        // This needs to be done asynchronously for Safari to work.
+        setTimeout(function() {
+            unzoom();
+            update_for_width_change();
+            restore_state();
+        }, 0);
+    } else {
+        restore_state();
+    }
+}
+// event listeners
+window.addEventListener("click", function(e) {
+    var target = find_group(e.target);
+    if (target) {
+        if (target.nodeName == "a") {
+            if (e.ctrlKey === false) return;
+            e.preventDefault();
+        }
+        if (target.classList.contains("parent")) unzoom();
+        zoom(target);
+
+        // set parameters for zoom state
+        var el = target.querySelector("rect");
+        if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
+            var params = get_params()
+            params.x = el.attributes._orig_x.value;
+            params.y = el.attributes.y.value;
+            history.replaceState(null, null, parse_params(params));
+        }
+    }
+    else if (e.target.id == "unzoom") {
+        unzoom();
+
+        // remove zoom state
+        var params = get_params();
+        if (params.x) delete params.x;
+        if (params.y) delete params.y;
+        history.replaceState(null, null, parse_params(params));
+    }
+    else if (e.target.id == "search") search_prompt();
+}, false)
+// mouse-over for info
+// show
+window.addEventListener("mouseover", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = nametype + " " + g_to_text(target);
+}, false)
+// clear
+window.addEventListener("mouseout", function(e) {
+    var target = find_group(e.target);
+    if (target) details.nodeValue = ' ';
+}, false)
+// ctrl-F for search
+window.addEventListener("keydown",function (e) {
+    if (e.keyCode === 114 || (e.ctrlKey && e.keyCode === 70)) {
+        e.preventDefault();
+        search_prompt();
+    }
+}, false)
+// functions
+function get_params() {
+    var params = {};
+    var paramsarr = window.location.search.substr(1).split('&');
+    for (var i = 0; i < paramsarr.length; ++i) {
+        var tmp = paramsarr[i].split("=");
+        if (!tmp[0] || !tmp[1]) continue;
+        params[tmp[0]]  = decodeURIComponent(tmp[1]);
+    }
+    return params;
+}
+function parse_params(params) {
+    var uri = "?";
+    for (var key in params) {
+        uri += key + '=' + encodeURIComponent(params[key]) + '&';
+    }
+    if (uri.slice(-1) == "&")
+        uri = uri.substring(0, uri.length - 1);
+    if (uri == '?')
+        uri = window.location.href.split('?')[0];
+    return uri;
+}
+function find_child(node, selector) {
+    var children = node.querySelectorAll(selector);
+    if (children.length) return children[0];
+    return;
+}
+function find_group(node) {
+    var parent = node.parentElement;
+    if (!parent) return;
+    if (parent.id == "frames") return node;
+    return find_group(parent);
+}
+function orig_save(e, attr, val) {
+    if (e.attributes["_orig_" + attr] != undefined) return;
+    if (e.attributes[attr] == undefined) return;
+    if (val == undefined) val = e.attributes[attr].value;
+    e.setAttribute("_orig_" + attr, val);
+}
+function orig_load(e, attr) {
+    if (e.attributes["_orig_"+attr] == undefined) return;
+    e.attributes[attr].value = e.attributes["_orig_" + attr].value;
+    e.removeAttribute("_orig_" + attr);
+}
+function g_to_text(e) {
+    var text = find_child(e, "title").firstChild.nodeValue;
+    return (text)
+}
+function g_to_func(e) {
+    var func = g_to_text(e);
+    // if there's any manipulation we want to do to the function
+    // name before it's searched, do it here before returning.
+    return (func);
+}
+function update_text(e) {
+    var r = find_child(e, "rect");
+    var t = find_child(e, "text");
+    var w = parseFloat(r.attributes.width.value) * frames.attributes.width.value / 100 - 3;
+    var txt = find_child(e, "title").textContent.replace(/\([^(]*\)$/,"");
+    t.attributes.x.value = format_percent((parseFloat(r.attributes.x.value) + (100 * 3 / frames.attributes.width.value)));
+    // Smaller than this size won't fit anything
+    if (w < 2 * fontsize * fontwidth) {
+        t.textContent = "";
+        return;
+    }
+    t.textContent = txt;
+    // Fit in full text width
+    if (/^ *\$/.test(txt) || t.getComputedTextLength() < w)
+        return;
+    if (truncate_text_right) {
+        // Truncate the right side of the text.
+        for (var x = txt.length - 2; x > 0; x--) {
+            if (t.getSubStringLength(0, x + 2) <= w) {
+                t.textContent = txt.substring(0, x) + "..";
+                return;
+            }
+        }
+    } else {
+        // Truncate the left side of the text.
+        for (var x = 2; x < txt.length; x++) {
+            if (t.getSubStringLength(x - 2, txt.length) <= w) {
+                t.textContent = ".." + txt.substring(x, txt.length);
+                return;
+            }
+        }
+    }
+    t.textContent = "";
+}
+// zoom
+function zoom_reset(e) {
+    if (e.attributes != undefined) {
+        orig_load(e, "x");
+        orig_load(e, "width");
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_reset(c[i]);
+    }
+}
+function zoom_child(e, x, ratio) {
+    if (e.attributes != undefined) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = format_percent((parseFloat(e.attributes.x.value) - x) * ratio);
+            if (e.tagName == "text") {
+                e.attributes.x.value = format_percent(parseFloat(find_child(e.parentNode, "rect[x]").attributes.x.value) + (100 * 3 / frames.attributes.width.value));
+            }
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = format_percent(parseFloat(e.attributes.width.value) * ratio);
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_child(c[i], x, ratio);
+    }
+}
+function zoom_parent(e) {
+    if (e.attributes) {
+        if (e.attributes.x != undefined) {
+            orig_save(e, "x");
+            e.attributes.x.value = "0.0%";
+        }
+        if (e.attributes.width != undefined) {
+            orig_save(e, "width");
+            e.attributes.width.value = "100.0%";
+        }
+    }
+    if (e.childNodes == undefined) return;
+    for(var i = 0, c = e.childNodes; i < c.length; i++) {
+        zoom_parent(c[i]);
+    }
+}
+function zoom(node) {
+    var attr = find_child(node, "rect").attributes;
+    var width = parseFloat(attr.width.value);
+    var xmin = parseFloat(attr.x.value);
+    var xmax = xmin + width;
+    var ymin = parseFloat(attr.y.value);
+    var ratio = 100 / width;
+    // XXX: Workaround for JavaScript float issues (fix me)
+    var fudge = 0.001;
+    unzoombtn.classList.remove("hide");
+    var el = frames.children;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var a = find_child(e, "rect").attributes;
+        var ex = parseFloat(a.x.value);
+        var ew = parseFloat(a.width.value);
+        // Is it an ancestor
+        if (!inverted) {
+            var upstack = parseFloat(a.y.value) > ymin;
+        } else {
+            var upstack = parseFloat(a.y.value) < ymin;
+        }
+        if (upstack) {
+            // Direct ancestor
+            if (ex <= xmin && (ex+ew+fudge) >= xmax) {
+                e.classList.add("parent");
+                zoom_parent(e);
+                update_text(e);
+            }
+            // not in current path
+            else
+                e.classList.add("hide");
+        }
+        // Children maybe
+        else {
+            // no common path
+            if (ex < xmin || ex + fudge >= xmax) {
+                e.classList.add("hide");
+            }
+            else {
+                zoom_child(e, xmin, ratio);
+                update_text(e);
+            }
+        }
+    }
+}
+function unzoom() {
+    unzoombtn.classList.add("hide");
+    var el = frames.children;
+    for(var i = 0; i < el.length; i++) {
+        el[i].classList.remove("parent");
+        el[i].classList.remove("hide");
+        zoom_reset(el[i]);
+        update_text(el[i]);
+    }
+}
+// search
+function reset_search() {
+    var el = document.querySelectorAll("#frames rect");
+    for (var i = 0; i < el.length; i++) {
+        orig_load(el[i], "fill")
+    }
+    var params = get_params();
+    delete params.s;
+    history.replaceState(null, null, parse_params(params));
+}
+function search_prompt() {
+    if (!searching) {
+        var term = prompt("Enter a search term (regexp " +
+            "allowed, eg: ^ext4_)", "");
+        if (term != null) {
+            search(term)
+        }
+    } else {
+        reset_search();
+        searching = 0;
+        searchbtn.classList.remove("show");
+        searchbtn.firstChild.nodeValue = "Search"
+        matchedtxt.classList.add("hide");
+        matchedtxt.firstChild.nodeValue = ""
+    }
+}
+function search(term) {
+    var re = new RegExp(term);
+    var el = frames.children;
+    var matches = new Object();
+    var maxwidth = 0;
+    for (var i = 0; i < el.length; i++) {
+        var e = el[i];
+        var func = g_to_func(e);
+        var rect = find_child(e, "rect");
+        if (func == null || rect == null)
+            continue;
+        // Save max width. Only works as we have a root frame
+        var w = parseFloat(rect.attributes.width.value);
+        if (w > maxwidth)
+            maxwidth = w;
+        if (func.match(re)) {
+            // highlight
+            var x = parseFloat(rect.attributes.x.value);
+            orig_save(rect, "fill");
+            rect.attributes.fill.value = searchcolor;
+            // remember matches
+            if (matches[x] == undefined) {
+                matches[x] = w;
+            } else {
+                if (w > matches[x]) {
+                    // overwrite with parent
+                    matches[x] = w;
+                }
+            }
+            searching = 1;
+        }
+    }
+    if (!searching)
+        return;
+    var params = get_params();
+    params.s = term;
+    history.replaceState(null, null, parse_params(params));
+
+    searchbtn.classList.add("show");
+    searchbtn.firstChild.nodeValue = "Reset Search";
+    // calculate percent matched, excluding vertical overlap
+    var count = 0;
+    var lastx = -1;
+    var lastw = 0;
+    var keys = Array();
+    for (k in matches) {
+        if (matches.hasOwnProperty(k))
+            keys.push(k);
+    }
+    // sort the matched frames by their x location
+    // ascending, then width descending
+    keys.sort(function(a, b){
+        return a - b;
+    });
+    // Step through frames saving only the biggest bottom-up frames
+    // thanks to the sort order. This relies on the tree property
+    // where children are always smaller than their parents.
+    var fudge = 0.0001;    // JavaScript floating point
+    for (var k in keys) {
+        var x = parseFloat(keys[k]);
+        var w = matches[keys[k]];
+        if (x >= lastx + lastw - fudge) {
+            count += w;
+            lastx = x;
+            lastw = w;
+        }
+    }
+    // display matched percent
+    matchedtxt.classList.remove("hide");
+    var pct = 100 * count / maxwidth;
+    if (pct != 100) pct = pct.toFixed(1);
+    matchedtxt.firstChild.nodeValue = "Matched: " + pct + "%";
+}
+function format_percent(n) {
+    return n.toFixed(4) + "%";
+}
+]]></script><rect x="0" y="0" width="100%" height="214" fill="url(#background)"/><text id="title" x="50.0000%" y="24.00">Flame Graph</text><text id="details" x="10" y="197.00"> </text><text id="unzoom" class="hide" x="10" y="24.00">Reset Zoom</text><text id="search" x="1090" y="24.00">Search</text><text id="matched" x="1090" y="197.00"> </text><svg id="frames" x="10" width="1180"><g><title>VTT for std::basic_ostream&lt;wchar_t, std::char_traits&lt;wchar_t&gt; &gt; (1 samples, 1.89%)</title><rect x="0.0000%" y="133" width="1.8868%" height="15" fill="rgb(227,0,7)"/><text x="0.2500%" y="143.50">V..</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (1 samples, 1.89%)</title><rect x="0.0000%" y="117" width="1.8868%" height="15" fill="rgb(217,0,24)"/><text x="0.2500%" y="127.50">s..</text></g><g><title>[libm-2.32.so] (1 samples, 1.89%)</title><rect x="1.8868%" y="117" width="1.8868%" height="15" fill="rgb(221,193,54)"/><text x="2.1368%" y="127.50">[..</text></g><g><title>main (1 samples, 1.89%)</title><rect x="3.7736%" y="117" width="1.8868%" height="15" fill="rgb(248,212,6)"/><text x="4.0236%" y="127.50">m..</text></g><g><title>[unknown] (4 samples, 7.55%)</title><rect x="1.8868%" y="133" width="7.5472%" height="15" fill="rgb(208,68,35)"/><text x="2.1368%" y="143.50">[unknown]</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (2 samples, 3.77%)</title><rect x="5.6604%" y="117" width="3.7736%" height="15" fill="rgb(232,128,0)"/><text x="5.9104%" y="127.50">std:..</text></g><g><title>[libm-2.32.so] (23 samples, 43.40%)</title><rect x="45.2830%" y="85" width="43.3962%" height="15" fill="rgb(207,160,47)"/><text x="45.5330%" y="95.50">[libm-2.32.so]</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::emplace_back&lt;unsigned long&gt; (4 samples, 7.55%)</title><rect x="88.6792%" y="85" width="7.5472%" height="15" fill="rgb(228,23,34)"/><text x="88.9292%" y="95.50">std::vecto..</text></g><g><title>__memmove_avx_unaligned_erms (1 samples, 1.89%)</title><rect x="94.3396%" y="69" width="1.8868%" height="15" fill="rgb(218,30,26)"/><text x="94.5896%" y="79.50">_..</text></g><g><title>__libc_start_main (43 samples, 81.13%)</title><rect x="16.9811%" y="117" width="81.1321%" height="15" fill="rgb(220,122,19)"/><text x="17.2311%" y="127.50">__libc_start_main</text></g><g><title>main (43 samples, 81.13%)</title><rect x="16.9811%" y="101" width="81.1321%" height="15" fill="rgb(250,228,42)"/><text x="17.2311%" y="111.50">main</text></g><g><title>std::vector&lt;unsigned long, std::allocator&lt;unsigned long&gt; &gt;::vector (1 samples, 1.89%)</title><rect x="96.2264%" y="85" width="1.8868%" height="15" fill="rgb(240,193,28)"/><text x="96.4764%" y="95.50">s..</text></g><g><title>__memset_avx2_erms (1 samples, 1.89%)</title><rect x="96.2264%" y="69" width="1.8868%" height="15" fill="rgb(216,20,37)"/><text x="96.4764%" y="79.50">_..</text></g><g><title>all (53 samples, 100%)</title><rect x="0.0000%" y="165" width="100.0000%" height="15" fill="rgb(206,188,39)"/><text x="0.2500%" y="175.50"></text></g><g><title>shuffle3-releas (53 samples, 100.00%)</title><rect x="0.0000%" y="149" width="100.0000%" height="15" fill="rgb(217,207,13)"/><text x="0.2500%" y="159.50">shuffle3-releas</text></g><g><title>_start (48 samples, 90.57%)</title><rect x="9.4340%" y="133" width="90.5660%" height="15" fill="rgb(231,73,38)"/><text x="9.6840%" y="143.50">_start</text></g><g><title>_dl_start (1 samples, 1.89%)</title><rect x="98.1132%" y="117" width="1.8868%" height="15" fill="rgb(225,20,46)"/><text x="98.3632%" y="127.50">_..</text></g><g><title>_dl_sysdep_start (1 samples, 1.89%)</title><rect x="98.1132%" y="101" width="1.8868%" height="15" fill="rgb(210,31,41)"/><text x="98.3632%" y="111.50">_..</text></g><g><title>dl_main (1 samples, 1.89%)</title><rect x="98.1132%" y="85" width="1.8868%" height="15" fill="rgb(221,200,47)"/><text x="98.3632%" y="95.50">d..</text></g><g><title>_dl_init_paths (1 samples, 1.89%)</title><rect x="98.1132%" y="69" width="1.8868%" height="15" fill="rgb(226,26,5)"/><text x="98.3632%" y="79.50">_..</text></g><g><title>fillin_rpath.isra.0 (1 samples, 1.89%)</title><rect x="98.1132%" y="53" width="1.8868%" height="15" fill="rgb(249,33,26)"/><text x="98.3632%" y="63.50">f..</text></g><g><title>__strsep (1 samples, 1.89%)</title><rect x="98.1132%" y="37" width="1.8868%" height="15" fill="rgb(235,183,28)"/><text x="98.3632%" y="47.50">_..</text></g></svg></svg>
\ No newline at end of file