koffi 0.9.3 → 0.9.6

package/vendor/dragonbox/subproject/test/source/verify_cache_precision.cpp

@@ -1,338 +1,338 @@
-// Copyright 2020-2022 Junekey Jeon
-//
-// The contents of this file may be used under the terms of
-// the Apache License v2.0 with LLVM Exceptions.
-//
-//    (See accompanying file LICENSE-Apache or copy at
-//     https://llvm.org/foundation/relicensing/LICENSE.txt)
-//
-// Alternatively, the contents of this file may be used under the terms of
-// the Boost Software License, Version 1.0.
-//    (See accompanying file LICENSE-Boost or copy at
-//     https://www.boost.org/LICENSE_1_0.txt)
-//
-// Unless required by applicable law or agreed to in writing, this software
-// is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-// KIND, either express or implied.
-
-#include "best_rational_approx.h"
-#include "good_rational_approx.h"
-#include "big_uint.h"
-#include "rational_continued_fractions.h"
-#include "dragonbox/dragonbox.h"
-#include <algorithm>
-#include <cstddef>
-#include <fstream>
-#include <iomanip>
-#include <iostream>
-#include <limits>
-#include <stdexcept>
-#include <utility>
-#include <vector>
-
-std::ostream& operator<<(std::ostream& out, jkj::big_uint const& n) {
-    auto dec = n.to_decimal();
-    assert(!dec.empty());
-
-    out << dec.back();
-
-    auto cur_fill = out.fill();
-    out << std::setfill('0');
-
-    for (std::size_t back_idx = 0; back_idx < dec.size() - 1; ++back_idx) {
-        out << std::setw(19) << dec[dec.size() - back_idx - 2];
-    }
-
-    out << std::setfill(cur_fill);
-    return out;
-}
-
-struct analysis_result {
-    struct result_per_cache_entry {
-        int sufficient_bits_for_multiplication;
-        int sufficient_bits_for_integer_checks;
-        jkj::unsigned_rational<jkj::big_uint> distance_to_upper_bound;
-    };
-    std::vector<result_per_cache_entry> results;
-
-    struct error_case {
-        int e;
-        int k;
-        jkj::unsigned_rational<jkj::big_uint> target;
-        jkj::unsigned_rational<jkj::big_uint> unit;
-        std::vector<jkj::big_uint> candidate_multipliers{};
-    };
-    std::vector<error_case> error_cases;
-};
-
-template <class FloatTraits>
-bool analyze(std::ostream& out) {
-    out << "e,bits_for_multiplication,bits_for_integer_check\n";
-
-    using impl = jkj::dragonbox::detail::impl<typename FloatTraits::type, FloatTraits>;
-    using namespace jkj::dragonbox::detail::log;
-
-    auto n_max = jkj::big_uint::power_of_2(impl::significand_bits + 2);
-
-    analysis_result result;
-    result.results.resize(impl::max_k - impl::min_k + 1);
-
-    jkj::unsigned_rational<jkj::big_uint> target{1, 1}, unit;
-    int prev_k = impl::max_k + 1;
-    for (int e = impl::min_exponent - impl::significand_bits;
-         e <= impl::max_exponent - impl::significand_bits; ++e) {
-        int k = impl::kappa - floor_log10_pow2(e);
-        auto exp_2 = k - floor_log2_pow10(k) - 1;
-        int beta = e + floor_log2_pow10(k);
-
-        auto& results_for_k = result.results[k - impl::min_k];
-
-        // target = 2^(k - klog2(10) - 1) * 5^k = phi_k / 2^Q in [1/2, 1).
-        if (k != prev_k) {
-            target.numerator = 1;
-            target.denominator = 1;
-            if (k >= 0) {
-                target.numerator = jkj::big_uint::pow(5, k);
-            }
-            else {
-                target.denominator = jkj::big_uint::pow(5, -k);
-            }
-            if (exp_2 >= 0) {
-                target.numerator *= jkj::big_uint::power_of_2(exp_2);
-            }
-            else {
-                target.denominator *= jkj::big_uint::power_of_2(-exp_2);
-            }
-        }
-
-        // unit = 2^(e + k - 1) * 5^k = a/b.
-        unit.numerator = 1;
-        unit.denominator = 1;
-        if (k >= 0) {
-            unit.numerator = jkj::big_uint::pow(5, k);
-        }
-        else {
-            unit.denominator = jkj::big_uint::pow(5, -k);
-        }
-        if (e + k - 1 >= 0) {
-            unit.numerator *= jkj::big_uint::power_of_2(e + k - 1);
-        }
-        else {
-            unit.denominator *= jkj::big_uint::power_of_2(-e - k + 1);
-        }
-
-
-        jkj::unsigned_rational<jkj::big_uint> upper_bound;
-        int sufficient_bits_for_integer_checks;
-        if (unit.denominator <= n_max) {
-            if (unit.denominator == 1) {
-                upper_bound = {unit.numerator * n_max + 1, n_max * jkj::big_uint::power_of_2(beta)};
-            }
-            else {
-                // We want to find the largest v <= n_max such that va == -1 (mod b).
-                // To obtain such v, we first find the smallest positive v0 such that
-                // v0 * a == -1 (mod b). Then v = v0 + floor((n_max - v0)/b) * b.
-                auto v0 = jkj::find_best_rational_approx<
-                              jkj::rational_continued_fractions<jkj::big_uint>>(
-                              unit, unit.denominator - 1)
-                              .above.denominator;
-                auto v = v0 + ((n_max - v0) / unit.denominator) * unit.denominator;
-
-                auto div_result = div(v * unit.numerator + 1, unit.denominator);
-                assert(div_result.rem.is_zero());
-                upper_bound = jkj::unsigned_rational<jkj::big_uint>{
-                    div_result.quot, v * jkj::big_uint::power_of_2(beta)};
-            }
-
-            sufficient_bits_for_integer_checks =
-                impl::carrier_bits + int(jkj::big_uint(1).multiply_2_until(unit.denominator));
-        }
-        else {
-            auto [below, above] =
-                jkj::find_best_rational_approx<jkj::rational_continued_fractions<jkj::big_uint>>(
-                    unit, n_max);
-
-            upper_bound = std::move(above);
-            upper_bound.denominator *= jkj::big_uint::power_of_2(beta);
-
-            sufficient_bits_for_integer_checks =
-                impl::carrier_bits +
-                int((unit.numerator * below.denominator - below.numerator * unit.denominator)
-                        .multiply_2_until(unit.denominator));
-
-            // Collect all cases where cache_bits seems insufficient.
-            if (sufficient_bits_for_integer_checks > impl::cache_bits) {
-                result.error_cases.push_back({e, k, target, unit});
-            }
-        }
-
-        // Compute the required number of bits for successful multiplication.
-        // The following is an upper bound.
-        auto div_result = div(upper_bound.denominator * target.denominator,
-                              upper_bound.numerator * target.denominator -
-                                  upper_bound.denominator * target.numerator);
-        if (!div_result.rem.is_zero()) {
-            div_result.quot += 1;
-        }
-        auto sufficient_bits_for_multiplication =
-            int(jkj::big_uint(1).multiply_2_until(div_result.quot));
-
-        // Tentatively decrease the above result to find the minimum admissible value.
-        while (sufficient_bits_for_multiplication > 0) {
-            auto r = (jkj::big_uint::power_of_2(sufficient_bits_for_multiplication - 1) *
-                      target.numerator) %
-                     target.denominator;
-            if (!r.is_zero()) {
-                r = target.denominator - r;
-            }
-
-            if (r * upper_bound.denominator >=
-                jkj::big_uint::power_of_2(sufficient_bits_for_multiplication - 1) *
-                    (upper_bound.numerator * target.denominator -
-                     upper_bound.denominator * target.numerator)) {
-                break;
-            }
-
-            --sufficient_bits_for_multiplication;
-        }
-
-        out << e << "," << sufficient_bits_for_multiplication << ","
-            << sufficient_bits_for_integer_checks << "\n";
-
-        // Update.
-        if (results_for_k.sufficient_bits_for_multiplication < sufficient_bits_for_multiplication) {
-            results_for_k.sufficient_bits_for_multiplication = sufficient_bits_for_multiplication;
-        }
-        if (results_for_k.sufficient_bits_for_integer_checks < sufficient_bits_for_integer_checks) {
-            results_for_k.sufficient_bits_for_integer_checks = sufficient_bits_for_integer_checks;
-        }
-        auto distance = jkj::unsigned_rational<jkj::big_uint>{
-            upper_bound.numerator * target.denominator - upper_bound.denominator * target.numerator,
-            upper_bound.denominator * target.denominator};
-        if (results_for_k.distance_to_upper_bound.denominator.is_zero()) {
-            results_for_k.distance_to_upper_bound = std::move(distance);
-        }
-        else if (results_for_k.distance_to_upper_bound.numerator * distance.denominator >
-                 distance.numerator * results_for_k.distance_to_upper_bound.denominator) {
-            results_for_k.distance_to_upper_bound = distance;
-        }
-    }
-
-    // Analyze all error cases.
-    auto reciprocal_error_threshold =
-        jkj::big_uint::power_of_2(impl::cache_bits - impl::carrier_bits);
-    for (auto& ec : result.error_cases) {
-        // We want to find all n such that
-        // d:= na/b - floor(na/b) < 2^(q-Q).
-
-        ec.candidate_multipliers = jkj::find_all_good_rational_approx_from_below_denoms<
-            jkj::rational_continued_fractions<jkj::big_uint>>(
-            ec.unit, n_max, jkj::unsigned_rational<jkj::big_uint>{1, reciprocal_error_threshold});
-    }
-
-    auto sufficient_bits_for_multiplication =
-        std::max_element(result.results.cbegin(), result.results.cend(),
-                         [](auto const& a, auto const& b) {
-                             return a.sufficient_bits_for_multiplication <
-                                    b.sufficient_bits_for_multiplication;
-                         })
-            ->sufficient_bits_for_multiplication;
-    auto sufficient_bits_for_integer_checks =
-        std::max_element(result.results.cbegin(), result.results.cend(),
-                         [](auto const& a, auto const& b) {
-                             return a.sufficient_bits_for_integer_checks <
-                                    b.sufficient_bits_for_integer_checks;
-                         })
-            ->sufficient_bits_for_integer_checks;
-    auto larger = std::max(sufficient_bits_for_multiplication, sufficient_bits_for_integer_checks);
-
-    auto distance_to_upper_bound =
-        std::min_element(
-            result.results.cbegin(), result.results.cend(),
-            [](auto const& a, auto const& b) {
-                if (a.distance_to_upper_bound.denominator == 0) {
-                    return false;
-                }
-                else if (b.distance_to_upper_bound.denominator == 0) {
-                    return true;
-                }
-                return a.distance_to_upper_bound.numerator * b.distance_to_upper_bound.denominator <
-                       b.distance_to_upper_bound.numerator * a.distance_to_upper_bound.denominator;
-            })
-            ->distance_to_upper_bound;
-
-    // Reduce the fraction.
-    distance_to_upper_bound =
-        jkj::find_best_rational_approx<jkj::rational_continued_fractions<jkj::big_uint>>(
-            distance_to_upper_bound, distance_to_upper_bound.denominator)
-            .below;
-
-    std::cout << "An upper bound on the minimum required bits for successful multiplication is "
-              << sufficient_bits_for_multiplication
-              << "-bits.\nAn upper bound on the minimum required bits for successful integer "
-                 "checks is "
-              << sufficient_bits_for_integer_checks << "-bits.\n";
-    std::cout << "A lower bound on the margin is " << distance_to_upper_bound.numerator << " / "
-              << distance_to_upper_bound.denominator << ".\n";
-
-    if (impl::cache_bits < larger) {
-        auto success = true;
-        std::cout << "Error cases:\n";
-        auto threshold = jkj::big_uint::power_of_2(impl::significand_bits + 1) - 1;
-        for (auto const& ec : result.error_cases) {
-            for (auto const& n : ec.candidate_multipliers) {
-                std::cout << "  e: " << ec.e << "  k: " << ec.k << "  n: " << n;
-
-                // When e != min_e and n != 1, 2, then
-                // n must be at least 2^(p+1)-2, otherwise this is a false
-                // positive.
-
-                if (ec.e != impl::min_exponent - impl::significand_bits && n != 1 && n != 2 &&
-                    n < threshold) {
-                    std::cout << "\n    n is smaller than " << threshold
-                              << ", so this case is a false positive.";
-                }
-                else if (ec.e == -81 && n == 29711844 || ec.e == -80 && n == 29711844) {
-                    std::cout << "\n    This case has been carefully addressed.";
-                }
-                else {
-                    success = false;
-                }
-
-                std::cout << "\n\n";
-            }
-        }
-
-        if (!success) {
-            std::cout << "Verification failed. " << impl::cache_bits
-                      << "-bits are not sufficient.\n\n";
-            return false;
-        }
-    }
-
-    std::cout << "Verified. " << impl::cache_bits << "-bits are sufficient.\n\n";
-    return true;
-}
-
-
-
-int main() {
-    bool success = true;
-    std::ofstream out;
-
-    std::cout << "[Verifying sufficiency of cache precision for binary32...]\n";
-    out.open("results/binary32.csv");
-    if (!analyze<jkj::dragonbox::default_float_traits<float>>(out)) {
-        success = false;
-    }
-    out.close();
-
-    std::cout << "[Verifying sufficiency of cache precision for binary64...]\n";
-    out.open("results/binary64.csv");
-    if (!analyze<jkj::dragonbox::default_float_traits<double>>(out)) {
-        success = false;
-    }
-    out.close();
-
-    return success ? 0 : -1;
+// Copyright 2020-2022 Junekey Jeon
+//
+// The contents of this file may be used under the terms of
+// the Apache License v2.0 with LLVM Exceptions.
+//
+//    (See accompanying file LICENSE-Apache or copy at
+//     https://llvm.org/foundation/relicensing/LICENSE.txt)
+//
+// Alternatively, the contents of this file may be used under the terms of
+// the Boost Software License, Version 1.0.
+//    (See accompanying file LICENSE-Boost or copy at
+//     https://www.boost.org/LICENSE_1_0.txt)
+//
+// Unless required by applicable law or agreed to in writing, this software
+// is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.
+
+#include "best_rational_approx.h"
+#include "good_rational_approx.h"
+#include "big_uint.h"
+#include "rational_continued_fractions.h"
+#include "dragonbox/dragonbox.h"
+#include <algorithm>
+#include <cstddef>
+#include <fstream>
+#include <iomanip>
+#include <iostream>
+#include <limits>
+#include <stdexcept>
+#include <utility>
+#include <vector>
+
+std::ostream& operator<<(std::ostream& out, jkj::big_uint const& n) {
+    auto dec = n.to_decimal();
+    assert(!dec.empty());
+
+    out << dec.back();
+
+    auto cur_fill = out.fill();
+    out << std::setfill('0');
+
+    for (std::size_t back_idx = 0; back_idx < dec.size() - 1; ++back_idx) {
+        out << std::setw(19) << dec[dec.size() - back_idx - 2];
+    }
+
+    out << std::setfill(cur_fill);
+    return out;
+}
+
+struct analysis_result {
+    struct result_per_cache_entry {
+        int sufficient_bits_for_multiplication;
+        int sufficient_bits_for_integer_checks;
+        jkj::unsigned_rational<jkj::big_uint> distance_to_upper_bound;
+    };
+    std::vector<result_per_cache_entry> results;
+
+    struct error_case {
+        int e;
+        int k;
+        jkj::unsigned_rational<jkj::big_uint> target;
+        jkj::unsigned_rational<jkj::big_uint> unit;
+        std::vector<jkj::big_uint> candidate_multipliers{};
+    };
+    std::vector<error_case> error_cases;
+};
+
+template <class FloatTraits>
+bool analyze(std::ostream& out) {
+    out << "e,bits_for_multiplication,bits_for_integer_check\n";
+
+    using impl = jkj::dragonbox::detail::impl<typename FloatTraits::type, FloatTraits>;
+    using namespace jkj::dragonbox::detail::log;
+
+    auto n_max = jkj::big_uint::power_of_2(impl::significand_bits + 2);
+
+    analysis_result result;
+    result.results.resize(impl::max_k - impl::min_k + 1);
+
+    jkj::unsigned_rational<jkj::big_uint> target{1, 1}, unit;
+    int prev_k = impl::max_k + 1;
+    for (int e = impl::min_exponent - impl::significand_bits;
+         e <= impl::max_exponent - impl::significand_bits; ++e) {
+        int k = impl::kappa - floor_log10_pow2(e);
+        auto exp_2 = k - floor_log2_pow10(k) - 1;
+        int beta = e + floor_log2_pow10(k);
+
+        auto& results_for_k = result.results[k - impl::min_k];
+
+        // target = 2^(k - klog2(10) - 1) * 5^k = phi_k / 2^Q in [1/2, 1).
+        if (k != prev_k) {
+            target.numerator = 1;
+            target.denominator = 1;
+            if (k >= 0) {
+                target.numerator = jkj::big_uint::pow(5, k);
+            }
+            else {
+                target.denominator = jkj::big_uint::pow(5, -k);
+            }
+            if (exp_2 >= 0) {
+                target.numerator *= jkj::big_uint::power_of_2(exp_2);
+            }
+            else {
+                target.denominator *= jkj::big_uint::power_of_2(-exp_2);
+            }
+        }
+
+        // unit = 2^(e + k - 1) * 5^k = a/b.
+        unit.numerator = 1;
+        unit.denominator = 1;
+        if (k >= 0) {
+            unit.numerator = jkj::big_uint::pow(5, k);
+        }
+        else {
+            unit.denominator = jkj::big_uint::pow(5, -k);
+        }
+        if (e + k - 1 >= 0) {
+            unit.numerator *= jkj::big_uint::power_of_2(e + k - 1);
+        }
+        else {
+            unit.denominator *= jkj::big_uint::power_of_2(-e - k + 1);
+        }
+
+
+        jkj::unsigned_rational<jkj::big_uint> upper_bound;
+        int sufficient_bits_for_integer_checks;
+        if (unit.denominator <= n_max) {
+            if (unit.denominator == 1) {
+                upper_bound = {unit.numerator * n_max + 1, n_max * jkj::big_uint::power_of_2(beta)};
+            }
+            else {
+                // We want to find the largest v <= n_max such that va == -1 (mod b).
+                // To obtain such v, we first find the smallest positive v0 such that
+                // v0 * a == -1 (mod b). Then v = v0 + floor((n_max - v0)/b) * b.
+                auto v0 = jkj::find_best_rational_approx<
+                              jkj::rational_continued_fractions<jkj::big_uint>>(
+                              unit, unit.denominator - 1)
+                              .above.denominator;
+                auto v = v0 + ((n_max - v0) / unit.denominator) * unit.denominator;
+
+                auto div_result = div(v * unit.numerator + 1, unit.denominator);
+                assert(div_result.rem.is_zero());
+                upper_bound = jkj::unsigned_rational<jkj::big_uint>{
+                    div_result.quot, v * jkj::big_uint::power_of_2(beta)};
+            }
+
+            sufficient_bits_for_integer_checks =
+                impl::carrier_bits + int(jkj::big_uint(1).multiply_2_until(unit.denominator));
+        }
+        else {
+            auto [below, above] =
+                jkj::find_best_rational_approx<jkj::rational_continued_fractions<jkj::big_uint>>(
+                    unit, n_max);
+
+            upper_bound = std::move(above);
+            upper_bound.denominator *= jkj::big_uint::power_of_2(beta);
+
+            sufficient_bits_for_integer_checks =
+                impl::carrier_bits +
+                int((unit.numerator * below.denominator - below.numerator * unit.denominator)
+                        .multiply_2_until(unit.denominator));
+
+            // Collect all cases where cache_bits seems insufficient.
+            if (sufficient_bits_for_integer_checks > impl::cache_bits) {
+                result.error_cases.push_back({e, k, target, unit});
+            }
+        }
+
+        // Compute the required number of bits for successful multiplication.
+        // The following is an upper bound.
+        auto div_result = div(upper_bound.denominator * target.denominator,
+                              upper_bound.numerator * target.denominator -
+                                  upper_bound.denominator * target.numerator);
+        if (!div_result.rem.is_zero()) {
+            div_result.quot += 1;
+        }
+        auto sufficient_bits_for_multiplication =
+            int(jkj::big_uint(1).multiply_2_until(div_result.quot));
+
+        // Tentatively decrease the above result to find the minimum admissible value.
+        while (sufficient_bits_for_multiplication > 0) {
+            auto r = (jkj::big_uint::power_of_2(sufficient_bits_for_multiplication - 1) *
+                      target.numerator) %
+                     target.denominator;
+            if (!r.is_zero()) {
+                r = target.denominator - r;
+            }
+
+            if (r * upper_bound.denominator >=
+                jkj::big_uint::power_of_2(sufficient_bits_for_multiplication - 1) *
+                    (upper_bound.numerator * target.denominator -
+                     upper_bound.denominator * target.numerator)) {
+                break;
+            }
+
+            --sufficient_bits_for_multiplication;
+        }
+
+        out << e << "," << sufficient_bits_for_multiplication << ","
+            << sufficient_bits_for_integer_checks << "\n";
+
+        // Update.
+        if (results_for_k.sufficient_bits_for_multiplication < sufficient_bits_for_multiplication) {
+            results_for_k.sufficient_bits_for_multiplication = sufficient_bits_for_multiplication;
+        }
+        if (results_for_k.sufficient_bits_for_integer_checks < sufficient_bits_for_integer_checks) {
+            results_for_k.sufficient_bits_for_integer_checks = sufficient_bits_for_integer_checks;
+        }
+        auto distance = jkj::unsigned_rational<jkj::big_uint>{
+            upper_bound.numerator * target.denominator - upper_bound.denominator * target.numerator,
+            upper_bound.denominator * target.denominator};
+        if (results_for_k.distance_to_upper_bound.denominator.is_zero()) {
+            results_for_k.distance_to_upper_bound = std::move(distance);
+        }
+        else if (results_for_k.distance_to_upper_bound.numerator * distance.denominator >
+                 distance.numerator * results_for_k.distance_to_upper_bound.denominator) {
+            results_for_k.distance_to_upper_bound = distance;
+        }
+    }
+
+    // Analyze all error cases.
+    auto reciprocal_error_threshold =
+        jkj::big_uint::power_of_2(impl::cache_bits - impl::carrier_bits);
+    for (auto& ec : result.error_cases) {
+        // We want to find all n such that
+        // d:= na/b - floor(na/b) < 2^(q-Q).
+
+        ec.candidate_multipliers = jkj::find_all_good_rational_approx_from_below_denoms<
+            jkj::rational_continued_fractions<jkj::big_uint>>(
+            ec.unit, n_max, jkj::unsigned_rational<jkj::big_uint>{1, reciprocal_error_threshold});
+    }
+
+    auto sufficient_bits_for_multiplication =
+        std::max_element(result.results.cbegin(), result.results.cend(),
+                         [](auto const& a, auto const& b) {
+                             return a.sufficient_bits_for_multiplication <
+                                    b.sufficient_bits_for_multiplication;
+                         })
+            ->sufficient_bits_for_multiplication;
+    auto sufficient_bits_for_integer_checks =
+        std::max_element(result.results.cbegin(), result.results.cend(),
+                         [](auto const& a, auto const& b) {
+                             return a.sufficient_bits_for_integer_checks <
+                                    b.sufficient_bits_for_integer_checks;
+                         })
+            ->sufficient_bits_for_integer_checks;
+    auto larger = std::max(sufficient_bits_for_multiplication, sufficient_bits_for_integer_checks);
+
+    auto distance_to_upper_bound =
+        std::min_element(
+            result.results.cbegin(), result.results.cend(),
+            [](auto const& a, auto const& b) {
+                if (a.distance_to_upper_bound.denominator == 0) {
+                    return false;
+                }
+                else if (b.distance_to_upper_bound.denominator == 0) {
+                    return true;
+                }
+                return a.distance_to_upper_bound.numerator * b.distance_to_upper_bound.denominator <
+                       b.distance_to_upper_bound.numerator * a.distance_to_upper_bound.denominator;
+            })
+            ->distance_to_upper_bound;
+
+    // Reduce the fraction.
+    distance_to_upper_bound =
+        jkj::find_best_rational_approx<jkj::rational_continued_fractions<jkj::big_uint>>(
+            distance_to_upper_bound, distance_to_upper_bound.denominator)
+            .below;
+
+    std::cout << "An upper bound on the minimum required bits for successful multiplication is "
+              << sufficient_bits_for_multiplication
+              << "-bits.\nAn upper bound on the minimum required bits for successful integer "
+                 "checks is "
+              << sufficient_bits_for_integer_checks << "-bits.\n";
+    std::cout << "A lower bound on the margin is " << distance_to_upper_bound.numerator << " / "
+              << distance_to_upper_bound.denominator << ".\n";
+
+    if (impl::cache_bits < larger) {
+        auto success = true;
+        std::cout << "Error cases:\n";
+        auto threshold = jkj::big_uint::power_of_2(impl::significand_bits + 1) - 1;
+        for (auto const& ec : result.error_cases) {
+            for (auto const& n : ec.candidate_multipliers) {
+                std::cout << "  e: " << ec.e << "  k: " << ec.k << "  n: " << n;
+
+                // When e != min_e and n != 1, 2, then
+                // n must be at least 2^(p+1)-2, otherwise this is a false
+                // positive.
+
+                if (ec.e != impl::min_exponent - impl::significand_bits && n != 1 && n != 2 &&
+                    n < threshold) {
+                    std::cout << "\n    n is smaller than " << threshold
+                              << ", so this case is a false positive.";
+                }
+                else if (ec.e == -81 && n == 29711844 || ec.e == -80 && n == 29711844) {
+                    std::cout << "\n    This case has been carefully addressed.";
+                }
+                else {
+                    success = false;
+                }
+
+                std::cout << "\n\n";
+            }
+        }
+
+        if (!success) {
+            std::cout << "Verification failed. " << impl::cache_bits
+                      << "-bits are not sufficient.\n\n";
+            return false;
+        }
+    }
+
+    std::cout << "Verified. " << impl::cache_bits << "-bits are sufficient.\n\n";
+    return true;
+}
+
+
+
+int main() {
+    bool success = true;
+    std::ofstream out;
+
+    std::cout << "[Verifying sufficiency of cache precision for binary32...]\n";
+    out.open("results/binary32.csv");
+    if (!analyze<jkj::dragonbox::default_float_traits<float>>(out)) {
+        success = false;
+    }
+    out.close();
+
+    std::cout << "[Verifying sufficiency of cache precision for binary64...]\n";
+    out.open("results/binary64.csv");
+    if (!analyze<jkj::dragonbox::default_float_traits<double>>(out)) {
+        success = false;
+    }
+    out.close();
+
+    return success ? 0 : -1;
 }