diff options
34 files changed, 382 insertions, 3840 deletions
@@ -39,7 +39,7 @@ Highlights - **Fully memory managed** - All containers will destruct keys/values via destructor passed as macro parameters to the ***using***-declaration. Also, shared pointers are supported and can be stored in containers, see ***csptr***.
- **Fully type safe** - Because of templating, it avoids error-prone casting of container types and elements back and forth from the containers.
- **Uniform, easy-to-learn API** - Methods to ***construct***, ***initialize***, ***iterate*** and ***destruct*** have uniform and intuitive usage across the various containers.
-- **Small footprint** - Small source code and generated executables. The executable from the example below using six different containers is *27 kb in size* compiled with TinyC.
+- **Small footprint** - Small source code and generated executables. The executable from the example below with six different containers is *22 kb in size* compiled with gcc -Os on linux.
- **Dual mode compilation** - By default it is a simple header-only library with inline and static methods only, but you can easily switch to create a traditional library with shared symbols, without changing existing source files. See the Installation section.
- **No callback functions** - All passed template argument functions/macros are directly called from the implementation, no slow callbacks which requires storage.
- **Compiles with C++ and C99** - C code can be compiled with C++.
@@ -64,18 +64,18 @@ templates in c++. A basic usage example: ```c
#include <stc/cvec.h>
-using_cvec(i, int);
+using_cvec(f32, float);
int main(void) {
- cvec_i vec = cvec_i_init();
- cvec_i_push_back(&vec, 10);
- cvec_i_push_back(&vec, 20);
- cvec_i_push_back(&vec, 30);
+ cvec_f32 vec = cvec_f32_init();
+ cvec_f32_push_back(&vec, 10.f);
+ cvec_f32_push_back(&vec, 20.f);
+ cvec_f32_push_back(&vec, 30.f);
- c_foreach (i, cvec_i, vec)
- printf(" %d", *i.ref);
+ c_foreach (i, cvec_f32, vec)
+ printf(" %g", *i.ref);
- cvec_i_del(&vec);
+ cvec_f32_del(&vec);
}
```
With six different containers:
@@ -91,66 +91,68 @@ With six different containers: struct Point { float x, y; };
int Point_compare(const struct Point* a, const struct Point* b) {
- if (a->x != b->x) return 1 - 2*(a->x < b->x);
- return (a->y > b->y) - (a->y < b->y);
+ int cmp = c_default_compare(&a->x, &b->x);
+ return cmp ? cmp : c_default_compare(&a->y, &b->y);
}
// declare container types
-using_cset(i, int); // unordered set
-using_cvec(p, struct Point, Point_compare); // vector, struct as elements
-using_cdeq(i, int); // deque
-using_clist(i, int); // singly linked list
-using_cqueue(i, cdeq_i); // queue, using deque as adapter
-using_csmap(i, int, int); // sorted map
+using_cset(i32, int); // unordered set
+using_cvec(pnt, struct Point, Point_compare); // vector, struct as elements
+using_cdeq(i32, int); // deque
+using_clist(i32, int); // singly linked list
+using_cqueue(i32, cdeq_i32); // queue, using deque as adapter
+using_csmap(i32, int, int); // sorted map
int main(void) {
- // shorthand for define + initialize containers with c_var():
- c_var (cset_i, set, {10, 20, 30});
- c_var (cvec_p, vec, { {10, 1}, {20, 2}, {30, 3} });
- c_var (cdeq_i, deq, {10, 20, 30});
- c_var (clist_i, lst, {10, 20, 30});
- c_var (cqueue_i, que, {10, 20, 30});
- c_var (csmap_i, map, { {20, 2}, {10, 1}, {30, 3} });
-
- // add one more element to each container
- cset_i_insert(&set, 40);
- cvec_p_push_back(&vec, (struct Point) {40, 4});
- cdeq_i_push_front(&deq, 5);
- clist_i_push_front(&lst, 5);
- cqueue_i_push(&que, 40);
- csmap_i_emplace(&map, 40, 4);
-
- // find an element in each container
- cset_i_iter_t i1 = cset_i_find(&set, 20);
- cvec_p_iter_t i2 = cvec_p_find(&vec, (struct Point) {20, 2});
- cdeq_i_iter_t i3 = cdeq_i_find(&deq, 20);
- clist_i_iter_t i4 = clist_i_find(&lst, 20);
- csmap_i_iter_t i5 = csmap_i_find(&map, 20);
- printf("\nFound: %d, (%g, %g), %d, %d, [%d: %d]\n", *i1.ref, i2.ref->x, i2.ref->y,
- *i3.ref, *i4.ref,
- i5.ref->first, i5.ref->second);
- // erase the elements found
- cset_i_erase_at(&set, i1);
- cvec_p_erase_at(&vec, i2);
- cdeq_i_erase_at(&deq, i3);
- clist_i_erase_at(&lst, i4);
- csmap_i_erase_at(&map, i5);
-
- printf("After erasing elements found:");
- printf("\n set:"); c_foreach (i, cset_i, set) printf(" %d", *i.ref);
- printf("\n vec:"); c_foreach (i, cvec_p, vec) printf(" (%g, %g)", i.ref->x, i.ref->y);
- printf("\n deq:"); c_foreach (i, cdeq_i, deq) printf(" %d", *i.ref);
- printf("\n lst:"); c_foreach (i, clist_i, lst) printf(" %d", *i.ref);
- printf("\n que:"); c_foreach (i, cqueue_i, que) printf(" %d", *i.ref);
- printf("\n map:"); c_foreach (i, csmap_i, map) printf(" [%d: %d]", i.ref->first, i.ref->second);
-
- // cleanup
- cset_i_del(&set);
- cvec_p_del(&vec);
- cdeq_i_del(&deq);
- clist_i_del(&lst);
- cqueue_i_del(&que);
- csmap_i_del(&map);
+ // define six containers with automatic call of init and del (destruction after scope exit)
+ c_forauto (cset_i32, set)
+ c_forauto (cvec_pnt, vec)
+ c_forauto (cdeq_i32, deq)
+ c_forauto (clist_i32, lst)
+ c_forauto (cqueue_i32, que)
+ c_forauto (csmap_i32, map)
+ {
+ // add some elements to each container
+ c_emplace(cset_i32, set, {10, 20, 30});
+ c_emplace(cvec_pnt, vec, { {10, 1}, {20, 2}, {30, 3} });
+ c_emplace(cdeq_i32, deq, {10, 20, 30});
+ c_emplace(clist_i32, lst, {10, 20, 30});
+ c_emplace(cqueue_i32, que, {10, 20, 30});
+ c_emplace(csmap_i32, map, { {20, 2}, {10, 1}, {30, 3} });
+
+ // add one more element to each container
+ cset_i32_insert(&set, 40);
+ cvec_pnt_push_back(&vec, (struct Point) {40, 4});
+ cdeq_i32_push_front(&deq, 5);
+ clist_i32_push_front(&lst, 5);
+ cqueue_i32_push(&que, 40);
+ csmap_i32_emplace(&map, 40, 4);
+
+ // find an element in each container
+ cset_i32_iter_t i1 = cset_i32_find(&set, 20);
+ cvec_pnt_iter_t i2 = cvec_pnt_find(&vec, (struct Point) {20, 2});
+ cdeq_i32_iter_t i3 = cdeq_i32_find(&deq, 20);
+ clist_i32_iter_t i4 = clist_i32_find(&lst, 20);
+ csmap_i32_iter_t i5 = csmap_i32_find(&map, 20);
+ printf("\nFound: %d, (%g, %g), %d, %d, [%d: %d]\n", *i1.ref, i2.ref->x, i2.ref->y,
+ *i3.ref, *i4.ref,
+ i5.ref->first, i5.ref->second);
+ // erase the elements found
+ cset_i32_erase_at(&set, i1);
+ cvec_pnt_erase_at(&vec, i2);
+ cdeq_i32_erase_at(&deq, i3);
+ clist_i32_erase_at(&lst, i4);
+ csmap_i32_erase_at(&map, i5);
+
+ printf("After erasing elements found:");
+ printf("\n set:"); c_foreach (i, cset_i32, set) printf(" %d", *i.ref);
+ printf("\n vec:"); c_foreach (i, cvec_pnt, vec) printf(" (%g, %g)", i.ref->x, i.ref->y);
+ printf("\n deq:"); c_foreach (i, cdeq_i32, deq) printf(" %d", *i.ref);
+ printf("\n lst:"); c_foreach (i, clist_i32, lst) printf(" %d", *i.ref);
+ printf("\n que:"); c_foreach (i, cqueue_i32, que) printf(" %d", *i.ref);
+ printf("\n map:"); c_foreach (i, csmap_i32, map) printf(" [%d: %d]", i.ref->first,
+ i.ref->second);
+ }
}
```
Output
diff --git a/benchmarks/others/clist_v1.h b/benchmarks/others/clist_v1.h index 00e04e9b..0dd8feef 100644 --- a/benchmarks/others/clist_v1.h +++ b/benchmarks/others/clist_v1.h @@ -69,7 +69,7 @@ _c_using_clist(clist_##X, Value, valueCompare, valueDel, valueFromRaw, valueToRaw, RawValue)
#define using_clist_str() \
- _c_using_clist(clist_str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*)
+ _c_using_clist(clist_str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*)
#define _c_using_clist_types(CX, Value) \
diff --git a/benchmarks/others/csmap_v1.h b/benchmarks/others/csmap_v1.h index 14114150..69f3f06d 100644 --- a/benchmarks/others/csmap_v1.h +++ b/benchmarks/others/csmap_v1.h @@ -87,8 +87,8 @@ int main(void) { #define using_csmap_str() \
_c_using_aatree(csmap_str, csmap_, cstr, cstr, c_rawstr_compare, \
- cstr_del, cstr_from, cstr_toraw, const char*, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ cstr_del, cstr_from, cstr_str, const char*, \
+ cstr_del, cstr_from, cstr_str, const char*)
#define using_csmap_strkey(...) c_MACRO_OVERLOAD(using_csmap_strkey, __VA_ARGS__)
@@ -105,7 +105,7 @@ int main(void) { #define _c_using_aatree_strkey(X, C, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped) \
_c_using_aatree(C##X, C, cstr, Mapped, c_rawstr_compare, \
mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ cstr_del, cstr_from, cstr_str, const char*)
#define using_csmap_strval(...) c_MACRO_OVERLOAD(using_csmap_strval, __VA_ARGS__)
@@ -121,7 +121,7 @@ int main(void) { #define using_csmap_strval_7(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey) \
_c_using_aatree(csmap_##X, csmap_, Key, cstr, keyCompareRaw, \
- cstr_del, cstr_from, cstr_toraw, const char*, \
+ cstr_del, cstr_from, cstr_str, const char*, \
keyDel, keyFromRaw, keyToRaw, RawKey)
#define SET_ONLY_csmap_(...)
diff --git a/benchmarks/others/pcg_extras.hpp b/benchmarks/others/pcg_extras.hpp deleted file mode 100644 index 8445ca24..00000000 --- a/benchmarks/others/pcg_extras.hpp +++ /dev/null @@ -1,664 +0,0 @@ -/* - * PCG Random Number Generation for C++ - * - * Copyright 2014-2017 Melissa O'Neill <[email protected]>, - * and the PCG Project contributors. - * - * SPDX-License-Identifier: (Apache-2.0 OR MIT) - * - * Licensed under the Apache License, Version 2.0 (provided in - * LICENSE-APACHE.txt and at http://www.apache.org/licenses/LICENSE-2.0) - * or under the MIT license (provided in LICENSE-MIT.txt and at - * http://opensource.org/licenses/MIT), at your option. This file may not - * be copied, modified, or distributed except according to those terms. - * - * Distributed on an "AS IS" BASIS, WITHOUT WARRANTY OF ANY KIND, either - * express or implied. See your chosen license for details. - * - * For additional information about the PCG random number generation scheme, - * visit http://www.pcg-random.org/. - */ - -/* - * This file provides support code that is useful for random-number generation - * but not specific to the PCG generation scheme, including: - * - 128-bit int support for platforms where it isn't available natively - * - bit twiddling operations - * - I/O of 128-bit and 8-bit integers - * - Handling the evilness of SeedSeq - * - Support for efficiently producing random numbers less than a given - * bound - */ - -#ifndef PCG_EXTRAS_HPP_INCLUDED -#define PCG_EXTRAS_HPP_INCLUDED 1 - -#include <cinttypes> -#include <cstddef> -#include <cstdlib> -#include <cstring> -#include <cassert> -#include <limits> -#include <iostream> -#include <type_traits> -#include <utility> -#include <locale> -#include <iterator> - -#ifdef __GNUC__ - #include <cxxabi.h> -#endif - -/* - * Abstractions for compiler-specific directives - */ - -#ifdef __GNUC__ - #define PCG_NOINLINE __attribute__((noinline)) -#else - #define PCG_NOINLINE -#endif - -/* - * Some members of the PCG library use 128-bit math. When compiling on 64-bit - * platforms, both GCC and Clang provide 128-bit integer types that are ideal - * for the job. - * - * On 32-bit platforms (or with other compilers), we fall back to a C++ - * class that provides 128-bit unsigned integers instead. It may seem - * like we're reinventing the wheel here, because libraries already exist - * that support large integers, but most existing libraries provide a very - * generic multiprecision code, but here we're operating at a fixed size. - * Also, most other libraries are fairly heavyweight. So we use a direct - * implementation. Sadly, it's much slower than hand-coded assembly or - * direct CPU support. - * - */ -#if __SIZEOF_INT128__ - namespace pcg_extras { - typedef __uint128_t pcg128_t; - } - #define PCG_128BIT_CONSTANT(high,low) \ - ((pcg_extras::pcg128_t(high) << 64) + low) -#else - #include "pcg_uint128.hpp" - namespace pcg_extras { - typedef pcg_extras::uint_x4<uint32_t,uint64_t> pcg128_t; - } - #define PCG_128BIT_CONSTANT(high,low) \ - pcg_extras::pcg128_t(high,low) - #define PCG_EMULATED_128BIT_MATH 1 -#endif - - -namespace pcg_extras { - -/* - * We often need to represent a "number of bits". When used normally, these - * numbers are never greater than 128, so an unsigned char is plenty. - * If you're using a nonstandard generator of a larger size, you can set - * PCG_BITCOUNT_T to have it define it as a larger size. (Some compilers - * might produce faster code if you set it to an unsigned int.) - */ - -#ifndef PCG_BITCOUNT_T - typedef uint8_t bitcount_t; -#else - typedef PCG_BITCOUNT_T bitcount_t; -#endif - -/* - * C++ requires us to be able to serialize RNG state by printing or reading - * it from a stream. Because we use 128-bit ints, we also need to be able - * ot print them, so here is code to do so. - * - * This code provides enough functionality to print 128-bit ints in decimal - * and zero-padded in hex. It's not a full-featured implementation. - */ - -template <typename CharT, typename Traits> -std::basic_ostream<CharT,Traits>& -operator<<(std::basic_ostream<CharT,Traits>& out, pcg128_t value) -{ - auto desired_base = out.flags() & out.basefield; - bool want_hex = desired_base == out.hex; - - if (want_hex) { - uint64_t highpart = uint64_t(value >> 64); - uint64_t lowpart = uint64_t(value); - auto desired_width = out.width(); - if (desired_width > 16) { - out.width(desired_width - 16); - } - if (highpart != 0 || desired_width > 16) - out << highpart; - CharT oldfill = '\0'; - if (highpart != 0) { - out.width(16); - oldfill = out.fill('0'); - } - auto oldflags = out.setf(decltype(desired_base){}, out.showbase); - out << lowpart; - out.setf(oldflags); - if (highpart != 0) { - out.fill(oldfill); - } - return out; - } - constexpr size_t MAX_CHARS_128BIT = 40; - - char buffer[MAX_CHARS_128BIT]; - char* pos = buffer+sizeof(buffer); - *(--pos) = '\0'; - constexpr auto BASE = pcg128_t(10ULL); - do { - auto div = value / BASE; - auto mod = uint32_t(value - (div * BASE)); - *(--pos) = '0' + char(mod); - value = div; - } while(value != pcg128_t(0ULL)); - return out << pos; -} - -template <typename CharT, typename Traits> -std::basic_istream<CharT,Traits>& -operator>>(std::basic_istream<CharT,Traits>& in, pcg128_t& value) -{ - typename std::basic_istream<CharT,Traits>::sentry s(in); - - if (!s) - return in; - - constexpr auto BASE = pcg128_t(10ULL); - pcg128_t current(0ULL); - bool did_nothing = true; - bool overflow = false; - for(;;) { - CharT wide_ch = in.get(); - if (!in.good()) - break; - auto ch = in.narrow(wide_ch, '\0'); - if (ch < '0' || ch > '9') { - in.unget(); - break; - } - did_nothing = false; - pcg128_t digit(uint32_t(ch - '0')); - pcg128_t timesbase = current*BASE; - overflow = overflow || timesbase < current; - current = timesbase + digit; - overflow = overflow || current < digit; - } - - if (did_nothing || overflow) { - in.setstate(std::ios::failbit); - if (overflow) - current = ~pcg128_t(0ULL); - } - - value = current; - - return in; -} - -/* - * Likewise, if people use tiny rngs, we'll be serializing uint8_t. - * If we just used the provided IO operators, they'd read/write chars, - * not ints, so we need to define our own. We *can* redefine this operator - * here because we're in our own namespace. - */ - -template <typename CharT, typename Traits> -std::basic_ostream<CharT,Traits>& -operator<<(std::basic_ostream<CharT,Traits>&out, uint8_t value) -{ - return out << uint32_t(value); -} - -template <typename CharT, typename Traits> -std::basic_istream<CharT,Traits>& -operator>>(std::basic_istream<CharT,Traits>& in, uint8_t& target) -{ - uint32_t value = 0xdecea5edU; - in >> value; - if (!in && value == 0xdecea5edU) - return in; - if (value > uint8_t(~0)) { - in.setstate(std::ios::failbit); - value = ~0U; - } - target = uint8_t(value); - return in; -} - -/* Unfortunately, the above functions don't get found in preference to the - * built in ones, so we create some more specific overloads that will. - * Ugh. - */ - -inline std::ostream& operator<<(std::ostream& out, uint8_t value) -{ - return pcg_extras::operator<< <char>(out, value); -} - -inline std::istream& operator>>(std::istream& in, uint8_t& value) -{ - return pcg_extras::operator>> <char>(in, value); -} - - - -/* - * Useful bitwise operations. - */ - -/* - * XorShifts are invertable, but they are someting of a pain to invert. - * This function backs them out. It's used by the whacky "inside out" - * generator defined later. - */ - -template <typename itype> -inline itype unxorshift(itype x, bitcount_t bits, bitcount_t shift) -{ - if (2*shift >= bits) { - return x ^ (x >> shift); - } - itype lowmask1 = (itype(1U) << (bits - shift*2)) - 1; - itype highmask1 = ~lowmask1; - itype top1 = x; - itype bottom1 = x & lowmask1; - top1 ^= top1 >> shift; - top1 &= highmask1; - x = top1 | bottom1; - itype lowmask2 = (itype(1U) << (bits - shift)) - 1; - itype bottom2 = x & lowmask2; - bottom2 = unxorshift(bottom2, bits - shift, shift); - bottom2 &= lowmask1; - return top1 | bottom2; -} - -/* - * Rotate left and right. - * - * In ideal world, compilers would spot idiomatic rotate code and convert it - * to a rotate instruction. Of course, opinions vary on what the correct - * idiom is and how to spot it. For clang, sometimes it generates better - * (but still crappy) code if you define PCG_USE_ZEROCHECK_ROTATE_IDIOM. - */ - -template <typename itype> -inline itype rotl(itype value, bitcount_t rot) -{ - constexpr bitcount_t bits = sizeof(itype) * 8; - constexpr bitcount_t mask = bits - 1; -#if PCG_USE_ZEROCHECK_ROTATE_IDIOM - return rot ? (value << rot) | (value >> (bits - rot)) : value; -#else - return (value << rot) | (value >> ((- rot) & mask)); -#endif -} - -template <typename itype> -inline itype rotr(itype value, bitcount_t rot) -{ - constexpr bitcount_t bits = sizeof(itype) * 8; - constexpr bitcount_t mask = bits - 1; -#if PCG_USE_ZEROCHECK_ROTATE_IDIOM - return rot ? (value >> rot) | (value << (bits - rot)) : value; -#else - return (value >> rot) | (value << ((- rot) & mask)); -#endif -} - -/* Unfortunately, both Clang and GCC sometimes perform poorly when it comes - * to properly recognizing idiomatic rotate code, so for we also provide - * assembler directives (enabled with PCG_USE_INLINE_ASM). Boo, hiss. - * (I hope that these compilers get better so that this code can die.) - * - * These overloads will be preferred over the general template code above. - */ -#if PCG_USE_INLINE_ASM && __GNUC__ && (__x86_64__ || __i386__) - -inline uint8_t rotr(uint8_t value, bitcount_t rot) -{ - asm ("rorb %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); - return value; -} - -inline uint16_t rotr(uint16_t value, bitcount_t rot) -{ - asm ("rorw %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); - return value; -} - -inline uint32_t rotr(uint32_t value, bitcount_t rot) -{ - asm ("rorl %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); - return value; -} - -#if __x86_64__ -inline uint64_t rotr(uint64_t value, bitcount_t rot) -{ - asm ("rorq %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); - return value; -} -#endif // __x86_64__ - -#elif defined(_MSC_VER) - // Use MSVC++ bit rotation intrinsics - -#pragma intrinsic(_rotr, _rotr64, _rotr8, _rotr16) - -inline uint8_t rotr(uint8_t value, bitcount_t rot) -{ - return _rotr8(value, rot); -} - -inline uint16_t rotr(uint16_t value, bitcount_t rot) -{ - return _rotr16(value, rot); -} - -inline uint32_t rotr(uint32_t value, bitcount_t rot) -{ - return _rotr(value, rot); -} - -inline uint64_t rotr(uint64_t value, bitcount_t rot) -{ - return _rotr64(value, rot); -} - -#endif // PCG_USE_INLINE_ASM - - -/* - * The C++ SeedSeq concept (modelled by seed_seq) can fill an array of - * 32-bit integers with seed data, but sometimes we want to produce - * larger or smaller integers. - * - * The following code handles this annoyance. - * - * uneven_copy will copy an array of 32-bit ints to an array of larger or - * smaller ints (actually, the code is general it only needing forward - * iterators). The copy is identical to the one that would be performed if - * we just did memcpy on a standard little-endian machine, but works - * regardless of the endian of the machine (or the weirdness of the ints - * involved). - * - * generate_to initializes an array of integers using a SeedSeq - * object. It is given the size as a static constant at compile time and - * tries to avoid memory allocation. If we're filling in 32-bit constants - * we just do it directly. If we need a separate buffer and it's small, - * we allocate it on the stack. Otherwise, we fall back to heap allocation. - * Ugh. - * - * generate_one produces a single value of some integral type using a - * SeedSeq object. - */ - - /* uneven_copy helper, case where destination ints are less than 32 bit. */ - -template<class SrcIter, class DestIter> -SrcIter uneven_copy_impl( - SrcIter src_first, DestIter dest_first, DestIter dest_last, - std::true_type) -{ - typedef typename std::iterator_traits<SrcIter>::value_type src_t; - typedef typename std::iterator_traits<DestIter>::value_type dest_t; - - constexpr bitcount_t SRC_SIZE = sizeof(src_t); - constexpr bitcount_t DEST_SIZE = sizeof(dest_t); - constexpr bitcount_t DEST_BITS = DEST_SIZE * 8; - constexpr bitcount_t SCALE = SRC_SIZE / DEST_SIZE; - - size_t count = 0; - src_t value = 0; - - while (dest_first != dest_last) { - if ((count++ % SCALE) == 0) - value = *src_first++; // Get more bits - else - value >>= DEST_BITS; // Move down bits - - *dest_first++ = dest_t(value); // Truncates, ignores high bits. - } - return src_first; -} - - /* uneven_copy helper, case where destination ints are more than 32 bit. */ - -template<class SrcIter, class DestIter> -SrcIter uneven_copy_impl( - SrcIter src_first, DestIter dest_first, DestIter dest_last, - std::false_type) -{ - typedef typename std::iterator_traits<SrcIter>::value_type src_t; - typedef typename std::iterator_traits<DestIter>::value_type dest_t; - - constexpr auto SRC_SIZE = sizeof(src_t); - constexpr auto SRC_BITS = SRC_SIZE * 8; - constexpr auto DEST_SIZE = sizeof(dest_t); - constexpr auto SCALE = (DEST_SIZE+SRC_SIZE-1) / SRC_SIZE; - - while (dest_first != dest_last) { - dest_t value(0UL); - unsigned int shift = 0; - - for (size_t i = 0; i < SCALE; ++i) { - value |= dest_t(*src_first++) << shift; - shift += SRC_BITS; - } - - *dest_first++ = value; - } - return src_first; -} - -/* uneven_copy, call the right code for larger vs. smaller */ - -template<class SrcIter, class DestIter> -inline SrcIter uneven_copy(SrcIter src_first, - DestIter dest_first, DestIter dest_last) -{ - typedef typename std::iterator_traits<SrcIter>::value_type src_t; - typedef typename std::iterator_traits<DestIter>::value_type dest_t; - - constexpr bool DEST_IS_SMALLER = sizeof(dest_t) < sizeof(src_t); - - return uneven_copy_impl(src_first, dest_first, dest_last, - std::integral_constant<bool, DEST_IS_SMALLER>{}); -} - -/* generate_to, fill in a fixed-size array of integral type using a SeedSeq - * (actually works for any random-access iterator) - */ - -template <size_t size, typename SeedSeq, typename DestIter> -inline void generate_to_impl(SeedSeq&& generator, DestIter dest, - std::true_type) -{ - generator.generate(dest, dest+size); -} - -template <size_t size, typename SeedSeq, typename DestIter> -void generate_to_impl(SeedSeq&& generator, DestIter dest, - std::false_type) -{ - typedef typename std::iterator_traits<DestIter>::value_type dest_t; - constexpr auto DEST_SIZE = sizeof(dest_t); - constexpr auto GEN_SIZE = sizeof(uint32_t); - - constexpr bool GEN_IS_SMALLER = GEN_SIZE < DEST_SIZE; - constexpr size_t FROM_ELEMS = - GEN_IS_SMALLER - ? size * ((DEST_SIZE+GEN_SIZE-1) / GEN_SIZE) - : (size + (GEN_SIZE / DEST_SIZE) - 1) - / ((GEN_SIZE / DEST_SIZE) + GEN_IS_SMALLER); - // this odd code ^^^^^^^^^^^^^^^^^ is work-around for - // a bug: http://llvm.org/bugs/show_bug.cgi?id=21287 - - if (FROM_ELEMS <= 1024) { - uint32_t buffer[FROM_ELEMS]; - generator.generate(buffer, buffer+FROM_ELEMS); - uneven_copy(buffer, dest, dest+size); - } else { - uint32_t* buffer = static_cast<uint32_t*>(malloc(GEN_SIZE * FROM_ELEMS)); - generator.generate(buffer, buffer+FROM_ELEMS); - uneven_copy(buffer, dest, dest+size); - free(static_cast<void*>(buffer)); - } -} - -template <size_t size, typename SeedSeq, typename DestIter> -inline void generate_to(SeedSeq&& generator, DestIter dest) -{ - typedef typename std::iterator_traits<DestIter>::value_type dest_t; - constexpr bool IS_32BIT = sizeof(dest_t) == sizeof(uint32_t); - - generate_to_impl<size>(std::forward<SeedSeq>(generator), dest, - std::integral_constant<bool, IS_32BIT>{}); -} - -/* generate_one, produce a value of integral type using a SeedSeq - * (optionally, we can have it produce more than one and pick which one - * we want) - */ - -template <typename UInt, size_t i = 0UL, size_t N = i+1UL, typename SeedSeq> -inline UInt generate_one(SeedSeq&& generator) -{ - UInt result[N]; - generate_to<N>(std::forward<SeedSeq>(generator), result); - return result[i]; -} - -template <typename RngType> -auto bounded_rand(RngType& rng, typename RngType::result_type upper_bound) - -> typename RngType::result_type -{ - typedef typename RngType::result_type rtype; - rtype threshold = (RngType::max() - RngType::min() + rtype(1) - upper_bound) - % upper_bound; - for (;;) { - rtype r = rng() - RngType::min(); - if (r >= threshold) - return r % upper_bound; - } -} - -template <typename Iter, typename RandType> -void shuffle(Iter from, Iter to, RandType&& rng) -{ - typedef typename std::iterator_traits<Iter>::difference_type delta_t; - typedef typename std::remove_reference<RandType>::type::result_type result_t; - auto count = to - from; - while (count > 1) { - delta_t chosen = delta_t(bounded_rand(rng, result_t(count))); - --count; - --to; - using std::swap; - swap(*(from + chosen), *to); - } -} - -/* - * Although std::seed_seq is useful, it isn't everything. Often we want to - * initialize a random-number generator some other way, such as from a random - * device. - * - * Technically, it does not meet the requirements of a SeedSequence because - * it lacks some of the rarely-used member functions (some of which would - * be impossible to provide). However the C++ standard is quite specific - * that actual engines only called the generate method, so it ought not to be - * a problem in practice. - */ - -template <typename RngType> -class seed_seq_from { -private: - RngType rng_; - - typedef uint_least32_t result_type; - -public: - template<typename... Args> - seed_seq_from(Args&&... args) : - rng_(std::forward<Args>(args)...) - { - // Nothing (else) to do... - } - - template<typename Iter> - void generate(Iter start, Iter finish) - { - for (auto i = start; i != finish; ++i) - *i = result_type(rng_()); - } - - constexpr size_t size() const - { - return (sizeof(typename RngType::result_type) > sizeof(result_type) - && RngType::max() > ~size_t(0UL)) - ? ~size_t(0UL) - : size_t(RngType::max()); - } -}; - -/* - * Sometimes you might want a distinct seed based on when the program - * was compiled. That way, a particular instance of the program will - * behave the same way, but when recompiled it'll produce a different - * value. - */ - -template <typename IntType> -struct static_arbitrary_seed { -private: - static constexpr IntType fnv(IntType hash, const char* pos) { - return *pos == '\0' - ? hash - : fnv((hash * IntType(16777619U)) ^ *pos, (pos+1)); - } - -public: - static constexpr IntType value = fnv(IntType(2166136261U ^ sizeof(IntType)), - __DATE__ __TIME__ __FILE__); -}; - -// Sometimes, when debugging or testing, it's handy to be able print the name -// of a (in human-readable form). This code allows the idiom: -// -// cout << printable_typename<my_foo_type_t>() -// -// to print out my_foo_type_t (or its concrete type if it is a synonym) - -#if __cpp_rtti || __GXX_RTTI - -template <typename T> -struct printable_typename {}; - -template <typename T> -std::ostream& operator<<(std::ostream& out, printable_typename<T>) { - const char *implementation_typename = typeid(T).name(); -#ifdef __GNUC__ - int status; - char* pretty_name = - abi::__cxa_demangle(implementation_typename, nullptr, nullptr, &status); - if (status == 0) - out << pretty_name; - free(static_cast<void*>(pretty_name)); - if (status == 0) - return out; -#endif - out << implementation_typename; - return out; -} - -#endif // __cpp_rtti || __GXX_RTTI - -} // namespace pcg_extras - -#endif // PCG_EXTRAS_HPP_INCLUDED diff --git a/benchmarks/others/pcg_random.hpp b/benchmarks/others/pcg_random.hpp deleted file mode 100644 index 4ab37cc4..00000000 --- a/benchmarks/others/pcg_random.hpp +++ /dev/null @@ -1,1947 +0,0 @@ -/* - * PCG Random Number Generation for C++ - * - * Copyright 2014-2019 Melissa O'Neill <[email protected]>, - * and the PCG Project contributors. - * - * SPDX-License-Identifier: (Apache-2.0 OR MIT) - * - * Licensed under the Apache License, Version 2.0 (provided in - * LICENSE-APACHE.txt and at http://www.apache.org/licenses/LICENSE-2.0) - * or under the MIT license (provided in LICENSE-MIT.txt and at - * http://opensource.org/licenses/MIT), at your option. This file may not - * be copied, modified, or distributed except according to those terms. - * - * Distributed on an "AS IS" BASIS, WITHOUT WARRANTY OF ANY KIND, either - * express or implied. See your chosen license for details. - * - * For additional information about the PCG random number generation scheme, - * visit http://www.pcg-random.org/. - */ - -/* - * This code provides the reference implementation of the PCG family of - * random number generators. The code is complex because it implements - * - * - several members of the PCG family, specifically members corresponding - * to the output functions: - * - XSH RR (good for 64-bit state, 32-bit output) - * - XSH RS (good for 64-bit state, 32-bit output) - * - XSL RR (good for 128-bit state, 64-bit output) - * - RXS M XS (statistically most powerful generator) - * - XSL RR RR (good for 128-bit state, 128-bit output) - * - and RXS, RXS M, XSH, XSL (mostly for testing) - * - at potentially *arbitrary* bit sizes - * - with four different techniques for random streams (MCG, one-stream - * LCG, settable-stream LCG, unique-stream LCG) - * - and the extended generation schemes allowing arbitrary periods - * - with all features of C++11 random number generation (and more), - * some of which are somewhat painful, including - * - initializing with a SeedSequence which writes 32-bit values - * to memory, even though the state of the generator may not - * use 32-bit values (it might use smaller or larger integers) - * - I/O for RNGs and a prescribed format, which needs to handle - * the issue that 8-bit and 128-bit integers don't have working - * I/O routines (e.g., normally 8-bit = char, not integer) - * - equality and inequality for RNGs - * - and a number of convenience typedefs to mask all the complexity - * - * The code employes a fairly heavy level of abstraction, and has to deal - * with various C++ minutia. If you're looking to learn about how the PCG - * scheme works, you're probably best of starting with one of the other - * codebases (see www.pcg-random.org). But if you're curious about the - * constants for the various output functions used in those other, simpler, - * codebases, this code shows how they are calculated. - * - * On the positive side, at least there are convenience typedefs so that you - * can say - * - * pcg32 myRNG; - * - * rather than: - * - * pcg_detail::engine< - * uint32_t, // Output Type - * uint64_t, // State Type - * pcg_detail::xsh_rr_mixin<uint32_t, uint64_t>, true, // Output Func - * pcg_detail::specific_stream<uint64_t>, // Stream Kind - * pcg_detail::default_multiplier<uint64_t> // LCG Mult - * > myRNG; - * - */ - -#ifndef PCG_RAND_HPP_INCLUDED -#define PCG_RAND_HPP_INCLUDED 1 - -#include <algorithm> -#include <cinttypes> -#include <cstddef> -#include <cstdlib> -#include <cstring> -#include <cassert> -#include <limits> -#include <iostream> -#include <iterator> -#include <type_traits> -#include <utility> -#include <locale> -#include <new> -#include <stdexcept> - -#ifdef _MSC_VER - #pragma warning(disable:4146) -#endif - -#ifdef _MSC_VER - #define PCG_ALWAYS_INLINE __forceinline -#elif __GNUC__ - #define PCG_ALWAYS_INLINE __attribute__((always_inline)) -#else - #define PCG_ALWAYS_INLINE inline -#endif - -/* - * The pcg_extras namespace contains some support code that is likley to - * be useful for a variety of RNGs, including: - * - 128-bit int support for platforms where it isn't available natively - * - bit twiddling operations - * - I/O of 128-bit and 8-bit integers - * - Handling the evilness of SeedSeq - * - Support for efficiently producing random numbers less than a given - * bound - */ - -#include "pcg_extras.hpp" - -namespace pcg_detail { - -using namespace pcg_extras; - -/* - * The LCG generators need some constants to function. This code lets you - * look up the constant by *type*. For example - * - * default_multiplier<uint32_t>::multiplier() - * - * gives you the default multipler for 32-bit integers. We use the name - * of the constant and not a generic word like value to allow these classes - * to be used as mixins. - */ - -template <typename T> -struct default_multiplier { - // Not defined for an arbitrary type -}; - -template <typename T> -struct default_increment { - // Not defined for an arbitrary type -}; - -#define PCG_DEFINE_CONSTANT(type, what, kind, constant) \ - template <> \ - struct what ## _ ## kind<type> { \ - static constexpr type kind() { \ - return constant; \ - } \ - }; - -PCG_DEFINE_CONSTANT(uint8_t, default, multiplier, 141U) -PCG_DEFINE_CONSTANT(uint8_t, default, increment, 77U) - -PCG_DEFINE_CONSTANT(uint16_t, default, multiplier, 12829U) -PCG_DEFINE_CONSTANT(uint16_t, default, increment, 47989U) - -PCG_DEFINE_CONSTANT(uint32_t, default, multiplier, 747796405U) -PCG_DEFINE_CONSTANT(uint32_t, default, increment, 2891336453U) - -PCG_DEFINE_CONSTANT(uint64_t, default, multiplier, 6364136223846793005ULL) -PCG_DEFINE_CONSTANT(uint64_t, default, increment, 1442695040888963407ULL) - -PCG_DEFINE_CONSTANT(pcg128_t, default, multiplier, - PCG_128BIT_CONSTANT(2549297995355413924ULL,4865540595714422341ULL)) -PCG_DEFINE_CONSTANT(pcg128_t, default, increment, - PCG_128BIT_CONSTANT(6364136223846793005ULL,1442695040888963407ULL)) - -/* Alternative (cheaper) multipliers for 128-bit */ - -template <typename T> -struct cheap_multiplier : public default_multiplier<T> { - // For most types just use the default. -}; - -template <> -struct cheap_multiplier<pcg128_t> { - static constexpr uint64_t multiplier() { - return 0xda942042e4dd58b5ULL; - } -}; - - -/* - * Each PCG generator is available in four variants, based on how it applies - * the additive constant for its underlying LCG; the variations are: - * - * single stream - all instances use the same fixed constant, thus - * the RNG always somewhere in same sequence - * mcg - adds zero, resulting in a single stream and reduced - * period - * specific stream - the constant can be changed at any time, selecting - * a different random sequence - * unique stream - the constant is based on the memory address of the - * object, thus every RNG has its own unique sequence - * - * This variation is provided though mixin classes which define a function - * value called increment() that returns the nesessary additive constant. - */ - - - -/* - * unique stream - */ - - -template <typename itype> -class unique_stream { -protected: - static constexpr bool is_mcg = false; - - // Is never called, but is provided for symmetry with specific_stream - void set_stream(...) - { - abort(); - } - -public: - typedef itype state_type; - - constexpr itype increment() const { - return itype(reinterpret_cast<uintptr_t>(this) | 1); - } - - constexpr itype stream() const - { - return increment() >> 1; - } - - static constexpr bool can_specify_stream = false; - - static constexpr size_t streams_pow2() - { - return (sizeof(itype) < sizeof(size_t) ? sizeof(itype) - : sizeof(size_t))*8 - 1u; - } - -protected: - constexpr unique_stream() = default; -}; - - -/* - * no stream (mcg) - */ - -template <typename itype> -class no_stream { -protected: - static constexpr bool is_mcg = true; - - // Is never called, but is provided for symmetry with specific_stream - void set_stream(...) - { - abort(); - } - -public: - typedef itype state_type; - - static constexpr itype increment() { - return 0; - } - - static constexpr bool can_specify_stream = false; - - static constexpr size_t streams_pow2() - { - return 0u; - } - -protected: - constexpr no_stream() = default; -}; - - -/* - * single stream/sequence (oneseq) - */ - -template <typename itype> -class oneseq_stream : public default_increment<itype> { -protected: - static constexpr bool is_mcg = false; - - // Is never called, but is provided for symmetry with specific_stream - void set_stream(...) - { - abort(); - } - -public: - typedef itype state_type; - - static constexpr itype stream() - { - return default_increment<itype>::increment() >> 1; - } - - static constexpr bool can_specify_stream = false; - - static constexpr size_t streams_pow2() - { - return 0u; - } - -protected: - constexpr oneseq_stream() = default; -}; - - -/* - * specific stream - */ - -template <typename itype> -class specific_stream { -protected: - static constexpr bool is_mcg = false; - - itype inc_ = default_increment<itype>::increment(); - -public: - typedef itype state_type; - typedef itype stream_state; - - constexpr itype increment() const { - return inc_; - } - - itype stream() - { - return inc_ >> 1; - } - - void set_stream(itype specific_seq) - { - inc_ = (specific_seq << 1) | 1; - } - - static constexpr bool can_specify_stream = true; - - static constexpr size_t streams_pow2() - { - return (sizeof(itype)*8) - 1u; - } - -protected: - specific_stream() = default; - - specific_stream(itype specific_seq) - : inc_(itype(specific_seq << 1) | itype(1U)) - { - // Nothing (else) to do. - } -}; - - -/* - * This is where it all comes together. This function joins together three - * mixin classes which define - * - the LCG additive constant (the stream) - * - the LCG multiplier - * - the output function - * in addition, we specify the type of the LCG state, and the result type, - * and whether to use the pre-advance version of the state for the output - * (increasing instruction-level parallelism) or the post-advance version - * (reducing register pressure). - * - * Given the high level of parameterization, the code has to use some - * template-metaprogramming tricks to handle some of the suble variations - * involved. - */ - -template <typename xtype, typename itype, - typename output_mixin, - bool output_previous = true, - typename stream_mixin = oneseq_stream<itype>, - typename multiplier_mixin = default_multiplier<itype> > -class engine : protected output_mixin, - public stream_mixin, - protected multiplier_mixin { -protected: - itype state_; - - struct can_specify_stream_tag {}; - struct no_specifiable_stream_tag {}; - - using stream_mixin::increment; - using multiplier_mixin::multiplier; - -public: - typedef xtype result_type; - typedef itype state_type; - - static constexpr size_t period_pow2() - { - return sizeof(state_type)*8 - 2*stream_mixin::is_mcg; - } - - // It would be nice to use std::numeric_limits for these, but - // we can't be sure that it'd be defined for the 128-bit types. - - static constexpr result_type min() - { - return result_type(0UL); - } - - static constexpr result_type max() - { - return result_type(~result_type(0UL)); - } - -protected: - itype bump(itype state) - { - return state * multiplier() + increment(); - } - - itype base_generate() - { - return state_ = bump(state_); - } - - itype base_generate0() - { - itype old_state = state_; - state_ = bump(state_); - return old_state; - } - -public: - result_type operator()() - { - if (output_previous) - return this->output(base_generate0()); - else - return this->output(base_generate()); - } - - result_type operator()(result_type upper_bound) - { - return bounded_rand(*this, upper_bound); - } - -protected: - static itype advance(itype state, itype delta, - itype cur_mult, itype cur_plus); - - static itype distance(itype cur_state, itype newstate, itype cur_mult, - itype cur_plus, itype mask = ~itype(0U)); - - itype distance(itype newstate, itype mask = itype(~itype(0U))) const - { - return distance(state_, newstate, multiplier(), increment(), mask); - } - -public: - void advance(itype delta) - { - state_ = advance(state_, delta, this->multiplier(), this->increment()); - } - - void backstep(itype delta) - { - advance(-delta); - } - - void discard(itype delta) - { - advance(delta); - } - - bool wrapped() - { - if (stream_mixin::is_mcg) { - // For MCGs, the low order two bits never change. In this - // implementation, we keep them fixed at 3 to make this test - // easier. - return state_ == 3; - } else { - return state_ == 0; - } - } - - engine(itype state = itype(0xcafef00dd15ea5e5ULL)) - : state_(this->is_mcg ? state|state_type(3U) - : bump(state + this->increment())) - { - // Nothing else to do. - } - - // This function may or may not exist. It thus has to be a template - // to use SFINAE; users don't have to worry about its template-ness. - - template <typename sm = stream_mixin> - engine(itype state, typename sm::stream_state stream_seed) - : stream_mixin(stream_seed), - state_(this->is_mcg ? state|state_type(3U) - : bump(state + this->increment())) - { - // Nothing else to do. - } - - template<typename SeedSeq> - engine(SeedSeq&& seedSeq, typename std::enable_if< - !stream_mixin::can_specify_stream - && !std::is_convertible<SeedSeq, itype>::value - && !std::is_convertible<SeedSeq, engine>::value, - no_specifiable_stream_tag>::type = {}) - : engine(generate_one<itype>(std::forward<SeedSeq>(seedSeq))) - { - // Nothing else to do. - } - - template<typename SeedSeq> - engine(SeedSeq&& seedSeq, typename std::enable_if< - stream_mixin::can_specify_stream - && !std::is_convertible<SeedSeq, itype>::value - && !std::is_convertible<SeedSeq, engine>::value, - can_specify_stream_tag>::type = {}) - : engine(generate_one<itype,1,2>(seedSeq), - generate_one<itype,0,2>(seedSeq)) - { - // Nothing else to do. - } - - - template<typename... Args> - void seed(Args&&... args) - { - new (this) engine(std::forward<Args>(args)...); - } - - template <typename xtype1, typename itype1, - typename output_mixin1, bool output_previous1, - typename stream_mixin_lhs, typename multiplier_mixin_lhs, - typename stream_mixin_rhs, typename multiplier_mixin_rhs> - friend bool operator==(const engine<xtype1,itype1, - output_mixin1,output_previous1, - stream_mixin_lhs, multiplier_mixin_lhs>&, - const engine<xtype1,itype1, - output_mixin1,output_previous1, - stream_mixin_rhs, multiplier_mixin_rhs>&); - - template <typename xtype1, typename itype1, - typename output_mixin1, bool output_previous1, - typename stream_mixin_lhs, typename multiplier_mixin_lhs, - typename stream_mixin_rhs, typename multiplier_mixin_rhs> - friend itype1 operator-(const engine<xtype1,itype1, - output_mixin1,output_previous1, - stream_mixin_lhs, multiplier_mixin_lhs>&, - const engine<xtype1,itype1, - output_mixin1,output_previous1, - stream_mixin_rhs, multiplier_mixin_rhs>&); - - template <typename CharT, typename Traits, - typename xtype1, typename itype1, - typename output_mixin1, bool output_previous1, - typename stream_mixin1, typename multiplier_mixin1> - friend std::basic_ostream<CharT,Traits>& - operator<<(std::basic_ostream<CharT,Traits>& out, - const engine<xtype1,itype1, - output_mixin1,output_previous1, - stream_mixin1, multiplier_mixin1>&); - - template <typename CharT, typename Traits, - typename xtype1, typename itype1, - typename output_mixin1, bool output_previous1, - typename stream_mixin1, typename multiplier_mixin1> - friend std::basic_istream<CharT,Traits>& - operator>>(std::basic_istream<CharT,Traits>& in, - engine<xtype1, itype1, - output_mixin1, output_previous1, - stream_mixin1, multiplier_mixin1>& rng); -}; - -template <typename CharT, typename Traits, - typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin, typename multiplier_mixin> -std::basic_ostream<CharT,Traits>& -operator<<(std::basic_ostream<CharT,Traits>& out, - const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin, multiplier_mixin>& rng) -{ - using pcg_extras::operator<<; - - auto orig_flags = out.flags(std::ios_base::dec | std::ios_base::left); - auto space = out.widen(' '); - auto orig_fill = out.fill(); - - out << rng.multiplier() << space - << rng.increment() << space - << rng.state_; - - out.flags(orig_flags); - out.fill(orig_fill); - return out; -} - - -template <typename CharT, typename Traits, - typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin, typename multiplier_mixin> -std::basic_istream<CharT,Traits>& -operator>>(std::basic_istream<CharT,Traits>& in, - engine<xtype,itype, - output_mixin,output_previous, - stream_mixin, multiplier_mixin>& rng) -{ - using pcg_extras::operator>>; - - auto orig_flags = in.flags(std::ios_base::dec | std::ios_base::skipws); - - itype multiplier, increment, state; - in >> multiplier >> increment >> state; - - if (!in.fail()) { - bool good = true; - if (multiplier != rng.multiplier()) { - good = false; - } else if (rng.can_specify_stream) { - rng.set_stream(increment >> 1); - } else if (increment != rng.increment()) { - good = false; - } - if (good) { - rng.state_ = state; - } else { - in.clear(std::ios::failbit); - } - } - - in.flags(orig_flags); - return in; -} - - -template <typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin, typename multiplier_mixin> -itype engine<xtype,itype,output_mixin,output_previous,stream_mixin, - multiplier_mixin>::advance( - itype state, itype delta, itype cur_mult, itype cur_plus) -{ - // The method used here is based on Brown, "Random Number Generation - // with Arbitrary Stride,", Transactions of the American Nuclear - // Society (Nov. 1994). The algorithm is very similar to fast - // exponentiation. - // - // Even though delta is an unsigned integer, we can pass a - // signed integer to go backwards, it just goes "the long way round". - - constexpr itype ZERO = 0u; // itype may be a non-trivial types, so - constexpr itype ONE = 1u; // we define some ugly constants. - itype acc_mult = 1; - itype acc_plus = 0; - while (delta > ZERO) { - if (delta & ONE) { - acc_mult *= cur_mult; - acc_plus = acc_plus*cur_mult + cur_plus; - } - cur_plus = (cur_mult+ONE)*cur_plus; - cur_mult *= cur_mult; - delta >>= 1; - } - return acc_mult * state + acc_plus; -} - -template <typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin, typename multiplier_mixin> -itype engine<xtype,itype,output_mixin,output_previous,stream_mixin, - multiplier_mixin>::distance( - itype cur_state, itype newstate, itype cur_mult, itype cur_plus, itype mask) -{ - constexpr itype ONE = 1u; // itype could be weird, so use constant - bool is_mcg = cur_plus == itype(0); - itype the_bit = is_mcg ? itype(4u) : itype(1u); - itype distance = 0u; - while ((cur_state & mask) != (newstate & mask)) { - if ((cur_state & the_bit) != (newstate & the_bit)) { - cur_state = cur_state * cur_mult + cur_plus; - distance |= the_bit; - } - assert((cur_state & the_bit) == (newstate & the_bit)); - the_bit <<= 1; - cur_plus = (cur_mult+ONE)*cur_plus; - cur_mult *= cur_mult; - } - return is_mcg ? distance >> 2 : distance; -} - -template <typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin_lhs, typename multiplier_mixin_lhs, - typename stream_mixin_rhs, typename multiplier_mixin_rhs> -itype operator-(const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_lhs, multiplier_mixin_lhs>& lhs, - const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_rhs, multiplier_mixin_rhs>& rhs) -{ - static_assert( - std::is_same<stream_mixin_lhs, stream_mixin_rhs>::value && - std::is_same<multiplier_mixin_lhs, multiplier_mixin_rhs>::value, - "Incomparable generators"); - if (lhs.increment() == rhs.increment()) { - return rhs.distance(lhs.state_); - } else { - constexpr itype ONE = 1u; - itype lhs_diff = lhs.increment() + (lhs.multiplier()-ONE) * lhs.state_; - itype rhs_diff = rhs.increment() + (rhs.multiplier()-ONE) * rhs.state_; - if ((lhs_diff & itype(3u)) != (rhs_diff & itype(3u))) { - rhs_diff = -rhs_diff; - } - return rhs.distance(rhs_diff, lhs_diff, rhs.multiplier(), itype(0u)); - } -} - - -template <typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin_lhs, typename multiplier_mixin_lhs, - typename stream_mixin_rhs, typename multiplier_mixin_rhs> -bool operator==(const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_lhs, multiplier_mixin_lhs>& lhs, - const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_rhs, multiplier_mixin_rhs>& rhs) -{ - return (lhs.multiplier() == rhs.multiplier()) - && (lhs.increment() == rhs.increment()) - && (lhs.state_ == rhs.state_); -} - -template <typename xtype, typename itype, - typename output_mixin, bool output_previous, - typename stream_mixin_lhs, typename multiplier_mixin_lhs, - typename stream_mixin_rhs, typename multiplier_mixin_rhs> -inline bool operator!=(const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_lhs, multiplier_mixin_lhs>& lhs, - const engine<xtype,itype, - output_mixin,output_previous, - stream_mixin_rhs, multiplier_mixin_rhs>& rhs) -{ - return !operator==(lhs,rhs); -} - - -template <typename xtype, typename itype, - template<typename XT,typename IT> class output_mixin, - bool output_previous = (sizeof(itype) <= 8), - template<typename IT> class multiplier_mixin = default_multiplier> -using oneseq_base = engine<xtype, itype, - output_mixin<xtype, itype>, output_previous, - oneseq_stream<itype>, - multiplier_mixin<itype> >; - -template <typename xtype, typename itype, - template<typename XT,typename IT> class output_mixin, - bool output_previous = (sizeof(itype) <= 8), - template<typename IT> class multiplier_mixin = default_multiplier> -using unique_base = engine<xtype, itype, - output_mixin<xtype, itype>, output_previous, - unique_stream<itype>, - multiplier_mixin<itype> >; - -template <typename xtype, typename itype, - template<typename XT,typename IT> class output_mixin, - bool output_previous = (sizeof(itype) <= 8), - template<typename IT> class multiplier_mixin = default_multiplier> -using setseq_base = engine<xtype, itype, - output_mixin<xtype, itype>, output_previous, - specific_stream<itype>, - multiplier_mixin<itype> >; - -template <typename xtype, typename itype, - template<typename XT,typename IT> class output_mixin, - bool output_previous = (sizeof(itype) <= 8), - template<typename IT> class multiplier_mixin = default_multiplier> -using mcg_base = engine<xtype, itype, - output_mixin<xtype, itype>, output_previous, - no_stream<itype>, - multiplier_mixin<itype> >; - -/* - * OUTPUT FUNCTIONS. - * - * These are the core of the PCG generation scheme. They specify how to - * turn the base LCG's internal state into the output value of the final - * generator. - * - * They're implemented as mixin classes. - * - * All of the classes have code that is written to allow it to be applied - * at *arbitrary* bit sizes, although in practice they'll only be used at - * standard sizes supported by C++. - */ - -/* - * XSH RS -- high xorshift, followed by a random shift - * - * Fast. A good performer. - */ - -template <typename xtype, typename itype> -struct xsh_rs_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t sparebits = bits - xtypebits; - constexpr bitcount_t opbits = - sparebits-5 >= 64 ? 5 - : sparebits-4 >= 32 ? 4 - : sparebits-3 >= 16 ? 3 - : sparebits-2 >= 4 ? 2 - : sparebits-1 >= 1 ? 1 - : 0; - constexpr bitcount_t mask = (1 << opbits) - 1; - constexpr bitcount_t maxrandshift = mask; - constexpr bitcount_t topspare = opbits; - constexpr bitcount_t bottomspare = sparebits - topspare; - constexpr bitcount_t xshift = topspare + (xtypebits+maxrandshift)/2; - bitcount_t rshift = - opbits ? bitcount_t(internal >> (bits - opbits)) & mask : 0; - internal ^= internal >> xshift; - xtype result = xtype(internal >> (bottomspare - maxrandshift + rshift)); - return result; - } -}; - -/* - * XSH RR -- high xorshift, followed by a random rotate - * - * Fast. A good performer. Slightly better statistically than XSH RS. - */ - -template <typename xtype, typename itype> -struct xsh_rr_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype)*8); - constexpr bitcount_t sparebits = bits - xtypebits; - constexpr bitcount_t wantedopbits = - xtypebits >= 128 ? 7 - : xtypebits >= 64 ? 6 - : xtypebits >= 32 ? 5 - : xtypebits >= 16 ? 4 - : 3; - constexpr bitcount_t opbits = - sparebits >= wantedopbits ? wantedopbits - : sparebits; - constexpr bitcount_t amplifier = wantedopbits - opbits; - constexpr bitcount_t mask = (1 << opbits) - 1; - constexpr bitcount_t topspare = opbits; - constexpr bitcount_t bottomspare = sparebits - topspare; - constexpr bitcount_t xshift = (topspare + xtypebits)/2; - bitcount_t rot = opbits ? bitcount_t(internal >> (bits - opbits)) & mask - : 0; - bitcount_t amprot = (rot << amplifier) & mask; - internal ^= internal >> xshift; - xtype result = xtype(internal >> bottomspare); - result = rotr(result, amprot); - return result; - } -}; - -/* - * RXS -- random xorshift - */ - -template <typename xtype, typename itype> -struct rxs_mixin { -static xtype output_rxs(itype internal) - { - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype)*8); - constexpr bitcount_t shift = bits - xtypebits; - constexpr bitcount_t extrashift = (xtypebits - shift)/2; - bitcount_t rshift = shift > 64+8 ? (internal >> (bits - 6)) & 63 - : shift > 32+4 ? (internal >> (bits - 5)) & 31 - : shift > 16+2 ? (internal >> (bits - 4)) & 15 - : shift > 8+1 ? (internal >> (bits - 3)) & 7 - : shift > 4+1 ? (internal >> (bits - 2)) & 3 - : shift > 2+1 ? (internal >> (bits - 1)) & 1 - : 0; - internal ^= internal >> (shift + extrashift - rshift); - xtype result = internal >> rshift; - return result; - } -}; - -/* - * RXS M XS -- random xorshift, mcg multiply, fixed xorshift - * - * The most statistically powerful generator, but all those steps - * make it slower than some of the others. We give it the rottenest jobs. - * - * Because it's usually used in contexts where the state type and the - * result type are the same, it is a permutation and is thus invertable. - * We thus provide a function to invert it. This function is used to - * for the "inside out" generator used by the extended generator. - */ - -/* Defined type-based concepts for the multiplication step. They're actually - * all derived by truncating the 128-bit, which was computed to be a good - * "universal" constant. - */ - -template <typename T> -struct mcg_multiplier { - // Not defined for an arbitrary type -}; - -template <typename T> -struct mcg_unmultiplier { - // Not defined for an arbitrary type -}; - -PCG_DEFINE_CONSTANT(uint8_t, mcg, multiplier, 217U) -PCG_DEFINE_CONSTANT(uint8_t, mcg, unmultiplier, 105U) - -PCG_DEFINE_CONSTANT(uint16_t, mcg, multiplier, 62169U) -PCG_DEFINE_CONSTANT(uint16_t, mcg, unmultiplier, 28009U) - -PCG_DEFINE_CONSTANT(uint32_t, mcg, multiplier, 277803737U) -PCG_DEFINE_CONSTANT(uint32_t, mcg, unmultiplier, 2897767785U) - -PCG_DEFINE_CONSTANT(uint64_t, mcg, multiplier, 12605985483714917081ULL) -PCG_DEFINE_CONSTANT(uint64_t, mcg, unmultiplier, 15009553638781119849ULL) - -PCG_DEFINE_CONSTANT(pcg128_t, mcg, multiplier, - PCG_128BIT_CONSTANT(17766728186571221404ULL, 12605985483714917081ULL)) -PCG_DEFINE_CONSTANT(pcg128_t, mcg, unmultiplier, - PCG_128BIT_CONSTANT(14422606686972528997ULL, 15009553638781119849ULL)) - - -template <typename xtype, typename itype> -struct rxs_m_xs_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t opbits = xtypebits >= 128 ? 6 - : xtypebits >= 64 ? 5 - : xtypebits >= 32 ? 4 - : xtypebits >= 16 ? 3 - : 2; - constexpr bitcount_t shift = bits - xtypebits; - constexpr bitcount_t mask = (1 << opbits) - 1; - bitcount_t rshift = - opbits ? bitcount_t(internal >> (bits - opbits)) & mask : 0; - internal ^= internal >> (opbits + rshift); - internal *= mcg_multiplier<itype>::multiplier(); - xtype result = internal >> shift; - result ^= result >> ((2U*xtypebits+2U)/3U); - return result; - } - - static itype unoutput(itype internal) - { - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t opbits = bits >= 128 ? 6 - : bits >= 64 ? 5 - : bits >= 32 ? 4 - : bits >= 16 ? 3 - : 2; - constexpr bitcount_t mask = (1 << opbits) - 1; - - internal = unxorshift(internal, bits, (2U*bits+2U)/3U); - - internal *= mcg_unmultiplier<itype>::unmultiplier(); - - bitcount_t rshift = opbits ? (internal >> (bits - opbits)) & mask : 0; - internal = unxorshift(internal, bits, opbits + rshift); - - return internal; - } -}; - - -/* - * RXS M -- random xorshift, mcg multiply - */ - -template <typename xtype, typename itype> -struct rxs_m_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t opbits = xtypebits >= 128 ? 6 - : xtypebits >= 64 ? 5 - : xtypebits >= 32 ? 4 - : xtypebits >= 16 ? 3 - : 2; - constexpr bitcount_t shift = bits - xtypebits; - constexpr bitcount_t mask = (1 << opbits) - 1; - bitcount_t rshift = opbits ? (internal >> (bits - opbits)) & mask : 0; - internal ^= internal >> (opbits + rshift); - internal *= mcg_multiplier<itype>::multiplier(); - xtype result = internal >> shift; - return result; - } -}; - - -/* - * DXSM -- double xorshift multiply - * - * This is a new, more powerful output permutation (added in 2019). It's - * a more comprehensive scrambling than RXS M, but runs faster on 128-bit - * types. Although primarily intended for use at large sizes, also works - * at smaller sizes as well. - * - * This permutation is similar to xorshift multiply hash functions, except - * that one of the multipliers is the LCG multiplier (to avoid needing to - * have a second constant) and the other is based on the low-order bits. - * This latter aspect means that the scrambling applied to the high bits - * depends on the low bits, and makes it (to my eye) impractical to back - * out the permutation without having the low-order bits. - */ - -template <typename xtype, typename itype> -struct dxsm_mixin { - inline xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t itypebits = bitcount_t(sizeof(itype) * 8); - static_assert(xtypebits <= itypebits/2, - "Output type must be half the size of the state type."); - - xtype hi = xtype(internal >> (itypebits - xtypebits)); - xtype lo = xtype(internal); - - lo |= 1; - hi ^= hi >> (xtypebits/2); - hi *= xtype(cheap_multiplier<itype>::multiplier()); - hi ^= hi >> (3*(xtypebits/4)); - hi *= lo; - return hi; - } -}; - - -/* - * XSL RR -- fixed xorshift (to low bits), random rotate - * - * Useful for 128-bit types that are split across two CPU registers. - */ - -template <typename xtype, typename itype> -struct xsl_rr_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t sparebits = bits - xtypebits; - constexpr bitcount_t wantedopbits = xtypebits >= 128 ? 7 - : xtypebits >= 64 ? 6 - : xtypebits >= 32 ? 5 - : xtypebits >= 16 ? 4 - : 3; - constexpr bitcount_t opbits = sparebits >= wantedopbits ? wantedopbits - : sparebits; - constexpr bitcount_t amplifier = wantedopbits - opbits; - constexpr bitcount_t mask = (1 << opbits) - 1; - constexpr bitcount_t topspare = sparebits; - constexpr bitcount_t bottomspare = sparebits - topspare; - constexpr bitcount_t xshift = (topspare + xtypebits) / 2; - - bitcount_t rot = - opbits ? bitcount_t(internal >> (bits - opbits)) & mask : 0; - bitcount_t amprot = (rot << amplifier) & mask; - internal ^= internal >> xshift; - xtype result = xtype(internal >> bottomspare); - result = rotr(result, amprot); - return result; - } -}; - - -/* - * XSL RR RR -- fixed xorshift (to low bits), random rotate (both parts) - * - * Useful for 128-bit types that are split across two CPU registers. - * If you really want an invertable 128-bit RNG, I guess this is the one. - */ - -template <typename T> struct halfsize_trait {}; -template <> struct halfsize_trait<pcg128_t> { typedef uint64_t type; }; -template <> struct halfsize_trait<uint64_t> { typedef uint32_t type; }; -template <> struct halfsize_trait<uint32_t> { typedef uint16_t type; }; -template <> struct halfsize_trait<uint16_t> { typedef uint8_t type; }; - -template <typename xtype, typename itype> -struct xsl_rr_rr_mixin { - typedef typename halfsize_trait<itype>::type htype; - - static itype output(itype internal) - { - constexpr bitcount_t htypebits = bitcount_t(sizeof(htype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t sparebits = bits - htypebits; - constexpr bitcount_t wantedopbits = htypebits >= 128 ? 7 - : htypebits >= 64 ? 6 - : htypebits >= 32 ? 5 - : htypebits >= 16 ? 4 - : 3; - constexpr bitcount_t opbits = sparebits >= wantedopbits ? wantedopbits - : sparebits; - constexpr bitcount_t amplifier = wantedopbits - opbits; - constexpr bitcount_t mask = (1 << opbits) - 1; - constexpr bitcount_t topspare = sparebits; - constexpr bitcount_t xshift = (topspare + htypebits) / 2; - - bitcount_t rot = - opbits ? bitcount_t(internal >> (bits - opbits)) & mask : 0; - bitcount_t amprot = (rot << amplifier) & mask; - internal ^= internal >> xshift; - htype lowbits = htype(internal); - lowbits = rotr(lowbits, amprot); - htype highbits = htype(internal >> topspare); - bitcount_t rot2 = lowbits & mask; - bitcount_t amprot2 = (rot2 << amplifier) & mask; - highbits = rotr(highbits, amprot2); - return (itype(highbits) << topspare) ^ itype(lowbits); - } -}; - - -/* - * XSH -- fixed xorshift (to high bits) - * - * You shouldn't use this at 64-bits or less. - */ - -template <typename xtype, typename itype> -struct xsh_mixin { - static xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t sparebits = bits - xtypebits; - constexpr bitcount_t topspare = 0; - constexpr bitcount_t bottomspare = sparebits - topspare; - constexpr bitcount_t xshift = (topspare + xtypebits) / 2; - - internal ^= internal >> xshift; - xtype result = internal >> bottomspare; - return result; - } -}; - -/* - * XSL -- fixed xorshift (to low bits) - * - * You shouldn't use this at 64-bits or less. - */ - -template <typename xtype, typename itype> -struct xsl_mixin { - inline xtype output(itype internal) - { - constexpr bitcount_t xtypebits = bitcount_t(sizeof(xtype) * 8); - constexpr bitcount_t bits = bitcount_t(sizeof(itype) * 8); - constexpr bitcount_t sparebits = bits - xtypebits; - constexpr bitcount_t topspare = sparebits; - constexpr bitcount_t bottomspare = sparebits - topspare; - constexpr bitcount_t xshift = (topspare + xtypebits) / 2; - - internal ^= internal >> xshift; - xtype result = internal >> bottomspare; - return result; - } -}; - - -/* ---- End of Output Functions ---- */ - - -template <typename baseclass> -struct inside_out : private baseclass { - inside_out() = delete; - - typedef typename baseclass::result_type result_type; - typedef typename baseclass::state_type state_type; - static_assert(sizeof(result_type) == sizeof(state_type), - "Require a RNG whose output function is a permutation"); - - static bool external_step(result_type& randval, size_t i) - { - state_type state = baseclass::unoutput(randval); - state = state * baseclass::multiplier() + baseclass::increment() - + state_type(i*2); - result_type result = baseclass::output(state); - randval = result; - state_type zero = - baseclass::is_mcg ? state & state_type(3U) : state_type(0U); - return result == zero; - } - - static bool external_advance(result_type& randval, size_t i, - result_type delta, bool forwards = true) - { - state_type state = baseclass::unoutput(randval); - state_type mult = baseclass::multiplier(); - state_type inc = baseclass::increment() + state_type(i*2); - state_type zero = - baseclass::is_mcg ? state & state_type(3U) : state_type(0U); - state_type dist_to_zero = baseclass::distance(state, zero, mult, inc); - bool crosses_zero = - forwards ? dist_to_zero <= delta - : (-dist_to_zero) <= delta; - if (!forwards) - delta = -delta; - state = baseclass::advance(state, delta, mult, inc); - randval = baseclass::output(state); - return crosses_zero; - } -}; - - -template <bitcount_t table_pow2, bitcount_t advance_pow2, typename baseclass, typename extvalclass, bool kdd = true> -class extended : public baseclass { -public: - typedef typename baseclass::state_type state_type; - typedef typename baseclass::result_type result_type; - typedef inside_out<extvalclass> insideout; - -private: - static constexpr bitcount_t rtypebits = sizeof(result_type)*8; - static constexpr bitcount_t stypebits = sizeof(state_type)*8; - - static constexpr bitcount_t tick_limit_pow2 = 64U; - - static constexpr size_t table_size = 1UL << table_pow2; - static constexpr size_t table_shift = stypebits - table_pow2; - static constexpr state_type table_mask = - (state_type(1U) << table_pow2) - state_type(1U); - - static constexpr bool may_tick = - (advance_pow2 < stypebits) && (advance_pow2 < tick_limit_pow2); - static constexpr size_t tick_shift = stypebits - advance_pow2; - static constexpr state_type tick_mask = - may_tick ? state_type( - (uint64_t(1) << (advance_pow2*may_tick)) - 1) - // ^-- stupidity to appease GCC warnings - : ~state_type(0U); - - static constexpr bool may_tock = stypebits < tick_limit_pow2; - - result_type data_[table_size]; - - PCG_NOINLINE void advance_table(); - - PCG_NOINLINE void advance_table(state_type delta, bool isForwards = true); - - result_type& get_extended_value() - { - state_type state = this->state_; - if (kdd && baseclass::is_mcg) { - // The low order bits of an MCG are constant, so drop them. - state >>= 2; - } - size_t index = kdd ? state & table_mask - : state >> table_shift; - - if (may_tick) { - bool tick = kdd ? (state & tick_mask) == state_type(0u) - : (state >> tick_shift) == state_type(0u); - if (tick) - advance_table(); - } - if (may_tock) { - bool tock = state == state_type(0u); - if (tock) - advance_table(); - } - return data_[index]; - } - -public: - static constexpr size_t period_pow2() - { - return baseclass::period_pow2() + table_size*extvalclass::period_pow2(); - } - - PCG_ALWAYS_INLINE result_type operator()() - { - result_type rhs = get_extended_value(); - result_type lhs = this->baseclass::operator()(); - return lhs ^ rhs; - } - - result_type operator()(result_type upper_bound) - { - return bounded_rand(*this, upper_bound); - } - - void set(result_type wanted) - { - result_type& rhs = get_extended_value(); - result_type lhs = this->baseclass::operator()(); - rhs = lhs ^ wanted; - } - - void advance(state_type distance, bool forwards = true); - - void backstep(state_type distance) - { - advance(distance, false); - } - - extended(const result_type* data) - : baseclass() - { - datainit(data); - } - - extended(const result_type* data, state_type seed) - : baseclass(seed) - { - datainit(data); - } - - // This function may or may not exist. It thus has to be a template - // to use SFINAE; users don't have to worry about its template-ness. - - template <typename bc = baseclass> - extended(const result_type* data, state_type seed, - typename bc::stream_state stream_seed) - : baseclass(seed, stream_seed) - { - datainit(data); - } - - extended() - : baseclass() - { - selfinit(); - } - - extended(state_type seed) - : baseclass(seed) - { - selfinit(); - } - - // This function may or may not exist. It thus has to be a template - // to use SFINAE; users don't have to worry about its template-ness. - - template <typename bc = baseclass> - extended(state_type seed, typename bc::stream_state stream_seed) - : baseclass(seed, stream_seed) - { - selfinit(); - } - -private: - void selfinit(); - void datainit(const result_type* data); - -public: - - template<typename SeedSeq, typename = typename std::enable_if< - !std::is_convertible<SeedSeq, result_type>::value - && !std::is_convertible<SeedSeq, extended>::value>::type> - extended(SeedSeq&& seedSeq) - : baseclass(seedSeq) - { - generate_to<table_size>(seedSeq, data_); - } - - template<typename... Args> - void seed(Args&&... args) - { - new (this) extended(std::forward<Args>(args)...); - } - - template <bitcount_t table_pow2_, bitcount_t advance_pow2_, - typename baseclass_, typename extvalclass_, bool kdd_> - friend bool operator==(const extended<table_pow2_, advance_pow2_, - baseclass_, extvalclass_, kdd_>&, - const extended<table_pow2_, advance_pow2_, - baseclass_, extvalclass_, kdd_>&); - - template <typename CharT, typename Traits, - bitcount_t table_pow2_, bitcount_t advance_pow2_, - typename baseclass_, typename extvalclass_, bool kdd_> - friend std::basic_ostream<CharT,Traits>& - operator<<(std::basic_ostream<CharT,Traits>& out, - const extended<table_pow2_, advance_pow2_, - baseclass_, extvalclass_, kdd_>&); - - template <typename CharT, typename Traits, - bitcount_t table_pow2_, bitcount_t advance_pow2_, - typename baseclass_, typename extvalclass_, bool kdd_> - friend std::basic_istream<CharT,Traits>& - operator>>(std::basic_istream<CharT,Traits>& in, - extended<table_pow2_, advance_pow2_, - baseclass_, extvalclass_, kdd_>&); - -}; - - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -void extended<table_pow2,advance_pow2,baseclass,extvalclass,kdd>::datainit( - const result_type* data) -{ - for (size_t i = 0; i < table_size; ++i) - data_[i] = data[i]; -} - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -void extended<table_pow2,advance_pow2,baseclass,extvalclass,kdd>::selfinit() -{ - // We need to fill the extended table with something, and we have - // very little provided data, so we use the base generator to - // produce values. Although not ideal (use a seed sequence, folks!), - // unexpected correlations are mitigated by - // - using XOR differences rather than the number directly - // - the way the table is accessed, its values *won't* be accessed - // in the same order the were written. - // - any strange correlations would only be apparent if we - // were to backstep the generator so that the base generator - // was generating the same values again - result_type lhs = baseclass::operator()(); - result_type rhs = baseclass::operator()(); - result_type xdiff = lhs - rhs; - for (size_t i = 0; i < table_size; ++i) { - data_[i] = baseclass::operator()() ^ xdiff; - } -} - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -bool operator==(const extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& lhs, - const extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& rhs) -{ - auto& base_lhs = static_cast<const baseclass&>(lhs); - auto& base_rhs = static_cast<const baseclass&>(rhs); - return base_lhs == base_rhs - && std::equal( - std::begin(lhs.data_), std::end(lhs.data_), - std::begin(rhs.data_) - ); -} - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -inline bool operator!=(const extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& lhs, - const extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& rhs) -{ - return !operator==(lhs, rhs); -} - -template <typename CharT, typename Traits, - bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -std::basic_ostream<CharT,Traits>& -operator<<(std::basic_ostream<CharT,Traits>& out, - const extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& rng) -{ - auto orig_flags = out.flags(std::ios_base::dec | std::ios_base::left); - auto space = out.widen(' '); - auto orig_fill = out.fill(); - - out << rng.multiplier() << space - << rng.increment() << space - << rng.state_; - - for (const auto& datum : rng.data_) - out << space << datum; - - out.flags(orig_flags); - out.fill(orig_fill); - return out; -} - -template <typename CharT, typename Traits, - bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -std::basic_istream<CharT,Traits>& -operator>>(std::basic_istream<CharT,Traits>& in, - extended<table_pow2, advance_pow2, - baseclass, extvalclass, kdd>& rng) -{ - extended<table_pow2, advance_pow2, baseclass, extvalclass> new_rng; - auto& base_rng = static_cast<baseclass&>(new_rng); - in >> base_rng; - - if (in.fail()) - return in; - - auto orig_flags = in.flags(std::ios_base::dec | std::ios_base::skipws); - - for (auto& datum : new_rng.data_) { - in >> datum; - if (in.fail()) - goto bail; - } - - rng = new_rng; - -bail: - in.flags(orig_flags); - return in; -} - - - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -void -extended<table_pow2,advance_pow2,baseclass,extvalclass,kdd>::advance_table() -{ - bool carry = false; - for (size_t i = 0; i < table_size; ++i) { - if (carry) { - carry = insideout::external_step(data_[i],i+1); - } - bool carry2 = insideout::external_step(data_[i],i+1); - carry = carry || carry2; - } -} - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -void -extended<table_pow2,advance_pow2,baseclass,extvalclass,kdd>::advance_table( - state_type delta, bool isForwards) -{ - typedef typename baseclass::state_type base_state_t; - typedef typename extvalclass::state_type ext_state_t; - constexpr bitcount_t basebits = sizeof(base_state_t)*8; - constexpr bitcount_t extbits = sizeof(ext_state_t)*8; - static_assert(basebits <= extbits || advance_pow2 > 0, - "Current implementation might overflow its carry"); - - base_state_t carry = 0; - for (size_t i = 0; i < table_size; ++i) { - base_state_t total_delta = carry + delta; - ext_state_t trunc_delta = ext_state_t(total_delta); - if (basebits > extbits) { - carry = total_delta >> extbits; - } else { - carry = 0; - } - carry += - insideout::external_advance(data_[i],i+1, trunc_delta, isForwards); - } -} - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename baseclass, typename extvalclass, bool kdd> -void extended<table_pow2,advance_pow2,baseclass,extvalclass,kdd>::advance( - state_type distance, bool forwards) -{ - static_assert(kdd, - "Efficient advance is too hard for non-kdd extension. " - "For a weak advance, cast to base class"); - state_type zero = - baseclass::is_mcg ? this->state_ & state_type(3U) : state_type(0U); - if (may_tick) { - state_type ticks = distance >> (advance_pow2*may_tick); - // ^-- stupidity to appease GCC - // warnings - state_type adv_mask = - baseclass::is_mcg ? tick_mask << 2 : tick_mask; - state_type next_advance_distance = this->distance(zero, adv_mask); - if (!forwards) - next_advance_distance = (-next_advance_distance) & tick_mask; - if (next_advance_distance < (distance & tick_mask)) { - ++ticks; - } - if (ticks) - advance_table(ticks, forwards); - } - if (forwards) { - if (may_tock && this->distance(zero) <= distance) - advance_table(); - baseclass::advance(distance); - } else { - if (may_tock && -(this->distance(zero)) <= distance) - advance_table(state_type(1U), false); - baseclass::advance(-distance); - } -} - -} // namespace pcg_detail - -namespace pcg_engines { - -using namespace pcg_detail; - -/* Predefined types for XSH RS */ - -typedef oneseq_base<uint8_t, uint16_t, xsh_rs_mixin> oneseq_xsh_rs_16_8; -typedef oneseq_base<uint16_t, uint32_t, xsh_rs_mixin> oneseq_xsh_rs_32_16; -typedef oneseq_base<uint32_t, uint64_t, xsh_rs_mixin> oneseq_xsh_rs_64_32; -typedef oneseq_base<uint64_t, pcg128_t, xsh_rs_mixin> oneseq_xsh_rs_128_64; -typedef oneseq_base<uint64_t, pcg128_t, xsh_rs_mixin, true, cheap_multiplier> - cm_oneseq_xsh_rs_128_64; - -typedef unique_base<uint8_t, uint16_t, xsh_rs_mixin> unique_xsh_rs_16_8; -typedef unique_base<uint16_t, uint32_t, xsh_rs_mixin> unique_xsh_rs_32_16; -typedef unique_base<uint32_t, uint64_t, xsh_rs_mixin> unique_xsh_rs_64_32; -typedef unique_base<uint64_t, pcg128_t, xsh_rs_mixin> unique_xsh_rs_128_64; -typedef unique_base<uint64_t, pcg128_t, xsh_rs_mixin, true, cheap_multiplier> - cm_unique_xsh_rs_128_64; - -typedef setseq_base<uint8_t, uint16_t, xsh_rs_mixin> setseq_xsh_rs_16_8; -typedef setseq_base<uint16_t, uint32_t, xsh_rs_mixin> setseq_xsh_rs_32_16; -typedef setseq_base<uint32_t, uint64_t, xsh_rs_mixin> setseq_xsh_rs_64_32; -typedef setseq_base<uint64_t, pcg128_t, xsh_rs_mixin> setseq_xsh_rs_128_64; -typedef setseq_base<uint64_t, pcg128_t, xsh_rs_mixin, true, cheap_multiplier> - cm_setseq_xsh_rs_128_64; - -typedef mcg_base<uint8_t, uint16_t, xsh_rs_mixin> mcg_xsh_rs_16_8; -typedef mcg_base<uint16_t, uint32_t, xsh_rs_mixin> mcg_xsh_rs_32_16; -typedef mcg_base<uint32_t, uint64_t, xsh_rs_mixin> mcg_xsh_rs_64_32; -typedef mcg_base<uint64_t, pcg128_t, xsh_rs_mixin> mcg_xsh_rs_128_64; -typedef mcg_base<uint64_t, pcg128_t, xsh_rs_mixin, true, cheap_multiplier> - cm_mcg_xsh_rs_128_64; - -/* Predefined types for XSH RR */ - -typedef oneseq_base<uint8_t, uint16_t, xsh_rr_mixin> oneseq_xsh_rr_16_8; -typedef oneseq_base<uint16_t, uint32_t, xsh_rr_mixin> oneseq_xsh_rr_32_16; -typedef oneseq_base<uint32_t, uint64_t, xsh_rr_mixin> oneseq_xsh_rr_64_32; -typedef oneseq_base<uint64_t, pcg128_t, xsh_rr_mixin> oneseq_xsh_rr_128_64; -typedef oneseq_base<uint64_t, pcg128_t, xsh_rr_mixin, true, cheap_multiplier> - cm_oneseq_xsh_rr_128_64; - -typedef unique_base<uint8_t, uint16_t, xsh_rr_mixin> unique_xsh_rr_16_8; -typedef unique_base<uint16_t, uint32_t, xsh_rr_mixin> unique_xsh_rr_32_16; -typedef unique_base<uint32_t, uint64_t, xsh_rr_mixin> unique_xsh_rr_64_32; -typedef unique_base<uint64_t, pcg128_t, xsh_rr_mixin> unique_xsh_rr_128_64; -typedef unique_base<uint64_t, pcg128_t, xsh_rr_mixin, true, cheap_multiplier> - cm_unique_xsh_rr_128_64; - -typedef setseq_base<uint8_t, uint16_t, xsh_rr_mixin> setseq_xsh_rr_16_8; -typedef setseq_base<uint16_t, uint32_t, xsh_rr_mixin> setseq_xsh_rr_32_16; -typedef setseq_base<uint32_t, uint64_t, xsh_rr_mixin> setseq_xsh_rr_64_32; -typedef setseq_base<uint64_t, pcg128_t, xsh_rr_mixin> setseq_xsh_rr_128_64; -typedef setseq_base<uint64_t, pcg128_t, xsh_rr_mixin, true, cheap_multiplier> - cm_setseq_xsh_rr_128_64; - -typedef mcg_base<uint8_t, uint16_t, xsh_rr_mixin> mcg_xsh_rr_16_8; -typedef mcg_base<uint16_t, uint32_t, xsh_rr_mixin> mcg_xsh_rr_32_16; -typedef mcg_base<uint32_t, uint64_t, xsh_rr_mixin> mcg_xsh_rr_64_32; -typedef mcg_base<uint64_t, pcg128_t, xsh_rr_mixin> mcg_xsh_rr_128_64; -typedef mcg_base<uint64_t, pcg128_t, xsh_rr_mixin, true, cheap_multiplier> - cm_mcg_xsh_rr_128_64; - - -/* Predefined types for RXS M XS */ - -typedef oneseq_base<uint8_t, uint8_t, rxs_m_xs_mixin> oneseq_rxs_m_xs_8_8; -typedef oneseq_base<uint16_t, uint16_t, rxs_m_xs_mixin> oneseq_rxs_m_xs_16_16; -typedef oneseq_base<uint32_t, uint32_t, rxs_m_xs_mixin> oneseq_rxs_m_xs_32_32; -typedef oneseq_base<uint64_t, uint64_t, rxs_m_xs_mixin> oneseq_rxs_m_xs_64_64; -typedef oneseq_base<pcg128_t, pcg128_t, rxs_m_xs_mixin> - oneseq_rxs_m_xs_128_128; -typedef oneseq_base<pcg128_t, pcg128_t, rxs_m_xs_mixin, true, cheap_multiplier> - cm_oneseq_rxs_m_xs_128_128; - -typedef unique_base<uint8_t, uint8_t, rxs_m_xs_mixin> unique_rxs_m_xs_8_8; -typedef unique_base<uint16_t, uint16_t, rxs_m_xs_mixin> unique_rxs_m_xs_16_16; -typedef unique_base<uint32_t, uint32_t, rxs_m_xs_mixin> unique_rxs_m_xs_32_32; -typedef unique_base<uint64_t, uint64_t, rxs_m_xs_mixin> unique_rxs_m_xs_64_64; -typedef unique_base<pcg128_t, pcg128_t, rxs_m_xs_mixin> unique_rxs_m_xs_128_128; -typedef unique_base<pcg128_t, pcg128_t, rxs_m_xs_mixin, true, cheap_multiplier> - cm_unique_rxs_m_xs_128_128; - -typedef setseq_base<uint8_t, uint8_t, rxs_m_xs_mixin> setseq_rxs_m_xs_8_8; -typedef setseq_base<uint16_t, uint16_t, rxs_m_xs_mixin> setseq_rxs_m_xs_16_16; -typedef setseq_base<uint32_t, uint32_t, rxs_m_xs_mixin> setseq_rxs_m_xs_32_32; -typedef setseq_base<uint64_t, uint64_t, rxs_m_xs_mixin> setseq_rxs_m_xs_64_64; -typedef setseq_base<pcg128_t, pcg128_t, rxs_m_xs_mixin> setseq_rxs_m_xs_128_128; -typedef setseq_base<pcg128_t, pcg128_t, rxs_m_xs_mixin, true, cheap_multiplier> - cm_setseq_rxs_m_xs_128_128; - - // MCG versions don't make sense here, so aren't defined. - -/* Predefined types for RXS M */ - -typedef oneseq_base<uint8_t, uint16_t, rxs_m_mixin> oneseq_rxs_m_16_8; -typedef oneseq_base<uint16_t, uint32_t, rxs_m_mixin> oneseq_rxs_m_32_16; -typedef oneseq_base<uint32_t, uint64_t, rxs_m_mixin> oneseq_rxs_m_64_32; -typedef oneseq_base<uint64_t, pcg128_t, rxs_m_mixin> oneseq_rxs_m_128_64; -typedef oneseq_base<uint64_t, pcg128_t, rxs_m_mixin, true, cheap_multiplier> - cm_oneseq_rxs_m_128_64; - -typedef unique_base<uint8_t, uint16_t, rxs_m_mixin> unique_rxs_m_16_8; -typedef unique_base<uint16_t, uint32_t, rxs_m_mixin> unique_rxs_m_32_16; -typedef unique_base<uint32_t, uint64_t, rxs_m_mixin> unique_rxs_m_64_32; -typedef unique_base<uint64_t, pcg128_t, rxs_m_mixin> unique_rxs_m_128_64; -typedef unique_base<uint64_t, pcg128_t, rxs_m_mixin, true, cheap_multiplier> - cm_unique_rxs_m_128_64; - -typedef setseq_base<uint8_t, uint16_t, rxs_m_mixin> setseq_rxs_m_16_8; -typedef setseq_base<uint16_t, uint32_t, rxs_m_mixin> setseq_rxs_m_32_16; -typedef setseq_base<uint32_t, uint64_t, rxs_m_mixin> setseq_rxs_m_64_32; -typedef setseq_base<uint64_t, pcg128_t, rxs_m_mixin> setseq_rxs_m_128_64; -typedef setseq_base<uint64_t, pcg128_t, rxs_m_mixin, true, cheap_multiplier> - cm_setseq_rxs_m_128_64; - -typedef mcg_base<uint8_t, uint16_t, rxs_m_mixin> mcg_rxs_m_16_8; -typedef mcg_base<uint16_t, uint32_t, rxs_m_mixin> mcg_rxs_m_32_16; -typedef mcg_base<uint32_t, uint64_t, rxs_m_mixin> mcg_rxs_m_64_32; -typedef mcg_base<uint64_t, pcg128_t, rxs_m_mixin> mcg_rxs_m_128_64; -typedef mcg_base<uint64_t, pcg128_t, rxs_m_mixin, true, cheap_multiplier> - cm_mcg_rxs_m_128_64; - -/* Predefined types for DXSM */ - -typedef oneseq_base<uint8_t, uint16_t, dxsm_mixin> oneseq_dxsm_16_8; -typedef oneseq_base<uint16_t, uint32_t, dxsm_mixin> oneseq_dxsm_32_16; -typedef oneseq_base<uint32_t, uint64_t, dxsm_mixin> oneseq_dxsm_64_32; -typedef oneseq_base<uint64_t, pcg128_t, dxsm_mixin> oneseq_dxsm_128_64; -typedef oneseq_base<uint64_t, pcg128_t, dxsm_mixin, true, cheap_multiplier> - cm_oneseq_dxsm_128_64; - -typedef unique_base<uint8_t, uint16_t, dxsm_mixin> unique_dxsm_16_8; -typedef unique_base<uint16_t, uint32_t, dxsm_mixin> unique_dxsm_32_16; -typedef unique_base<uint32_t, uint64_t, dxsm_mixin> unique_dxsm_64_32; -typedef unique_base<uint64_t, pcg128_t, dxsm_mixin> unique_dxsm_128_64; -typedef unique_base<uint64_t, pcg128_t, dxsm_mixin, true, cheap_multiplier> - cm_unique_dxsm_128_64; - -typedef setseq_base<uint8_t, uint16_t, dxsm_mixin> setseq_dxsm_16_8; -typedef setseq_base<uint16_t, uint32_t, dxsm_mixin> setseq_dxsm_32_16; -typedef setseq_base<uint32_t, uint64_t, dxsm_mixin> setseq_dxsm_64_32; -typedef setseq_base<uint64_t, pcg128_t, dxsm_mixin> setseq_dxsm_128_64; -typedef setseq_base<uint64_t, pcg128_t, dxsm_mixin, true, cheap_multiplier> - cm_setseq_dxsm_128_64; - -typedef mcg_base<uint8_t, uint16_t, dxsm_mixin> mcg_dxsm_16_8; -typedef mcg_base<uint16_t, uint32_t, dxsm_mixin> mcg_dxsm_32_16; -typedef mcg_base<uint32_t, uint64_t, dxsm_mixin> mcg_dxsm_64_32; -typedef mcg_base<uint64_t, pcg128_t, dxsm_mixin> mcg_dxsm_128_64; -typedef mcg_base<uint64_t, pcg128_t, dxsm_mixin, true, cheap_multiplier> - cm_mcg_dxsm_128_64; - -/* Predefined types for XSL RR (only defined for "large" types) */ - -typedef oneseq_base<uint32_t, uint64_t, xsl_rr_mixin> oneseq_xsl_rr_64_32; -typedef oneseq_base<uint64_t, pcg128_t, xsl_rr_mixin> oneseq_xsl_rr_128_64; -typedef oneseq_base<uint64_t, pcg128_t, xsl_rr_mixin, true, cheap_multiplier> - cm_oneseq_xsl_rr_128_64; - -typedef unique_base<uint32_t, uint64_t, xsl_rr_mixin> unique_xsl_rr_64_32; -typedef unique_base<uint64_t, pcg128_t, xsl_rr_mixin> unique_xsl_rr_128_64; -typedef unique_base<uint64_t, pcg128_t, xsl_rr_mixin, true, cheap_multiplier> - cm_unique_xsl_rr_128_64; - -typedef setseq_base<uint32_t, uint64_t, xsl_rr_mixin> setseq_xsl_rr_64_32; -typedef setseq_base<uint64_t, pcg128_t, xsl_rr_mixin> setseq_xsl_rr_128_64; -typedef setseq_base<uint64_t, pcg128_t, xsl_rr_mixin, true, cheap_multiplier> - cm_setseq_xsl_rr_128_64; - -typedef mcg_base<uint32_t, uint64_t, xsl_rr_mixin> mcg_xsl_rr_64_32; -typedef mcg_base<uint64_t, pcg128_t, xsl_rr_mixin> mcg_xsl_rr_128_64; -typedef mcg_base<uint64_t, pcg128_t, xsl_rr_mixin, true, cheap_multiplier> - cm_mcg_xsl_rr_128_64; - - -/* Predefined types for XSL RR RR (only defined for "large" types) */ - -typedef oneseq_base<uint64_t, uint64_t, xsl_rr_rr_mixin> - oneseq_xsl_rr_rr_64_64; -typedef oneseq_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin> - oneseq_xsl_rr_rr_128_128; -typedef oneseq_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin, true, cheap_multiplier> - cm_oneseq_xsl_rr_rr_128_128; - -typedef unique_base<uint64_t, uint64_t, xsl_rr_rr_mixin> - unique_xsl_rr_rr_64_64; -typedef unique_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin> - unique_xsl_rr_rr_128_128; -typedef unique_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin, true, cheap_multiplier> - cm_unique_xsl_rr_rr_128_128; - -typedef setseq_base<uint64_t, uint64_t, xsl_rr_rr_mixin> - setseq_xsl_rr_rr_64_64; -typedef setseq_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin> - setseq_xsl_rr_rr_128_128; -typedef setseq_base<pcg128_t, pcg128_t, xsl_rr_rr_mixin, true, cheap_multiplier> - cm_setseq_xsl_rr_rr_128_128; - - // MCG versions don't make sense here, so aren't defined. - -/* Extended generators */ - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename BaseRNG, bool kdd = true> -using ext_std8 = extended<table_pow2, advance_pow2, BaseRNG, - oneseq_rxs_m_xs_8_8, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename BaseRNG, bool kdd = true> -using ext_std16 = extended<table_pow2, advance_pow2, BaseRNG, - oneseq_rxs_m_xs_16_16, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename BaseRNG, bool kdd = true> -using ext_std32 = extended<table_pow2, advance_pow2, BaseRNG, - oneseq_rxs_m_xs_32_32, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, - typename BaseRNG, bool kdd = true> -using ext_std64 = extended<table_pow2, advance_pow2, BaseRNG, - oneseq_rxs_m_xs_64_64, kdd>; - - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_oneseq_rxs_m_xs_32_32 = - ext_std32<table_pow2, advance_pow2, oneseq_rxs_m_xs_32_32, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_mcg_xsh_rs_64_32 = - ext_std32<table_pow2, advance_pow2, mcg_xsh_rs_64_32, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_oneseq_xsh_rs_64_32 = - ext_std32<table_pow2, advance_pow2, oneseq_xsh_rs_64_32, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_setseq_xsh_rr_64_32 = - ext_std32<table_pow2, advance_pow2, setseq_xsh_rr_64_32, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_mcg_xsl_rr_128_64 = - ext_std64<table_pow2, advance_pow2, mcg_xsl_rr_128_64, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_oneseq_xsl_rr_128_64 = - ext_std64<table_pow2, advance_pow2, oneseq_xsl_rr_128_64, kdd>; - -template <bitcount_t table_pow2, bitcount_t advance_pow2, bool kdd = true> -using ext_setseq_xsl_rr_128_64 = - ext_std64<table_pow2, advance_pow2, setseq_xsl_rr_128_64, kdd>; - -} // namespace pcg_engines - -typedef pcg_engines::setseq_xsh_rr_64_32 pcg32; -typedef pcg_engines::oneseq_xsh_rr_64_32 pcg32_oneseq; -typedef pcg_engines::unique_xsh_rr_64_32 pcg32_unique; -typedef pcg_engines::mcg_xsh_rs_64_32 pcg32_fast; - -typedef pcg_engines::setseq_xsl_rr_128_64 pcg64; -typedef pcg_engines::oneseq_xsl_rr_128_64 pcg64_oneseq; -typedef pcg_engines::unique_xsl_rr_128_64 pcg64_unique; -typedef pcg_engines::mcg_xsl_rr_128_64 pcg64_fast; - -typedef pcg_engines::setseq_rxs_m_xs_8_8 pcg8_once_insecure; -typedef pcg_engines::setseq_rxs_m_xs_16_16 pcg16_once_insecure; -typedef pcg_engines::setseq_rxs_m_xs_32_32 pcg32_once_insecure; -typedef pcg_engines::setseq_rxs_m_xs_64_64 pcg64_once_insecure; -typedef pcg_engines::setseq_xsl_rr_rr_128_128 pcg128_once_insecure; - -typedef pcg_engines::oneseq_rxs_m_xs_8_8 pcg8_oneseq_once_insecure; -typedef pcg_engines::oneseq_rxs_m_xs_16_16 pcg16_oneseq_once_insecure; -typedef pcg_engines::oneseq_rxs_m_xs_32_32 pcg32_oneseq_once_insecure; -typedef pcg_engines::oneseq_rxs_m_xs_64_64 pcg64_oneseq_once_insecure; -typedef pcg_engines::oneseq_xsl_rr_rr_128_128 pcg128_oneseq_once_insecure; - - -// These two extended RNGs provide two-dimensionally equidistributed -// 32-bit generators. pcg32_k2_fast occupies the same space as pcg64, -// and can be called twice to generate 64 bits, but does not required -// 128-bit math; on 32-bit systems, it's faster than pcg64 as well. - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<1,16,true> pcg32_k2; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<1,32,true> pcg32_k2_fast; - -// These eight extended RNGs have about as much state as arc4random -// -// - the k variants are k-dimensionally equidistributed -// - the c variants offer better crypographic security -// -// (just how good the cryptographic security is is an open question) - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<6,16,true> pcg32_k64; -typedef pcg_engines::ext_mcg_xsh_rs_64_32<6,32,true> pcg32_k64_oneseq; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<6,32,true> pcg32_k64_fast; - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<6,16,false> pcg32_c64; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<6,32,false> pcg32_c64_oneseq; -typedef pcg_engines::ext_mcg_xsh_rs_64_32<6,32,false> pcg32_c64_fast; - -typedef pcg_engines::ext_setseq_xsl_rr_128_64<5,16,true> pcg64_k32; -typedef pcg_engines::ext_oneseq_xsl_rr_128_64<5,128,true> pcg64_k32_oneseq; -typedef pcg_engines::ext_mcg_xsl_rr_128_64<5,128,true> pcg64_k32_fast; - -typedef pcg_engines::ext_setseq_xsl_rr_128_64<5,16,false> pcg64_c32; -typedef pcg_engines::ext_oneseq_xsl_rr_128_64<5,128,false> pcg64_c32_oneseq; -typedef pcg_engines::ext_mcg_xsl_rr_128_64<5,128,false> pcg64_c32_fast; - -// These eight extended RNGs have more state than the Mersenne twister -// -// - the k variants are k-dimensionally equidistributed -// - the c variants offer better crypographic security -// -// (just how good the cryptographic security is is an open question) - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<10,16,true> pcg32_k1024; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<10,32,true> pcg32_k1024_fast; - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<10,16,false> pcg32_c1024; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<10,32,false> pcg32_c1024_fast; - -typedef pcg_engines::ext_setseq_xsl_rr_128_64<10,16,true> pcg64_k1024; -typedef pcg_engines::ext_oneseq_xsl_rr_128_64<10,128,true> pcg64_k1024_fast; - -typedef pcg_engines::ext_setseq_xsl_rr_128_64<10,16,false> pcg64_c1024; -typedef pcg_engines::ext_oneseq_xsl_rr_128_64<10,128,false> pcg64_c1024_fast; - -// These generators have an insanely huge period (2^524352), and is suitable -// for silly party tricks, such as dumping out 64 KB ZIP files at an arbitrary -// point in the future. [Actually, over the full period of the generator, it -// will produce every 64 KB ZIP file 2^64 times!] - -typedef pcg_engines::ext_setseq_xsh_rr_64_32<14,16,true> pcg32_k16384; -typedef pcg_engines::ext_oneseq_xsh_rs_64_32<14,32,true> pcg32_k16384_fast; - -#ifdef _MSC_VER - #pragma warning(default:4146) -#endif - -#endif // PCG_RAND_HPP_INCLUDED diff --git a/benchmarks/others/pcg_uint128.hpp b/benchmarks/others/pcg_uint128.hpp deleted file mode 100644 index 9131b21d..00000000 --- a/benchmarks/others/pcg_uint128.hpp +++ /dev/null @@ -1,884 +0,0 @@ -/* - * PCG Random Number Generation for C++ - * - * Copyright 2014-2017 Melissa O'Neill <[email protected]>, - * and the PCG Project contributors. - * - * SPDX-License-Identifier: (Apache-2.0 OR MIT) - * - * Licensed under the Apache License, Version 2.0 (provided in - * LICENSE-APACHE.txt and at http://www.apache.org/licenses/LICENSE-2.0) - * or under the MIT license (provided in LICENSE-MIT.txt and at - * http://opensource.org/licenses/MIT), at your option. This file may not - * be copied, modified, or distributed except according to those terms. - * - * Distributed on an "AS IS" BASIS, WITHOUT WARRANTY OF ANY KIND, either - * express or implied. See your chosen license for details. - * - * For additional information about the PCG random number generation scheme, - * visit http://www.pcg-random.org/. - */ - -/* - * This code provides a a C++ class that can provide 128-bit (or higher) - * integers. To produce 2K-bit integers, it uses two K-bit integers, - * placed in a union that allowes the code to also see them as four K/2 bit - * integers (and access them either directly name, or by index). - * - * It may seem like we're reinventing the wheel here, because several - * libraries already exist that support large integers, but most existing - * libraries provide a very generic multiprecision code, but here we're - * operating at a fixed size. Also, most other libraries are fairly - * heavyweight. So we use a direct implementation. Sadly, it's much slower - * than hand-coded assembly or direct CPU support. - */ - -#ifndef PCG_UINT128_HPP_INCLUDED -#define PCG_UINT128_HPP_INCLUDED 1 - -#include <cstdint> -#include <cstdio> -#include <cassert> -#include <climits> -#include <utility> -#include <initializer_list> -#include <type_traits> - -#if defined(_MSC_VER) // Use MSVC++ intrinsics -#include <intrin.h> -#endif - -/* - * We want to lay the type out the same way that a native type would be laid - * out, which means we must know the machine's endian, at compile time. - * This ugliness attempts to do so. - */ - -#ifndef PCG_LITTLE_ENDIAN - #if defined(__BYTE_ORDER__) - #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ - #define PCG_LITTLE_ENDIAN 1 - #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ - #define PCG_LITTLE_ENDIAN 0 - #else - #error __BYTE_ORDER__ does not match a standard endian, pick a side - #endif - #elif __LITTLE_ENDIAN__ || _LITTLE_ENDIAN - #define PCG_LITTLE_ENDIAN 1 - #elif __BIG_ENDIAN__ || _BIG_ENDIAN - #define PCG_LITTLE_ENDIAN 0 - #elif __x86_64 || __x86_64__ || _M_X64 || __i386 || __i386__ || _M_IX86 - #define PCG_LITTLE_ENDIAN 1 - #elif __powerpc__ || __POWERPC__ || __ppc__ || __PPC__ \ - || __m68k__ || __mc68000__ - #define PCG_LITTLE_ENDIAN 0 - #else - #error Unable to determine target endianness - #endif -#endif - -namespace pcg_extras { - -// Recent versions of GCC have intrinsics we can use to quickly calculate -// the number of leading and trailing zeros in a number. If possible, we -// use them, otherwise we fall back to old-fashioned bit twiddling to figure -// them out. - -#ifndef PCG_BITCOUNT_T - typedef uint8_t bitcount_t; -#else - typedef PCG_BITCOUNT_T bitcount_t; -#endif - -/* - * Provide some useful helper functions - * * flog2 floor(log2(x)) - * * trailingzeros number of trailing zero bits - */ - -#if defined(__GNUC__) // Any GNU-compatible compiler supporting C++11 has - // some useful intrinsics we can use. - -inline bitcount_t flog2(uint32_t v) -{ - return 31 - __builtin_clz(v); -} - -inline bitcount_t trailingzeros(uint32_t v) -{ - return __builtin_ctz(v); -} - -inline bitcount_t flog2(uint64_t v) -{ -#if UINT64_MAX == ULONG_MAX - return 63 - __builtin_clzl(v); -#elif UINT64_MAX == ULLONG_MAX - return 63 - __builtin_clzll(v); -#else - #error Cannot find a function for uint64_t -#endif -} - -inline bitcount_t trailingzeros(uint64_t v) -{ -#if UINT64_MAX == ULONG_MAX - return __builtin_ctzl(v); -#elif UINT64_MAX == ULLONG_MAX - return __builtin_ctzll(v); -#else - #error Cannot find a function for uint64_t -#endif -} - -#elif defined(_MSC_VER) // Use MSVC++ intrinsics - -#pragma intrinsic(_BitScanReverse, _BitScanForward) -#if defined(_M_X64) || defined(_M_ARM) || defined(_M_ARM64) -#pragma intrinsic(_BitScanReverse64, _BitScanForward64) -#endif - -inline bitcount_t flog2(uint32_t v) -{ - unsigned long i; - _BitScanReverse(&i, v); - return bitcount_t(i); -} - -inline bitcount_t trailingzeros(uint32_t v) -{ - unsigned long i; - _BitScanForward(&i, v); - return bitcount_t(i); -} - -inline bitcount_t flog2(uint64_t v) -{ -#if defined(_M_X64) || defined(_M_ARM) || defined(_M_ARM64) - unsigned long i; - _BitScanReverse64(&i, v); - return bitcount_t(i); -#else - // 32-bit x86 - uint32_t high = v >> 32; - uint32_t low = uint32_t(v); - return high ? 32+flog2(high) : flog2(low); -#endif -} - -inline bitcount_t trailingzeros(uint64_t v) -{ -#if defined(_M_X64) || defined(_M_ARM) || defined(_M_ARM64) - unsigned long i; - _BitScanForward64(&i, v); - return bitcount_t(i); -#else - // 32-bit x86 - uint32_t high = v >> 32; - uint32_t low = uint32_t(v); - return low ? trailingzeros(low) : trailingzeros(high)+32; -#endif -} - -#else // Otherwise, we fall back to bit twiddling - // implementations - -inline bitcount_t flog2(uint32_t v) -{ - // Based on code by Eric Cole and Mark Dickinson, which appears at - // https://graphics.stanford.edu/~seander/bithacks.html#IntegerLogDeBruijn - - static const uint8_t multiplyDeBruijnBitPos[32] = { - 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, - 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 - }; - - v |= v >> 1; // first round down to one less than a power of 2 - v |= v >> 2; - v |= v >> 4; - v |= v >> 8; - v |= v >> 16; - - return multiplyDeBruijnBitPos[(uint32_t)(v * 0x07C4ACDDU) >> 27]; -} - -inline bitcount_t trailingzeros(uint32_t v) -{ - static const uint8_t multiplyDeBruijnBitPos[32] = { - 0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8, - 31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 - }; - - return multiplyDeBruijnBitPos[((uint32_t)((v & -v) * 0x077CB531U)) >> 27]; -} - -inline bitcount_t flog2(uint64_t v) -{ - uint32_t high = v >> 32; - uint32_t low = uint32_t(v); - - return high ? 32+flog2(high) : flog2(low); -} - -inline bitcount_t trailingzeros(uint64_t v) -{ - uint32_t high = v >> 32; - uint32_t low = uint32_t(v); - - return low ? trailingzeros(low) : trailingzeros(high)+32; -} - -#endif - -inline bitcount_t flog2(uint8_t v) -{ - return flog2(uint32_t(v)); -} - -inline bitcount_t flog2(uint16_t v) -{ - return flog2(uint32_t(v)); -} - -#if __SIZEOF_INT128__ -inline bitcount_t flog2(__uint128_t v) -{ - uint64_t high = uint64_t(v >> 64); - uint64_t low = uint64_t(v); - - return high ? 64+flog2(high) : flog2(low); -} -#endif - -inline bitcount_t trailingzeros(uint8_t v) -{ - return trailingzeros(uint32_t(v)); -} - -inline bitcount_t trailingzeros(uint16_t v) -{ - return trailingzeros(uint32_t(v)); -} - -#if __SIZEOF_INT128__ -inline bitcount_t trailingzeros(__uint128_t v) -{ - uint64_t high = uint64_t(v >> 64); - uint64_t low = uint64_t(v); - return low ? trailingzeros(low) : trailingzeros(high)+64; -} -#endif - -template <typename UInt> -inline bitcount_t clog2(UInt v) -{ - return flog2(v) + ((v & (-v)) != v); -} - -template <typename UInt> -inline UInt addwithcarry(UInt x, UInt y, bool carryin, bool* carryout) -{ - UInt half_result = y + carryin; - UInt result = x + half_result; - *carryout = (half_result < y) || (result < x); - return result; -} - -template <typename UInt> -inline UInt subwithcarry(UInt x, UInt y, bool carryin, bool* carryout) -{ - UInt half_result = y + carryin; - UInt result = x - half_result; - *carryout = (half_result < y) || (result > x); - return result; -} - - -template <typename UInt, typename UIntX2> -class uint_x4 { -// private: - static constexpr unsigned int UINT_BITS = sizeof(UInt) * CHAR_BIT; -public: - union { -#if PCG_LITTLE_ENDIAN - struct { - UInt v0, v1, v2, v3; - } w; - struct { - UIntX2 v01, v23; - } d; -#else - struct { - UInt v3, v2, v1, v0; - } w; - struct { - UIntX2 v23, v01; - } d; -#endif - // For the array access versions, the code that uses the array - // must handle endian itself. Yuck. - UInt wa[4]; - UIntX2 da[2]; - }; - -public: - uint_x4() = default; - - constexpr uint_x4(UInt v3, UInt v2, UInt v1, UInt v0) -#if PCG_LITTLE_ENDIAN - : w{v0, v1, v2, v3} -#else - : w{v3, v2, v1, v0} -#endif - { - // Nothing (else) to do - } - - constexpr uint_x4(UIntX2 v23, UIntX2 v01) -#if PCG_LITTLE_ENDIAN - : d{v01,v23} -#else - : d{v23,v01} -#endif - { - // Nothing (else) to do - } - - constexpr uint_x4(UIntX2 v01) -#if PCG_LITTLE_ENDIAN - : d{v01, UIntX2(0)} -#else - : d{UIntX2(0),v01} -#endif - { - // Nothing (else) to do - } - - template<class Integral, - typename std::enable_if<(std::is_integral<Integral>::value - && sizeof(Integral) <= sizeof(UIntX2)) - >::type* = nullptr> - constexpr uint_x4(Integral v01) -#if PCG_LITTLE_ENDIAN - : d{UIntX2(v01), UIntX2(0)} -#else - : d{UIntX2(0), UIntX2(v01)} -#endif - { - // Nothing (else) to do - } - - explicit constexpr operator UIntX2() const - { - return d.v01; - } - - template<class Integral, - typename std::enable_if<(std::is_integral<Integral>::value - && sizeof(Integral) <= sizeof(UIntX2)) - >::type* = nullptr> - explicit constexpr operator Integral() const - { - return Integral(d.v01); - } - - explicit constexpr operator bool() const - { - return d.v01 || d.v23; - } - - template<typename U, typename V> - friend uint_x4<U,V> operator*(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator*(const uint_x4<U,V>&, V); - - template<typename U, typename V> - friend std::pair< uint_x4<U,V>,uint_x4<U,V> > - divmod(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator+(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator-(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator<<(const uint_x4<U,V>&, const bitcount_t shift); - - template<typename U, typename V> - friend uint_x4<U,V> operator>>(const uint_x4<U,V>&, const bitcount_t shift); - - template<typename U, typename V> - friend uint_x4<U,V> operator&(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator|(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator^(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator==(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator!=(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator<(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator<=(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator>(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend bool operator>=(const uint_x4<U,V>&, const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator~(const uint_x4<U,V>&); - - template<typename U, typename V> - friend uint_x4<U,V> operator-(const uint_x4<U,V>&); - - template<typename U, typename V> - friend bitcount_t flog2(const uint_x4<U,V>&); - - template<typename U, typename V> - friend bitcount_t trailingzeros(const uint_x4<U,V>&); - - uint_x4& operator*=(const uint_x4& rhs) - { - uint_x4 result = *this * rhs; - return *this = result; - } - - uint_x4& operator*=(UIntX2 rhs) - { - uint_x4 result = *this * rhs; - return *this = result; - } - - uint_x4& operator/=(const uint_x4& rhs) - { - uint_x4 result = *this / rhs; - return *this = result; - } - - uint_x4& operator%=(const uint_x4& rhs) - { - uint_x4 result = *this % rhs; - return *this = result; - } - - uint_x4& operator+=(const uint_x4& rhs) - { - uint_x4 result = *this + rhs; - return *this = result; - } - - uint_x4& operator-=(const uint_x4& rhs) - { - uint_x4 result = *this - rhs; - return *this = result; - } - - uint_x4& operator&=(const uint_x4& rhs) - { - uint_x4 result = *this & rhs; - return *this = result; - } - - uint_x4& operator|=(const uint_x4& rhs) - { - uint_x4 result = *this | rhs; - return *this = result; - } - - uint_x4& operator^=(const uint_x4& rhs) - { - uint_x4 result = *this ^ rhs; - return *this = result; - } - - uint_x4& operator>>=(bitcount_t shift) - { - uint_x4 result = *this >> shift; - return *this = result; - } - - uint_x4& operator<<=(bitcount_t shift) - { - uint_x4 result = *this << shift; - return *this = result; - } - -}; - -template<typename U, typename V> -bitcount_t flog2(const uint_x4<U,V>& v) -{ -#if PCG_LITTLE_ENDIAN - for (uint8_t i = 4; i !=0; /* dec in loop */) { - --i; -#else - for (uint8_t i = 0; i < 4; ++i) { -#endif - if (v.wa[i] == 0) - continue; - return flog2(v.wa[i]) + uint_x4<U,V>::UINT_BITS*i; - } - abort(); -} - -template<typename U, typename V> -bitcount_t trailingzeros(const uint_x4<U,V>& v) -{ -#if PCG_LITTLE_ENDIAN - for (uint8_t i = 0; i < 4; ++i) { -#else - for (uint8_t i = 4; i !=0; /* dec in loop */) { - --i; -#endif - if (v.wa[i] != 0) - return trailingzeros(v.wa[i]) + uint_x4<U,V>::UINT_BITS*i; - } - return uint_x4<U,V>::UINT_BITS*4; -} - -template <typename UInt, typename UIntX2> -std::pair< uint_x4<UInt,UIntX2>, uint_x4<UInt,UIntX2> > - divmod(const uint_x4<UInt,UIntX2>& orig_dividend, - const uint_x4<UInt,UIntX2>& divisor) -{ - // If the dividend is less than the divisor, the answer is always zero. - // This takes care of boundary cases like 0/x (which would otherwise be - // problematic because we can't take the log of zero. (The boundary case - // of division by zero is undefined.) - if (orig_dividend < divisor) - return { uint_x4<UInt,UIntX2>(UIntX2(0)), orig_dividend }; - - auto dividend = orig_dividend; - - auto log2_divisor = flog2(divisor); - auto log2_dividend = flog2(dividend); - // assert(log2_dividend >= log2_divisor); - bitcount_t logdiff = log2_dividend - log2_divisor; - - constexpr uint_x4<UInt,UIntX2> ONE(UIntX2(1)); - if (logdiff == 0) - return { ONE, dividend - divisor }; - - // Now we change the log difference to - // floor(log2(divisor)) - ceil(log2(dividend)) - // to ensure that we *underestimate* the result. - logdiff -= 1; - - uint_x4<UInt,UIntX2> quotient(UIntX2(0)); - - auto qfactor = ONE << logdiff; - auto factor = divisor << logdiff; - - do { - dividend -= factor; - quotient += qfactor; - while (dividend < factor) { - factor >>= 1; - qfactor >>= 1; - } - } while (dividend >= divisor); - - return { quotient, dividend }; -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator/(const uint_x4<UInt,UIntX2>& dividend, - const uint_x4<UInt,UIntX2>& divisor) -{ - return divmod(dividend, divisor).first; -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator%(const uint_x4<UInt,UIntX2>& dividend, - const uint_x4<UInt,UIntX2>& divisor) -{ - return divmod(dividend, divisor).second; -} - - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator*(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - constexpr auto UINT_BITS = uint_x4<UInt,UIntX2>::UINT_BITS; - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - bool carryin = false; - bool carryout; - UIntX2 a0b0 = UIntX2(a.w.v0) * UIntX2(b.w.v0); - r.w.v0 = UInt(a0b0); - r.w.v1 = UInt(a0b0 >> UINT_BITS); - - UIntX2 a1b0 = UIntX2(a.w.v1) * UIntX2(b.w.v0); - r.w.v2 = UInt(a1b0 >> UINT_BITS); - r.w.v1 = addwithcarry(r.w.v1, UInt(a1b0), carryin, &carryout); - carryin = carryout; - r.w.v2 = addwithcarry(r.w.v2, UInt(0U), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - UIntX2 a0b1 = UIntX2(a.w.v0) * UIntX2(b.w.v1); - carryin = false; - r.w.v2 = addwithcarry(r.w.v2, UInt(a0b1 >> UINT_BITS), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - carryin = false; - r.w.v1 = addwithcarry(r.w.v1, UInt(a0b1), carryin, &carryout); - carryin = carryout; - r.w.v2 = addwithcarry(r.w.v2, UInt(0U), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - UIntX2 a1b1 = UIntX2(a.w.v1) * UIntX2(b.w.v1); - carryin = false; - r.w.v2 = addwithcarry(r.w.v2, UInt(a1b1), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(a1b1 >> UINT_BITS), carryin, &carryout); - - r.d.v23 += a.d.v01 * b.d.v23 + a.d.v23 * b.d.v01; - - return r; -} - - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator*(const uint_x4<UInt,UIntX2>& a, - UIntX2 b01) -{ - constexpr auto UINT_BITS = uint_x4<UInt,UIntX2>::UINT_BITS; - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - bool carryin = false; - bool carryout; - UIntX2 a0b0 = UIntX2(a.w.v0) * UIntX2(UInt(b01)); - r.w.v0 = UInt(a0b0); - r.w.v1 = UInt(a0b0 >> UINT_BITS); - - UIntX2 a1b0 = UIntX2(a.w.v1) * UIntX2(UInt(b01)); - r.w.v2 = UInt(a1b0 >> UINT_BITS); - r.w.v1 = addwithcarry(r.w.v1, UInt(a1b0), carryin, &carryout); - carryin = carryout; - r.w.v2 = addwithcarry(r.w.v2, UInt(0U), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - UIntX2 a0b1 = UIntX2(a.w.v0) * UIntX2(b01 >> UINT_BITS); - carryin = false; - r.w.v2 = addwithcarry(r.w.v2, UInt(a0b1 >> UINT_BITS), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - carryin = false; - r.w.v1 = addwithcarry(r.w.v1, UInt(a0b1), carryin, &carryout); - carryin = carryout; - r.w.v2 = addwithcarry(r.w.v2, UInt(0U), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(0U), carryin, &carryout); - - UIntX2 a1b1 = UIntX2(a.w.v1) * UIntX2(b01 >> UINT_BITS); - carryin = false; - r.w.v2 = addwithcarry(r.w.v2, UInt(a1b1), carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(r.w.v3, UInt(a1b1 >> UINT_BITS), carryin, &carryout); - - r.d.v23 += a.d.v23 * b01; - - return r; -} - - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator+(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - - bool carryin = false; - bool carryout; - r.w.v0 = addwithcarry(a.w.v0, b.w.v0, carryin, &carryout); - carryin = carryout; - r.w.v1 = addwithcarry(a.w.v1, b.w.v1, carryin, &carryout); - carryin = carryout; - r.w.v2 = addwithcarry(a.w.v2, b.w.v2, carryin, &carryout); - carryin = carryout; - r.w.v3 = addwithcarry(a.w.v3, b.w.v3, carryin, &carryout); - - return r; -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator-(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - - bool carryin = false; - bool carryout; - r.w.v0 = subwithcarry(a.w.v0, b.w.v0, carryin, &carryout); - carryin = carryout; - r.w.v1 = subwithcarry(a.w.v1, b.w.v1, carryin, &carryout); - carryin = carryout; - r.w.v2 = subwithcarry(a.w.v2, b.w.v2, carryin, &carryout); - carryin = carryout; - r.w.v3 = subwithcarry(a.w.v3, b.w.v3, carryin, &carryout); - - return r; -} - - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator&(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - return uint_x4<UInt,UIntX2>(a.d.v23 & b.d.v23, a.d.v01 & b.d.v01); -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator|(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - return uint_x4<UInt,UIntX2>(a.d.v23 | b.d.v23, a.d.v01 | b.d.v01); -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator^(const uint_x4<UInt,UIntX2>& a, - const uint_x4<UInt,UIntX2>& b) -{ - return uint_x4<UInt,UIntX2>(a.d.v23 ^ b.d.v23, a.d.v01 ^ b.d.v01); -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator~(const uint_x4<UInt,UIntX2>& v) -{ - return uint_x4<UInt,UIntX2>(~v.d.v23, ~v.d.v01); -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator-(const uint_x4<UInt,UIntX2>& v) -{ - return uint_x4<UInt,UIntX2>(0UL,0UL) - v; -} - -template <typename UInt, typename UIntX2> -bool operator==(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return (a.d.v01 == b.d.v01) && (a.d.v23 == b.d.v23); -} - -template <typename UInt, typename UIntX2> -bool operator!=(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return !operator==(a,b); -} - - -template <typename UInt, typename UIntX2> -bool operator<(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return (a.d.v23 < b.d.v23) - || ((a.d.v23 == b.d.v23) && (a.d.v01 < b.d.v01)); -} - -template <typename UInt, typename UIntX2> -bool operator>(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return operator<(b,a); -} - -template <typename UInt, typename UIntX2> -bool operator<=(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return !(operator<(b,a)); -} - -template <typename UInt, typename UIntX2> -bool operator>=(const uint_x4<UInt,UIntX2>& a, const uint_x4<UInt,UIntX2>& b) -{ - return !(operator<(a,b)); -} - - - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator<<(const uint_x4<UInt,UIntX2>& v, - const bitcount_t shift) -{ - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - const bitcount_t bits = uint_x4<UInt,UIntX2>::UINT_BITS; - const bitcount_t bitmask = bits - 1; - const bitcount_t shiftdiv = shift / bits; - const bitcount_t shiftmod = shift & bitmask; - - if (shiftmod) { - UInt carryover = 0; -#if PCG_LITTLE_ENDIAN - for (uint8_t out = shiftdiv, in = 0; out < 4; ++out, ++in) { -#else - for (uint8_t out = 4-shiftdiv, in = 4; out != 0; /* dec in loop */) { - --out, --in; -#endif - r.wa[out] = (v.wa[in] << shiftmod) | carryover; - carryover = (v.wa[in] >> (bits - shiftmod)); - } - } else { -#if PCG_LITTLE_ENDIAN - for (uint8_t out = shiftdiv, in = 0; out < 4; ++out, ++in) { -#else - for (uint8_t out = 4-shiftdiv, in = 4; out != 0; /* dec in loop */) { - --out, --in; -#endif - r.wa[out] = v.wa[in]; - } - } - - return r; -} - -template <typename UInt, typename UIntX2> -uint_x4<UInt,UIntX2> operator>>(const uint_x4<UInt,UIntX2>& v, - const bitcount_t shift) -{ - uint_x4<UInt,UIntX2> r = {0U, 0U, 0U, 0U}; - const bitcount_t bits = uint_x4<UInt,UIntX2>::UINT_BITS; - const bitcount_t bitmask = bits - 1; - const bitcount_t shiftdiv = shift / bits; - const bitcount_t shiftmod = shift & bitmask; - - if (shiftmod) { - UInt carryover = 0; -#if PCG_LITTLE_ENDIAN - for (uint8_t out = 4-shiftdiv, in = 4; out != 0; /* dec in loop */) { - --out, --in; -#else - for (uint8_t out = shiftdiv, in = 0; out < 4; ++out, ++in) { -#endif - r.wa[out] = (v.wa[in] >> shiftmod) | carryover; - carryover = (v.wa[in] << (bits - shiftmod)); - } - } else { -#if PCG_LITTLE_ENDIAN - for (uint8_t out = 4-shiftdiv, in = 4; out != 0; /* dec in loop */) { - --out, --in; -#else - for (uint8_t out = shiftdiv, in = 0; out < 4; ++out, ++in) { -#endif - r.wa[out] = v.wa[in]; - } - } - - return r; -} - -} // namespace pcg_extras - -#endif // PCG_UINT128_HPP_INCLUDED diff --git a/benchmarks/others/sstr.h b/benchmarks/others/sstr.h index 7947870a..5301978a 100644 --- a/benchmarks/others/sstr.h +++ b/benchmarks/others/sstr.h @@ -163,19 +163,19 @@ STC_INLINE sstr_size_t sstr_capacity(sstr s) { return sstr_select_(&s, cap(&s));
}
-STC_INLINE bool sstr_equals(sstr s1, const char* str) {
+STC_INLINE bool sstr_equalto(sstr s1, const char* str) {
return strcmp(sstr_str(&s1), str) == 0;
}
-STC_INLINE bool sstr_equals_s(sstr s1, sstr s2) {
+STC_INLINE bool sstr_equalto_s(sstr s1, sstr s2) {
return strcmp(sstr_str(&s1), sstr_str(&s2)) == 0;
}
-STC_INLINE int sstr_equals_ref(const sstr* s1, const sstr* s2) {
+STC_INLINE int sstr_equals(const sstr* s1, const sstr* s2) {
return strcmp(sstr_str(s1), sstr_str(s2)) == 0;
}
-STC_INLINE int sstr_compare_ref(const sstr* s1, const sstr* s2) {
+STC_INLINE int sstr_compare(const sstr* s1, const sstr* s2) {
return strcmp(sstr_str(s1), sstr_str(s2));
}
diff --git a/docs/ccommon_api.md b/docs/ccommon_api.md index ddf3783c..0eb11f82 100644 --- a/docs/ccommon_api.md +++ b/docs/ccommon_api.md @@ -2,25 +2,53 @@ The following handy macros are safe to use, i.e. have no side-effects. -### c_fordefer, c_forscope, c_forvar, c_forvar_initdel +### c_forvar, c_forauto, c_forscope, c_fordefer General ***defer*** mechanics for resource acquisition. These macros allows to specify the release of the resource where the resource acquisition takes place. Makes it easier to verify that resources are released. -**Note**: These macros are one-time executed **for-loops**. Use ***only*** `continue` in order to break out -of a `c_for*`-block. ***Do not*** use `return`, `goto` or `break`, as they will prevent the `end`-statement to -be executed when leaving scope. This is not particular to the `c_for*()` macros, as one must always make sure -to unwind temporary allocated resources before a `return` in C. +**NB**: These macros are one-time executed **for-loops**. Use ***only*** `continue` in order to break out +of these four `c_for`-blocks. ***Do not*** use `return` or `goto` (or `break`) inside them, as they will +prevent the `end`-statement to be executed when leaving scope. This is not particular to the `c_for*()` +macros, as one must always make sure to unwind temporary allocated resources before a `return` in C. | Usage | Description | |:---------------------------------------|:--------------------------------------------------| -| `c_fordefer (end...)` | Defer execution of `end` to end of block | -| `c_forscope (start, end)` | Execute `start`. Defer `end` to end of block | | `c_forvar (Type var=init, end...)` | Declare `var`. Defer `end` to end of block | -| `c_forvar_initdel (Type, var...)` | `c_forvar (Type var=Type_init(), Type_del(&var))` | +| `c_forauto (Type, var...)` | `c_forvar (Type var=Type_init(), Type_del(&var))` | +| `c_forscope (start, end)` | Execute `start`. Defer `end` to end of block | +| `c_fordefer (end...)` | Defer execution of `end` to end of block | -For multiple variables, use either multiple **c_forvar** in sequence, or declare variable outside -scope and use **c_fordefer** or **c_forscope**. Also, **c_forvar_initdel** support up to 3 vars. +For multiple variables, use either multiple **c_forvar** in sequence, or declare variable outside +scope and use **c_forscope** or **c_fordefer**. Also, **c_forauto** support up to 3 vars. ```c +c_forvar (cstr s = cstr_lit("Hello"), cstr_del(&s)) +{ + cstr_append(&s, " world"); + printf("%s\n", s.str); +} + +c_forauto (cstr, s1, s2) +{ + cstr_append(&s1, "Hello"); + cstr_append(&s1, " world"); + + cstr_append(&s2, "Cool"); + cstr_append(&s2, " stuff"); + + printf("%s %s\n", s1.str, s2.str); +} + +using_cvec(i32, int32_t); +... +cvec_i32 vec; +c_forscope (vec = cvec_i32_init(), cvec_i32_del(&vec)) +{ + cvec_i32_push_back(&vec, 123); + cvec_i32_push_back(&vec, 321); + + c_foreach (i, cvec_i32, vec) printf(" %d", *i.ref); +} + cstr s1 = cstr_lit("Hello"), s2 = cstr_lit("world"); c_fordefer (cstr_del(&s1), cstr_del(&s2)) { @@ -66,10 +94,12 @@ int main() using_csset(x, int); ... c_var (csset_x, set, {23, 3, 7, 5, 12}); -c_foreach (i, csset_x, set) printf(" %d", *i.ref); +c_foreach (i, csset_x, set) + printf(" %d", *i.ref); // 3 5 7 12 23 csset_x_iter_t it = csset_x_find(&set, 7); -c_foreach (i, csset_x, it, csset_x_end(&set)) printf(" %d", *i.ref); +c_foreach (i, csset_x, it, csset_x_end(&set)) + printf(" %d", *i.ref); // 7 12 23 ``` diff --git a/docs/cdeq_api.md b/docs/cdeq_api.md index 42e46240..83c0f19f 100644 --- a/docs/cdeq_api.md +++ b/docs/cdeq_api.md @@ -23,7 +23,7 @@ be replaced by `i` in all of the following documentation. `using_cdeq_str()` is a shorthand for: ``` -using_cdeq(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*) +using_cdeq(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*) ``` ## Methods diff --git a/docs/clist_api.md b/docs/clist_api.md index 2028e764..b57662ca 100644 --- a/docs/clist_api.md +++ b/docs/clist_api.md @@ -35,7 +35,7 @@ The macro `using_clist()` must be instantiated in the global scope. `X` is a typ will affect the names of all clist types and methods. E.g. declaring `using_clist(i, int);`, `X` should be replaced by `i` in all of the following documentation. `using_clist_str()` is a shorthand for ```c -using_clist(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*) +using_clist(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*) ``` ## Methods diff --git a/docs/cmap_api.md b/docs/cmap_api.md index ea9efb39..59afa5ab 100644 --- a/docs/cmap_api.md +++ b/docs/cmap_api.md @@ -41,7 +41,7 @@ using_cmap_str() // using_cmap(str, cstr, cstr, . ``` The `using_cmap()` macro family must be instantiated in the global scope. `X` is a type tag name and will affect the names of all cmap types and methods. E.g. declaring `using_cmap(ii, int, int);`, `X` should -be replaced by `ii` in all of the following documentation. +be replaced by `ii` in all of the following documentation. Argument `flag` is `c_true` by default. ## Methods @@ -275,7 +275,7 @@ static int Viking_equals(const Viking* a, const Viking* b) { } static uint32_t Viking_hash(const Viking* a, int ignored) { - return cstr_hash(a->name) ^ (cstr_hash(a->country) >> 15); + return cstr_hash(&a->name) ^ (cstr_hash(&a->country) >> 15); } static void Viking_del(Viking* v) { diff --git a/docs/cset_api.md b/docs/cset_api.md index 2c0f4611..dc628750 100644 --- a/docs/cset_api.md +++ b/docs/cset_api.md @@ -12,9 +12,9 @@ A **cset** is an associative container that contains a set of unique objects of using_cset(X, Key);
using_cset(X, Key, keyEquals, keyHash);
using_cset(X, Key, keyEquals, keyHash, keyDel, keyClone = c_no_clone);
-using_cset(X, Key, keyEqualsRaw, keyHashRaw, keyDel, keyFromRaw, keyToRaw, RawKey);
+using_cset(X, Key, keyEqualsRaw, keyHashRaw, keyDel, keyFromRaw, keyToRaw, RawKey, flag);
-using_cset_str(); // using_cset(str, cstr, ...)
+using_cset_str(); // using_cset(str, cstr, c_rawstr_equals, c_rawstr_hash, cstr_del, ...)
```
The macro `using_cset()` must be instantiated in the global scope. `X` is a type tag name and
will affect the names of all cset types and methods. E.g. declaring `using_cset(i, int);`, `X` should
diff --git a/docs/csmap_api.md b/docs/csmap_api.md index 8e152497..dba4fecd 100644 --- a/docs/csmap_api.md +++ b/docs/csmap_api.md @@ -19,27 +19,27 @@ See the c++ class [std::map](https://en.cppreference.com/w/cpp/container/map) fo using_csmap(X, Key, Mapped); using_csmap(X, Key, Mapped, keyCompare); -using_csmap(X, Key, Mapped, keyCompare, mappedDel, mappedClone); +using_csmap(X, Key, Mapped, keyCompare, mappedDel, mappedClone = c_no_clone); using_csmap(X, Key, Mapped, keyCompare, mappedDel, mappedFromRaw, mappedToRaw, RawMapped); using_csmap(X, Key, Mapped, keyCompareRaw, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, - keyDel, keyFromRaw, keyToRaw, RawKey); + keyDel, keyFromRaw, keyToRaw, RawKey, flag); using_csmap_keydef(X, Key, Mapped, keyCompare, keyDel, keyClone); -using_csmap_keydef(X, Key, Mapped, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey); +using_csmap_keydef(X, Key, Mapped, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey, flag); using_csmap_strkey(X, Mapped); // using_csmap(X, cstr, Mapped, ...) using_csmap_strkey(X, Mapped, mappedDel, mappedClone); -using_csmap_strkey(X, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped); +using_csmap_strkey(X, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, flag); using_csmap_strval(X, Key); // using_csmap(X, Key, cstr, ...) using_csmap_strval(X, Key, keyCompare); using_csmap_strval(X, Key, keyCompare, keyDel, keyClone); -using_csmap_strval(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey); +using_csmap_strval(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey, flag); using_csmap_str(); // using_csmap(str, cstr, cstr, ...) ``` The `using_csmap()` macro family must be instantiated in the global scope. `X` is a type tag name and will affect the names of all csmap types and methods. E.g. declaring `using_csmap(ii, int, int);`, `X` should -be replaced by `ii` in all of the following documentation. +be replaced by `ii` in all of the following documentation. Argument `flag` is `c_true` by default. ## Methods diff --git a/docs/csset_api.md b/docs/csset_api.md index 38e9072e..d308fefd 100644 --- a/docs/csset_api.md +++ b/docs/csset_api.md @@ -13,9 +13,9 @@ See the c++ class [std::set](https://en.cppreference.com/w/cpp/container/set) fo using_csset(X, Key);
using_csset(X, Key, keyCompare);
using_csset(X, Key, keyCompare, keyDel, keyClone = c_no_clone);
-using_csset(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey);
+using_csset(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey, flag);
-using_csset_str(); // using_csset(str, cstr, ...)
+using_csset_str(); // using_csset(str, cstr, cstr_del, cstr_from, cstr_str, const char*, c_true)
```
The macro `using_csset()` must be instantiated in the global scope. `X` is a type tag name and
will affect the names of all csset types and methods. E.g. declaring `using_csset(i, int);`, `X` should
diff --git a/docs/cstr_api.md b/docs/cstr_api.md index adf57209..fd2b241f 100644 --- a/docs/cstr_api.md +++ b/docs/cstr_api.md @@ -28,6 +28,8 @@ cstr* cstr_take(cstr* self, cstr s); // take th cstr cstr_move(cstr* self); // move string to caller, leave empty string void cstr_del(cstr *self); // destructor +const char* cstr_str(const cstr* self); // self->str +char* cstr_data(cstr* self); // self->str size_t cstr_size(cstr s); size_t cstr_length(cstr s); size_t cstr_capacity(cstr s); @@ -58,15 +60,15 @@ void cstr_replace_all(cstr* self, const char* find, const char* replace) void cstr_erase(cstr* self, size_t pos); void cstr_erase_n(cstr* self, size_t pos, size_t n); -bool cstr_equals(cstr s, const char* str); -bool cstr_equals_s(cstr s, cstr s2); +bool cstr_equalto(cstr s, const char* str); +bool cstr_equalto_s(cstr s, cstr s2); size_t cstr_find(cstr s, const char* needle); size_t cstr_find_n(cstr s, const char* needle, size_t pos, size_t nmax); bool cstr_contains(cstr s, const char* needle); bool cstr_starts_with(cstr s, const char* str); bool cstr_ends_with(cstr s, const char* str); -bool cstr_iequals(cstr s, const char* str); // prefix i = case-insensitive +bool cstr_iequalto(cstr s, const char* str); // prefix i = case-insensitive size_t cstr_ifind_n(cstr s, const char* needle, size_t pos, size_t nmax); bool cstr_icontains(cstr s, const char* needle); bool cstr_istarts_with(cstr s, const char* str); @@ -89,14 +91,13 @@ Note that all methods with arguments `(..., const char* str, size_t n)`, `n` mus #### Helper methods: ```c -const char* cstr_toraw(const cstr* self); - -int cstr_compare_ref(const cstr *s1, const cstr *s2); -bool cstr_equals_ref(const cstr *s1, const cstr *s2); +int cstr_compare(const cstr *s1, const cstr *s2); +bool cstr_equals(const cstr *s1, const cstr *s2); +bool cstr_hash(const cstr *s, ...); int c_rawstr_compare(const char** x, const char** y); bool c_rawstr_equals(const char** x, const char** y); -uint64_t c_rawstr_hash(const char* const* x, size_t ignored); +uint64_t c_rawstr_hash(const char* const* x, ...); int c_strncasecmp(const char* str1, const char* str2, size_t n); char* c_strnstrn(const char* str, const char* needle, size_t slen, size_t nlen); diff --git a/docs/csview_api.md b/docs/csview_api.md index 2906acdc..dfca37c5 100644 --- a/docs/csview_api.md +++ b/docs/csview_api.md @@ -45,7 +45,7 @@ csview csview_slice(csview sv, intptr_t p1, intptr_t p2); // negative p csview csview_first_token(csview sv, csview sep); // see split example below. csview csview_next_token(csview sv, csview sep, csview token); -bool csview_equals(csview sv, csview sv2); +bool csview_equalto(csview sv, csview sv2); size_t csview_find(csview sv, csview needle); bool csview_contains(csview sv, csview needle); bool csview_starts_with(csview sv, csview sub); @@ -78,9 +78,9 @@ bool cstr_ends_with_v(cstr s, csview sub); ``` #### Helper methods ```c -int csview_compare_ref(const csview* x, const csview* y); -bool csview_equals_ref(const csview* x, const csview* y); -uint64_t csview_hash_ref(const csview* x, size_t ignored); +int csview_compare(const csview* x, const csview* y); +bool csview_equals(const csview* x, const csview* y); +uint64_t csview_hash(const csview* x, ...); ``` ## Types @@ -220,7 +220,7 @@ int main() csview suffix = csview_substr(text, -12, cstr_npos); // from pos -12 to end printf("%.*s\n", csview_ARG(suffix)); - c_forvar_initdel (cmap_si, map) + c_forauto (cmap_si, map) { cmap_si_emplace(&map, c_sv("hello"), 100); cmap_si_emplace(&map, c_sv("world"), 200); diff --git a/docs/cvec_api.md b/docs/cvec_api.md index 016d82e3..5a98b9ca 100644 --- a/docs/cvec_api.md +++ b/docs/cvec_api.md @@ -27,7 +27,7 @@ be replaced by `i` in all of the following documentation. `using_cvec_str()` is a shorthand for: ``` -using_cvec(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*) +using_cvec(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*) ``` ## Methods diff --git a/examples/csmap_erase.c b/examples/csmap_erase.c index 35afbe3a..6cebffb7 100644 --- a/examples/csmap_erase.c +++ b/examples/csmap_erase.c @@ -1,77 +1,77 @@ -// map_erase.c
-// https://docs.microsoft.com/en-us/cpp/standard-library/map-class?view=msvc-160#example-16
-#include <stc/csmap.h>
-#include <stc/cstr.h>
-#include <stdio.h>
-
-using_csmap_strval(my, int);
-
-void printmap(csmap_my map)
-{
- c_foreach (e, csmap_my, map)
- printf(" [%d, %s]", e.ref->first, e.ref->second.str);
- printf("\nsize() == %zu\n\n", csmap_my_size(map));
-}
-
-int main()
-{
- c_forvar (csmap_my m1 = csmap_my_init(), csmap_my_del(&m1))
- {
- // Fill in some data to test with, one at a time
- csmap_my_emplace(&m1, 1, "A");
- csmap_my_emplace(&m1, 2, "B");
- csmap_my_emplace(&m1, 3, "C");
- csmap_my_emplace(&m1, 4, "D");
- csmap_my_emplace(&m1, 5, "E");
-
- puts("Starting data of map m1 is:");
- printmap(m1);
- // The 1st member function removes an element at a given position
- csmap_my_erase_at(&m1, csmap_my_fwd(csmap_my_begin(&m1), 1));
- puts("After the 2nd element is deleted, the map m1 is:");
- printmap(m1);
- }
-
- c_forvar (csmap_my m2 = csmap_my_init(), csmap_my_del(&m2))
- {
- // Fill in some data to test with, one at a time, using emplace
- c_emplace(csmap_my, m2, {
- {10, "Bob"},
- {11, "Rob"},
- {12, "Robert"},
- {13, "Bert"},
- {14, "Bobby"}
- });
-
- puts("Starting data of map m2 is:");
- printmap(m2);
- csmap_my_iter_t it1 = csmap_my_fwd(csmap_my_begin(&m2), 1);
- csmap_my_iter_t it2 = csmap_my_find(&m2, csmap_my_back(&m2)->first);
- // The 2nd member function removes elements
- // in the range [First, Last)
- csmap_my_erase_range(&m2, it1, it2);
- puts("After the middle elements are deleted, the map m2 is:");
- printmap(m2);
- }
-
- c_forvar (csmap_my m3 = csmap_my_init(), csmap_my_del(&m3))
- {
- // Fill in some data to test with, one at a time, using emplace
- csmap_my_emplace(&m3, 1, "red");
- csmap_my_emplace(&m3, 2, "yellow");
- csmap_my_emplace(&m3, 3, "blue");
- csmap_my_emplace(&m3, 4, "green");
- csmap_my_emplace(&m3, 5, "orange");
- csmap_my_emplace(&m3, 5, "purple");
- csmap_my_emplace(&m3, 5, "pink");
-
- puts("Starting data of map m3 is:");
- printmap(m3);
- // The 3rd member function removes elements with a given Key
- size_t count = csmap_my_erase(&m3, 2);
- // The 3rd member function also returns the number of elements removed
- printf("The number of elements removed from m3 is: %zu\n", count);
- puts("After the element with a key of 2 is deleted, the map m3 is:");
- printmap(m3);
- }
-}
\ No newline at end of file +// map_erase.c +// From C++ example: https://docs.microsoft.com/en-us/cpp/standard-library/map-class?view=msvc-160#example-16 +#include <stc/csmap.h> +#include <stc/cstr.h> +#include <stdio.h> + +using_csmap_strval(my, int); + +void printmap(csmap_my map) +{ + c_foreach (e, csmap_my, map) + printf(" [%d, %s]", e.ref->first, e.ref->second.str); + printf("\nsize() == %zu\n\n", csmap_my_size(map)); +} + +int main() +{ + c_forauto (csmap_my, m1) + { + // Fill in some data to test with, one at a time + csmap_my_insert(&m1, 1, cstr_lit("A")); + csmap_my_insert(&m1, 2, cstr_lit("B")); + csmap_my_insert(&m1, 3, cstr_lit("C")); + csmap_my_insert(&m1, 4, cstr_lit("D")); + csmap_my_insert(&m1, 5, cstr_lit("E")); + + puts("Starting data of map m1 is:"); + printmap(m1); + // The 1st member function removes an element at a given position + csmap_my_erase_at(&m1, csmap_my_fwd(csmap_my_begin(&m1), 1)); + puts("After the 2nd element is deleted, the map m1 is:"); + printmap(m1); + } + + c_forauto (csmap_my, m2) + { + // Fill in some data to test with, one at a time, using c_emplace() + c_emplace(csmap_my, m2, { + {10, "Bob"}, + {11, "Rob"}, + {12, "Robert"}, + {13, "Bert"}, + {14, "Bobby"} + }); + + puts("Starting data of map m2 is:"); + printmap(m2); + csmap_my_iter_t it1 = csmap_my_fwd(csmap_my_begin(&m2), 1); + csmap_my_iter_t it2 = csmap_my_find(&m2, csmap_my_back(&m2)->first); + // The 2nd member function removes elements + // in the range [First, Last) + csmap_my_erase_range(&m2, it1, it2); + puts("After the middle elements are deleted, the map m2 is:"); + printmap(m2); + } + + c_forauto (csmap_my, m3) + { + // Fill in some data to test with, one at a time, using emplace + csmap_my_emplace(&m3, 1, "red"); + csmap_my_emplace(&m3, 2, "yellow"); + csmap_my_emplace(&m3, 3, "blue"); + csmap_my_emplace(&m3, 4, "green"); + csmap_my_emplace(&m3, 5, "orange"); + csmap_my_emplace(&m3, 6, "purple"); + csmap_my_emplace(&m3, 7, "pink"); + + puts("Starting data of map m3 is:"); + printmap(m3); + // The 3rd member function removes elements with a given Key + size_t count = csmap_my_erase(&m3, 2); + // The 3rd member function also returns the number of elements removed + printf("The number of elements removed from m3 is: %zu\n", count); + puts("After the element with a key of 2 is deleted, the map m3 is:"); + printmap(m3); + } +} diff --git a/examples/csmap_find.c b/examples/csmap_find.c index 314e702e..46ae01fa 100644 --- a/examples/csmap_find.c +++ b/examples/csmap_find.c @@ -1,65 +1,68 @@ -// This implements the c++ std::map::find example at:
-// https://docs.microsoft.com/en-us/cpp/standard-library/map-class?view=msvc-160#example-17
-#include <stc/csmap.h>
-#include <stc/cvec.h>
-#include <stc/cstr.h>
-#include <stdio.h>
-
-using_csmap_strval(istr, int);
-using_cvec(istr, csmap_istr_rawvalue_t, c_no_compare);
-
-void print_elem(csmap_istr_rawvalue_t p) {
- printf("(%d, %s) ", p.first, p.second);
-}
-
-#define using_print_collection(CX) \
- void print_collection_##CX(CX t) { \
- printf("%zu elements: ", CX##_size(t)); \
- \
- c_foreach (p, CX, t) { \
- print_elem(CX##_value_toraw(p.ref)); \
- } \
- puts(""); \
- }
-
-using_print_collection(csmap_istr);
-using_print_collection(cvec_istr);
-
-
-void findit(csmap_istr c, csmap_istr_key_t val) {
- printf("Trying find() on value %d\n", val);
- csmap_istr_value_t* result = csmap_istr_get(&c, val); // easier with get than find.
- if (result) {
- printf("Element found: "); print_elem(csmap_istr_value_toraw(result)); puts("");
- } else {
- puts("Element not found.");
- }
-}
-
-int main()
-{
- c_forvar (csmap_istr m1 = csmap_istr_init(), csmap_istr_del(&m1))
- c_forvar (cvec_istr v = cvec_istr_init(), cvec_istr_del(&v)) {
- c_emplace(csmap_istr, m1, { { 40, "Zr" }, { 45, "Rh" } });
- puts("The starting map m1 is (key, value):");
- print_collection_csmap_istr(m1);
-
- typedef cvec_istr_value_t pair;
- cvec_istr_emplace_back(&v, (pair){43, "Tc"});
- cvec_istr_emplace_back(&v, (pair){41, "Nb"});
- cvec_istr_emplace_back(&v, (pair){46, "Pd"});
- cvec_istr_emplace_back(&v, (pair){42, "Mo"});
- cvec_istr_emplace_back(&v, (pair){44, "Ru"});
- cvec_istr_emplace_back(&v, (pair){44, "Ru"}); // attempt a duplicate
-
- puts("Inserting the following vector data into m1:");
- print_collection_cvec_istr(v);
- c_foreach (i, cvec_istr, v) csmap_istr_emplace(&m1, i.ref->first, i.ref->second);
-
- puts("The modified map m1 is (key, value):");
- print_collection_csmap_istr(m1);
- puts("");
- findit(m1, 45);
- findit(m1, 6);
- }
-}
\ No newline at end of file +// This implements the c++ std::map::find example at: +// https://docs.microsoft.com/en-us/cpp/standard-library/map-class?view=msvc-160#example-17 +#include <stc/csmap.h> +#include <stc/cvec.h> +#include <stc/cstr.h> +#include <stdio.h> + +using_csmap_strval(istr, int); +using_cvec(istr, csmap_istr_rawvalue_t, c_no_compare); + +void print_elem(csmap_istr_rawvalue_t p) { + printf("(%d, %s) ", p.first, p.second); +} + +#define using_print_collection(CX) \ + void print_collection_##CX(CX t) { \ + printf("%zu elements: ", CX##_size(t)); \ + \ + c_foreach (p, CX, t) { \ + print_elem(CX##_value_toraw(p.ref)); \ + } \ + puts(""); \ + } + +using_print_collection(csmap_istr); +using_print_collection(cvec_istr); + + +void findit(csmap_istr c, csmap_istr_key_t val) +{ + printf("Trying find() on value %d\n", val); + csmap_istr_value_t* result = csmap_istr_get(&c, val); // easier with get than find. + if (result) { + printf("Element found: "); print_elem(csmap_istr_value_toraw(result)); puts(""); + } else { + puts("Element not found."); + } +} + +int main() +{ + c_forauto (csmap_istr, m1) + c_forauto (cvec_istr, v) + { + c_emplace(csmap_istr, m1, { { 40, "Zr" }, { 45, "Rh" } }); + puts("The starting map m1 is (key, value):"); + print_collection_csmap_istr(m1); + + typedef cvec_istr_value_t pair; + cvec_istr_emplace_back(&v, (pair){43, "Tc"}); + cvec_istr_emplace_back(&v, (pair){41, "Nb"}); + cvec_istr_emplace_back(&v, (pair){46, "Pd"}); + cvec_istr_emplace_back(&v, (pair){42, "Mo"}); + cvec_istr_emplace_back(&v, (pair){44, "Ru"}); + cvec_istr_emplace_back(&v, (pair){44, "Ru"}); // attempt a duplicate + + puts("Inserting the following vector data into m1:"); + print_collection_cvec_istr(v); + + c_foreach (i, cvec_istr, v) csmap_istr_emplace(&m1, i.ref->first, i.ref->second); + + puts("The modified map m1 is (key, value):"); + print_collection_csmap_istr(m1); + puts(""); + findit(m1, 45); + findit(m1, 6); + } +} diff --git a/examples/cstr_match.c b/examples/cstr_match.c index 72bdd2d2..a4dbe8d8 100644 --- a/examples/cstr_match.c +++ b/examples/cstr_match.c @@ -6,7 +6,7 @@ int main() c_forvar (cstr ss = cstr_lit("The quick brown fox jumps over the lazy dog.JPG"), cstr_del(&ss)) {
size_t pos = cstr_find_n(ss, "brown", 0, 5);
printf("%zu [%s]\n", pos, pos == cstr_npos ? "<NULL>" : &ss.str[pos]);
- printf("iequals: %d\n", cstr_iequals(ss, "the quick brown fox jumps over the lazy dog.jpg"));
+ printf("iequals: %d\n", cstr_iequalto(ss, "the quick brown fox jumps over the lazy dog.jpg"));
printf("icontains: %d\n", cstr_icontains(ss, "uMPS Ove"));
printf("istarts_with: %d\n", cstr_istarts_with(ss, "The Quick Brown"));
printf("iends_with: %d\n", cstr_iends_with(ss, ".Jpg"));
diff --git a/examples/ex_gauss1.c b/examples/ex_gauss1.c index 9eb5982d..d089de89 100644 --- a/examples/ex_gauss1.c +++ b/examples/ex_gauss1.c @@ -6,8 +6,8 @@ #include <stc/cmap.h>
#include <stc/cvec.h>
-// Declare int -> int hashmap. Uses typetag 'i' for ints.
-using_cmap(i, int, size_t);
+// Declare int -> int hashmap. Uses typetag 'ii' for ints.
+using_cmap(ii, int32_t, size_t);
// Declare int vector with map entries that can be sorted by map keys.
typedef struct {int first; size_t second;} mapval;
@@ -15,7 +15,7 @@ static int compare(mapval *a, mapval *b) { return c_default_compare(&a->first, &b->first);
}
-using_cvec(e, mapval, compare);
+using_cvec(pair, mapval, compare);
int main()
{
@@ -30,23 +30,23 @@ int main() stc64_normalf_t dist = stc64_normalf_init(Mean, StdDev);
// Create and init histogram vec and map with defered destructors:
- c_forvar (cvec_e vhist = cvec_e_init(), cvec_e_del(&vhist))
- c_forvar (cmap_i mhist = cmap_i_init(), cmap_i_del(&mhist))
+ c_forauto (cvec_pair, histvec)
+ c_forauto (cmap_ii, histmap)
{
c_forrange (N) {
int index = (int) round( stc64_normalf(&rng, &dist) );
- cmap_i_emplace(&mhist, index, 0).ref->second += 1;
+ cmap_ii_emplace(&histmap, index, 0).ref->second += 1;
}
// Transfer map to vec and sort it by map keys.
- c_foreach (i, cmap_i, mhist)
- cvec_e_push_back(&vhist, c_make(mapval){i.ref->first, i.ref->second});
+ c_foreach (i, cmap_ii, histmap)
+ cvec_pair_push_back(&histvec, (mapval){i.ref->first, i.ref->second});
- cvec_e_sort(&vhist);
+ cvec_pair_sort(&histvec);
// Print the gaussian bar chart
- c_forvar (cstr bar = cstr_null, cstr_del(&bar))
- c_foreach (i, cvec_e, vhist) {
+ c_forauto (cstr, bar)
+ c_foreach (i, cvec_pair, histvec) {
size_t n = (size_t) (i.ref->second * StdDev * Scale * 2.5 / (float)N);
if (n > 0) {
cstr_resize(&bar, n, '*');
diff --git a/examples/mmap.c b/examples/mmap.c index 04e501aa..9c92b19b 100644 --- a/examples/mmap.c +++ b/examples/mmap.c @@ -7,26 +7,26 @@ // Map of int => clist_str. Note the negation of c_default_compare
using_clist_str();
-using_csmap(m, int, clist_str, -c_default_compare, clist_str_del);
+using_csmap(mult, int, clist_str, -c_default_compare, clist_str_del);
-void print(const csmap_m mmap)
+void print(const csmap_mult mmap)
{
- c_foreach (e, csmap_m, mmap) {
+ c_foreach (e, csmap_mult, mmap) {
c_foreach (s, clist_str, e.ref->second)
printf("{%d,%s} ", e.ref->first, s.ref->str);
}
puts("");
}
-void insert(csmap_m* mmap, int key, const char* str)
+void insert(csmap_mult* mmap, int key, const char* str)
{
- clist_str *list = &csmap_m_insert(mmap, key, clist_str_init()).ref->second;
+ clist_str *list = &csmap_mult_insert(mmap, key, clist_str_init()).ref->second;
clist_str_emplace_back(list, str);
}
int main()
{
- c_forvar (csmap_m mmap = csmap_m_init(), csmap_m_del(&mmap))
+ c_forauto (csmap_mult, mmap)
{
// list-initialize
struct {int i; const char* s;} vals[] = {{2, "foo"}, {2, "bar"}, {3, "baz"}, {1, "abc"}, {5, "def"}};
@@ -49,13 +49,13 @@ int main() print(mmap);
// erase all entries with key 5
- csmap_m_erase(&mmap, 5);
+ csmap_mult_erase(&mmap, 5);
print(mmap);
// find and erase a specific entry
clist_str_iter_t pos;
- c_foreach (e, csmap_m, mmap)
- if ((pos = clist_str_find(&e.ref->second, "bar")).ref) {
+ c_foreach (e, csmap_mult, mmap)
+ if ((pos = clist_str_find(&e.ref->second, "bar")).ref != clist_str_end(&e.ref->second).ref) {
clist_str_erase(&e.ref->second, pos);
break;
}
diff --git a/examples/priority.c b/examples/priority.c index ee64bf42..4824a294 100644 --- a/examples/priority.c +++ b/examples/priority.c @@ -13,7 +13,7 @@ int main() { size_t N = 10000000;
stc64_t rng = stc64_init(time(NULL));
stc64_uniform_t dist = stc64_uniform_init(0, N * 10);
- c_forvar (cpque_i heap = cpque_i_init(), cpque_i_del(&heap))
+ c_forauto (cpque_i, heap)
{
// Push ten million random numbers to priority queue
c_forrange (N)
diff --git a/include/stc/ccommon.h b/include/stc/ccommon.h index cdfb1e61..e901f499 100644 --- a/include/stc/ccommon.h +++ b/include/stc/ccommon.h @@ -53,9 +53,11 @@ /* Macro overloading feature support based on: https://rextester.com/ONP80107 */
#define c_MACRO_OVERLOAD(name, ...) \
- c_SELECT(name, c_NUM_ARGS(__VA_ARGS__))(__VA_ARGS__)
-#define c_SELECT(name, num) c_CONCAT(name ## _, num)
+ c_PASTE3(name, _, c_NUM_ARGS(__VA_ARGS__))(__VA_ARGS__)
#define c_CONCAT(a, b) a ## b
+#define c_CONCAT3(a, b, c) a ## b ## c
+#define c_PASTE(a, b) c_CONCAT(a, b)
+#define c_PASTE3(a, b, c) c_CONCAT3(a, b, c)
#define c_EXPAND(...) __VA_ARGS__
#define c_NUM_ARGS(...) _c_APPLY_ARG_N((__VA_ARGS__, _c_RSEQ_N))
@@ -99,7 +101,7 @@ #define c_rawstr_compare(x, y) strcmp(*(x), *(y))
#define c_rawstr_equals(x, y) (strcmp(*(x), *(y)) == 0)
-#define c_rawstr_hash(p, none) c_default_hash(*(p), strlen(*(p)))
+#define c_rawstr_hash(p, ...) c_default_hash(*(p), strlen(*(p)))
#define c_no_clone(x) (assert(!"c_no_clone() called"), x)
#define c_default_fromraw(x) (x)
@@ -133,13 +135,13 @@ #define c_forscope(start, end) for (int _c_ii = (start, 0); !_c_ii; ++_c_ii, end)
#define c_fordefer(...) for (int _c_ii = 0; !_c_ii; ++_c_ii, __VA_ARGS__)
-#define c_forvar_initdel(...) c_MACRO_OVERLOAD(c_forvar_initdel, __VA_ARGS__)
-#define c_forvar_initdel_2(CX, a) \
+#define c_forauto(...) c_MACRO_OVERLOAD(c_forauto, __VA_ARGS__)
+#define c_forauto_2(CX, a) \
c_forvar(CX a = CX##_init(), CX##_del(&a))
-#define c_forvar_initdel_3(CX, a, b) \
+#define c_forauto_3(CX, a, b) \
c_forvar(c_EXPAND(CX a = CX##_init(), b = CX##_init()), \
CX##_del(&b), CX##_del(&a))
-#define c_forvar_initdel_4(CX, a, b, c) \
+#define c_forauto_4(CX, a, b, c) \
c_forvar(c_EXPAND(CX a = CX##_init(), b = CX##_init(), c = CX##_init()), \
CX##_del(&c), CX##_del(&b), CX##_del(&a))
diff --git a/include/stc/cdeq.h b/include/stc/cdeq.h index 49de8662..d052e9eb 100644 --- a/include/stc/cdeq.h +++ b/include/stc/cdeq.h @@ -45,7 +45,7 @@ _c_using_cdeq(cdeq_##X, Value, valueCompareRaw, valueDel, valueFromRaw, valueToRaw, RawValue, defTypes)
#define using_cdeq_str() \
- using_cdeq_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*)
+ using_cdeq_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*)
struct cdeq_rep { size_t size, cap; void* base[]; };
diff --git a/include/stc/clist.h b/include/stc/clist.h index 4166fffe..75364a4f 100644 --- a/include/stc/clist.h +++ b/include/stc/clist.h @@ -74,7 +74,7 @@ _c_using_clist(clist_##X, Value, valueCompareRaw, valueDel, valueFromRaw, valueToRaw, RawValue, defTypes)
#define using_clist_str() \
- using_clist_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*)
+ using_clist_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*)
#define _c_clist_types(CX, Value) \
diff --git a/include/stc/cmap.h b/include/stc/cmap.h index 89975374..476558c7 100644 --- a/include/stc/cmap.h +++ b/include/stc/cmap.h @@ -67,14 +67,14 @@ int main(void) { #define using_cmap_10(X, Key, Mapped, keyEquals, keyHash, \
mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes) \
using_cmap_14(X, Key, Mapped, keyEquals, keyHash, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- c_default_del, c_default_fromraw, c_default_toraw, Key)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ c_default_del, c_default_fromraw, c_default_toraw, Key, defTypes)
#define using_cmap_14(X, Key, Mapped, keyEqualsRaw, keyHashRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey) \
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_chash(cmap_##X, cmap_, Key, Mapped, keyEqualsRaw, keyHashRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define using_cmap_keydef(...) c_MACRO_OVERLOAD(using_cmap_keydef, __VA_ARGS__)
@@ -85,13 +85,13 @@ int main(void) { #define using_cmap_keydef_10(X, Key, Mapped, keyEqualsRaw, keyHashRaw, \
keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_chash(cmap_##X, cmap_, Key, Mapped, keyEqualsRaw, keyHashRaw, \
- c_default_del, c_default_fromraw, c_default_toraw, Mapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ c_default_del, c_default_fromraw, c_default_toraw, Mapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define using_cmap_str() \
_c_using_chash(cmap_str, cmap_, cstr, cstr, c_rawstr_equals, c_rawstr_hash, \
- cstr_del, cstr_from, cstr_toraw, const char*, c_true, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ cstr_del, cstr_from, cstr_str, const char*, \
+ cstr_del, cstr_from, cstr_str, const char*, c_true)
#define using_cmap_strkey(...) c_MACRO_OVERLOAD(using_cmap_strkey, __VA_ARGS__)
@@ -106,8 +106,8 @@ int main(void) { _c_using_chash_strkey(X, cmap_, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes)
#define _c_using_chash_strkey(X, C, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes) \
_c_using_chash(C##X, C, cstr, Mapped, c_rawstr_equals, c_rawstr_hash, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ cstr_del, cstr_from, cstr_str, const char*, defTypes)
#define using_cmap_strval(...) c_MACRO_OVERLOAD(using_cmap_strval, __VA_ARGS__)
@@ -122,8 +122,8 @@ int main(void) { using_cmap_strval_9(X, Key, keyEquals, keyHash, keyDel, keyClone, c_default_toraw, Key, c_true)
#define using_cmap_strval_9(X, Key, keyEqualsRaw, keyHashRaw, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_chash(cmap_##X, cmap_, Key, cstr, keyEqualsRaw, keyHashRaw, \
- cstr_del, cstr_from, cstr_toraw, const char*, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ cstr_del, cstr_from, cstr_str, const char*, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define SET_ONLY_cmap_(...)
#define MAP_ONLY_cmap_(...) __VA_ARGS__
@@ -169,8 +169,8 @@ STC_INLINE uint64_t c_default_hash64(const void* data, size_t ignored) } CX
#define _c_using_chash(CX, C, Key, Mapped, keyEqualsRaw, keyHashRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey) \
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
defTypes( _c_chash_types(CX, C, Key, Mapped); ) \
\
MAP_ONLY_##C( struct CX##_value_t { \
@@ -368,7 +368,7 @@ STC_INLINE size_t fastrange_uint64_t(uint64_t x, uint64_t n) \ CX##_bucket_(const CX* self, const CX##_rawkey_t* rkeyptr) { \
const uint64_t _hash = keyHashRaw(rkeyptr, sizeof *rkeyptr); \
uint_fast8_t _hx; size_t _cap = self->bucket_count; \
- chash_bucket_t b = {c_SELECT(fastrange,CMAP_SIZE_T)(_hash, _cap), (uint_fast8_t)(_hash | 0x80)}; \
+ chash_bucket_t b = {c_PASTE(fastrange_,CMAP_SIZE_T)(_hash, _cap), (uint_fast8_t)(_hash | 0x80)}; \
const uint8_t* _hashx = self->_hashx; \
while ((_hx = _hashx[b.idx])) { \
if (_hx == b.hx) { \
@@ -444,7 +444,7 @@ STC_INLINE size_t fastrange_uint64_t(uint64_t x, uint64_t n) \ if (! _hashx[j]) \
break; \
CX##_rawkey_t _raw = keyToRaw(KEY_REF_##C(_slot+j)); \
- k = c_SELECT(fastrange,CMAP_SIZE_T)(keyHashRaw(&_raw, sizeof _raw), _cap); \
+ k = c_PASTE(fastrange_,CMAP_SIZE_T)(keyHashRaw(&_raw, sizeof _raw), _cap); \
if ((j < i) ^ (k <= i) ^ (k > j)) /* is k outside (i, j]? */ \
memcpy((void *) &_slot[i], &_slot[j], sizeof *_slot), _hashx[i] = _hashx[j], i = j; \
} \
diff --git a/include/stc/cset.h b/include/stc/cset.h index 071a7cbc..6979311a 100644 --- a/include/stc/cset.h +++ b/include/stc/cset.h @@ -57,11 +57,11 @@ int main(void) { using_cset_9(X, Key, keyEquals, keyHash, keyDel, keyClone, c_default_toraw, Key, c_true)
#define using_cset_9(X, Key, keyEqualsRaw, keyHashRaw, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_chash(cset_##X, cset_, Key, Key, keyEqualsRaw, keyHashRaw, \
- @@, @@, @@, void, defTypes, keyDel, keyFromRaw, keyToRaw, RawKey)
+ @@, @@, @@, void, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
-/* cset_str: */
#define using_cset_str() \
- _c_using_chash_strkey(str, cset_, cstr, @@, @@, @@, void, c_true)
+ using_cset_9(str, cstr, c_rawstr_equals, c_rawstr_hash, cstr_del, \
+ cstr_from, cstr_str, const char*, c_true)
#define SET_ONLY_cset_(...) __VA_ARGS__
#define MAP_ONLY_cset_(...)
diff --git a/include/stc/csmap.h b/include/stc/csmap.h index b61af839..bf7fd6ab 100644 --- a/include/stc/csmap.h +++ b/include/stc/csmap.h @@ -52,6 +52,7 @@ int main(void) { #define forward_csmap(X, Key, Mapped) _c_aatree_types(csmap_##X, csmap_, Key, Mapped)
+
#define using_csmap(...) c_MACRO_OVERLOAD(using_csmap, __VA_ARGS__)
#define using_csmap_3(X, Key, Mapped) \
@@ -64,14 +65,14 @@ int main(void) { using_csmap_9(X, Key, Mapped, keyCompare, mappedDel, mappedClone, c_default_toraw, Mapped, c_true)
#define using_csmap_9(X, Key, Mapped, keyCompare, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes) \
using_csmap_13(X, Key, Mapped, keyCompare, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- c_default_del, c_default_fromraw, c_default_toraw, Key)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ c_default_del, c_default_fromraw, c_default_toraw, Key, defTypes)
#define using_csmap_13(X, Key, Mapped, keyCompareRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey) \
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_aatree(csmap_##X, csmap_, Key, Mapped, keyCompareRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define using_csmap_keydef(...) c_MACRO_OVERLOAD(using_csmap_keydef, __VA_ARGS__)
@@ -82,13 +83,13 @@ int main(void) { #define using_csmap_keydef_9(X, Key, Mapped, keyCompareRaw, \
keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_aatree(csmap_##X, csmap_, Key, Mapped, keyCompareRaw, \
- c_default_del, c_default_fromraw, c_default_toraw, Mapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ c_default_del, c_default_fromraw, c_default_toraw, Mapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define using_csmap_str() \
_c_using_aatree(csmap_str, csmap_, cstr, cstr, c_rawstr_compare, \
- cstr_del, cstr_from, cstr_toraw, const char*, c_true, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ cstr_del, cstr_from, cstr_str, const char*, \
+ cstr_del, cstr_from, cstr_str, const char*, c_true)
#define using_csmap_strkey(...) c_MACRO_OVERLOAD(using_csmap_strkey, __VA_ARGS__)
@@ -103,8 +104,8 @@ int main(void) { _c_using_aatree_strkey(X, csmap_, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes)
#define _c_using_aatree_strkey(X, C, Mapped, mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes) \
_c_using_aatree(C##X, C, cstr, Mapped, c_rawstr_compare, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- cstr_del, cstr_from, cstr_toraw, const char*)
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ cstr_del, cstr_from, cstr_str, const char*, defTypes)
#define using_csmap_strval(...) c_MACRO_OVERLOAD(using_csmap_strval, __VA_ARGS__)
@@ -119,8 +120,8 @@ int main(void) { using_csmap_strval_8(X, Key, keyCompare, keyDel, keyClone, c_default_toraw, Key, c_true)
#define using_csmap_strval_8(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_aatree(csmap_##X, csmap_, Key, cstr, keyCompareRaw, \
- cstr_del, cstr_from, cstr_toraw, const char*, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey)
+ cstr_del, cstr_from, cstr_str, const char*, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define SET_ONLY_csmap_(...)
#define MAP_ONLY_csmap_(...) __VA_ARGS__
@@ -159,8 +160,8 @@ struct csmap_rep { size_t root, disp, head, size, cap; void* nodes[]; }; } CX
#define _c_using_aatree(CX, C, Key, Mapped, keyCompareRaw, \
- mappedDel, mappedFromRaw, mappedToRaw, RawMapped, defTypes, \
- keyDel, keyFromRaw, keyToRaw, RawKey) \
+ mappedDel, mappedFromRaw, mappedToRaw, RawMapped, \
+ keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
defTypes( _c_aatree_types(CX, C, Key, Mapped); ) \
\
MAP_ONLY_##C( struct CX##_value_t { \
diff --git a/include/stc/csptr.h b/include/stc/csptr.h index c6c36fb0..4f183498 100644 --- a/include/stc/csptr.h +++ b/include/stc/csptr.h @@ -80,8 +80,8 @@ typedef long atomic_count_t; using_csptr_4(X, Value, valueCompare, c_default_del)
#define using_csptr_4(X, Value, valueCompare, valueDel) \
_c_using_csptr(csptr_##X, Value, valueCompare, valueDel, c_true)
-#define using_csptr_5(X, Value, valueCompare, valueDel, defineTypes) \
- _c_using_csptr(csptr_##X, Value, valueCompare, valueDel, defineTypes)
+#define using_csptr_5(X, Value, valueCompare, valueDel, defTypes) \
+ _c_using_csptr(csptr_##X, Value, valueCompare, valueDel, defTypes)
#define _csptr_types(CX, Value) \
typedef Value CX##_value_t; \
@@ -91,8 +91,8 @@ typedef long atomic_count_t; atomic_count_t* use_count; \
} CX
-#define _c_using_csptr(CX, Value, valueCompare, valueDel, defineTypes) \
- defineTypes( _csptr_types(CX, Value); ) \
+#define _c_using_csptr(CX, Value, valueCompare, valueDel, defTypes) \
+ defTypes( _csptr_types(CX, Value); ) \
struct CX##_rep_ {atomic_count_t cnt; CX##_value_t val;}; \
\
STC_INLINE CX \
diff --git a/include/stc/csset.h b/include/stc/csset.h index 2056009b..747307e4 100644 --- a/include/stc/csset.h +++ b/include/stc/csset.h @@ -59,10 +59,11 @@ int main(void) { using_csset_8(X, Key, keyCompare, keyDel, keyClone, c_default_toraw, Key, c_true)
#define using_csset_8(X, Key, keyCompareRaw, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes) \
_c_using_aatree(csset_##X, csset_, Key, Key, keyCompareRaw, \
- @@, @@, @@, void, defTypes, keyDel, keyFromRaw, keyToRaw, RawKey)
+ @@, @@, @@, void, keyDel, keyFromRaw, keyToRaw, RawKey, defTypes)
#define using_csset_str() \
- _c_using_aatree_strkey(str, csset_, cstr, @@, @@, @@, void, c_true)
+ using_csset_8(str, cstr, c_rawstr_compare, cstr_del, \
+ cstr_from, cstr_str, const char*, c_true)
#define SET_ONLY_csset_(...) __VA_ARGS__
#define MAP_ONLY_csset_(...)
diff --git a/include/stc/cstr.h b/include/stc/cstr.h index d5c2dd86..67f1a0a5 100644 --- a/include/stc/cstr.h +++ b/include/stc/cstr.h @@ -118,11 +118,11 @@ STC_INLINE cstr_iter_t cstr_begin(cstr* self) STC_INLINE cstr_iter_t cstr_end(cstr* self)
{ return c_make(cstr_iter_t){self->str + _cstr_rep(self)->size}; }
STC_INLINE void cstr_next(cstr_iter_t* it) {++it->ref; }
-STC_INLINE bool cstr_equals(cstr s, const char* str)
+STC_INLINE bool cstr_equalto(cstr s, const char* str)
{ return strcmp(s.str, str) == 0; }
-STC_INLINE bool cstr_equals_s(cstr s1, cstr s2)
+STC_INLINE bool cstr_equalto_s(cstr s1, cstr s2)
{ return strcmp(s1.str, s2.str) == 0; }
-STC_INLINE bool cstr_iequals(cstr s, const char* str)
+STC_INLINE bool cstr_iequalto(cstr s, const char* str)
{ return c_strncasecmp(s.str, str, cstr_npos) == 0; }
STC_INLINE bool cstr_contains(cstr s, const char* needle)
{ return strstr(s.str, needle) != NULL; }
@@ -184,12 +184,11 @@ cstr_iends_with(cstr s, const char* sub) { return n <= sz && !c_strncasecmp(s.str + sz - n, sub, n);
}
-/* cvec/cmap adaption functions: */
-#define cstr_toraw(xp) ((xp)->str)
-#define cstr_compare_ref(xp, yp) strcmp((xp)->str, (yp)->str)
-#define cstr_equals_ref(xp, yp) (strcmp((xp)->str, (yp)->str) == 0)
-#define cstr_hash_ref(xp, none) c_default_hash((xp)->str, cstr_size(*(xp)))
-#define cstr_hash(x) c_default_hash((x).str, cstr_size(x))
+/* container adaptor functions: */
+#define cstr_toraw(xp) ((xp)->str) // deprecated
+#define cstr_compare(xp, yp) strcmp((xp)->str, (yp)->str)
+#define cstr_equals(xp, yp) (strcmp((xp)->str, (yp)->str) == 0)
+#define cstr_hash(xp, ...) c_default_hash((xp)->str, cstr_size(*(xp)))
/* -------------------------- IMPLEMENTATION ------------------------- */
diff --git a/include/stc/csview.h b/include/stc/csview.h index c631a0b7..d5d6d246 100644 --- a/include/stc/csview.h +++ b/include/stc/csview.h @@ -57,8 +57,7 @@ STC_INLINE char csview_back(csview sv) { return sv.str[sv.size - 1]; } STC_INLINE void csview_clear(csview* self) { *self = csview_null; }
-#define csview_hash(sv) c_default_hash((sv).str, (sv).size)
-STC_INLINE bool csview_equals(csview sv, csview sv2)
+STC_INLINE bool csview_equalto(csview sv, csview sv2)
{ return sv.size == sv2.size && !memcmp(sv.str, sv2.str, sv.size); }
STC_INLINE size_t csview_find(csview sv, csview needle)
{ char* res = c_strnstrn(sv.str, needle.str, sv.size, needle.size);
@@ -115,10 +114,9 @@ STC_INLINE bool cstr_ends_with_v(cstr s, csview sub) /* ---- Container helper functions ---- */
-STC_INLINE bool csview_equals_ref(const csview* a, const csview* b)
- { return a->size == b->size && !memcmp(a->str, b->str, a->size); }
-#define csview_hash_ref(xp, none) c_default_hash((xp)->str, (xp)->size)
-#define csview_compare_ref(xp, yp) strcmp((xp)->str, (yp)->str)
+#define csview_compare(xp, yp) strcmp((xp)->str, (yp)->str)
+#define csview_hash(xp, ...) c_default_hash((xp)->str, (xp)->size)
+#define csview_equals(xp, yp) (strcmp((xp)->str, (yp)->str) == 0)
/* -------------------------- IMPLEMENTATION ------------------------- */
diff --git a/include/stc/cvec.h b/include/stc/cvec.h index 75a8008b..567c22ae 100644 --- a/include/stc/cvec.h +++ b/include/stc/cvec.h @@ -45,7 +45,7 @@ _c_using_cvec(cvec_##X, Value, valueCompareRaw, valueDel, valueFromRaw, valueToRaw, RawValue, defTypes)
#define using_cvec_str() \
- using_cvec_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_toraw, const char*)
+ using_cvec_7(str, cstr, c_rawstr_compare, cstr_del, cstr_from, cstr_str, const char*)
struct cvec_rep { size_t size, cap; void* data[]; };
#define _cvec_rep(self) c_container_of((self)->data, struct cvec_rep, data)
|