diff options
Diffstat (limited to 'src/haversine_generator/libs')
21 files changed, 5984 insertions, 0 deletions
diff --git a/src/haversine_generator/libs/listing_065.cpp b/src/haversine_generator/libs/listing_065.cpp new file mode 100644 index 0000000..86e087c --- /dev/null +++ b/src/haversine_generator/libs/listing_065.cpp @@ -0,0 +1,39 @@ +#include <math.h> + +static f64 Square(f64 A) +{ + f64 Result = (A*A); + return Result; +} + +static f64 RadiansFromDegrees(f64 Degrees) +{ + f64 Result = 0.01745329251994329577 * Degrees; + return Result; +} + +// NOTE(casey): EarthRadius is generally expected to be 6372.8 +static f64 ReferenceHaversine(f64 X0, f64 Y0, f64 X1, f64 Y1, f64 EarthRadius) +{ + /* NOTE(casey): This is not meant to be a "good" way to calculate the Haversine distance. + Instead, it attempts to follow, as closely as possible, the formula used in the real-world + question on which these homework exercises are loosely based. + */ + + f64 lat1 = Y0; + f64 lat2 = Y1; + f64 lon1 = X0; + f64 lon2 = X1; + + f64 dLat = RadiansFromDegrees(lat2 - lat1); + f64 dLon = RadiansFromDegrees(lon2 - lon1); + lat1 = RadiansFromDegrees(lat1); + lat2 = RadiansFromDegrees(lat2); + + f64 a = Square(sin(dLat/2.0)) + cos(lat1)*cos(lat2)*Square(sin(dLon/2)); + f64 c = 2.0*asin(sqrt(a)); + + f64 Result = EarthRadius * c; + + return Result; +} diff --git a/src/haversine_generator/libs/pcg/pcg-advance-128.c b/src/haversine_generator/libs/pcg/pcg-advance-128.c new file mode 100644 index 0000000..be72009 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-advance-128.c @@ -0,0 +1,64 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Repetative C code is derived using C preprocessor metaprogramming + * techniques. + */ + +#include "pcg_variants.h" + +/* Multi-step advance functions (jump-ahead, jump-back) + * + * 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". + */ + +#if PCG_HAS_128BIT_OPS +pcg128_t pcg_advance_lcg_128(pcg128_t state, pcg128_t delta, pcg128_t cur_mult, + pcg128_t cur_plus) +{ + pcg128_t acc_mult = 1u; + pcg128_t acc_plus = 0u; + while (delta > 0) { + if (delta & 1) { + acc_mult *= cur_mult; + acc_plus = acc_plus * cur_mult + cur_plus; + } + cur_plus = (cur_mult + 1) * cur_plus; + cur_mult *= cur_mult; + delta /= 2; + } + return acc_mult * state + acc_plus; +} +#endif + diff --git a/src/haversine_generator/libs/pcg/pcg-advance-16.c b/src/haversine_generator/libs/pcg/pcg-advance-16.c new file mode 100644 index 0000000..11461d9 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-advance-16.c @@ -0,0 +1,62 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Repetative C code is derived using C preprocessor metaprogramming + * techniques. + */ + +#include "pcg_variants.h" + +/* Multi-step advance functions (jump-ahead, jump-back) + * + * 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". + */ + +uint16_t pcg_advance_lcg_16(uint16_t state, uint16_t delta, uint16_t cur_mult, + uint16_t cur_plus) +{ + uint16_t acc_mult = 1u; + uint16_t acc_plus = 0u; + while (delta > 0) { + if (delta & 1) { + acc_mult *= cur_mult; + acc_plus = acc_plus * cur_mult + cur_plus; + } + cur_plus = (cur_mult + 1) * cur_plus; + cur_mult *= cur_mult; + delta /= 2; + } + return acc_mult * state + acc_plus; +} + diff --git a/src/haversine_generator/libs/pcg/pcg-advance-32.c b/src/haversine_generator/libs/pcg/pcg-advance-32.c new file mode 100644 index 0000000..76f35fc --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-advance-32.c @@ -0,0 +1,62 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Repetative C code is derived using C preprocessor metaprogramming + * techniques. + */ + +#include "pcg_variants.h" + +/* Multi-step advance functions (jump-ahead, jump-back) + * + * 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". + */ + +uint32_t pcg_advance_lcg_32(uint32_t state, uint32_t delta, uint32_t cur_mult, + uint32_t cur_plus) +{ + uint32_t acc_mult = 1u; + uint32_t acc_plus = 0u; + while (delta > 0) { + if (delta & 1) { + acc_mult *= cur_mult; + acc_plus = acc_plus * cur_mult + cur_plus; + } + cur_plus = (cur_mult + 1) * cur_plus; + cur_mult *= cur_mult; + delta /= 2; + } + return acc_mult * state + acc_plus; +} + diff --git a/src/haversine_generator/libs/pcg/pcg-advance-64.c b/src/haversine_generator/libs/pcg/pcg-advance-64.c new file mode 100644 index 0000000..8210e75 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-advance-64.c @@ -0,0 +1,62 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Repetative C code is derived using C preprocessor metaprogramming + * techniques. + */ + +#include "pcg_variants.h" + +/* Multi-step advance functions (jump-ahead, jump-back) + * + * 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". + */ + +uint64_t pcg_advance_lcg_64(uint64_t state, uint64_t delta, uint64_t cur_mult, + uint64_t cur_plus) +{ + uint64_t acc_mult = 1u; + uint64_t acc_plus = 0u; + while (delta > 0) { + if (delta & 1) { + acc_mult *= cur_mult; + acc_plus = acc_plus * cur_mult + cur_plus; + } + cur_plus = (cur_mult + 1) * cur_plus; + cur_mult *= cur_mult; + delta /= 2; + } + return acc_mult * state + acc_plus; +} + diff --git a/src/haversine_generator/libs/pcg/pcg-advance-8.c b/src/haversine_generator/libs/pcg/pcg-advance-8.c new file mode 100644 index 0000000..8280958 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-advance-8.c @@ -0,0 +1,62 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Repetative C code is derived using C preprocessor metaprogramming + * techniques. + */ + +#include "pcg_variants.h" + +/* Multi-step advance functions (jump-ahead, jump-back) + * + * 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". + */ + +uint8_t pcg_advance_lcg_8(uint8_t state, uint8_t delta, uint8_t cur_mult, + uint8_t cur_plus) +{ + uint8_t acc_mult = 1u; + uint8_t acc_plus = 0u; + while (delta > 0) { + if (delta & 1) { + acc_mult *= cur_mult; + acc_plus = acc_plus * cur_mult + cur_plus; + } + cur_plus = (cur_mult + 1) * cur_plus; + cur_mult *= cur_mult; + delta /= 2; + } + return acc_mult * state + acc_plus; +} + diff --git a/src/haversine_generator/libs/pcg/pcg-global-32.c b/src/haversine_generator/libs/pcg/pcg-global-32.c new file mode 100644 index 0000000..8c18e48 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-global-32.c @@ -0,0 +1,56 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +static pcg32_random_t pcg32_global = PCG32_INITIALIZER; + +uint32_t pcg32_random() +{ + return pcg32_random_r(&pcg32_global); +} + +uint32_t pcg32_boundedrand(uint32_t bound) +{ + return pcg32_boundedrand_r(&pcg32_global, bound); +} + +void pcg32_srandom(uint64_t seed, uint64_t seq) +{ + pcg32_srandom_r(&pcg32_global, seed, seq); +} + +void pcg32_advance(uint64_t delta) +{ + return pcg32_advance_r(&pcg32_global, delta); +} + diff --git a/src/haversine_generator/libs/pcg/pcg-global-64.c b/src/haversine_generator/libs/pcg/pcg-global-64.c new file mode 100644 index 0000000..26aa677 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-global-64.c @@ -0,0 +1,59 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +#if PCG_HAS_128BIT_OPS + +static pcg64_random_t pcg64_global = PCG64_INITIALIZER; + +uint64_t pcg64_random() +{ + return pcg64_random_r(&pcg64_global); +} + +uint64_t pcg64_boundedrand(uint64_t bound) +{ + return pcg64_boundedrand_r(&pcg64_global, bound); +} + +void pcg64_srandom(pcg128_t seed, pcg128_t seq) +{ + pcg64_srandom_r(&pcg64_global, seed, seq); +} + +void pcg64_advance(pcg128_t delta) +{ + pcg64_advance_r(&pcg64_global, delta); +} + +#endif diff --git a/src/haversine_generator/libs/pcg/pcg-output-128.c b/src/haversine_generator/libs/pcg/pcg-output-128.c new file mode 100644 index 0000000..cb2142e --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-output-128.c @@ -0,0 +1,64 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* + * Rotate helper functions. + */ + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t pcg_rotr_128(pcg128_t value, unsigned int rot); +#endif + +/* + * Output functions. These are the core of the PCG generation scheme. + */ + +// XSH RS + +// XSH RR + +// RXS M XS + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t pcg_output_rxs_m_xs_128_128(pcg128_t state); +#endif + +// XSL RR (only defined for >= 64 bits) + +// XSL RR RR (only defined for >= 64 bits) + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t pcg_output_xsl_rr_rr_128_128(pcg128_t state); +#endif + diff --git a/src/haversine_generator/libs/pcg/pcg-output-16.c b/src/haversine_generator/libs/pcg/pcg-output-16.c new file mode 100644 index 0000000..c593f67 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-output-16.c @@ -0,0 +1,60 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* + * Rotate helper functions. + */ + +extern inline uint16_t pcg_rotr_16(uint16_t value, unsigned int rot); + +/* + * Output functions. These are the core of the PCG generation scheme. + */ + +// XSH RS + +extern inline uint16_t pcg_output_xsh_rs_32_16(uint32_t state); + +// XSH RR + +extern inline uint16_t pcg_output_xsh_rr_32_16(uint32_t state); + +// RXS M XS + +extern inline uint16_t pcg_output_rxs_m_xs_16_16(uint16_t state); + +// XSL RR (only defined for >= 64 bits) + +// XSL RR RR (only defined for >= 64 bits) + diff --git a/src/haversine_generator/libs/pcg/pcg-output-32.c b/src/haversine_generator/libs/pcg/pcg-output-32.c new file mode 100644 index 0000000..e291c36 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-output-32.c @@ -0,0 +1,62 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* + * Rotate helper functions. + */ + +extern inline uint32_t pcg_rotr_32(uint32_t value, unsigned int rot); + +/* + * Output functions. These are the core of the PCG generation scheme. + */ + +// XSH RS + +extern inline uint32_t pcg_output_xsh_rs_64_32(uint64_t state); + +// XSH RR + +extern inline uint32_t pcg_output_xsh_rr_64_32(uint64_t state); + +// RXS M XS + +extern inline uint32_t pcg_output_rxs_m_xs_32_32(uint32_t state); + +// XSL RR (only defined for >= 64 bits) + +extern inline uint32_t pcg_output_xsl_rr_64_32(uint64_t state); + +// XSL RR RR (only defined for >= 64 bits) + diff --git a/src/haversine_generator/libs/pcg/pcg-output-64.c b/src/haversine_generator/libs/pcg/pcg-output-64.c new file mode 100644 index 0000000..8c6b7e4 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-output-64.c @@ -0,0 +1,70 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* + * Rotate helper functions. + */ + +extern inline uint64_t pcg_rotr_64(uint64_t value, unsigned int rot); + +/* + * Output functions. These are the core of the PCG generation scheme. + */ + +// XSH RS + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t pcg_output_xsh_rs_128_64(pcg128_t state); +#endif + +// XSH RR + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t pcg_output_xsh_rr_128_64(pcg128_t state); +#endif + +// RXS M XS + +extern inline uint64_t pcg_output_rxs_m_xs_64_64(uint64_t state); + +// XSL RR (only defined for >= 64 bits) + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t pcg_output_xsl_rr_128_64(pcg128_t state); +#endif + +// XSL RR RR (only defined for >= 64 bits) + +extern inline uint64_t pcg_output_xsl_rr_rr_64_64(uint64_t state); + diff --git a/src/haversine_generator/libs/pcg/pcg-output-8.c b/src/haversine_generator/libs/pcg/pcg-output-8.c new file mode 100644 index 0000000..83fe449 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-output-8.c @@ -0,0 +1,60 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* + * Rotate helper functions. + */ + +extern inline uint8_t pcg_rotr_8(uint8_t value, unsigned int rot); + +/* + * Output functions. These are the core of the PCG generation scheme. + */ + +// XSH RS + +extern inline uint8_t pcg_output_xsh_rs_16_8(uint16_t state); + +// XSH RR + +extern inline uint8_t pcg_output_xsh_rr_16_8(uint16_t state); + +// RXS M XS + +extern inline uint8_t pcg_output_rxs_m_xs_8_8(uint8_t state); + +// XSL RR (only defined for >= 64 bits) + +// XSL RR RR (only defined for >= 64 bits) + diff --git a/src/haversine_generator/libs/pcg/pcg-rngs-128.c b/src/haversine_generator/libs/pcg/pcg-rngs-128.c new file mode 100644 index 0000000..8023589 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-rngs-128.c @@ -0,0 +1,337 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_oneseq_128_step_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_oneseq_128_advance_r(struct pcg_state_128* rng, + pcg128_t delta); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_mcg_128_step_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_mcg_128_advance_r(struct pcg_state_128* rng, + pcg128_t delta); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_unique_128_step_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_unique_128_advance_r(struct pcg_state_128* rng, + pcg128_t delta); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_setseq_128_step_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_setseq_128_advance_r(struct pcg_state_setseq_128* rng, + pcg128_t delta); +#endif + +/* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_oneseq_128_srandom_r(struct pcg_state_128* rng, + pcg128_t initstate); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_mcg_128_srandom_r(struct pcg_state_128* rng, + pcg128_t initstate); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_unique_128_srandom_r(struct pcg_state_128* rng, + pcg128_t initstate); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline void pcg_setseq_128_srandom_r(struct pcg_state_setseq_128* rng, + pcg128_t initstate, + pcg128_t initseq); +#endif + +/* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + +/* Generation functions for XSH RS */ + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsh_rs_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsh_rs_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsh_rs_64_random_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsh_rs_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsh_rs_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, uint64_t bound); +#endif + +/* Generation functions for XSH RR */ + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsh_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsh_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsh_rr_64_random_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsh_rr_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsh_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, uint64_t bound); +#endif + +/* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_oneseq_128_rxs_m_xs_128_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_oneseq_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_unique_128_rxs_m_xs_128_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_unique_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_setseq_128_rxs_m_xs_128_random_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_setseq_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_setseq_128* rng, + pcg128_t bound); +#endif + +/* Generation functions for XSL RR (only defined for "large" types) */ + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsl_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_oneseq_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsl_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_unique_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsl_rr_64_random_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_setseq_128_xsl_rr_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsl_rr_64_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline uint64_t +pcg_mcg_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, uint64_t bound); +#endif + +/* Generation functions for XSL RR RR (only defined for "large" types) */ + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_oneseq_128_xsl_rr_rr_128_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_oneseq_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_unique_128_xsl_rr_rr_128_random_r(struct pcg_state_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_unique_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_setseq_128_xsl_rr_rr_128_random_r(struct pcg_state_setseq_128* rng); +#endif + +#if PCG_HAS_128BIT_OPS +extern inline pcg128_t +pcg_setseq_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_setseq_128* rng, + pcg128_t bound); +#endif + diff --git a/src/haversine_generator/libs/pcg/pcg-rngs-16.c b/src/haversine_generator/libs/pcg/pcg-rngs-16.c new file mode 100644 index 0000000..6d4e9b6 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-rngs-16.c @@ -0,0 +1,183 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + +extern inline void pcg_oneseq_16_step_r(struct pcg_state_16* rng); + +extern inline void pcg_oneseq_16_advance_r(struct pcg_state_16* rng, + uint16_t delta); + +extern inline void pcg_mcg_16_step_r(struct pcg_state_16* rng); + +extern inline void pcg_mcg_16_advance_r(struct pcg_state_16* rng, + uint16_t delta); + +extern inline void pcg_unique_16_step_r(struct pcg_state_16* rng); + +extern inline void pcg_unique_16_advance_r(struct pcg_state_16* rng, + uint16_t delta); + +extern inline void pcg_setseq_16_step_r(struct pcg_state_setseq_16* rng); + +extern inline void pcg_setseq_16_advance_r(struct pcg_state_setseq_16* rng, + uint16_t delta); + +/* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + +extern inline void pcg_oneseq_16_srandom_r(struct pcg_state_16* rng, + uint16_t initstate); + +extern inline void pcg_mcg_16_srandom_r(struct pcg_state_16* rng, + uint16_t initstate); + +extern inline void pcg_unique_16_srandom_r(struct pcg_state_16* rng, + uint16_t initstate); + +extern inline void pcg_setseq_16_srandom_r(struct pcg_state_setseq_16* rng, + uint16_t initstate, + uint16_t initseq); + +/* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + +/* Generation functions for XSH RS */ + +extern inline uint8_t pcg_oneseq_16_xsh_rs_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_oneseq_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +extern inline uint8_t pcg_unique_16_xsh_rs_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_unique_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +extern inline uint8_t +pcg_setseq_16_xsh_rs_8_random_r(struct pcg_state_setseq_16* rng); + +extern inline uint8_t +pcg_setseq_16_xsh_rs_8_boundedrand_r(struct pcg_state_setseq_16* rng, + uint8_t bound); + +extern inline uint8_t pcg_mcg_16_xsh_rs_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_mcg_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +/* Generation functions for XSH RR */ + +extern inline uint8_t pcg_oneseq_16_xsh_rr_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_oneseq_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +extern inline uint8_t pcg_unique_16_xsh_rr_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_unique_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +extern inline uint8_t +pcg_setseq_16_xsh_rr_8_random_r(struct pcg_state_setseq_16* rng); + +extern inline uint8_t +pcg_setseq_16_xsh_rr_8_boundedrand_r(struct pcg_state_setseq_16* rng, + uint8_t bound); + +extern inline uint8_t pcg_mcg_16_xsh_rr_8_random_r(struct pcg_state_16* rng); + +extern inline uint8_t +pcg_mcg_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, uint8_t bound); + +/* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + +extern inline uint16_t +pcg_oneseq_16_rxs_m_xs_16_random_r(struct pcg_state_16* rng); + +extern inline uint16_t +pcg_oneseq_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_16* rng, + uint16_t bound); + +extern inline uint16_t +pcg_unique_16_rxs_m_xs_16_random_r(struct pcg_state_16* rng); + +extern inline uint16_t +pcg_unique_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_16* rng, + uint16_t bound); + +extern inline uint16_t +pcg_setseq_16_rxs_m_xs_16_random_r(struct pcg_state_setseq_16* rng); + +extern inline uint16_t +pcg_setseq_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_setseq_16* rng, + uint16_t bound); + +/* Generation functions for XSL RR (only defined for "large" types) */ + +/* Generation functions for XSL RR RR (only defined for "large" types) */ + diff --git a/src/haversine_generator/libs/pcg/pcg-rngs-32.c b/src/haversine_generator/libs/pcg/pcg-rngs-32.c new file mode 100644 index 0000000..1c8da7e --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-rngs-32.c @@ -0,0 +1,187 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + +extern inline void pcg_oneseq_32_step_r(struct pcg_state_32* rng); + +extern inline void pcg_oneseq_32_advance_r(struct pcg_state_32* rng, + uint32_t delta); + +extern inline void pcg_mcg_32_step_r(struct pcg_state_32* rng); + +extern inline void pcg_mcg_32_advance_r(struct pcg_state_32* rng, + uint32_t delta); + +extern inline void pcg_unique_32_step_r(struct pcg_state_32* rng); + +extern inline void pcg_unique_32_advance_r(struct pcg_state_32* rng, + uint32_t delta); + +extern inline void pcg_setseq_32_step_r(struct pcg_state_setseq_32* rng); + +extern inline void pcg_setseq_32_advance_r(struct pcg_state_setseq_32* rng, + uint32_t delta); + +/* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + +extern inline void pcg_oneseq_32_srandom_r(struct pcg_state_32* rng, + uint32_t initstate); + +extern inline void pcg_mcg_32_srandom_r(struct pcg_state_32* rng, + uint32_t initstate); + +extern inline void pcg_unique_32_srandom_r(struct pcg_state_32* rng, + uint32_t initstate); + +extern inline void pcg_setseq_32_srandom_r(struct pcg_state_setseq_32* rng, + uint32_t initstate, + uint32_t initseq); + +/* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + +/* Generation functions for XSH RS */ + +extern inline uint16_t +pcg_oneseq_32_xsh_rs_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_oneseq_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +extern inline uint16_t +pcg_unique_32_xsh_rs_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_unique_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +extern inline uint16_t +pcg_setseq_32_xsh_rs_16_random_r(struct pcg_state_setseq_32* rng); + +extern inline uint16_t +pcg_setseq_32_xsh_rs_16_boundedrand_r(struct pcg_state_setseq_32* rng, + uint16_t bound); + +extern inline uint16_t pcg_mcg_32_xsh_rs_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_mcg_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +/* Generation functions for XSH RR */ + +extern inline uint16_t +pcg_oneseq_32_xsh_rr_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_oneseq_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +extern inline uint16_t +pcg_unique_32_xsh_rr_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_unique_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +extern inline uint16_t +pcg_setseq_32_xsh_rr_16_random_r(struct pcg_state_setseq_32* rng); + +extern inline uint16_t +pcg_setseq_32_xsh_rr_16_boundedrand_r(struct pcg_state_setseq_32* rng, + uint16_t bound); + +extern inline uint16_t pcg_mcg_32_xsh_rr_16_random_r(struct pcg_state_32* rng); + +extern inline uint16_t +pcg_mcg_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, uint16_t bound); + +/* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + +extern inline uint32_t +pcg_oneseq_32_rxs_m_xs_32_random_r(struct pcg_state_32* rng); + +extern inline uint32_t +pcg_oneseq_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_32* rng, + uint32_t bound); + +extern inline uint32_t +pcg_unique_32_rxs_m_xs_32_random_r(struct pcg_state_32* rng); + +extern inline uint32_t +pcg_unique_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_32* rng, + uint32_t bound); + +extern inline uint32_t +pcg_setseq_32_rxs_m_xs_32_random_r(struct pcg_state_setseq_32* rng); + +extern inline uint32_t +pcg_setseq_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_setseq_32* rng, + uint32_t bound); + +/* Generation functions for XSL RR (only defined for "large" types) */ + +/* Generation functions for XSL RR RR (only defined for "large" types) */ + diff --git a/src/haversine_generator/libs/pcg/pcg-rngs-64.c b/src/haversine_generator/libs/pcg/pcg-rngs-64.c new file mode 100644 index 0000000..cc0ff2c --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-rngs-64.c @@ -0,0 +1,232 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + +extern inline void pcg_oneseq_64_step_r(struct pcg_state_64* rng); + +extern inline void pcg_oneseq_64_advance_r(struct pcg_state_64* rng, + uint64_t delta); + +extern inline void pcg_mcg_64_step_r(struct pcg_state_64* rng); + +extern inline void pcg_mcg_64_advance_r(struct pcg_state_64* rng, + uint64_t delta); + +extern inline void pcg_unique_64_step_r(struct pcg_state_64* rng); + +extern inline void pcg_unique_64_advance_r(struct pcg_state_64* rng, + uint64_t delta); + +extern inline void pcg_setseq_64_step_r(struct pcg_state_setseq_64* rng); + +extern inline void pcg_setseq_64_advance_r(struct pcg_state_setseq_64* rng, + uint64_t delta); + +/* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + +extern inline void pcg_oneseq_64_srandom_r(struct pcg_state_64* rng, + uint64_t initstate); + +extern inline void pcg_mcg_64_srandom_r(struct pcg_state_64* rng, + uint64_t initstate); + +extern inline void pcg_unique_64_srandom_r(struct pcg_state_64* rng, + uint64_t initstate); + +extern inline void pcg_setseq_64_srandom_r(struct pcg_state_setseq_64* rng, + uint64_t initstate, + uint64_t initseq); + +/* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + +/* Generation functions for XSH RS */ + +extern inline uint32_t +pcg_oneseq_64_xsh_rs_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_oneseq_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_unique_64_xsh_rs_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_unique_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_setseq_64_xsh_rs_32_random_r(struct pcg_state_setseq_64* rng); + +extern inline uint32_t +pcg_setseq_64_xsh_rs_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound); + +extern inline uint32_t pcg_mcg_64_xsh_rs_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_mcg_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +/* Generation functions for XSH RR */ + +extern inline uint32_t +pcg_oneseq_64_xsh_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_oneseq_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_unique_64_xsh_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_unique_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_setseq_64_xsh_rr_32_random_r(struct pcg_state_setseq_64* rng); + +extern inline uint32_t +pcg_setseq_64_xsh_rr_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound); + +extern inline uint32_t pcg_mcg_64_xsh_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_mcg_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +/* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + +extern inline uint64_t +pcg_oneseq_64_rxs_m_xs_64_random_r(struct pcg_state_64* rng); + +extern inline uint64_t +pcg_oneseq_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound); + +extern inline uint64_t +pcg_unique_64_rxs_m_xs_64_random_r(struct pcg_state_64* rng); + +extern inline uint64_t +pcg_unique_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound); + +extern inline uint64_t +pcg_setseq_64_rxs_m_xs_64_random_r(struct pcg_state_setseq_64* rng); + +extern inline uint64_t +pcg_setseq_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_setseq_64* rng, + uint64_t bound); + +/* Generation functions for XSL RR (only defined for "large" types) */ + +extern inline uint32_t +pcg_oneseq_64_xsl_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_oneseq_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_unique_64_xsl_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_unique_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +extern inline uint32_t +pcg_setseq_64_xsl_rr_32_random_r(struct pcg_state_setseq_64* rng); + +extern inline uint32_t +pcg_setseq_64_xsl_rr_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound); + +extern inline uint32_t pcg_mcg_64_xsl_rr_32_random_r(struct pcg_state_64* rng); + +extern inline uint32_t +pcg_mcg_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, uint32_t bound); + +/* Generation functions for XSL RR RR (only defined for "large" types) */ + +extern inline uint64_t +pcg_oneseq_64_xsl_rr_rr_64_random_r(struct pcg_state_64* rng); + +extern inline uint64_t +pcg_oneseq_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound); + +extern inline uint64_t +pcg_unique_64_xsl_rr_rr_64_random_r(struct pcg_state_64* rng); + +extern inline uint64_t +pcg_unique_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound); + +extern inline uint64_t +pcg_setseq_64_xsl_rr_rr_64_random_r(struct pcg_state_setseq_64* rng); + +extern inline uint64_t +pcg_setseq_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_setseq_64* rng, + uint64_t bound); + diff --git a/src/haversine_generator/libs/pcg/pcg-rngs-8.c b/src/haversine_generator/libs/pcg/pcg-rngs-8.c new file mode 100644 index 0000000..8779aac --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg-rngs-8.c @@ -0,0 +1,128 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * The contents of this file were mechanically derived from pcg_variants.h + * (every inline function defined there gets an exern declaration here). + */ + +#include "pcg_variants.h" + +/* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + +extern inline void pcg_oneseq_8_step_r(struct pcg_state_8* rng); + +extern inline void pcg_oneseq_8_advance_r(struct pcg_state_8* rng, + uint8_t delta); + +extern inline void pcg_mcg_8_step_r(struct pcg_state_8* rng); + +extern inline void pcg_mcg_8_advance_r(struct pcg_state_8* rng, uint8_t delta); + +extern inline void pcg_unique_8_step_r(struct pcg_state_8* rng); + +extern inline void pcg_unique_8_advance_r(struct pcg_state_8* rng, + uint8_t delta); + +extern inline void pcg_setseq_8_step_r(struct pcg_state_setseq_8* rng); + +extern inline void pcg_setseq_8_advance_r(struct pcg_state_setseq_8* rng, + uint8_t delta); + +/* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + +extern inline void pcg_oneseq_8_srandom_r(struct pcg_state_8* rng, + uint8_t initstate); + +extern inline void pcg_mcg_8_srandom_r(struct pcg_state_8* rng, + uint8_t initstate); + +extern inline void pcg_unique_8_srandom_r(struct pcg_state_8* rng, + uint8_t initstate); + +extern inline void pcg_setseq_8_srandom_r(struct pcg_state_setseq_8* rng, + uint8_t initstate, uint8_t initseq); + +/* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + +/* Generation functions for XSH RS */ + +/* Generation functions for XSH RR */ + +/* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + +extern inline uint8_t pcg_oneseq_8_rxs_m_xs_8_random_r(struct pcg_state_8* rng); + +extern inline uint8_t +pcg_oneseq_8_rxs_m_xs_8_boundedrand_r(struct pcg_state_8* rng, uint8_t bound); + +extern inline uint8_t +pcg_setseq_8_rxs_m_xs_8_random_r(struct pcg_state_setseq_8* rng); + +extern inline uint8_t +pcg_setseq_8_rxs_m_xs_8_boundedrand_r(struct pcg_state_setseq_8* rng, + uint8_t bound); + +/* Generation functions for XSL RR (only defined for "large" types) */ + +/* Generation functions for XSL RR RR (only defined for "large" types) */ + diff --git a/src/haversine_generator/libs/pcg/pcg.c b/src/haversine_generator/libs/pcg/pcg.c new file mode 100644 index 0000000..cf29e6d --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg.c @@ -0,0 +1,16 @@ +#include "pcg_variants.h" +#include "pcg-advance-128.c" +#include "pcg-advance-16.c" +#include "pcg-advance-32.c" +#include "pcg-advance-64.c" +#include "pcg-advance-8.c" +#include "pcg-output-128.c" +#include "pcg-output-16.c" +#include "pcg-output-32.c" +#include "pcg-output-64.c" +#include "pcg-output-8.c" +#include "pcg-rngs-128.c" +#include "pcg-rngs-16.c" +#include "pcg-rngs-32.c" +#include "pcg-rngs-64.c" +#include "pcg-rngs-8.c" diff --git a/src/haversine_generator/libs/pcg/pcg_variants.h b/src/haversine_generator/libs/pcg/pcg_variants.h new file mode 100644 index 0000000..83edae8 --- /dev/null +++ b/src/haversine_generator/libs/pcg/pcg_variants.h @@ -0,0 +1,2213 @@ +/* + * PCG Random Number Generation for C. + * + * Copyright 2014 Melissa O'Neill <oneill@pcg-random.org> + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + * + * For additional information about the PCG random number generation scheme, + * including its license and other licensing options, visit + * + * http://www.pcg-random.org + */ + +/* + * This code is derived from the canonical C++ PCG implementation, which + * has many additional features and is preferable if you can use C++ in + * your project. + * + * Much of the derivation was performed mechanically. In particular, the + * output functions were generated by compiling the C++ output functions + * into LLVM bitcode and then transforming that using the LLVM C backend + * (from https://github.com/draperlaboratory/llvm-cbe), and then + * postprocessing and hand editing the output. + * + * Much of the remaining code was generated by C-preprocessor metaprogramming. + */ + +#ifndef PCG_VARIANTS_H_INCLUDED +#define PCG_VARIANTS_H_INCLUDED 1 + +#include <inttypes.h> + +#if __SIZEOF_INT128__ +typedef __uint128_t pcg128_t; +#define PCG_128BIT_CONSTANT(high,low) \ +((((pcg128_t)high) << 64) + low) +#define PCG_HAS_128BIT_OPS 1 +#else +#error "non" +#endif + +#if __GNUC_GNU_INLINE__ && !defined(__cplusplus) +#error Nonstandard GNU inlining semanatics. Compile with -std=c99 or better. +// We could instead use macros PCG_INLINE and PCG_EXTERN_INLINE +// but better to just reject ancient C code. +#endif + +#if __cplusplus +extern "C" { +#endif + + /* + * Rotate helper functions. + */ + + inline uint8_t pcg_rotr_8(uint8_t value, unsigned int rot) + { + /* Unfortunately, clang is kinda pathetic when it comes to properly + * recognizing idiomatic rotate code, so for clang we actually provide + * assembler directives (enabled with PCG_USE_INLINE_ASM). Boo, hiss. + */ +#if PCG_USE_INLINE_ASM && __clang__ && (__x86_64__ || __i386__) + asm ("rorb %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); + return value; +#else + return (value >> rot) | (value << ((- rot) & 7)); +#endif + } + + inline uint16_t pcg_rotr_16(uint16_t value, unsigned int rot) + { +#if PCG_USE_INLINE_ASM && __clang__ && (__x86_64__ || __i386__) + asm ("rorw %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); + return value; +#else + return (value >> rot) | (value << ((- rot) & 15)); +#endif + } + + inline uint32_t pcg_rotr_32(uint32_t value, unsigned int rot) + { +#if PCG_USE_INLINE_ASM && __clang__ && (__x86_64__ || __i386__) + asm ("rorl %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); + return value; +#else + return (value >> rot) | (value << ((- rot) & 31)); +#endif + } + + inline uint64_t pcg_rotr_64(uint64_t value, unsigned int rot) + { +#if 0 && PCG_USE_INLINE_ASM && __clang__ && __x86_64__ + // For whatever reason, clang actually *does* generator rotq by + // itself, so we don't need this code. + asm ("rorq %%cl, %0" : "=r" (value) : "0" (value), "c" (rot)); + return value; +#else + return (value >> rot) | (value << ((- rot) & 63)); +#endif + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_rotr_128(pcg128_t value, unsigned int rot) + { + return (value >> rot) | (value << ((- rot) & 127)); + } +#endif + + /* + * Output functions. These are the core of the PCG generation scheme. + */ + + // XSH RS + + inline uint8_t pcg_output_xsh_rs_16_8(uint16_t state) + { + return (uint8_t)(((state >> 7u) ^ state) >> ((state >> 14u) + 3u)); + } + + inline uint16_t pcg_output_xsh_rs_32_16(uint32_t state) + { + return (uint16_t)(((state >> 11u) ^ state) >> ((state >> 30u) + 11u)); + } + + inline uint32_t pcg_output_xsh_rs_64_32(uint64_t state) + { + + return (uint32_t)(((state >> 22u) ^ state) >> ((state >> 61u) + 22u)); + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_output_xsh_rs_128_64(pcg128_t state) + { + return (uint64_t)(((state >> 43u) ^ state) >> ((state >> 124u) + 45u)); + } +#endif + + // XSH RR + + inline uint8_t pcg_output_xsh_rr_16_8(uint16_t state) + { + return pcg_rotr_8(((state >> 5u) ^ state) >> 5u, state >> 13u); + } + + inline uint16_t pcg_output_xsh_rr_32_16(uint32_t state) + { + return pcg_rotr_16(((state >> 10u) ^ state) >> 12u, state >> 28u); + } + + inline uint32_t pcg_output_xsh_rr_64_32(uint64_t state) + { + return pcg_rotr_32(((state >> 18u) ^ state) >> 27u, state >> 59u); + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_output_xsh_rr_128_64(pcg128_t state) + { + return pcg_rotr_64(((state >> 29u) ^ state) >> 58u, state >> 122u); + } +#endif + + // RXS M XS + + inline uint8_t pcg_output_rxs_m_xs_8_8(uint8_t state) + { + uint8_t word = ((state >> ((state >> 6u) + 2u)) ^ state) * 217u; + return (word >> 6u) ^ word; + } + + inline uint16_t pcg_output_rxs_m_xs_16_16(uint16_t state) + { + uint16_t word = ((state >> ((state >> 13u) + 3u)) ^ state) * 62169u; + return (word >> 11u) ^ word; + } + + inline uint32_t pcg_output_rxs_m_xs_32_32(uint32_t state) + { + uint32_t word = ((state >> ((state >> 28u) + 4u)) ^ state) * 277803737u; + return (word >> 22u) ^ word; + } + + inline uint64_t pcg_output_rxs_m_xs_64_64(uint64_t state) + { + uint64_t word = ((state >> ((state >> 59u) + 5u)) ^ state) + * 12605985483714917081ull; + return (word >> 43u) ^ word; + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_output_rxs_m_xs_128_128(pcg128_t state) + { + pcg128_t word = ((state >> ((state >> 122u) + 6u)) ^ state) + * (PCG_128BIT_CONSTANT(17766728186571221404ULL, + 12605985483714917081ULL)); + // 327738287884841127335028083622016905945 + return (word >> 86u) ^ word; + } +#endif + + // XSL RR (only defined for >= 64 bits) + + inline uint32_t pcg_output_xsl_rr_64_32(uint64_t state) + { + return pcg_rotr_32(((uint32_t)(state >> 32u)) ^ (uint32_t)state, + state >> 59u); + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_output_xsl_rr_128_64(pcg128_t state) + { + return pcg_rotr_64(((uint64_t)(state >> 64u)) ^ (uint64_t)state, + state >> 122u); + } +#endif + + // XSL RR RR (only defined for >= 64 bits) + + inline uint64_t pcg_output_xsl_rr_rr_64_64(uint64_t state) + { + uint32_t rot1 = (uint32_t)(state >> 59u); + uint32_t high = (uint32_t)(state >> 32u); + uint32_t low = (uint32_t)state; + uint32_t xored = high ^ low; + uint32_t newlow = pcg_rotr_32(xored, rot1); + uint32_t newhigh = pcg_rotr_32(high, newlow & 31u); + return (((uint64_t)newhigh) << 32u) | newlow; + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_output_xsl_rr_rr_128_128(pcg128_t state) + { + uint32_t rot1 = (uint32_t)(state >> 122u); + uint64_t high = (uint64_t)(state >> 64u); + uint64_t low = (uint64_t)state; + uint64_t xored = high ^ low; + uint64_t newlow = pcg_rotr_64(xored, rot1); + uint64_t newhigh = pcg_rotr_64(high, newlow & 63u); + return (((pcg128_t)newhigh) << 64u) | newlow; + } +#endif + +#define PCG_DEFAULT_MULTIPLIER_8 141U +#define PCG_DEFAULT_MULTIPLIER_16 12829U +#define PCG_DEFAULT_MULTIPLIER_32 747796405U +#define PCG_DEFAULT_MULTIPLIER_64 6364136223846793005ULL + +#define PCG_DEFAULT_INCREMENT_8 77U +#define PCG_DEFAULT_INCREMENT_16 47989U +#define PCG_DEFAULT_INCREMENT_32 2891336453U +#define PCG_DEFAULT_INCREMENT_64 1442695040888963407ULL + +#if PCG_HAS_128BIT_OPS +#define PCG_DEFAULT_MULTIPLIER_128 \ +PCG_128BIT_CONSTANT(2549297995355413924ULL,4865540595714422341ULL) +#define PCG_DEFAULT_INCREMENT_128 \ +PCG_128BIT_CONSTANT(6364136223846793005ULL,1442695040888963407ULL) +#endif + + /* + * Static initialization constants (if you can't call srandom for some + * bizarre reason). + */ + +#if PCG_HAS_128BIT_OPS +#define PCG_STATE_ONESEQ_8_INITIALIZER { 0xd7U } +#define PCG_STATE_ONESEQ_16_INITIALIZER { 0x20dfU } +#define PCG_STATE_ONESEQ_32_INITIALIZER { 0x46b56677U } +#define PCG_STATE_ONESEQ_64_INITIALIZER { 0x4d595df4d0f33173ULL } +#define PCG_STATE_ONESEQ_128_INITIALIZER \ +{ PCG_128BIT_CONSTANT(0xb8dc10e158a92392ULL, 0x98046df007ec0a53ULL) } +#endif + +#if PCG_HAS_128BIT_OPS +#define PCG_STATE_UNIQUE_8_INITIALIZER PCG_STATE_ONESEQ_8_INITIALIZER +#define PCG_STATE_UNIQUE_16_INITIALIZER PCG_STATE_ONESEQ_16_INITIALIZER +#define PCG_STATE_UNIQUE_32_INITIALIZER PCG_STATE_ONESEQ_32_INITIALIZER +#define PCG_STATE_UNIQUE_64_INITIALIZER PCG_STATE_ONESEQ_64_INITIALIZER +#define PCG_STATE_UNIQUE_128_INITIALIZER PCG_STATE_ONESEQ_128_INITIALIZER +#endif + +#if PCG_HAS_128BIT_OPS +#define PCG_STATE_MCG_8_INITIALIZER { 0xe5U } +#define PCG_STATE_MCG_16_INITIALIZER { 0xa5e5U } +#define PCG_STATE_MCG_32_INITIALIZER { 0xd15ea5e5U } +#define PCG_STATE_MCG_64_INITIALIZER { 0xcafef00dd15ea5e5ULL } +#define PCG_STATE_MCG_128_INITIALIZER \ +{ PCG_128BIT_CONSTANT(0x0000000000000000ULL, 0xcafef00dd15ea5e5ULL) } +#endif + +#if PCG_HAS_128BIT_OPS +#define PCG_STATE_SETSEQ_8_INITIALIZER { 0x9bU, 0xdbU } +#define PCG_STATE_SETSEQ_16_INITIALIZER { 0xe39bU, 0x5bdbU } +#define PCG_STATE_SETSEQ_32_INITIALIZER { 0xec02d89bU, 0x94b95bdbU } +#define PCG_STATE_SETSEQ_64_INITIALIZER \ +{ 0x853c49e6748fea9bULL, 0xda3e39cb94b95bdbULL } +#define PCG_STATE_SETSEQ_128_INITIALIZER \ +{ PCG_128BIT_CONSTANT(0x979c9a98d8462005ULL, 0x7d3e9cb6cfe0549bULL), \ +PCG_128BIT_CONSTANT(0x0000000000000001ULL, 0xda3e39cb94b95bdbULL) } +#endif + + /* Representations for the oneseq, mcg, and unique variants */ + + struct pcg_state_8 { + uint8_t state; + }; + + struct pcg_state_16 { + uint16_t state; + }; + + struct pcg_state_32 { + uint32_t state; + }; + + struct pcg_state_64 { + uint64_t state; + }; + +#if PCG_HAS_128BIT_OPS + struct pcg_state_128 { + pcg128_t state; + }; +#endif + + /* Representations setseq variants */ + + struct pcg_state_setseq_8 { + uint8_t state; + uint8_t inc; + }; + + struct pcg_state_setseq_16 { + uint16_t state; + uint16_t inc; + }; + + struct pcg_state_setseq_32 { + uint32_t state; + uint32_t inc; + }; + + struct pcg_state_setseq_64 { + uint64_t state; + uint64_t inc; + }; + +#if PCG_HAS_128BIT_OPS + struct pcg_state_setseq_128 { + pcg128_t state; + pcg128_t inc; + }; +#endif + + /* Multi-step advance functions (jump-ahead, jump-back) */ + + extern uint8_t pcg_advance_lcg_8(uint8_t state, uint8_t delta, uint8_t cur_mult, + uint8_t cur_plus); + extern uint16_t pcg_advance_lcg_16(uint16_t state, uint16_t delta, + uint16_t cur_mult, uint16_t cur_plus); + extern uint32_t pcg_advance_lcg_32(uint32_t state, uint32_t delta, + uint32_t cur_mult, uint32_t cur_plus); + extern uint64_t pcg_advance_lcg_64(uint64_t state, uint64_t delta, + uint64_t cur_mult, uint64_t cur_plus); + +#if PCG_HAS_128BIT_OPS + extern pcg128_t pcg_advance_lcg_128(pcg128_t state, pcg128_t delta, + pcg128_t cur_mult, pcg128_t cur_plus); +#endif + + /* Functions to advance the underlying LCG, one version for each size and + * each style. These functions are considered semi-private. There is rarely + * a good reason to call them directly. + */ + + inline void pcg_oneseq_8_step_r(struct pcg_state_8* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_8 + + PCG_DEFAULT_INCREMENT_8; + } + + inline void pcg_oneseq_8_advance_r(struct pcg_state_8* rng, uint8_t delta) + { + rng->state = pcg_advance_lcg_8(rng->state, delta, PCG_DEFAULT_MULTIPLIER_8, + PCG_DEFAULT_INCREMENT_8); + } + + inline void pcg_mcg_8_step_r(struct pcg_state_8* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_8; + } + + inline void pcg_mcg_8_advance_r(struct pcg_state_8* rng, uint8_t delta) + { + rng->state + = pcg_advance_lcg_8(rng->state, delta, PCG_DEFAULT_MULTIPLIER_8, 0u); + } + + inline void pcg_unique_8_step_r(struct pcg_state_8* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_8 + + (uint8_t)(((intptr_t)rng) | 1u); + } + + inline void pcg_unique_8_advance_r(struct pcg_state_8* rng, uint8_t delta) + { + rng->state = pcg_advance_lcg_8(rng->state, delta, PCG_DEFAULT_MULTIPLIER_8, + (uint8_t)(((intptr_t)rng) | 1u)); + } + + inline void pcg_setseq_8_step_r(struct pcg_state_setseq_8* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_8 + rng->inc; + } + + inline void pcg_setseq_8_advance_r(struct pcg_state_setseq_8* rng, + uint8_t delta) + { + rng->state = pcg_advance_lcg_8(rng->state, delta, PCG_DEFAULT_MULTIPLIER_8, + rng->inc); + } + + inline void pcg_oneseq_16_step_r(struct pcg_state_16* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_16 + + PCG_DEFAULT_INCREMENT_16; + } + + inline void pcg_oneseq_16_advance_r(struct pcg_state_16* rng, uint16_t delta) + { + rng->state = pcg_advance_lcg_16( + rng->state, delta, PCG_DEFAULT_MULTIPLIER_16, PCG_DEFAULT_INCREMENT_16); + } + + inline void pcg_mcg_16_step_r(struct pcg_state_16* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_16; + } + + inline void pcg_mcg_16_advance_r(struct pcg_state_16* rng, uint16_t delta) + { + rng->state + = pcg_advance_lcg_16(rng->state, delta, PCG_DEFAULT_MULTIPLIER_16, 0u); + } + + inline void pcg_unique_16_step_r(struct pcg_state_16* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_16 + + (uint16_t)(((intptr_t)rng) | 1u); + } + + inline void pcg_unique_16_advance_r(struct pcg_state_16* rng, uint16_t delta) + { + rng->state + = pcg_advance_lcg_16(rng->state, delta, PCG_DEFAULT_MULTIPLIER_16, + (uint16_t)(((intptr_t)rng) | 1u)); + } + + inline void pcg_setseq_16_step_r(struct pcg_state_setseq_16* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_16 + rng->inc; + } + + inline void pcg_setseq_16_advance_r(struct pcg_state_setseq_16* rng, + uint16_t delta) + { + rng->state = pcg_advance_lcg_16(rng->state, delta, + PCG_DEFAULT_MULTIPLIER_16, rng->inc); + } + + inline void pcg_oneseq_32_step_r(struct pcg_state_32* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_32 + + PCG_DEFAULT_INCREMENT_32; + } + + inline void pcg_oneseq_32_advance_r(struct pcg_state_32* rng, uint32_t delta) + { + rng->state = pcg_advance_lcg_32( + rng->state, delta, PCG_DEFAULT_MULTIPLIER_32, PCG_DEFAULT_INCREMENT_32); + } + + inline void pcg_mcg_32_step_r(struct pcg_state_32* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_32; + } + + inline void pcg_mcg_32_advance_r(struct pcg_state_32* rng, uint32_t delta) + { + rng->state + = pcg_advance_lcg_32(rng->state, delta, PCG_DEFAULT_MULTIPLIER_32, 0u); + } + + inline void pcg_unique_32_step_r(struct pcg_state_32* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_32 + + (uint32_t)(((intptr_t)rng) | 1u); + } + + inline void pcg_unique_32_advance_r(struct pcg_state_32* rng, uint32_t delta) + { + rng->state + = pcg_advance_lcg_32(rng->state, delta, PCG_DEFAULT_MULTIPLIER_32, + (uint32_t)(((intptr_t)rng) | 1u)); + } + + inline void pcg_setseq_32_step_r(struct pcg_state_setseq_32* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_32 + rng->inc; + } + + inline void pcg_setseq_32_advance_r(struct pcg_state_setseq_32* rng, + uint32_t delta) + { + rng->state = pcg_advance_lcg_32(rng->state, delta, + PCG_DEFAULT_MULTIPLIER_32, rng->inc); + } + + inline void pcg_oneseq_64_step_r(struct pcg_state_64* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_64 + + PCG_DEFAULT_INCREMENT_64; + } + + inline void pcg_oneseq_64_advance_r(struct pcg_state_64* rng, uint64_t delta) + { + rng->state = pcg_advance_lcg_64( + rng->state, delta, PCG_DEFAULT_MULTIPLIER_64, PCG_DEFAULT_INCREMENT_64); + } + + inline void pcg_mcg_64_step_r(struct pcg_state_64* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_64; + } + + inline void pcg_mcg_64_advance_r(struct pcg_state_64* rng, uint64_t delta) + { + rng->state + = pcg_advance_lcg_64(rng->state, delta, PCG_DEFAULT_MULTIPLIER_64, 0u); + } + + inline void pcg_unique_64_step_r(struct pcg_state_64* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_64 + + (uint64_t)(((intptr_t)rng) | 1u); + } + + inline void pcg_unique_64_advance_r(struct pcg_state_64* rng, uint64_t delta) + { + rng->state + = pcg_advance_lcg_64(rng->state, delta, PCG_DEFAULT_MULTIPLIER_64, + (uint64_t)(((intptr_t)rng) | 1u)); + } + + inline void pcg_setseq_64_step_r(struct pcg_state_setseq_64* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_64 + rng->inc; + } + + inline void pcg_setseq_64_advance_r(struct pcg_state_setseq_64* rng, + uint64_t delta) + { + rng->state = pcg_advance_lcg_64(rng->state, delta, + PCG_DEFAULT_MULTIPLIER_64, rng->inc); + } + +#if PCG_HAS_128BIT_OPS + inline void pcg_oneseq_128_step_r(struct pcg_state_128* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_128 + + PCG_DEFAULT_INCREMENT_128; + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_oneseq_128_advance_r(struct pcg_state_128* rng, pcg128_t delta) + { + rng->state + = pcg_advance_lcg_128(rng->state, delta, PCG_DEFAULT_MULTIPLIER_128, + PCG_DEFAULT_INCREMENT_128); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_mcg_128_step_r(struct pcg_state_128* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_128; + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_mcg_128_advance_r(struct pcg_state_128* rng, pcg128_t delta) + { + rng->state = pcg_advance_lcg_128(rng->state, delta, + PCG_DEFAULT_MULTIPLIER_128, 0u); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_unique_128_step_r(struct pcg_state_128* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_128 + + (pcg128_t)(((intptr_t)rng) | 1u); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_unique_128_advance_r(struct pcg_state_128* rng, pcg128_t delta) + { + rng->state + = pcg_advance_lcg_128(rng->state, delta, PCG_DEFAULT_MULTIPLIER_128, + (pcg128_t)(((intptr_t)rng) | 1u)); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_setseq_128_step_r(struct pcg_state_setseq_128* rng) + { + rng->state = rng->state * PCG_DEFAULT_MULTIPLIER_128 + rng->inc; + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_setseq_128_advance_r(struct pcg_state_setseq_128* rng, + pcg128_t delta) + { + rng->state = pcg_advance_lcg_128(rng->state, delta, + PCG_DEFAULT_MULTIPLIER_128, rng->inc); + } +#endif + + /* Functions to seed the RNG state, one version for each size and each + * style. Unlike the step functions, regular users can and should call + * these functions. + */ + + inline void pcg_oneseq_8_srandom_r(struct pcg_state_8* rng, uint8_t initstate) + { + rng->state = 0U; + pcg_oneseq_8_step_r(rng); + rng->state += initstate; + pcg_oneseq_8_step_r(rng); + } + + inline void pcg_mcg_8_srandom_r(struct pcg_state_8* rng, uint8_t initstate) + { + rng->state = initstate | 1u; + } + + inline void pcg_unique_8_srandom_r(struct pcg_state_8* rng, uint8_t initstate) + { + rng->state = 0U; + pcg_unique_8_step_r(rng); + rng->state += initstate; + pcg_unique_8_step_r(rng); + } + + inline void pcg_setseq_8_srandom_r(struct pcg_state_setseq_8* rng, + uint8_t initstate, uint8_t initseq) + { + rng->state = 0U; + rng->inc = (initseq << 1u) | 1u; + pcg_setseq_8_step_r(rng); + rng->state += initstate; + pcg_setseq_8_step_r(rng); + } + + inline void pcg_oneseq_16_srandom_r(struct pcg_state_16* rng, + uint16_t initstate) + { + rng->state = 0U; + pcg_oneseq_16_step_r(rng); + rng->state += initstate; + pcg_oneseq_16_step_r(rng); + } + + inline void pcg_mcg_16_srandom_r(struct pcg_state_16* rng, uint16_t initstate) + { + rng->state = initstate | 1u; + } + + inline void pcg_unique_16_srandom_r(struct pcg_state_16* rng, + uint16_t initstate) + { + rng->state = 0U; + pcg_unique_16_step_r(rng); + rng->state += initstate; + pcg_unique_16_step_r(rng); + } + + inline void pcg_setseq_16_srandom_r(struct pcg_state_setseq_16* rng, + uint16_t initstate, uint16_t initseq) + { + rng->state = 0U; + rng->inc = (initseq << 1u) | 1u; + pcg_setseq_16_step_r(rng); + rng->state += initstate; + pcg_setseq_16_step_r(rng); + } + + inline void pcg_oneseq_32_srandom_r(struct pcg_state_32* rng, + uint32_t initstate) + { + rng->state = 0U; + pcg_oneseq_32_step_r(rng); + rng->state += initstate; + pcg_oneseq_32_step_r(rng); + } + + inline void pcg_mcg_32_srandom_r(struct pcg_state_32* rng, uint32_t initstate) + { + rng->state = initstate | 1u; + } + + inline void pcg_unique_32_srandom_r(struct pcg_state_32* rng, + uint32_t initstate) + { + rng->state = 0U; + pcg_unique_32_step_r(rng); + rng->state += initstate; + pcg_unique_32_step_r(rng); + } + + inline void pcg_setseq_32_srandom_r(struct pcg_state_setseq_32* rng, + uint32_t initstate, uint32_t initseq) + { + rng->state = 0U; + rng->inc = (initseq << 1u) | 1u; + pcg_setseq_32_step_r(rng); + rng->state += initstate; + pcg_setseq_32_step_r(rng); + } + + inline void pcg_oneseq_64_srandom_r(struct pcg_state_64* rng, + uint64_t initstate) + { + rng->state = 0U; + pcg_oneseq_64_step_r(rng); + rng->state += initstate; + pcg_oneseq_64_step_r(rng); + } + + inline void pcg_mcg_64_srandom_r(struct pcg_state_64* rng, uint64_t initstate) + { + rng->state = initstate | 1u; + } + + inline void pcg_unique_64_srandom_r(struct pcg_state_64* rng, + uint64_t initstate) + { + rng->state = 0U; + pcg_unique_64_step_r(rng); + rng->state += initstate; + pcg_unique_64_step_r(rng); + } + + inline void pcg_setseq_64_srandom_r(struct pcg_state_setseq_64* rng, + uint64_t initstate, uint64_t initseq) + { + rng->state = 0U; + rng->inc = (initseq << 1u) | 1u; + pcg_setseq_64_step_r(rng); + rng->state += initstate; + pcg_setseq_64_step_r(rng); + } + +#if PCG_HAS_128BIT_OPS + inline void pcg_oneseq_128_srandom_r(struct pcg_state_128* rng, + pcg128_t initstate) + { + rng->state = 0U; + pcg_oneseq_128_step_r(rng); + rng->state += initstate; + pcg_oneseq_128_step_r(rng); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_mcg_128_srandom_r(struct pcg_state_128* rng, pcg128_t initstate) + { + rng->state = initstate | 1u; + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_unique_128_srandom_r(struct pcg_state_128* rng, + pcg128_t initstate) + { + rng->state = 0U; + pcg_unique_128_step_r(rng); + rng->state += initstate; + pcg_unique_128_step_r(rng); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline void pcg_setseq_128_srandom_r(struct pcg_state_setseq_128* rng, + pcg128_t initstate, pcg128_t initseq) + { + rng->state = 0U; + rng->inc = (initseq << 1u) | 1u; + pcg_setseq_128_step_r(rng); + rng->state += initstate; + pcg_setseq_128_step_r(rng); + } +#endif + + /* Now, finally we create each of the individual generators. We provide + * a random_r function that provides a random number of the appropriate + * type (using the full range of the type) and a boundedrand_r version + * that provides + * + * Implementation notes for boundedrand_r: + * + * To avoid bias, we need to make the range of the RNG a multiple of + * bound, which we do by dropping output less than a threshold. + * Let's consider a 32-bit case... A naive scheme to calculate the + * threshold would be to do + * + * uint32_t threshold = 0x100000000ull % bound; + * + * but 64-bit div/mod is slower than 32-bit div/mod (especially on + * 32-bit platforms). In essence, we do + * + * uint32_t threshold = (0x100000000ull-bound) % bound; + * + * because this version will calculate the same modulus, but the LHS + * value is less than 2^32. + * + * (Note that using modulo is only wise for good RNGs, poorer RNGs + * such as raw LCGs do better using a technique based on division.) + * Empricical tests show that division is preferable to modulus for + * reducting the range of an RNG. It's faster, and sometimes it can + * even be statistically prefereable. + */ + + /* Generation functions for XSH RS */ + + inline uint8_t pcg_oneseq_16_xsh_rs_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_oneseq_16_step_r(rng); + return pcg_output_xsh_rs_16_8(oldstate); + } + + inline uint8_t pcg_oneseq_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_oneseq_16_xsh_rs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_oneseq_32_xsh_rs_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_oneseq_32_step_r(rng); + return pcg_output_xsh_rs_32_16(oldstate); + } + + inline uint16_t pcg_oneseq_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_oneseq_32_xsh_rs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_oneseq_64_xsh_rs_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_oneseq_64_step_r(rng); + return pcg_output_xsh_rs_64_32(oldstate); + } + + inline uint32_t pcg_oneseq_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_oneseq_64_xsh_rs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_oneseq_128_xsh_rs_64_random_r(struct pcg_state_128* rng) + { + pcg_oneseq_128_step_r(rng); + return pcg_output_xsh_rs_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_oneseq_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_oneseq_128_xsh_rs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_unique_16_xsh_rs_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_unique_16_step_r(rng); + return pcg_output_xsh_rs_16_8(oldstate); + } + + inline uint8_t pcg_unique_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_unique_16_xsh_rs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_unique_32_xsh_rs_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_unique_32_step_r(rng); + return pcg_output_xsh_rs_32_16(oldstate); + } + + inline uint16_t pcg_unique_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_unique_32_xsh_rs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_unique_64_xsh_rs_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_unique_64_step_r(rng); + return pcg_output_xsh_rs_64_32(oldstate); + } + + inline uint32_t pcg_unique_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_unique_64_xsh_rs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_unique_128_xsh_rs_64_random_r(struct pcg_state_128* rng) + { + pcg_unique_128_step_r(rng); + return pcg_output_xsh_rs_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_unique_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_unique_128_xsh_rs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_setseq_16_xsh_rs_8_random_r(struct pcg_state_setseq_16* rng) + { + uint16_t oldstate = rng->state; + pcg_setseq_16_step_r(rng); + return pcg_output_xsh_rs_16_8(oldstate); + } + + inline uint8_t + pcg_setseq_16_xsh_rs_8_boundedrand_r(struct pcg_state_setseq_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_setseq_16_xsh_rs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t + pcg_setseq_32_xsh_rs_16_random_r(struct pcg_state_setseq_32* rng) + { + uint32_t oldstate = rng->state; + pcg_setseq_32_step_r(rng); + return pcg_output_xsh_rs_32_16(oldstate); + } + + inline uint16_t + pcg_setseq_32_xsh_rs_16_boundedrand_r(struct pcg_state_setseq_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_setseq_32_xsh_rs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t + pcg_setseq_64_xsh_rs_32_random_r(struct pcg_state_setseq_64* rng) + { + uint64_t oldstate = rng->state; + pcg_setseq_64_step_r(rng); + return pcg_output_xsh_rs_64_32(oldstate); + } + + inline uint32_t + pcg_setseq_64_xsh_rs_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_setseq_64_xsh_rs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsh_rs_64_random_r(struct pcg_state_setseq_128* rng) + { + pcg_setseq_128_step_r(rng); + return pcg_output_xsh_rs_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsh_rs_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_setseq_128_xsh_rs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_mcg_16_xsh_rs_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_mcg_16_step_r(rng); + return pcg_output_xsh_rs_16_8(oldstate); + } + + inline uint8_t pcg_mcg_16_xsh_rs_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_mcg_16_xsh_rs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_mcg_32_xsh_rs_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_mcg_32_step_r(rng); + return pcg_output_xsh_rs_32_16(oldstate); + } + + inline uint16_t pcg_mcg_32_xsh_rs_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_mcg_32_xsh_rs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_mcg_64_xsh_rs_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_mcg_64_step_r(rng); + return pcg_output_xsh_rs_64_32(oldstate); + } + + inline uint32_t pcg_mcg_64_xsh_rs_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_mcg_64_xsh_rs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsh_rs_64_random_r(struct pcg_state_128* rng) + { + pcg_mcg_128_step_r(rng); + return pcg_output_xsh_rs_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsh_rs_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_mcg_128_xsh_rs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + /* Generation functions for XSH RR */ + + inline uint8_t pcg_oneseq_16_xsh_rr_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_oneseq_16_step_r(rng); + return pcg_output_xsh_rr_16_8(oldstate); + } + + inline uint8_t pcg_oneseq_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_oneseq_16_xsh_rr_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_oneseq_32_xsh_rr_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_oneseq_32_step_r(rng); + return pcg_output_xsh_rr_32_16(oldstate); + } + + inline uint16_t pcg_oneseq_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_oneseq_32_xsh_rr_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_oneseq_64_xsh_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_oneseq_64_step_r(rng); + return pcg_output_xsh_rr_64_32(oldstate); + } + + inline uint32_t pcg_oneseq_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_oneseq_64_xsh_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_oneseq_128_xsh_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_oneseq_128_step_r(rng); + return pcg_output_xsh_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_oneseq_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_oneseq_128_xsh_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_unique_16_xsh_rr_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_unique_16_step_r(rng); + return pcg_output_xsh_rr_16_8(oldstate); + } + + inline uint8_t pcg_unique_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_unique_16_xsh_rr_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_unique_32_xsh_rr_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_unique_32_step_r(rng); + return pcg_output_xsh_rr_32_16(oldstate); + } + + inline uint16_t pcg_unique_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_unique_32_xsh_rr_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_unique_64_xsh_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_unique_64_step_r(rng); + return pcg_output_xsh_rr_64_32(oldstate); + } + + inline uint32_t pcg_unique_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_unique_64_xsh_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_unique_128_xsh_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_unique_128_step_r(rng); + return pcg_output_xsh_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_unique_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_unique_128_xsh_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_setseq_16_xsh_rr_8_random_r(struct pcg_state_setseq_16* rng) + { + uint16_t oldstate = rng->state; + pcg_setseq_16_step_r(rng); + return pcg_output_xsh_rr_16_8(oldstate); + } + + inline uint8_t + pcg_setseq_16_xsh_rr_8_boundedrand_r(struct pcg_state_setseq_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_setseq_16_xsh_rr_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t + pcg_setseq_32_xsh_rr_16_random_r(struct pcg_state_setseq_32* rng) + { + uint32_t oldstate = rng->state; + pcg_setseq_32_step_r(rng); + return pcg_output_xsh_rr_32_16(oldstate); + } + + inline uint16_t + pcg_setseq_32_xsh_rr_16_boundedrand_r(struct pcg_state_setseq_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_setseq_32_xsh_rr_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t + pcg_setseq_64_xsh_rr_32_random_r(struct pcg_state_setseq_64* rng) + { + uint64_t oldstate = rng->state; + pcg_setseq_64_step_r(rng); + return pcg_output_xsh_rr_64_32(oldstate); + } + + inline uint32_t + pcg_setseq_64_xsh_rr_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_setseq_64_xsh_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsh_rr_64_random_r(struct pcg_state_setseq_128* rng) + { + pcg_setseq_128_step_r(rng); + return pcg_output_xsh_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsh_rr_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_setseq_128_xsh_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_mcg_16_xsh_rr_8_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_mcg_16_step_r(rng); + return pcg_output_xsh_rr_16_8(oldstate); + } + + inline uint8_t pcg_mcg_16_xsh_rr_8_boundedrand_r(struct pcg_state_16* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_mcg_16_xsh_rr_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_mcg_32_xsh_rr_16_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_mcg_32_step_r(rng); + return pcg_output_xsh_rr_32_16(oldstate); + } + + inline uint16_t pcg_mcg_32_xsh_rr_16_boundedrand_r(struct pcg_state_32* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_mcg_32_xsh_rr_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_mcg_64_xsh_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_mcg_64_step_r(rng); + return pcg_output_xsh_rr_64_32(oldstate); + } + + inline uint32_t pcg_mcg_64_xsh_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_mcg_64_xsh_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsh_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_mcg_128_step_r(rng); + return pcg_output_xsh_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsh_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_mcg_128_xsh_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + /* Generation functions for RXS M XS (no MCG versions because they + * don't make sense when you want to use the entire state) + */ + + inline uint8_t pcg_oneseq_8_rxs_m_xs_8_random_r(struct pcg_state_8* rng) + { + uint8_t oldstate = rng->state; + pcg_oneseq_8_step_r(rng); + return pcg_output_rxs_m_xs_8_8(oldstate); + } + + inline uint8_t pcg_oneseq_8_rxs_m_xs_8_boundedrand_r(struct pcg_state_8* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_oneseq_8_rxs_m_xs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t pcg_oneseq_16_rxs_m_xs_16_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_oneseq_16_step_r(rng); + return pcg_output_rxs_m_xs_16_16(oldstate); + } + + inline uint16_t + pcg_oneseq_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_16* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_oneseq_16_rxs_m_xs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_oneseq_32_rxs_m_xs_32_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_oneseq_32_step_r(rng); + return pcg_output_rxs_m_xs_32_32(oldstate); + } + + inline uint32_t + pcg_oneseq_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_32* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_oneseq_32_rxs_m_xs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint64_t pcg_oneseq_64_rxs_m_xs_64_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_oneseq_64_step_r(rng); + return pcg_output_rxs_m_xs_64_64(oldstate); + } + + inline uint64_t + pcg_oneseq_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_oneseq_64_rxs_m_xs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_oneseq_128_rxs_m_xs_128_random_r(struct pcg_state_128* rng) + { + pcg_oneseq_128_step_r(rng); + return pcg_output_rxs_m_xs_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_oneseq_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_oneseq_128_rxs_m_xs_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint16_t pcg_unique_16_rxs_m_xs_16_random_r(struct pcg_state_16* rng) + { + uint16_t oldstate = rng->state; + pcg_unique_16_step_r(rng); + return pcg_output_rxs_m_xs_16_16(oldstate); + } + + inline uint16_t + pcg_unique_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_16* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_unique_16_rxs_m_xs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t pcg_unique_32_rxs_m_xs_32_random_r(struct pcg_state_32* rng) + { + uint32_t oldstate = rng->state; + pcg_unique_32_step_r(rng); + return pcg_output_rxs_m_xs_32_32(oldstate); + } + + inline uint32_t + pcg_unique_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_32* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_unique_32_rxs_m_xs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint64_t pcg_unique_64_rxs_m_xs_64_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_unique_64_step_r(rng); + return pcg_output_rxs_m_xs_64_64(oldstate); + } + + inline uint64_t + pcg_unique_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_unique_64_rxs_m_xs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_unique_128_rxs_m_xs_128_random_r(struct pcg_state_128* rng) + { + pcg_unique_128_step_r(rng); + return pcg_output_rxs_m_xs_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_unique_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_unique_128_rxs_m_xs_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint8_t pcg_setseq_8_rxs_m_xs_8_random_r(struct pcg_state_setseq_8* rng) + { + uint8_t oldstate = rng->state; + pcg_setseq_8_step_r(rng); + return pcg_output_rxs_m_xs_8_8(oldstate); + } + + inline uint8_t + pcg_setseq_8_rxs_m_xs_8_boundedrand_r(struct pcg_state_setseq_8* rng, + uint8_t bound) + { + uint8_t threshold = ((uint8_t)(-bound)) % bound; + for (;;) { + uint8_t r = pcg_setseq_8_rxs_m_xs_8_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint16_t + pcg_setseq_16_rxs_m_xs_16_random_r(struct pcg_state_setseq_16* rng) + { + uint16_t oldstate = rng->state; + pcg_setseq_16_step_r(rng); + return pcg_output_rxs_m_xs_16_16(oldstate); + } + + inline uint16_t + pcg_setseq_16_rxs_m_xs_16_boundedrand_r(struct pcg_state_setseq_16* rng, + uint16_t bound) + { + uint16_t threshold = ((uint16_t)(-bound)) % bound; + for (;;) { + uint16_t r = pcg_setseq_16_rxs_m_xs_16_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint32_t + pcg_setseq_32_rxs_m_xs_32_random_r(struct pcg_state_setseq_32* rng) + { + uint32_t oldstate = rng->state; + pcg_setseq_32_step_r(rng); + return pcg_output_rxs_m_xs_32_32(oldstate); + } + + inline uint32_t + pcg_setseq_32_rxs_m_xs_32_boundedrand_r(struct pcg_state_setseq_32* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_setseq_32_rxs_m_xs_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + + inline uint64_t + pcg_setseq_64_rxs_m_xs_64_random_r(struct pcg_state_setseq_64* rng) + { + uint64_t oldstate = rng->state; + pcg_setseq_64_step_r(rng); + return pcg_output_rxs_m_xs_64_64(oldstate); + } + + inline uint64_t + pcg_setseq_64_rxs_m_xs_64_boundedrand_r(struct pcg_state_setseq_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_setseq_64_rxs_m_xs_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_setseq_128_rxs_m_xs_128_random_r(struct pcg_state_setseq_128* rng) + { + pcg_setseq_128_step_r(rng); + return pcg_output_rxs_m_xs_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_setseq_128_rxs_m_xs_128_boundedrand_r(struct pcg_state_setseq_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_setseq_128_rxs_m_xs_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + /* Generation functions for XSL RR (only defined for "large" types) */ + + inline uint32_t pcg_oneseq_64_xsl_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_oneseq_64_step_r(rng); + return pcg_output_xsl_rr_64_32(oldstate); + } + + inline uint32_t pcg_oneseq_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_oneseq_64_xsl_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_oneseq_128_xsl_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_oneseq_128_step_r(rng); + return pcg_output_xsl_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_oneseq_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_oneseq_128_xsl_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint32_t pcg_unique_64_xsl_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_unique_64_step_r(rng); + return pcg_output_xsl_rr_64_32(oldstate); + } + + inline uint32_t pcg_unique_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_unique_64_xsl_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_unique_128_xsl_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_unique_128_step_r(rng); + return pcg_output_xsl_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_unique_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_unique_128_xsl_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint32_t + pcg_setseq_64_xsl_rr_32_random_r(struct pcg_state_setseq_64* rng) + { + uint64_t oldstate = rng->state; + pcg_setseq_64_step_r(rng); + return pcg_output_xsl_rr_64_32(oldstate); + } + + inline uint32_t + pcg_setseq_64_xsl_rr_32_boundedrand_r(struct pcg_state_setseq_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_setseq_64_xsl_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsl_rr_64_random_r(struct pcg_state_setseq_128* rng) + { + pcg_setseq_128_step_r(rng); + return pcg_output_xsl_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t + pcg_setseq_128_xsl_rr_64_boundedrand_r(struct pcg_state_setseq_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_setseq_128_xsl_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint32_t pcg_mcg_64_xsl_rr_32_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_mcg_64_step_r(rng); + return pcg_output_xsl_rr_64_32(oldstate); + } + + inline uint32_t pcg_mcg_64_xsl_rr_32_boundedrand_r(struct pcg_state_64* rng, + uint32_t bound) + { + uint32_t threshold = -bound % bound; + for (;;) { + uint32_t r = pcg_mcg_64_xsl_rr_32_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsl_rr_64_random_r(struct pcg_state_128* rng) + { + pcg_mcg_128_step_r(rng); + return pcg_output_xsl_rr_128_64(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline uint64_t pcg_mcg_128_xsl_rr_64_boundedrand_r(struct pcg_state_128* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_mcg_128_xsl_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + /* Generation functions for XSL RR RR (only defined for "large" types) */ + + inline uint64_t pcg_oneseq_64_xsl_rr_rr_64_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_oneseq_64_step_r(rng); + return pcg_output_xsl_rr_rr_64_64(oldstate); + } + + inline uint64_t + pcg_oneseq_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_oneseq_64_xsl_rr_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_oneseq_128_xsl_rr_rr_128_random_r(struct pcg_state_128* rng) + { + pcg_oneseq_128_step_r(rng); + return pcg_output_xsl_rr_rr_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_oneseq_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_oneseq_128_xsl_rr_rr_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint64_t pcg_unique_64_xsl_rr_rr_64_random_r(struct pcg_state_64* rng) + { + uint64_t oldstate = rng->state; + pcg_unique_64_step_r(rng); + return pcg_output_xsl_rr_rr_64_64(oldstate); + } + + inline uint64_t + pcg_unique_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_unique_64_xsl_rr_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t pcg_unique_128_xsl_rr_rr_128_random_r(struct pcg_state_128* rng) + { + pcg_unique_128_step_r(rng); + return pcg_output_xsl_rr_rr_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_unique_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_unique_128_xsl_rr_rr_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + inline uint64_t + pcg_setseq_64_xsl_rr_rr_64_random_r(struct pcg_state_setseq_64* rng) + { + uint64_t oldstate = rng->state; + pcg_setseq_64_step_r(rng); + return pcg_output_xsl_rr_rr_64_64(oldstate); + } + + inline uint64_t + pcg_setseq_64_xsl_rr_rr_64_boundedrand_r(struct pcg_state_setseq_64* rng, + uint64_t bound) + { + uint64_t threshold = -bound % bound; + for (;;) { + uint64_t r = pcg_setseq_64_xsl_rr_rr_64_random_r(rng); + if (r >= threshold) + return r % bound; + } + } + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_setseq_128_xsl_rr_rr_128_random_r(struct pcg_state_setseq_128* rng) + { + pcg_setseq_128_step_r(rng); + return pcg_output_xsl_rr_rr_128_128(rng->state); + } +#endif + +#if PCG_HAS_128BIT_OPS + inline pcg128_t + pcg_setseq_128_xsl_rr_rr_128_boundedrand_r(struct pcg_state_setseq_128* rng, + pcg128_t bound) + { + pcg128_t threshold = -bound % bound; + for (;;) { + pcg128_t r = pcg_setseq_128_xsl_rr_rr_128_random_r(rng); + if (r >= threshold) + return r % bound; + } + } +#endif + + //// Typedefs + typedef struct pcg_state_setseq_64 pcg32_random_t; + typedef struct pcg_state_64 pcg32s_random_t; + typedef struct pcg_state_64 pcg32u_random_t; + typedef struct pcg_state_64 pcg32f_random_t; + //// random_r +#define pcg32_random_r pcg_setseq_64_xsh_rr_32_random_r +#define pcg32s_random_r pcg_oneseq_64_xsh_rr_32_random_r +#define pcg32u_random_r pcg_unique_64_xsh_rr_32_random_r +#define pcg32f_random_r pcg_mcg_64_xsh_rs_32_random_r + //// boundedrand_r +#define pcg32_boundedrand_r pcg_setseq_64_xsh_rr_32_boundedrand_r +#define pcg32s_boundedrand_r pcg_oneseq_64_xsh_rr_32_boundedrand_r +#define pcg32u_boundedrand_r pcg_unique_64_xsh_rr_32_boundedrand_r +#define pcg32f_boundedrand_r pcg_mcg_64_xsh_rs_32_boundedrand_r + //// srandom_r +#define pcg32_srandom_r pcg_setseq_64_srandom_r +#define pcg32s_srandom_r pcg_oneseq_64_srandom_r +#define pcg32u_srandom_r pcg_unique_64_srandom_r +#define pcg32f_srandom_r pcg_mcg_64_srandom_r + //// advance_r +#define pcg32_advance_r pcg_setseq_64_advance_r +#define pcg32s_advance_r pcg_oneseq_64_advance_r +#define pcg32u_advance_r pcg_unique_64_advance_r +#define pcg32f_advance_r pcg_mcg_64_advance_r + +#if PCG_HAS_128BIT_OPS + //// Typedefs + typedef struct pcg_state_setseq_128 pcg64_random_t; + typedef struct pcg_state_128 pcg64s_random_t; + typedef struct pcg_state_128 pcg64u_random_t; + typedef struct pcg_state_128 pcg64f_random_t; + //// random_r +#define pcg64_random_r pcg_setseq_128_xsl_rr_64_random_r +#define pcg64s_random_r pcg_oneseq_128_xsl_rr_64_random_r +#define pcg64u_random_r pcg_unique_128_xsl_rr_64_random_r +#define pcg64f_random_r pcg_mcg_128_xsl_rr_64_random_r + //// boundedrand_r +#define pcg64_boundedrand_r pcg_setseq_128_xsl_rr_64_boundedrand_r +#define pcg64s_boundedrand_r pcg_oneseq_128_xsl_rr_64_boundedrand_r +#define pcg64u_boundedrand_r pcg_unique_128_xsl_rr_64_boundedrand_r +#define pcg64f_boundedrand_r pcg_mcg_128_xsl_rr_64_boundedrand_r + //// srandom_r +#define pcg64_srandom_r pcg_setseq_128_srandom_r +#define pcg64s_srandom_r pcg_oneseq_128_srandom_r +#define pcg64u_srandom_r pcg_unique_128_srandom_r +#define pcg64f_srandom_r pcg_mcg_128_srandom_r + //// advance_r +#define pcg64_advance_r pcg_setseq_128_advance_r +#define pcg64s_advance_r pcg_oneseq_128_advance_r +#define pcg64u_advance_r pcg_unique_128_advance_r +#define pcg64f_advance_r pcg_mcg_128_advance_r +#endif + + //// Typedefs + typedef struct pcg_state_8 pcg8si_random_t; + typedef struct pcg_state_16 pcg16si_random_t; + typedef struct pcg_state_32 pcg32si_random_t; + typedef struct pcg_state_64 pcg64si_random_t; + //// random_r +#define pcg8si_random_r pcg_oneseq_8_rxs_m_xs_8_random_r +#define pcg16si_random_r pcg_oneseq_16_rxs_m_xs_16_random_r +#define pcg32si_random_r pcg_oneseq_32_rxs_m_xs_32_random_r +#define pcg64si_random_r pcg_oneseq_64_rxs_m_xs_64_random_r + //// boundedrand_r +#define pcg8si_boundedrand_r pcg_oneseq_8_rxs_m_xs_8_boundedrand_r +#define pcg16si_boundedrand_r pcg_oneseq_16_rxs_m_xs_16_boundedrand_r +#define pcg32si_boundedrand_r pcg_oneseq_32_rxs_m_xs_32_boundedrand_r +#define pcg64si_boundedrand_r pcg_oneseq_64_rxs_m_xs_64_boundedrand_r + //// srandom_r +#define pcg8si_srandom_r pcg_oneseq_8_srandom_r +#define pcg16si_srandom_r pcg_oneseq_16_srandom_r +#define pcg32si_srandom_r pcg_oneseq_32_srandom_r +#define pcg64si_srandom_r pcg_oneseq_64_srandom_r + //// advance_r +#define pcg8si_advance_r pcg_oneseq_8_advance_r +#define pcg16si_advance_r pcg_oneseq_16_advance_r +#define pcg32si_advance_r pcg_oneseq_32_advance_r +#define pcg64si_advance_r pcg_oneseq_64_advance_r + +#if PCG_HAS_128BIT_OPS + typedef struct pcg_state_128 pcg128si_random_t; +#define pcg128si_random_r pcg_oneseq_128_rxs_m_xs_128_random_r +#define pcg128si_boundedrand_r pcg_oneseq_128_rxs_m_xs_128_boundedrand_r +#define pcg128si_srandom_r pcg_oneseq_128_srandom_r +#define pcg128si_advance_r pcg_oneseq_128_advance_r +#endif + + //// Typedefs + typedef struct pcg_state_setseq_8 pcg8i_random_t; + typedef struct pcg_state_setseq_16 pcg16i_random_t; + typedef struct pcg_state_setseq_32 pcg32i_random_t; + typedef struct pcg_state_setseq_64 pcg64i_random_t; + //// random_r +#define pcg8i_random_r pcg_setseq_8_rxs_m_xs_8_random_r +#define pcg16i_random_r pcg_setseq_16_rxs_m_xs_16_random_r +#define pcg32i_random_r pcg_setseq_32_rxs_m_xs_32_random_r +#define pcg64i_random_r pcg_setseq_64_rxs_m_xs_64_random_r + //// boundedrand_r +#define pcg8i_boundedrand_r pcg_setseq_8_rxs_m_xs_8_boundedrand_r +#define pcg16i_boundedrand_r pcg_setseq_16_rxs_m_xs_16_boundedrand_r +#define pcg32i_boundedrand_r pcg_setseq_32_rxs_m_xs_32_boundedrand_r +#define pcg64i_boundedrand_r pcg_setseq_64_rxs_m_xs_64_boundedrand_r + //// srandom_r +#define pcg8i_srandom_r pcg_setseq_8_srandom_r +#define pcg16i_srandom_r pcg_setseq_16_srandom_r +#define pcg32i_srandom_r pcg_setseq_32_srandom_r +#define pcg64i_srandom_r pcg_setseq_64_srandom_r + //// advance_r +#define pcg8i_advance_r pcg_setseq_8_advance_r +#define pcg16i_advance_r pcg_setseq_16_advance_r +#define pcg32i_advance_r pcg_setseq_32_advance_r +#define pcg64i_advance_r pcg_setseq_64_advance_r + +#if PCG_HAS_128BIT_OPS + typedef struct pcg_state_setseq_128 pcg128i_random_t; +#define pcg128i_random_r pcg_setseq_128_rxs_m_xs_128_random_r +#define pcg128i_boundedrand_r pcg_setseq_128_rxs_m_xs_128_boundedrand_r +#define pcg128i_srandom_r pcg_setseq_128_srandom_r +#define pcg128i_advance_r pcg_setseq_128_advance_r +#endif + + extern uint32_t pcg32_random(); + extern uint32_t pcg32_boundedrand(uint32_t bound); + extern void pcg32_srandom(uint64_t seed, uint64_t seq); + extern void pcg32_advance(uint64_t delta); + +#if PCG_HAS_128BIT_OPS + extern uint64_t pcg64_random(); + extern uint64_t pcg64_boundedrand(uint64_t bound); + extern void pcg64_srandom(pcg128_t seed, pcg128_t seq); + extern void pcg64_advance(pcg128_t delta); +#endif + + /* + * Static initialization constants (if you can't call srandom for some + * bizarre reason). + */ + +#define PCG32_INITIALIZER PCG_STATE_SETSEQ_64_INITIALIZER +#define PCG32U_INITIALIZER PCG_STATE_UNIQUE_64_INITIALIZER +#define PCG32S_INITIALIZER PCG_STATE_ONESEQ_64_INITIALIZER +#define PCG32F_INITIALIZER PCG_STATE_MCG_64_INITIALIZER + +#if PCG_HAS_128BIT_OPS +#define PCG64_INITIALIZER PCG_STATE_SETSEQ_128_INITIALIZER +#define PCG64U_INITIALIZER PCG_STATE_UNIQUE_128_INITIALIZER +#define PCG64S_INITIALIZER PCG_STATE_ONESEQ_128_INITIALIZER +#define PCG64F_INITIALIZER PCG_STATE_MCG_128_INITIALIZER +#endif + +#if PCG_HAS_128BIT_OPS +#define PCG8SI_INITIALIZER PCG_STATE_ONESEQ_8_INITIALIZER +#define PCG16SI_INITIALIZER PCG_STATE_ONESEQ_16_INITIALIZER +#define PCG32SI_INITIALIZER PCG_STATE_ONESEQ_32_INITIALIZER +#define PCG64SI_INITIALIZER PCG_STATE_ONESEQ_64_INITIALIZER +#define PCG128SI_INITIALIZER PCG_STATE_ONESEQ_128_INITIALIZER +#endif + +#if PCG_HAS_128BIT_OPS +#define PCG8I_INITIALIZER PCG_STATE_SETSEQ_8_INITIALIZER +#define PCG16I_INITIALIZER PCG_STATE_SETSEQ_16_INITIALIZER +#define PCG32I_INITIALIZER PCG_STATE_SETSEQ_32_INITIALIZER +#define PCG64I_INITIALIZER PCG_STATE_SETSEQ_64_INITIALIZER +#define PCG128I_INITIALIZER PCG_STATE_SETSEQ_128_INITIALIZER +#endif + +#if __cplusplus +} +#endif + +#endif // PCG_VARIANTS_H_INCLUDED + diff --git a/src/haversine_generator/libs/stb_sprintf.h b/src/haversine_generator/libs/stb_sprintf.h new file mode 100644 index 0000000..ca432a6 --- /dev/null +++ b/src/haversine_generator/libs/stb_sprintf.h @@ -0,0 +1,1906 @@ +// stb_sprintf - v1.10 - public domain snprintf() implementation +// originally by Jeff Roberts / RAD Game Tools, 2015/10/20 +// http://github.com/nothings/stb +// +// allowed types: sc uidBboXx p AaGgEef n +// lengths : hh h ll j z t I64 I32 I +// +// Contributors: +// Fabian "ryg" Giesen (reformatting) +// github:aganm (attribute format) +// +// Contributors (bugfixes): +// github:d26435 +// github:trex78 +// github:account-login +// Jari Komppa (SI suffixes) +// Rohit Nirmal +// Marcin Wojdyr +// Leonard Ritter +// Stefano Zanotti +// Adam Allison +// Arvid Gerstmann +// Markus Kolb +// +// LICENSE: +// +// See end of file for license information. + +#ifndef STB_SPRINTF_H_INCLUDE +#define STB_SPRINTF_H_INCLUDE + +/* +Single file sprintf replacement. + +Originally written by Jeff Roberts at RAD Game Tools - 2015/10/20. +Hereby placed in public domain. + +This is a full sprintf replacement that supports everything that +the C runtime sprintfs support, including float/double, 64-bit integers, +hex floats, field parameters (%*.*d stuff), length reads backs, etc. + +Why would you need this if sprintf already exists? Well, first off, +it's *much* faster (see below). It's also much smaller than the CRT +versions code-space-wise. We've also added some simple improvements +that are super handy (commas in thousands, callbacks at buffer full, +for example). Finally, the format strings for MSVC and GCC differ +for 64-bit integers (among other small things), so this lets you use +the same format strings in cross platform code. + +It uses the standard single file trick of being both the header file +and the source itself. If you just include it normally, you just get +the header file function definitions. To get the code, you include +it from a C or C++ file and define STB_SPRINTF_IMPLEMENTATION first. + +It only uses va_args macros from the C runtime to do it's work. It +does cast doubles to S64s and shifts and divides U64s, which does +drag in CRT code on most platforms. + +It compiles to roughly 8K with float support, and 4K without. +As a comparison, when using MSVC static libs, calling sprintf drags +in 16K. + +API: +==== +int stbsp_sprintf( char * buf, char const * fmt, ... ) +int stbsp_snprintf( char * buf, int count, char const * fmt, ... ) + Convert an arg list into a buffer. stbsp_snprintf always returns + a zero-terminated string (unlike regular snprintf). + +int stbsp_vsprintf( char * buf, char const * fmt, va_list va ) +int stbsp_vsnprintf( char * buf, int count, char const * fmt, va_list va ) + Convert a va_list arg list into a buffer. stbsp_vsnprintf always returns + a zero-terminated string (unlike regular snprintf). + +int stbsp_vsprintfcb( STBSP_SPRINTFCB * callback, void * user, char * buf, char const * fmt, va_list va ) + typedef char * STBSP_SPRINTFCB( char const * buf, void * user, int len ); + Convert into a buffer, calling back every STB_SPRINTF_MIN chars. + Your callback can then copy the chars out, print them or whatever. + This function is actually the workhorse for everything else. + The buffer you pass in must hold at least STB_SPRINTF_MIN characters. + // you return the next buffer to use or 0 to stop converting + +void stbsp_set_separators( char comma, char period ) + Set the comma and period characters to use. + +FLOATS/DOUBLES: +=============== +This code uses a internal float->ascii conversion method that uses +doubles with error correction (double-doubles, for ~105 bits of +precision). This conversion is round-trip perfect - that is, an atof +of the values output here will give you the bit-exact double back. + +One difference is that our insignificant digits will be different than +with MSVC or GCC (but they don't match each other either). We also +don't attempt to find the minimum length matching float (pre-MSVC15 +doesn't either). + +If you don't need float or doubles at all, define STB_SPRINTF_NOFLOAT +and you'll save 4K of code space. + +64-BIT INTS: +============ +This library also supports 64-bit integers and you can use MSVC style or +GCC style indicators (%I64d or %lld). It supports the C99 specifiers +for size_t and ptr_diff_t (%jd %zd) as well. + +EXTRAS: +======= +Like some GCCs, for integers and floats, you can use a ' (single quote) +specifier and commas will be inserted on the thousands: "%'d" on 12345 +would print 12,345. + +For integers and floats, you can use a "$" specifier and the number +will be converted to float and then divided to get kilo, mega, giga or +tera and then printed, so "%$d" 1000 is "1.0 k", "%$.2d" 2536000 is +"2.53 M", etc. For byte values, use two $:s, like "%$$d" to turn +2536000 to "2.42 Mi". If you prefer JEDEC suffixes to SI ones, use three +$:s: "%$$$d" -> "2.42 M". To remove the space between the number and the +suffix, add "_" specifier: "%_$d" -> "2.53M". + +In addition to octal and hexadecimal conversions, you can print +integers in binary: "%b" for 256 would print 100. + +PERFORMANCE vs MSVC 2008 32-/64-bit (GCC is even slower than MSVC): +=================================================================== +"%d" across all 32-bit ints (4.8x/4.0x faster than 32-/64-bit MSVC) +"%24d" across all 32-bit ints (4.5x/4.2x faster) +"%x" across all 32-bit ints (4.5x/3.8x faster) +"%08x" across all 32-bit ints (4.3x/3.8x faster) +"%f" across e-10 to e+10 floats (7.3x/6.0x faster) +"%e" across e-10 to e+10 floats (8.1x/6.0x faster) +"%g" across e-10 to e+10 floats (10.0x/7.1x faster) +"%f" for values near e-300 (7.9x/6.5x faster) +"%f" for values near e+300 (10.0x/9.1x faster) +"%e" for values near e-300 (10.1x/7.0x faster) +"%e" for values near e+300 (9.2x/6.0x faster) +"%.320f" for values near e-300 (12.6x/11.2x faster) +"%a" for random values (8.6x/4.3x faster) +"%I64d" for 64-bits with 32-bit values (4.8x/3.4x faster) +"%I64d" for 64-bits > 32-bit values (4.9x/5.5x faster) +"%s%s%s" for 64 char strings (7.1x/7.3x faster) +"...512 char string..." ( 35.0x/32.5x faster!) +*/ + +#if defined(__clang__) + #if defined(__has_feature) && defined(__has_attribute) + #if __has_feature(address_sanitizer) + #if __has_attribute(__no_sanitize__) + #define STBSP__ASAN __attribute__((__no_sanitize__("address"))) + #elif __has_attribute(__no_sanitize_address__) + #define STBSP__ASAN __attribute__((__no_sanitize_address__)) + #elif __has_attribute(__no_address_safety_analysis__) + #define STBSP__ASAN __attribute__((__no_address_safety_analysis__)) + #endif + #endif + #endif +#elif defined(__GNUC__) && (__GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) + #if defined(__SANITIZE_ADDRESS__) && __SANITIZE_ADDRESS__ + #define STBSP__ASAN __attribute__((__no_sanitize_address__)) + #endif +#endif + +#ifndef STBSP__ASAN +#define STBSP__ASAN +#endif + +#ifdef STB_SPRINTF_STATIC +#define STBSP__PUBLICDEC static +#define STBSP__PUBLICDEF static STBSP__ASAN +#else +#ifdef __cplusplus +#define STBSP__PUBLICDEC extern "C" +#define STBSP__PUBLICDEF extern "C" STBSP__ASAN +#else +#define STBSP__PUBLICDEC extern +#define STBSP__PUBLICDEF STBSP__ASAN +#endif +#endif + +#if defined(__has_attribute) + #if __has_attribute(format) + #define STBSP__ATTRIBUTE_FORMAT(fmt,va) __attribute__((format(printf,fmt,va))) + #endif +#endif + +#ifndef STBSP__ATTRIBUTE_FORMAT +#define STBSP__ATTRIBUTE_FORMAT(fmt,va) +#endif + +#ifdef _MSC_VER +#define STBSP__NOTUSED(v) (void)(v) +#else +#define STBSP__NOTUSED(v) (void)sizeof(v) +#endif + +#include <stdarg.h> // for va_arg(), va_list() +#include <stddef.h> // size_t, ptrdiff_t + +#ifndef STB_SPRINTF_MIN +#define STB_SPRINTF_MIN 512 // how many characters per callback +#endif +typedef char *STBSP_SPRINTFCB(const char *buf, void *user, int len); + +#ifndef STB_SPRINTF_DECORATE +#define STB_SPRINTF_DECORATE(name) stbsp_##name // define this before including if you want to change the names +#endif + +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsprintf)(char *buf, char const *fmt, va_list va); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsnprintf)(char *buf, int count, char const *fmt, va_list va); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(sprintf)(char *buf, char const *fmt, ...) STBSP__ATTRIBUTE_FORMAT(2,3); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(snprintf)(char *buf, int count, char const *fmt, ...) STBSP__ATTRIBUTE_FORMAT(3,4); + +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsprintfcb)(STBSP_SPRINTFCB *callback, void *user, char *buf, char const *fmt, va_list va); +STBSP__PUBLICDEC void STB_SPRINTF_DECORATE(set_separators)(char comma, char period); + +#endif // STB_SPRINTF_H_INCLUDE + +#ifdef STB_SPRINTF_IMPLEMENTATION + +#define stbsp__uint32 unsigned int +#define stbsp__int32 signed int + +#ifdef _MSC_VER +#define stbsp__uint64 unsigned __int64 +#define stbsp__int64 signed __int64 +#else +#define stbsp__uint64 unsigned long long +#define stbsp__int64 signed long long +#endif +#define stbsp__uint16 unsigned short + +#ifndef stbsp__uintptr +#if defined(__ppc64__) || defined(__powerpc64__) || defined(__aarch64__) || defined(_M_X64) || defined(__x86_64__) || defined(__x86_64) || defined(__s390x__) +#define stbsp__uintptr stbsp__uint64 +#else +#define stbsp__uintptr stbsp__uint32 +#endif +#endif + +#ifndef STB_SPRINTF_MSVC_MODE // used for MSVC2013 and earlier (MSVC2015 matches GCC) +#if defined(_MSC_VER) && (_MSC_VER < 1900) +#define STB_SPRINTF_MSVC_MODE +#endif +#endif + +#ifdef STB_SPRINTF_NOUNALIGNED // define this before inclusion to force stbsp_sprintf to always use aligned accesses +#define STBSP__UNALIGNED(code) +#else +#define STBSP__UNALIGNED(code) code +#endif + +#ifndef STB_SPRINTF_NOFLOAT +// internal float utility functions +static stbsp__int32 stbsp__real_to_str(char const **start, stbsp__uint32 *len, char *out, stbsp__int32 *decimal_pos, double value, stbsp__uint32 frac_digits); +static stbsp__int32 stbsp__real_to_parts(stbsp__int64 *bits, stbsp__int32 *expo, double value); +#define STBSP__SPECIAL 0x7000 +#endif + +static char stbsp__period = '.'; +static char stbsp__comma = ','; +static struct +{ + short temp; // force next field to be 2-byte aligned + char pair[201]; +} stbsp__digitpair = +{ + 0, + "00010203040506070809101112131415161718192021222324" + "25262728293031323334353637383940414243444546474849" + "50515253545556575859606162636465666768697071727374" + "75767778798081828384858687888990919293949596979899" +}; + +STBSP__PUBLICDEF void STB_SPRINTF_DECORATE(set_separators)(char pcomma, char pperiod) +{ + stbsp__period = pperiod; + stbsp__comma = pcomma; +} + +#define STBSP__LEFTJUST 1 +#define STBSP__LEADINGPLUS 2 +#define STBSP__LEADINGSPACE 4 +#define STBSP__LEADING_0X 8 +#define STBSP__LEADINGZERO 16 +#define STBSP__INTMAX 32 +#define STBSP__TRIPLET_COMMA 64 +#define STBSP__NEGATIVE 128 +#define STBSP__METRIC_SUFFIX 256 +#define STBSP__HALFWIDTH 512 +#define STBSP__METRIC_NOSPACE 1024 +#define STBSP__METRIC_1024 2048 +#define STBSP__METRIC_JEDEC 4096 + +static void stbsp__lead_sign(stbsp__uint32 fl, char *sign) +{ + sign[0] = 0; + if (fl & STBSP__NEGATIVE) { + sign[0] = 1; + sign[1] = '-'; + } else if (fl & STBSP__LEADINGSPACE) { + sign[0] = 1; + sign[1] = ' '; + } else if (fl & STBSP__LEADINGPLUS) { + sign[0] = 1; + sign[1] = '+'; + } +} + +static STBSP__ASAN stbsp__uint32 stbsp__strlen_limited(char const *s, stbsp__uint32 limit) +{ + char const * sn = s; + + // get up to 4-byte alignment + for (;;) { + if (((stbsp__uintptr)sn & 3) == 0) + break; + + if (!limit || *sn == 0) + return (stbsp__uint32)(sn - s); + + ++sn; + --limit; + } + + // scan over 4 bytes at a time to find terminating 0 + // this will intentionally scan up to 3 bytes past the end of buffers, + // but becase it works 4B aligned, it will never cross page boundaries + // (hence the STBSP__ASAN markup; the over-read here is intentional + // and harmless) + while (limit >= 4) { + stbsp__uint32 v = *(stbsp__uint32 *)sn; + // bit hack to find if there's a 0 byte in there + if ((v - 0x01010101) & (~v) & 0x80808080UL) + break; + + sn += 4; + limit -= 4; + } + + // handle the last few characters to find actual size + while (limit && *sn) { + ++sn; + --limit; + } + + return (stbsp__uint32)(sn - s); +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(vsprintfcb)(STBSP_SPRINTFCB *callback, void *user, char *buf, char const *fmt, va_list va) +{ + static char hex[] = "0123456789abcdefxp"; + static char hexu[] = "0123456789ABCDEFXP"; + char *bf; + char const *f; + int tlen = 0; + + bf = buf; + f = fmt; + for (;;) { + stbsp__int32 fw, pr, tz; + stbsp__uint32 fl; + + // macros for the callback buffer stuff + #define stbsp__chk_cb_bufL(bytes) \ + { \ + int len = (int)(bf - buf); \ + if ((len + (bytes)) >= STB_SPRINTF_MIN) { \ + tlen += len; \ + if (0 == (bf = buf = callback(buf, user, len))) \ + goto done; \ + } \ + } + #define stbsp__chk_cb_buf(bytes) \ + { \ + if (callback) { \ + stbsp__chk_cb_bufL(bytes); \ + } \ + } + #define stbsp__flush_cb() \ + { \ + stbsp__chk_cb_bufL(STB_SPRINTF_MIN - 1); \ + } // flush if there is even one byte in the buffer + #define stbsp__cb_buf_clamp(cl, v) \ + cl = v; \ + if (callback) { \ + int lg = STB_SPRINTF_MIN - (int)(bf - buf); \ + if (cl > lg) \ + cl = lg; \ + } + + // fast copy everything up to the next % (or end of string) + for (;;) { + while (((stbsp__uintptr)f) & 3) { + schk1: + if (f[0] == '%') + goto scandd; + schk2: + if (f[0] == 0) + goto endfmt; + stbsp__chk_cb_buf(1); + *bf++ = f[0]; + ++f; + } + for (;;) { + // Check if the next 4 bytes contain %(0x25) or end of string. + // Using the 'hasless' trick: + // https://graphics.stanford.edu/~seander/bithacks.html#HasLessInWord + stbsp__uint32 v, c; + v = *(stbsp__uint32 *)f; + c = (~v) & 0x80808080; + if (((v ^ 0x25252525) - 0x01010101) & c) + goto schk1; + if ((v - 0x01010101) & c) + goto schk2; + if (callback) + if ((STB_SPRINTF_MIN - (int)(bf - buf)) < 4) + goto schk1; + #ifdef STB_SPRINTF_NOUNALIGNED + if(((stbsp__uintptr)bf) & 3) { + bf[0] = f[0]; + bf[1] = f[1]; + bf[2] = f[2]; + bf[3] = f[3]; + } else + #endif + { + *(stbsp__uint32 *)bf = v; + } + bf += 4; + f += 4; + } + } + scandd: + + ++f; + + // ok, we have a percent, read the modifiers first + fw = 0; + pr = -1; + fl = 0; + tz = 0; + + // flags + for (;;) { + switch (f[0]) { + // if we have left justify + case '-': + fl |= STBSP__LEFTJUST; + ++f; + continue; + // if we have leading plus + case '+': + fl |= STBSP__LEADINGPLUS; + ++f; + continue; + // if we have leading space + case ' ': + fl |= STBSP__LEADINGSPACE; + ++f; + continue; + // if we have leading 0x + case '#': + fl |= STBSP__LEADING_0X; + ++f; + continue; + // if we have thousand commas + case '\'': + fl |= STBSP__TRIPLET_COMMA; + ++f; + continue; + // if we have kilo marker (none->kilo->kibi->jedec) + case '$': + if (fl & STBSP__METRIC_SUFFIX) { + if (fl & STBSP__METRIC_1024) { + fl |= STBSP__METRIC_JEDEC; + } else { + fl |= STBSP__METRIC_1024; + } + } else { + fl |= STBSP__METRIC_SUFFIX; + } + ++f; + continue; + // if we don't want space between metric suffix and number + case '_': + fl |= STBSP__METRIC_NOSPACE; + ++f; + continue; + // if we have leading zero + case '0': + fl |= STBSP__LEADINGZERO; + ++f; + goto flags_done; + default: goto flags_done; + } + } + flags_done: + + // get the field width + if (f[0] == '*') { + fw = va_arg(va, stbsp__uint32); + ++f; + } else { + while ((f[0] >= '0') && (f[0] <= '9')) { + fw = fw * 10 + f[0] - '0'; + f++; + } + } + // get the precision + if (f[0] == '.') { + ++f; + if (f[0] == '*') { + pr = va_arg(va, stbsp__uint32); + ++f; + } else { + pr = 0; + while ((f[0] >= '0') && (f[0] <= '9')) { + pr = pr * 10 + f[0] - '0'; + f++; + } + } + } + + // handle integer size overrides + switch (f[0]) { + // are we halfwidth? + case 'h': + fl |= STBSP__HALFWIDTH; + ++f; + if (f[0] == 'h') + ++f; // QUARTERWIDTH + break; + // are we 64-bit (unix style) + case 'l': + fl |= ((sizeof(long) == 8) ? STBSP__INTMAX : 0); + ++f; + if (f[0] == 'l') { + fl |= STBSP__INTMAX; + ++f; + } + break; + // are we 64-bit on intmax? (c99) + case 'j': + fl |= (sizeof(size_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + // are we 64-bit on size_t or ptrdiff_t? (c99) + case 'z': + fl |= (sizeof(ptrdiff_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + case 't': + fl |= (sizeof(ptrdiff_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + // are we 64-bit (msft style) + case 'I': + if ((f[1] == '6') && (f[2] == '4')) { + fl |= STBSP__INTMAX; + f += 3; + } else if ((f[1] == '3') && (f[2] == '2')) { + f += 3; + } else { + fl |= ((sizeof(void *) == 8) ? STBSP__INTMAX : 0); + ++f; + } + break; + default: break; + } + + // handle each replacement + switch (f[0]) { + #define STBSP__NUMSZ 512 // big enough for e308 (with commas) or e-307 + char num[STBSP__NUMSZ]; + char lead[8]; + char tail[8]; + char *s; + char const *h; + stbsp__uint32 l, n, cs; + stbsp__uint64 n64; +#ifndef STB_SPRINTF_NOFLOAT + double fv; +#endif + stbsp__int32 dp; + char const *sn; + + case 's': + // get the string + s = va_arg(va, char *); + if (s == 0) + s = (char *)"null"; + // get the length, limited to desired precision + // always limit to ~0u chars since our counts are 32b + l = stbsp__strlen_limited(s, (pr >= 0) ? pr : ~0u); + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + // copy the string in + goto scopy; + + case 'c': // char + // get the character + s = num + STBSP__NUMSZ - 1; + *s = (char)va_arg(va, int); + l = 1; + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + goto scopy; + + case 'n': // weird write-bytes specifier + { + int *d = va_arg(va, int *); + *d = tlen + (int)(bf - buf); + } break; + +#ifdef STB_SPRINTF_NOFLOAT + case 'A': // float + case 'a': // hex float + case 'G': // float + case 'g': // float + case 'E': // float + case 'e': // float + case 'f': // float + va_arg(va, double); // eat it + s = (char *)"No float"; + l = 8; + lead[0] = 0; + tail[0] = 0; + pr = 0; + cs = 0; + STBSP__NOTUSED(dp); + goto scopy; +#else + case 'A': // hex float + case 'a': // hex float + h = (f[0] == 'A') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_parts((stbsp__int64 *)&n64, &dp, fv)) + fl |= STBSP__NEGATIVE; + + s = num + 64; + + stbsp__lead_sign(fl, lead); + + if (dp == -1023) + dp = (n64) ? -1022 : 0; + else + n64 |= (((stbsp__uint64)1) << 52); + n64 <<= (64 - 56); + if (pr < 15) + n64 += ((((stbsp__uint64)8) << 56) >> (pr * 4)); +// add leading chars + +#ifdef STB_SPRINTF_MSVC_MODE + *s++ = '0'; + *s++ = 'x'; +#else + lead[1 + lead[0]] = '0'; + lead[2 + lead[0]] = 'x'; + lead[0] += 2; +#endif + *s++ = h[(n64 >> 60) & 15]; + n64 <<= 4; + if (pr) + *s++ = stbsp__period; + sn = s; + + // print the bits + n = pr; + if (n > 13) + n = 13; + if (pr > (stbsp__int32)n) + tz = pr - n; + pr = 0; + while (n--) { + *s++ = h[(n64 >> 60) & 15]; + n64 <<= 4; + } + + // print the expo + tail[1] = h[17]; + if (dp < 0) { + tail[2] = '-'; + dp = -dp; + } else + tail[2] = '+'; + n = (dp >= 1000) ? 6 : ((dp >= 100) ? 5 : ((dp >= 10) ? 4 : 3)); + tail[0] = (char)n; + for (;;) { + tail[n] = '0' + dp % 10; + if (n <= 3) + break; + --n; + dp /= 10; + } + + dp = (int)(s - sn); + l = (int)(s - (num + 64)); + s = num + 64; + cs = 1 + (3 << 24); + goto scopy; + + case 'G': // float + case 'g': // float + h = (f[0] == 'G') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; + else if (pr == 0) + pr = 1; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, (pr - 1) | 0x80000000)) + fl |= STBSP__NEGATIVE; + + // clamp the precision and delete extra zeros after clamp + n = pr; + if (l > (stbsp__uint32)pr) + l = pr; + while ((l > 1) && (pr) && (sn[l - 1] == '0')) { + --pr; + --l; + } + + // should we use %e + if ((dp <= -4) || (dp > (stbsp__int32)n)) { + if (pr > (stbsp__int32)l) + pr = l - 1; + else if (pr) + --pr; // when using %e, there is one digit before the decimal + goto doexpfromg; + } + // this is the insane action to get the pr to match %g semantics for %f + if (dp > 0) { + pr = (dp < (stbsp__int32)l) ? l - dp : 0; + } else { + pr = -dp + ((pr > (stbsp__int32)l) ? (stbsp__int32) l : pr); + } + goto dofloatfromg; + + case 'E': // float + case 'e': // float + h = (f[0] == 'E') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, pr | 0x80000000)) + fl |= STBSP__NEGATIVE; + doexpfromg: + tail[0] = 0; + stbsp__lead_sign(fl, lead); + if (dp == STBSP__SPECIAL) { + s = (char *)sn; + cs = 0; + pr = 0; + goto scopy; + } + s = num + 64; + // handle leading chars + *s++ = sn[0]; + + if (pr) + *s++ = stbsp__period; + + // handle after decimal + if ((l - 1) > (stbsp__uint32)pr) + l = pr + 1; + for (n = 1; n < l; n++) + *s++ = sn[n]; + // trailing zeros + tz = pr - (l - 1); + pr = 0; + // dump expo + tail[1] = h[0xe]; + dp -= 1; + if (dp < 0) { + tail[2] = '-'; + dp = -dp; + } else + tail[2] = '+'; +#ifdef STB_SPRINTF_MSVC_MODE + n = 5; +#else + n = (dp >= 100) ? 5 : 4; +#endif + tail[0] = (char)n; + for (;;) { + tail[n] = '0' + dp % 10; + if (n <= 3) + break; + --n; + dp /= 10; + } + cs = 1 + (3 << 24); // how many tens + goto flt_lead; + + case 'f': // float + fv = va_arg(va, double); + doafloat: + // do kilos + if (fl & STBSP__METRIC_SUFFIX) { + double divisor; + divisor = 1000.0f; + if (fl & STBSP__METRIC_1024) + divisor = 1024.0; + while (fl < 0x4000000) { + if ((fv < divisor) && (fv > -divisor)) + break; + fv /= divisor; + fl += 0x1000000; + } + } + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, pr)) + fl |= STBSP__NEGATIVE; + dofloatfromg: + tail[0] = 0; + stbsp__lead_sign(fl, lead); + if (dp == STBSP__SPECIAL) { + s = (char *)sn; + cs = 0; + pr = 0; + goto scopy; + } + s = num + 64; + + // handle the three decimal varieties + if (dp <= 0) { + stbsp__int32 i; + // handle 0.000*000xxxx + *s++ = '0'; + if (pr) + *s++ = stbsp__period; + n = -dp; + if ((stbsp__int32)n > pr) + n = pr; + i = n; + while (i) { + if ((((stbsp__uintptr)s) & 3) == 0) + break; + *s++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)s = 0x30303030; + s += 4; + i -= 4; + } + while (i) { + *s++ = '0'; + --i; + } + if ((stbsp__int32)(l + n) > pr) + l = pr - n; + i = l; + while (i) { + *s++ = *sn++; + --i; + } + tz = pr - (n + l); + cs = 1 + (3 << 24); // how many tens did we write (for commas below) + } else { + cs = (fl & STBSP__TRIPLET_COMMA) ? ((600 - (stbsp__uint32)dp) % 3) : 0; + if ((stbsp__uint32)dp >= l) { + // handle xxxx000*000.0 + n = 0; + for (;;) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = sn[n]; + ++n; + if (n >= l) + break; + } + } + if (n < (stbsp__uint32)dp) { + n = dp - n; + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + while (n) { + if ((((stbsp__uintptr)s) & 3) == 0) + break; + *s++ = '0'; + --n; + } + while (n >= 4) { + *(stbsp__uint32 *)s = 0x30303030; + s += 4; + n -= 4; + } + } + while (n) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = '0'; + --n; + } + } + } + cs = (int)(s - (num + 64)) + (3 << 24); // cs is how many tens + if (pr) { + *s++ = stbsp__period; + tz = pr; + } + } else { + // handle xxxxx.xxxx000*000 + n = 0; + for (;;) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = sn[n]; + ++n; + if (n >= (stbsp__uint32)dp) + break; + } + } + cs = (int)(s - (num + 64)) + (3 << 24); // cs is how many tens + if (pr) + *s++ = stbsp__period; + if ((l - dp) > (stbsp__uint32)pr) + l = pr + dp; + while (n < l) { + *s++ = sn[n]; + ++n; + } + tz = pr - (l - dp); + } + } + pr = 0; + + // handle k,m,g,t + if (fl & STBSP__METRIC_SUFFIX) { + char idx; + idx = 1; + if (fl & STBSP__METRIC_NOSPACE) + idx = 0; + tail[0] = idx; + tail[1] = ' '; + { + if (fl >> 24) { // SI kilo is 'k', JEDEC and SI kibits are 'K'. + if (fl & STBSP__METRIC_1024) + tail[idx + 1] = "_KMGT"[fl >> 24]; + else + tail[idx + 1] = "_kMGT"[fl >> 24]; + idx++; + // If printing kibits and not in jedec, add the 'i'. + if (fl & STBSP__METRIC_1024 && !(fl & STBSP__METRIC_JEDEC)) { + tail[idx + 1] = 'i'; + idx++; + } + tail[0] = idx; + } + } + }; + + flt_lead: + // get the length that we copied + l = (stbsp__uint32)(s - (num + 64)); + s = num + 64; + goto scopy; +#endif + + case 'B': // upper binary + case 'b': // lower binary + h = (f[0] == 'B') ? hexu : hex; + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 2; + lead[1] = '0'; + lead[2] = h[0xb]; + } + l = (8 << 4) | (1 << 8); + goto radixnum; + + case 'o': // octal + h = hexu; + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 1; + lead[1] = '0'; + } + l = (3 << 4) | (3 << 8); + goto radixnum; + + case 'p': // pointer + fl |= (sizeof(void *) == 8) ? STBSP__INTMAX : 0; + pr = sizeof(void *) * 2; + fl &= ~STBSP__LEADINGZERO; // 'p' only prints the pointer with zeros + // fall through - to X + + case 'X': // upper hex + case 'x': // lower hex + h = (f[0] == 'X') ? hexu : hex; + l = (4 << 4) | (4 << 8); + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 2; + lead[1] = '0'; + lead[2] = h[16]; + } + radixnum: + // get the number + if (fl & STBSP__INTMAX) + n64 = va_arg(va, stbsp__uint64); + else + n64 = va_arg(va, stbsp__uint32); + + s = num + STBSP__NUMSZ; + dp = 0; + // clear tail, and clear leading if value is zero + tail[0] = 0; + if (n64 == 0) { + lead[0] = 0; + if (pr == 0) { + l = 0; + cs = 0; + goto scopy; + } + } + // convert to string + for (;;) { + *--s = h[n64 & ((1 << (l >> 8)) - 1)]; + n64 >>= (l >> 8); + if (!((n64) || ((stbsp__int32)((num + STBSP__NUMSZ) - s) < pr))) + break; + if (fl & STBSP__TRIPLET_COMMA) { + ++l; + if ((l & 15) == ((l >> 4) & 15)) { + l &= ~15; + *--s = stbsp__comma; + } + } + }; + // get the tens and the comma pos + cs = (stbsp__uint32)((num + STBSP__NUMSZ) - s) + ((((l >> 4) & 15)) << 24); + // get the length that we copied + l = (stbsp__uint32)((num + STBSP__NUMSZ) - s); + // copy it + goto scopy; + + case 'u': // unsigned + case 'i': + case 'd': // integer + // get the integer and abs it + if (fl & STBSP__INTMAX) { + stbsp__int64 i64 = va_arg(va, stbsp__int64); + n64 = (stbsp__uint64)i64; + if ((f[0] != 'u') && (i64 < 0)) { + n64 = (stbsp__uint64)-i64; + fl |= STBSP__NEGATIVE; + } + } else { + stbsp__int32 i = va_arg(va, stbsp__int32); + n64 = (stbsp__uint32)i; + if ((f[0] != 'u') && (i < 0)) { + n64 = (stbsp__uint32)-i; + fl |= STBSP__NEGATIVE; + } + } + +#ifndef STB_SPRINTF_NOFLOAT + if (fl & STBSP__METRIC_SUFFIX) { + if (n64 < 1024) + pr = 0; + else if (pr == -1) + pr = 1; + fv = (double)(stbsp__int64)n64; + goto doafloat; + } +#endif + + // convert to string + s = num + STBSP__NUMSZ; + l = 0; + + for (;;) { + // do in 32-bit chunks (avoid lots of 64-bit divides even with constant denominators) + char *o = s - 8; + if (n64 >= 100000000) { + n = (stbsp__uint32)(n64 % 100000000); + n64 /= 100000000; + } else { + n = (stbsp__uint32)n64; + n64 = 0; + } + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + do { + s -= 2; + *(stbsp__uint16 *)s = *(stbsp__uint16 *)&stbsp__digitpair.pair[(n % 100) * 2]; + n /= 100; + } while (n); + } + while (n) { + if ((fl & STBSP__TRIPLET_COMMA) && (l++ == 3)) { + l = 0; + *--s = stbsp__comma; + --o; + } else { + *--s = (char)(n % 10) + '0'; + n /= 10; + } + } + if (n64 == 0) { + if ((s[0] == '0') && (s != (num + STBSP__NUMSZ))) + ++s; + break; + } + while (s != o) + if ((fl & STBSP__TRIPLET_COMMA) && (l++ == 3)) { + l = 0; + *--s = stbsp__comma; + --o; + } else { + *--s = '0'; + } + } + + tail[0] = 0; + stbsp__lead_sign(fl, lead); + + // get the length that we copied + l = (stbsp__uint32)((num + STBSP__NUMSZ) - s); + if (l == 0) { + *--s = '0'; + l = 1; + } + cs = l + (3 << 24); + if (pr < 0) + pr = 0; + + scopy: + // get fw=leading/trailing space, pr=leading zeros + if (pr < (stbsp__int32)l) + pr = l; + n = pr + lead[0] + tail[0] + tz; + if (fw < (stbsp__int32)n) + fw = n; + fw -= n; + pr -= l; + + // handle right justify and leading zeros + if ((fl & STBSP__LEFTJUST) == 0) { + if (fl & STBSP__LEADINGZERO) // if leading zeros, everything is in pr + { + pr = (fw > pr) ? fw : pr; + fw = 0; + } else { + fl &= ~STBSP__TRIPLET_COMMA; // if no leading zeros, then no commas + } + } + + // copy the spaces and/or zeros + if (fw + pr) { + stbsp__int32 i; + stbsp__uint32 c; + + // copy leading spaces (or when doing %8.4d stuff) + if ((fl & STBSP__LEFTJUST) == 0) + while (fw > 0) { + stbsp__cb_buf_clamp(i, fw); + fw -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = ' '; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x20202020; + bf += 4; + i -= 4; + } + while (i) { + *bf++ = ' '; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy leader + sn = lead + 1; + while (lead[0]) { + stbsp__cb_buf_clamp(i, lead[0]); + lead[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy leading zeros + c = cs >> 24; + cs &= 0xffffff; + cs = (fl & STBSP__TRIPLET_COMMA) ? ((stbsp__uint32)(c - ((pr + cs) % (c + 1)))) : 0; + while (pr > 0) { + stbsp__cb_buf_clamp(i, pr); + pr -= i; + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x30303030; + bf += 4; + i -= 4; + } + } + while (i) { + if ((fl & STBSP__TRIPLET_COMMA) && (cs++ == c)) { + cs = 0; + *bf++ = stbsp__comma; + } else + *bf++ = '0'; + --i; + } + stbsp__chk_cb_buf(1); + } + } + + // copy leader if there is still one + sn = lead + 1; + while (lead[0]) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, lead[0]); + lead[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy the string + n = l; + while (n) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, n); + n -= i; + STBSP__UNALIGNED(while (i >= 4) { + *(stbsp__uint32 volatile *)bf = *(stbsp__uint32 volatile *)s; + bf += 4; + s += 4; + i -= 4; + }) + while (i) { + *bf++ = *s++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy trailing zeros + while (tz) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, tz); + tz -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x30303030; + bf += 4; + i -= 4; + } + while (i) { + *bf++ = '0'; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy tail if there is one + sn = tail + 1; + while (tail[0]) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, tail[0]); + tail[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // handle the left justify + if (fl & STBSP__LEFTJUST) + if (fw > 0) { + while (fw) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, fw); + fw -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = ' '; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x20202020; + bf += 4; + i -= 4; + } + while (i--) + *bf++ = ' '; + stbsp__chk_cb_buf(1); + } + } + break; + + default: // unknown, just copy code + s = num + STBSP__NUMSZ - 1; + *s = f[0]; + l = 1; + fw = fl = 0; + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + goto scopy; + } + ++f; + } +endfmt: + + if (!callback) + *bf = 0; + else + stbsp__flush_cb(); + +done: + return tlen + (int)(bf - buf); +} + +// cleanup +#undef STBSP__LEFTJUST +#undef STBSP__LEADINGPLUS +#undef STBSP__LEADINGSPACE +#undef STBSP__LEADING_0X +#undef STBSP__LEADINGZERO +#undef STBSP__INTMAX +#undef STBSP__TRIPLET_COMMA +#undef STBSP__NEGATIVE +#undef STBSP__METRIC_SUFFIX +#undef STBSP__NUMSZ +#undef stbsp__chk_cb_bufL +#undef stbsp__chk_cb_buf +#undef stbsp__flush_cb +#undef stbsp__cb_buf_clamp + +// ============================================================================ +// wrapper functions + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(sprintf)(char *buf, char const *fmt, ...) +{ + int result; + va_list va; + va_start(va, fmt); + result = STB_SPRINTF_DECORATE(vsprintfcb)(0, 0, buf, fmt, va); + va_end(va); + return result; +} + +typedef struct stbsp__context { + char *buf; + int count; + int length; + char tmp[STB_SPRINTF_MIN]; +} stbsp__context; + +static char *stbsp__clamp_callback(const char *buf, void *user, int len) +{ + stbsp__context *c = (stbsp__context *)user; + c->length += len; + + if (len > c->count) + len = c->count; + + if (len) { + if (buf != c->buf) { + const char *s, *se; + char *d; + d = c->buf; + s = buf; + se = buf + len; + do { + *d++ = *s++; + } while (s < se); + } + c->buf += len; + c->count -= len; + } + + if (c->count <= 0) + return c->tmp; + return (c->count >= STB_SPRINTF_MIN) ? c->buf : c->tmp; // go direct into buffer if you can +} + +static char * stbsp__count_clamp_callback( const char * buf, void * user, int len ) +{ + stbsp__context * c = (stbsp__context*)user; + (void) sizeof(buf); + + c->length += len; + return c->tmp; // go direct into buffer if you can +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE( vsnprintf )( char * buf, int count, char const * fmt, va_list va ) +{ + stbsp__context c; + + if ( (count == 0) && !buf ) + { + c.length = 0; + + STB_SPRINTF_DECORATE( vsprintfcb )( stbsp__count_clamp_callback, &c, c.tmp, fmt, va ); + } + else + { + int l; + + c.buf = buf; + c.count = count; + c.length = 0; + + STB_SPRINTF_DECORATE( vsprintfcb )( stbsp__clamp_callback, &c, stbsp__clamp_callback(0,&c,0), fmt, va ); + + // zero-terminate + l = (int)( c.buf - buf ); + if ( l >= count ) // should never be greater, only equal (or less) than count + l = count - 1; + buf[l] = 0; + } + + return c.length; +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(snprintf)(char *buf, int count, char const *fmt, ...) +{ + int result; + va_list va; + va_start(va, fmt); + + result = STB_SPRINTF_DECORATE(vsnprintf)(buf, count, fmt, va); + va_end(va); + + return result; +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(vsprintf)(char *buf, char const *fmt, va_list va) +{ + return STB_SPRINTF_DECORATE(vsprintfcb)(0, 0, buf, fmt, va); +} + +// ======================================================================= +// low level float utility functions + +#ifndef STB_SPRINTF_NOFLOAT + +// copies d to bits w/ strict aliasing (this compiles to nothing on /Ox) +#define STBSP__COPYFP(dest, src) \ + { \ + int cn; \ + for (cn = 0; cn < 8; cn++) \ + ((char *)&dest)[cn] = ((char *)&src)[cn]; \ + } + +// get float info +static stbsp__int32 stbsp__real_to_parts(stbsp__int64 *bits, stbsp__int32 *expo, double value) +{ + double d; + stbsp__int64 b = 0; + + // load value and round at the frac_digits + d = value; + + STBSP__COPYFP(b, d); + + *bits = b & ((((stbsp__uint64)1) << 52) - 1); + *expo = (stbsp__int32)(((b >> 52) & 2047) - 1023); + + return (stbsp__int32)((stbsp__uint64) b >> 63); +} + +static double const stbsp__bot[23] = { + 1e+000, 1e+001, 1e+002, 1e+003, 1e+004, 1e+005, 1e+006, 1e+007, 1e+008, 1e+009, 1e+010, 1e+011, + 1e+012, 1e+013, 1e+014, 1e+015, 1e+016, 1e+017, 1e+018, 1e+019, 1e+020, 1e+021, 1e+022 +}; +static double const stbsp__negbot[22] = { + 1e-001, 1e-002, 1e-003, 1e-004, 1e-005, 1e-006, 1e-007, 1e-008, 1e-009, 1e-010, 1e-011, + 1e-012, 1e-013, 1e-014, 1e-015, 1e-016, 1e-017, 1e-018, 1e-019, 1e-020, 1e-021, 1e-022 +}; +static double const stbsp__negboterr[22] = { + -5.551115123125783e-018, -2.0816681711721684e-019, -2.0816681711721686e-020, -4.7921736023859299e-021, -8.1803053914031305e-022, 4.5251888174113741e-023, + 4.5251888174113739e-024, -2.0922560830128471e-025, -6.2281591457779853e-026, -3.6432197315497743e-027, 6.0503030718060191e-028, 2.0113352370744385e-029, + -3.0373745563400371e-030, 1.1806906454401013e-032, -7.7705399876661076e-032, 2.0902213275965398e-033, -7.1542424054621921e-034, -7.1542424054621926e-035, + 2.4754073164739869e-036, 5.4846728545790429e-037, 9.2462547772103625e-038, -4.8596774326570872e-039 +}; +static double const stbsp__top[13] = { + 1e+023, 1e+046, 1e+069, 1e+092, 1e+115, 1e+138, 1e+161, 1e+184, 1e+207, 1e+230, 1e+253, 1e+276, 1e+299 +}; +static double const stbsp__negtop[13] = { + 1e-023, 1e-046, 1e-069, 1e-092, 1e-115, 1e-138, 1e-161, 1e-184, 1e-207, 1e-230, 1e-253, 1e-276, 1e-299 +}; +static double const stbsp__toperr[13] = { + 8388608, + 6.8601809640529717e+028, + -7.253143638152921e+052, + -4.3377296974619174e+075, + -1.5559416129466825e+098, + -3.2841562489204913e+121, + -3.7745893248228135e+144, + -1.7356668416969134e+167, + -3.8893577551088374e+190, + -9.9566444326005119e+213, + 6.3641293062232429e+236, + -5.2069140800249813e+259, + -5.2504760255204387e+282 +}; +static double const stbsp__negtoperr[13] = { + 3.9565301985100693e-040, -2.299904345391321e-063, 3.6506201437945798e-086, 1.1875228833981544e-109, + -5.0644902316928607e-132, -6.7156837247865426e-155, -2.812077463003139e-178, -5.7778912386589953e-201, + 7.4997100559334532e-224, -4.6439668915134491e-247, -6.3691100762962136e-270, -9.436808465446358e-293, + 8.0970921678014997e-317 +}; + +#if defined(_MSC_VER) && (_MSC_VER <= 1200) +static stbsp__uint64 const stbsp__powten[20] = { + 1, + 10, + 100, + 1000, + 10000, + 100000, + 1000000, + 10000000, + 100000000, + 1000000000, + 10000000000, + 100000000000, + 1000000000000, + 10000000000000, + 100000000000000, + 1000000000000000, + 10000000000000000, + 100000000000000000, + 1000000000000000000, + 10000000000000000000U +}; +#define stbsp__tento19th ((stbsp__uint64)1000000000000000000) +#else +static stbsp__uint64 const stbsp__powten[20] = { + 1, + 10, + 100, + 1000, + 10000, + 100000, + 1000000, + 10000000, + 100000000, + 1000000000, + 10000000000ULL, + 100000000000ULL, + 1000000000000ULL, + 10000000000000ULL, + 100000000000000ULL, + 1000000000000000ULL, + 10000000000000000ULL, + 100000000000000000ULL, + 1000000000000000000ULL, + 10000000000000000000ULL +}; +#define stbsp__tento19th (1000000000000000000ULL) +#endif + +#define stbsp__ddmulthi(oh, ol, xh, yh) \ + { \ + double ahi = 0, alo, bhi = 0, blo; \ + stbsp__int64 bt; \ + oh = xh * yh; \ + STBSP__COPYFP(bt, xh); \ + bt &= ((~(stbsp__uint64)0) << 27); \ + STBSP__COPYFP(ahi, bt); \ + alo = xh - ahi; \ + STBSP__COPYFP(bt, yh); \ + bt &= ((~(stbsp__uint64)0) << 27); \ + STBSP__COPYFP(bhi, bt); \ + blo = yh - bhi; \ + ol = ((ahi * bhi - oh) + ahi * blo + alo * bhi) + alo * blo; \ + } + +#define stbsp__ddtoS64(ob, xh, xl) \ + { \ + double ahi = 0, alo, vh, t; \ + ob = (stbsp__int64)xh; \ + vh = (double)ob; \ + ahi = (xh - vh); \ + t = (ahi - xh); \ + alo = (xh - (ahi - t)) - (vh + t); \ + ob += (stbsp__int64)(ahi + alo + xl); \ + } + +#define stbsp__ddrenorm(oh, ol) \ + { \ + double s; \ + s = oh + ol; \ + ol = ol - (s - oh); \ + oh = s; \ + } + +#define stbsp__ddmultlo(oh, ol, xh, xl, yh, yl) ol = ol + (xh * yl + xl * yh); + +#define stbsp__ddmultlos(oh, ol, xh, yl) ol = ol + (xh * yl); + +static void stbsp__raise_to_power10(double *ohi, double *olo, double d, stbsp__int32 power) // power can be -323 to +350 +{ + double ph, pl; + if ((power >= 0) && (power <= 22)) { + stbsp__ddmulthi(ph, pl, d, stbsp__bot[power]); + } else { + stbsp__int32 e, et, eb; + double p2h, p2l; + + e = power; + if (power < 0) + e = -e; + et = (e * 0x2c9) >> 14; /* %23 */ + if (et > 13) + et = 13; + eb = e - (et * 23); + + ph = d; + pl = 0.0; + if (power < 0) { + if (eb) { + --eb; + stbsp__ddmulthi(ph, pl, d, stbsp__negbot[eb]); + stbsp__ddmultlos(ph, pl, d, stbsp__negboterr[eb]); + } + if (et) { + stbsp__ddrenorm(ph, pl); + --et; + stbsp__ddmulthi(p2h, p2l, ph, stbsp__negtop[et]); + stbsp__ddmultlo(p2h, p2l, ph, pl, stbsp__negtop[et], stbsp__negtoperr[et]); + ph = p2h; + pl = p2l; + } + } else { + if (eb) { + e = eb; + if (eb > 22) + eb = 22; + e -= eb; + stbsp__ddmulthi(ph, pl, d, stbsp__bot[eb]); + if (e) { + stbsp__ddrenorm(ph, pl); + stbsp__ddmulthi(p2h, p2l, ph, stbsp__bot[e]); + stbsp__ddmultlos(p2h, p2l, stbsp__bot[e], pl); + ph = p2h; + pl = p2l; + } + } + if (et) { + stbsp__ddrenorm(ph, pl); + --et; + stbsp__ddmulthi(p2h, p2l, ph, stbsp__top[et]); + stbsp__ddmultlo(p2h, p2l, ph, pl, stbsp__top[et], stbsp__toperr[et]); + ph = p2h; + pl = p2l; + } + } + } + stbsp__ddrenorm(ph, pl); + *ohi = ph; + *olo = pl; +} + +// given a float value, returns the significant bits in bits, and the position of the +// decimal point in decimal_pos. +/-INF and NAN are specified by special values +// returned in the decimal_pos parameter. +// frac_digits is absolute normally, but if you want from first significant digits (got %g and %e), or in 0x80000000 +static stbsp__int32 stbsp__real_to_str(char const **start, stbsp__uint32 *len, char *out, stbsp__int32 *decimal_pos, double value, stbsp__uint32 frac_digits) +{ + double d; + stbsp__int64 bits = 0; + stbsp__int32 expo, e, ng, tens; + + d = value; + STBSP__COPYFP(bits, d); + expo = (stbsp__int32)((bits >> 52) & 2047); + ng = (stbsp__int32)((stbsp__uint64) bits >> 63); + if (ng) + d = -d; + + if (expo == 2047) // is nan or inf? + { + *start = (bits & ((((stbsp__uint64)1) << 52) - 1)) ? "NaN" : "Inf"; + *decimal_pos = STBSP__SPECIAL; + *len = 3; + return ng; + } + + if (expo == 0) // is zero or denormal + { + if (((stbsp__uint64) bits << 1) == 0) // do zero + { + *decimal_pos = 1; + *start = out; + out[0] = '0'; + *len = 1; + return ng; + } + // find the right expo for denormals + { + stbsp__int64 v = ((stbsp__uint64)1) << 51; + while ((bits & v) == 0) { + --expo; + v >>= 1; + } + } + } + + // find the decimal exponent as well as the decimal bits of the value + { + double ph, pl; + + // log10 estimate - very specifically tweaked to hit or undershoot by no more than 1 of log10 of all expos 1..2046 + tens = expo - 1023; + tens = (tens < 0) ? ((tens * 617) / 2048) : (((tens * 1233) / 4096) + 1); + + // move the significant bits into position and stick them into an int + stbsp__raise_to_power10(&ph, &pl, d, 18 - tens); + + // get full as much precision from double-double as possible + stbsp__ddtoS64(bits, ph, pl); + + // check if we undershot + if (((stbsp__uint64)bits) >= stbsp__tento19th) + ++tens; + } + + // now do the rounding in integer land + frac_digits = (frac_digits & 0x80000000) ? ((frac_digits & 0x7ffffff) + 1) : (tens + frac_digits); + if ((frac_digits < 24)) { + stbsp__uint32 dg = 1; + if ((stbsp__uint64)bits >= stbsp__powten[9]) + dg = 10; + while ((stbsp__uint64)bits >= stbsp__powten[dg]) { + ++dg; + if (dg == 20) + goto noround; + } + if (frac_digits < dg) { + stbsp__uint64 r; + // add 0.5 at the right position and round + e = dg - frac_digits; + if ((stbsp__uint32)e >= 24) + goto noround; + r = stbsp__powten[e]; + bits = bits + (r / 2); + if ((stbsp__uint64)bits >= stbsp__powten[dg]) + ++tens; + bits /= r; + } + noround:; + } + + // kill long trailing runs of zeros + if (bits) { + stbsp__uint32 n; + for (;;) { + if (bits <= 0xffffffff) + break; + if (bits % 1000) + goto donez; + bits /= 1000; + } + n = (stbsp__uint32)bits; + while ((n % 1000) == 0) + n /= 1000; + bits = n; + donez:; + } + + // convert to string + out += 64; + e = 0; + for (;;) { + stbsp__uint32 n; + char *o = out - 8; + // do the conversion in chunks of U32s (avoid most 64-bit divides, worth it, constant denomiators be damned) + if (bits >= 100000000) { + n = (stbsp__uint32)(bits % 100000000); + bits /= 100000000; + } else { + n = (stbsp__uint32)bits; + bits = 0; + } + while (n) { + out -= 2; + *(stbsp__uint16 *)out = *(stbsp__uint16 *)&stbsp__digitpair.pair[(n % 100) * 2]; + n /= 100; + e += 2; + } + if (bits == 0) { + if ((e) && (out[0] == '0')) { + ++out; + --e; + } + break; + } + while (out != o) { + *--out = '0'; + ++e; + } + } + + *decimal_pos = tens; + *start = out; + *len = e; + return ng; +} + +#undef stbsp__ddmulthi +#undef stbsp__ddrenorm +#undef stbsp__ddmultlo +#undef stbsp__ddmultlos +#undef STBSP__SPECIAL +#undef STBSP__COPYFP + +#endif // STB_SPRINTF_NOFLOAT + +// clean up +#undef stbsp__uint16 +#undef stbsp__uint32 +#undef stbsp__int32 +#undef stbsp__uint64 +#undef stbsp__int64 +#undef STBSP__UNALIGNED + +#endif // STB_SPRINTF_IMPLEMENTATION + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ |