| 1 | /* |
| 2 | * Copyright (c) Martin Liska, SUSE, Meta Platforms, Inc. and affiliates. |
| 3 | * All rights reserved. |
| 4 | * |
| 5 | * This source code is licensed under both the BSD-style license (found in the |
| 6 | * LICENSE file in the root directory of this source tree) and the GPLv2 (found |
| 7 | * in the COPYING file in the root directory of this source tree). |
| 8 | * You may select, at your option, one of the above-listed licenses. |
| 9 | */ |
| 10 | |
| 11 | |
| 12 | #include <stdio.h> // printf |
| 13 | #include <stdlib.h> // free |
| 14 | #include <string.h> // memset, strcat, strlen |
| 15 | #include <zstd.h> // presumes zstd library is installed |
| 16 | #include "common.h" // Helper functions, CHECK(), and CHECK_ZSTD() |
| 17 | #include <pthread.h> |
| 18 | |
| 19 | typedef struct compress_args |
| 20 | { |
| 21 | const char *fname; |
| 22 | char *outName; |
| 23 | int cLevel; |
| 24 | #if defined(ZSTD_STATIC_LINKING_ONLY) |
| 25 | ZSTD_threadPool *pool; |
| 26 | #endif |
| 27 | } compress_args_t; |
| 28 | |
| 29 | static void *compressFile_orDie(void *data) |
| 30 | { |
| 31 | const int nbThreads = 16; |
| 32 | |
| 33 | compress_args_t *args = (compress_args_t *)data; |
| 34 | fprintf (stderr, "Starting compression of %s with level %d, using %d threads\n", args->fname, args->cLevel, nbThreads); |
| 35 | /* Open the input and output files. */ |
| 36 | FILE* const fin = fopen_orDie(args->fname, "rb"); |
| 37 | FILE* const fout = fopen_orDie(args->outName, "wb"); |
| 38 | /* Create the input and output buffers. |
| 39 | * They may be any size, but we recommend using these functions to size them. |
| 40 | * Performance will only suffer significantly for very tiny buffers. |
| 41 | */ |
| 42 | size_t const buffInSize = ZSTD_CStreamInSize(); |
| 43 | void* const buffIn = malloc_orDie(buffInSize); |
| 44 | size_t const buffOutSize = ZSTD_CStreamOutSize(); |
| 45 | void* const buffOut = malloc_orDie(buffOutSize); |
| 46 | |
| 47 | /* Create the context. */ |
| 48 | ZSTD_CCtx* const cctx = ZSTD_createCCtx(); |
| 49 | CHECK(cctx != NULL, "ZSTD_createCCtx() failed!"); |
| 50 | |
| 51 | #if defined(ZSTD_STATIC_LINKING_ONLY) |
| 52 | size_t r = ZSTD_CCtx_refThreadPool(cctx, args->pool); |
| 53 | CHECK(r == 0, "ZSTD_CCtx_refThreadPool failed!"); |
| 54 | #endif |
| 55 | |
| 56 | /* Set any parameters you want. |
| 57 | * Here we set the compression level, and enable the checksum. |
| 58 | */ |
| 59 | CHECK_ZSTD( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, args->cLevel) ); |
| 60 | CHECK_ZSTD( ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1) ); |
| 61 | ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbThreads); |
| 62 | |
| 63 | /* This loop reads from the input file, compresses that entire chunk, |
| 64 | * and writes all output produced to the output file. |
| 65 | */ |
| 66 | size_t const toRead = buffInSize; |
| 67 | for (;;) { |
| 68 | size_t read = fread_orDie(buffIn, toRead, fin); |
| 69 | /* Select the flush mode. |
| 70 | * If the read may not be finished (read == toRead) we use |
| 71 | * ZSTD_e_continue. If this is the last chunk, we use ZSTD_e_end. |
| 72 | * Zstd optimizes the case where the first flush mode is ZSTD_e_end, |
| 73 | * since it knows it is compressing the entire source in one pass. |
| 74 | */ |
| 75 | int const lastChunk = (read < toRead); |
| 76 | ZSTD_EndDirective const mode = lastChunk ? ZSTD_e_end : ZSTD_e_continue; |
| 77 | /* Set the input buffer to what we just read. |
| 78 | * We compress until the input buffer is empty, each time flushing the |
| 79 | * output. |
| 80 | */ |
| 81 | ZSTD_inBuffer input = { buffIn, read, 0 }; |
| 82 | int finished; |
| 83 | do { |
| 84 | /* Compress into the output buffer and write all of the output to |
| 85 | * the file so we can reuse the buffer next iteration. |
| 86 | */ |
| 87 | ZSTD_outBuffer output = { buffOut, buffOutSize, 0 }; |
| 88 | size_t const remaining = ZSTD_compressStream2(cctx, &output , &input, mode); |
| 89 | CHECK_ZSTD(remaining); |
| 90 | fwrite_orDie(buffOut, output.pos, fout); |
| 91 | /* If we're on the last chunk we're finished when zstd returns 0, |
| 92 | * which means its consumed all the input AND finished the frame. |
| 93 | * Otherwise, we're finished when we've consumed all the input. |
| 94 | */ |
| 95 | finished = lastChunk ? (remaining == 0) : (input.pos == input.size); |
| 96 | } while (!finished); |
| 97 | CHECK(input.pos == input.size, |
| 98 | "Impossible: zstd only returns 0 when the input is completely consumed!"); |
| 99 | |
| 100 | if (lastChunk) { |
| 101 | break; |
| 102 | } |
| 103 | } |
| 104 | |
| 105 | fprintf (stderr, "Finishing compression of %s\n", args->outName); |
| 106 | |
| 107 | ZSTD_freeCCtx(cctx); |
| 108 | fclose_orDie(fout); |
| 109 | fclose_orDie(fin); |
| 110 | free(buffIn); |
| 111 | free(buffOut); |
| 112 | free(args->outName); |
| 113 | |
| 114 | return NULL; |
| 115 | } |
| 116 | |
| 117 | |
| 118 | static char* createOutFilename_orDie(const char* filename) |
| 119 | { |
| 120 | size_t const inL = strlen(filename); |
| 121 | size_t const outL = inL + 5; |
| 122 | void* const outSpace = malloc_orDie(outL); |
| 123 | memset(outSpace, 0, outL); |
| 124 | strcat(outSpace, filename); |
| 125 | strcat(outSpace, ".zst"); |
| 126 | return (char*)outSpace; |
| 127 | } |
| 128 | |
| 129 | int main(int argc, const char** argv) |
| 130 | { |
| 131 | const char* const exeName = argv[0]; |
| 132 | |
| 133 | if (argc<=3) { |
| 134 | printf("wrong arguments\n"); |
| 135 | printf("usage:\n"); |
| 136 | printf("%s POOL_SIZE LEVEL FILES\n", exeName); |
| 137 | return 1; |
| 138 | } |
| 139 | |
| 140 | int pool_size = atoi (argv[1]); |
| 141 | CHECK(pool_size != 0, "can't parse POOL_SIZE!"); |
| 142 | |
| 143 | int level = atoi (argv[2]); |
| 144 | CHECK(level != 0, "can't parse LEVEL!"); |
| 145 | |
| 146 | argc -= 3; |
| 147 | argv += 3; |
| 148 | |
| 149 | #if defined(ZSTD_STATIC_LINKING_ONLY) |
| 150 | ZSTD_threadPool *pool = ZSTD_createThreadPool (pool_size); |
| 151 | CHECK(pool != NULL, "ZSTD_createThreadPool() failed!"); |
| 152 | fprintf (stderr, "Using shared thread pool of size %d\n", pool_size); |
| 153 | #else |
| 154 | fprintf (stderr, "All threads use its own thread pool\n"); |
| 155 | #endif |
| 156 | |
| 157 | pthread_t *threads = malloc_orDie(argc * sizeof(pthread_t)); |
| 158 | compress_args_t *args = malloc_orDie(argc * sizeof(compress_args_t)); |
| 159 | |
| 160 | for (unsigned i = 0; i < argc; i++) |
| 161 | { |
| 162 | args[i].fname = argv[i]; |
| 163 | args[i].outName = createOutFilename_orDie(args[i].fname); |
| 164 | args[i].cLevel = level; |
| 165 | #if defined(ZSTD_STATIC_LINKING_ONLY) |
| 166 | args[i].pool = pool; |
| 167 | #endif |
| 168 | |
| 169 | pthread_create (&threads[i], NULL, compressFile_orDie, &args[i]); |
| 170 | } |
| 171 | |
| 172 | for (unsigned i = 0; i < argc; i++) |
| 173 | pthread_join (threads[i], NULL); |
| 174 | |
| 175 | #if defined(ZSTD_STATIC_LINKING_ONLY) |
| 176 | ZSTD_freeThreadPool (pool); |
| 177 | #endif |
| 178 | |
| 179 | return 0; |
| 180 | } |