]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | ************************************************************************** | |
3 | * © 2016 and later: Unicode, Inc. and others. | |
4 | * License & terms of use: http://www.unicode.org/copyright.html#License | |
5 | ************************************************************************* | |
6 | ************************************************************************* | |
7 | * Copyright (C) 2002-2014, International Business Machines | |
8 | * Corporation and others. All Rights Reserved. | |
9 | ************************************************************************* | |
10 | * file name: utfperf.cpp | |
11 | * encoding: UTF-8 | |
12 | * tab size: 8 (not used) | |
13 | * indentation:4 | |
14 | * | |
15 | * created on: 2005Nov17 | |
16 | * created by: Raymond Yang | |
17 | * | |
18 | * Ported from utfper.c created by Markus W. Scherer | |
19 | * Performance test program for Unicode converters | |
20 | */ | |
21 | ||
22 | #include <stdio.h> | |
23 | #include <stdlib.h> | |
24 | #include "unicode/uperf.h" | |
25 | #include "cmemory.h" // for UPRV_LENGTHOF | |
26 | #include "uoptions.h" | |
27 | ||
28 | /* definitions and text buffers */ | |
29 | ||
30 | #define INPUT_CAPACITY (1024*1024) | |
31 | #define INTERMEDIATE_CAPACITY 4096 | |
32 | #define INTERMEDIATE_SMALL_CAPACITY 20 | |
33 | #define PIVOT_CAPACITY 1024 | |
34 | #define OUTPUT_CAPACITY INPUT_CAPACITY | |
35 | ||
36 | static char utf8[INPUT_CAPACITY]; | |
37 | static UChar pivot[INTERMEDIATE_CAPACITY]; | |
38 | ||
39 | static UChar output[OUTPUT_CAPACITY]; | |
40 | static char intermediate[OUTPUT_CAPACITY]; | |
41 | ||
42 | static int32_t utf8Length, encodedLength, outputLength, countInputCodePoints; | |
43 | ||
44 | static int32_t fromUCallbackCount; | |
45 | ||
46 | // Command-line options specific to utfperf. | |
47 | // Options do not have abbreviations: Force readable command lines. | |
48 | // (Using U+0001 for abbreviation characters.) | |
49 | enum { | |
50 | CHARSET, | |
51 | CHUNK_LENGTH, | |
52 | PIVOT_LENGTH, | |
53 | UTFPERF_OPTIONS_COUNT | |
54 | }; | |
55 | ||
56 | static UOption options[UTFPERF_OPTIONS_COUNT]={ | |
57 | UOPTION_DEF("charset", '\x01', UOPT_REQUIRES_ARG), | |
58 | UOPTION_DEF("chunk", '\x01', UOPT_REQUIRES_ARG), | |
59 | UOPTION_DEF("pivot", '\x01', UOPT_REQUIRES_ARG) | |
60 | }; | |
61 | ||
62 | static const char *const utfperf_usage = | |
63 | "\t--charset Charset for which to test performance, e.g. windows-1251.\n" | |
64 | "\t Default: UTF-8\n" | |
65 | "\t--chunk Length (in bytes) of charset output chunks. [4096]\n" | |
66 | "\t--pivot Length (in UChars) of the UTF-16 pivot buffer, if applicable.\n" | |
67 | "\t [1024]\n"; | |
68 | ||
69 | // Test object. | |
70 | class UtfPerformanceTest : public UPerfTest{ | |
71 | public: | |
72 | UtfPerformanceTest(int32_t argc, const char *argv[], UErrorCode &status) | |
73 | : UPerfTest(argc, argv, options, UPRV_LENGTHOF(options), utfperf_usage, status) { | |
74 | if (U_SUCCESS(status)) { | |
75 | charset = options[CHARSET].value; | |
76 | ||
77 | chunkLength = atoi(options[CHUNK_LENGTH].value); | |
78 | if (chunkLength < 1 || OUTPUT_CAPACITY < chunkLength) { | |
79 | fprintf(stderr, "error: chunk length must be 1..%ld\n", (long)OUTPUT_CAPACITY); | |
80 | status = U_ILLEGAL_ARGUMENT_ERROR; | |
81 | } | |
82 | ||
83 | pivotLength = atoi(options[PIVOT_LENGTH].value); | |
84 | if (pivotLength < 1 || PIVOT_CAPACITY < pivotLength) { | |
85 | fprintf(stderr, "error: pivot length must be 1..%ld\n", (long)PIVOT_CAPACITY); | |
86 | status = U_ILLEGAL_ARGUMENT_ERROR; | |
87 | } | |
88 | ||
89 | int32_t inputLength; | |
90 | UPerfTest::getBuffer(inputLength, status); | |
91 | countInputCodePoints = u_countChar32(buffer, bufferLen); | |
92 | u_strToUTF8(utf8, (int32_t)sizeof(utf8), &utf8Length, buffer, bufferLen, &status); | |
93 | } | |
94 | } | |
95 | ||
96 | virtual UPerfFunction* runIndexedTest(int32_t index, UBool exec, const char* &name, char* par = NULL); | |
97 | ||
98 | const UChar *getBuffer() const { return buffer; } | |
99 | int32_t getBufferLen() const { return bufferLen; } | |
100 | ||
101 | const char *charset; | |
102 | int32_t chunkLength, pivotLength; | |
103 | }; | |
104 | ||
105 | U_CDECL_BEGIN | |
106 | // Custom callback for counting callback calls. | |
107 | static void U_CALLCONV | |
108 | fromUCallback(const void *context, | |
109 | UConverterFromUnicodeArgs *fromUArgs, | |
110 | const UChar *codeUnits, | |
111 | int32_t length, | |
112 | UChar32 codePoint, | |
113 | UConverterCallbackReason reason, | |
114 | UErrorCode *pErrorCode) { | |
115 | if (reason <= UCNV_IRREGULAR) { | |
116 | ++fromUCallbackCount; | |
117 | } | |
118 | UCNV_FROM_U_CALLBACK_SUBSTITUTE(context, fromUArgs, codeUnits, length, codePoint, reason, pErrorCode); | |
119 | } | |
120 | U_CDECL_END | |
121 | ||
122 | // Base class for Roundtrip, FromUnicode and FromUTF8 with common setup. | |
123 | class Command : public UPerfFunction { | |
124 | protected: | |
125 | Command(const UtfPerformanceTest &testcase) | |
126 | : testcase(testcase), | |
127 | input(testcase.getBuffer()), inputLength(testcase.getBufferLen()), | |
128 | errorCode(U_ZERO_ERROR) { | |
129 | cnv=ucnv_open(testcase.charset, &errorCode); | |
130 | if (U_FAILURE(errorCode)) { | |
131 | fprintf(stderr, "error opening converter for \"%s\" - %s\n", testcase.charset, u_errorName(errorCode)); | |
132 | } | |
133 | ucnv_setFromUCallBack(cnv, fromUCallback, NULL, NULL, NULL, &errorCode); | |
134 | } | |
135 | public: | |
136 | virtual ~Command(){ | |
137 | if(U_SUCCESS(errorCode)) { | |
138 | ucnv_close(cnv); | |
139 | } | |
140 | } | |
141 | // virtual void call(UErrorCode* pErrorCode) { ... } | |
142 | virtual long getOperationsPerIteration(){ | |
143 | return countInputCodePoints; | |
144 | } | |
145 | ||
146 | const UtfPerformanceTest &testcase; | |
147 | const UChar *input; | |
148 | int32_t inputLength; | |
149 | UErrorCode errorCode; | |
150 | UConverter *cnv; | |
151 | }; | |
152 | ||
153 | // Test roundtrip UTF-16->encoding->UTF-16. | |
154 | class Roundtrip : public Command { | |
155 | protected: | |
156 | Roundtrip(const UtfPerformanceTest &testcase) : Command(testcase) {} | |
157 | public: | |
158 | static UPerfFunction* get(const UtfPerformanceTest &testcase) { | |
159 | Roundtrip * t = new Roundtrip(testcase); | |
160 | if (U_SUCCESS(t->errorCode)){ | |
161 | return t; | |
162 | } else { | |
163 | delete t; | |
164 | return NULL; | |
165 | } | |
166 | } | |
167 | virtual void call(UErrorCode* pErrorCode){ | |
168 | const UChar *pIn, *pInLimit; | |
169 | UChar *pOut, *pOutLimit; | |
170 | char *pInter, *pInterLimit; | |
171 | const char *p; | |
172 | UBool flush; | |
173 | ||
174 | ucnv_reset(cnv); | |
175 | fromUCallbackCount=0; | |
176 | ||
177 | pIn=input; | |
178 | pInLimit=input+inputLength; | |
179 | ||
180 | pOut=output; | |
181 | pOutLimit=output+OUTPUT_CAPACITY; | |
182 | ||
183 | pInterLimit=intermediate+testcase.chunkLength; | |
184 | ||
185 | encodedLength=outputLength=0; | |
186 | flush=FALSE; | |
187 | ||
188 | do { | |
189 | /* convert a block of [pIn..pInLimit[ to the encoding in intermediate[] */ | |
190 | pInter=intermediate; | |
191 | ucnv_fromUnicode(cnv, &pInter, pInterLimit, &pIn, pInLimit, NULL, TRUE, pErrorCode); | |
192 | encodedLength+=(int32_t)(pInter-intermediate); | |
193 | ||
194 | if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { | |
195 | /* make sure that we convert once more to really flush */ | |
196 | *pErrorCode=U_ZERO_ERROR; | |
197 | } else if(U_FAILURE(*pErrorCode)) { | |
198 | return; | |
199 | } else if(pIn==pInLimit) { | |
200 | flush=TRUE; | |
201 | } | |
202 | ||
203 | /* convert the block [intermediate..pInter[ back to UTF-16 */ | |
204 | p=intermediate; | |
205 | ucnv_toUnicode(cnv, &pOut, pOutLimit,&p, pInter,NULL, flush,pErrorCode); | |
206 | if(U_FAILURE(*pErrorCode)) { | |
207 | return; | |
208 | } | |
209 | /* intermediate must have been consumed (p==pInter) because of the converter semantics */ | |
210 | } while(!flush); | |
211 | ||
212 | outputLength=pOut-output; | |
213 | if(inputLength!=outputLength) { | |
214 | fprintf(stderr, "error: roundtrip failed, inputLength %d!=outputLength %d\n", inputLength, outputLength); | |
215 | *pErrorCode=U_INTERNAL_PROGRAM_ERROR; | |
216 | } | |
217 | } | |
218 | }; | |
219 | ||
220 | // Test one-way conversion UTF-16->encoding. | |
221 | class FromUnicode : public Command { | |
222 | protected: | |
223 | FromUnicode(const UtfPerformanceTest &testcase) : Command(testcase) {} | |
224 | public: | |
225 | static UPerfFunction* get(const UtfPerformanceTest &testcase) { | |
226 | FromUnicode * t = new FromUnicode(testcase); | |
227 | if (U_SUCCESS(t->errorCode)){ | |
228 | return t; | |
229 | } else { | |
230 | delete t; | |
231 | return NULL; | |
232 | } | |
233 | } | |
234 | virtual void call(UErrorCode* pErrorCode){ | |
235 | const UChar *pIn, *pInLimit; | |
236 | char *pInter, *pInterLimit; | |
237 | ||
238 | ucnv_resetFromUnicode(cnv); | |
239 | fromUCallbackCount=0; | |
240 | ||
241 | pIn=input; | |
242 | pInLimit=input+inputLength; | |
243 | ||
244 | pInterLimit=intermediate+testcase.chunkLength; | |
245 | ||
246 | encodedLength=0; | |
247 | ||
248 | for(;;) { | |
249 | pInter=intermediate; | |
250 | ucnv_fromUnicode(cnv, &pInter, pInterLimit, &pIn, pInLimit, NULL, TRUE, pErrorCode); | |
251 | encodedLength+=(int32_t)(pInter-intermediate); | |
252 | ||
253 | if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { | |
254 | /* make sure that we convert once more to really flush */ | |
255 | *pErrorCode=U_ZERO_ERROR; | |
256 | } else if(U_FAILURE(*pErrorCode)) { | |
257 | return; | |
258 | } else { | |
259 | break; // all done | |
260 | } | |
261 | } | |
262 | } | |
263 | }; | |
264 | ||
265 | // Test one-way conversion UTF-8->encoding. | |
266 | class FromUTF8 : public Command { | |
267 | protected: | |
268 | FromUTF8(const UtfPerformanceTest &testcase) | |
269 | : Command(testcase), | |
270 | utf8Cnv(NULL), | |
271 | input8(utf8), input8Length(utf8Length) { | |
272 | utf8Cnv=ucnv_open("UTF-8", &errorCode); | |
273 | } | |
274 | public: | |
275 | static UPerfFunction* get(const UtfPerformanceTest &testcase) { | |
276 | FromUTF8 * t = new FromUTF8(testcase); | |
277 | if (U_SUCCESS(t->errorCode)){ | |
278 | return t; | |
279 | } else { | |
280 | delete t; | |
281 | return NULL; | |
282 | } | |
283 | } | |
284 | ~FromUTF8() { | |
285 | ucnv_close(utf8Cnv); | |
286 | } | |
287 | virtual void call(UErrorCode* pErrorCode){ | |
288 | const char *pIn, *pInLimit; | |
289 | char *pInter, *pInterLimit; | |
290 | UChar *pivotSource, *pivotTarget, *pivotLimit; | |
291 | ||
292 | ucnv_resetToUnicode(utf8Cnv); | |
293 | ucnv_resetFromUnicode(cnv); | |
294 | fromUCallbackCount=0; | |
295 | ||
296 | pIn=input8; | |
297 | pInLimit=input8+input8Length; | |
298 | ||
299 | pInterLimit=intermediate+testcase.chunkLength; | |
300 | ||
301 | pivotSource=pivotTarget=pivot; | |
302 | pivotLimit=pivot+testcase.pivotLength; | |
303 | ||
304 | encodedLength=0; | |
305 | ||
306 | for(;;) { | |
307 | pInter=intermediate; | |
308 | ucnv_convertEx(cnv, utf8Cnv, | |
309 | &pInter, pInterLimit, | |
310 | &pIn, pInLimit, | |
311 | pivot, &pivotSource, &pivotTarget, pivotLimit, | |
312 | FALSE, TRUE, pErrorCode); | |
313 | encodedLength+=(int32_t)(pInter-intermediate); | |
314 | ||
315 | if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { | |
316 | /* make sure that we convert once more to really flush */ | |
317 | *pErrorCode=U_ZERO_ERROR; | |
318 | } else if(U_FAILURE(*pErrorCode)) { | |
319 | return; | |
320 | } else { | |
321 | break; // all done | |
322 | } | |
323 | } | |
324 | } | |
325 | protected: | |
326 | UConverter *utf8Cnv; | |
327 | const char *input8; | |
328 | int32_t input8Length; | |
329 | }; | |
330 | ||
331 | UPerfFunction* UtfPerformanceTest::runIndexedTest(int32_t index, UBool exec, const char* &name, char* par) { | |
332 | switch (index) { | |
333 | case 0: name = "Roundtrip"; if (exec) return Roundtrip::get(*this); break; | |
334 | case 1: name = "FromUnicode"; if (exec) return FromUnicode::get(*this); break; | |
335 | case 2: name = "FromUTF8"; if (exec) return FromUTF8::get(*this); break; | |
336 | default: name = ""; break; | |
337 | } | |
338 | return NULL; | |
339 | } | |
340 | ||
341 | int main(int argc, const char *argv[]) | |
342 | { | |
343 | // Default values for command-line options. | |
344 | options[CHARSET].value = "UTF-8"; | |
345 | options[CHUNK_LENGTH].value = "4096"; | |
346 | options[PIVOT_LENGTH].value = "1024"; | |
347 | ||
348 | UErrorCode status = U_ZERO_ERROR; | |
349 | UtfPerformanceTest test(argc, argv, status); | |
350 | ||
351 | if (U_FAILURE(status)){ | |
352 | printf("The error is %s\n", u_errorName(status)); | |
353 | test.usage(); | |
354 | return status; | |
355 | } | |
356 | ||
357 | if (test.run() == FALSE){ | |
358 | fprintf(stderr, "FAILED: Tests could not be run please check the " | |
359 | "arguments.\n"); | |
360 | return -1; | |
361 | } | |
362 | ||
363 | if (fromUCallbackCount > 0) { | |
364 | printf("Number of fromUnicode callback calls in the last iteration: %ld\n", (long)fromUCallbackCount); | |
365 | } | |
366 | ||
367 | return 0; | |
368 | } |