├── LICENSE ├── README.md ├── exercise ├── 00-hello-world │ └── hello.cu ├── 1A-func-eval-1D │ ├── Makefile │ └── func_eval.cu ├── 1B-daxpy │ ├── Makefile │ └── daxpy.cu ├── 20-hello-world-2D │ ├── Makefile │ └── hello2D.cu ├── 3A-func-eval-2D │ ├── Makefile │ └── func_eval.cu ├── 3B-image-filtering │ ├── Makefile │ ├── caonima.jpg │ ├── filter-cpu.cu │ ├── filter.cu │ ├── original.png │ ├── reference.png │ └── visualize.m ├── 40-reduction │ ├── Makefile │ └── reduction.cu ├── Makefile.inc ├── matlab │ ├── fourier.m │ └── iron-man.jpg └── util │ ├── png.h │ └── util.h └── solution ├── 00-hello-world └── hello.cu ├── 1A-func-eval-1D ├── .cproject ├── .project ├── Makefile ├── func_eval.cu └── func_eval_0copy.cu ├── 1B-daxpy └── daxpy.cu ├── 20-hello-world-2D ├── Makefile └── hello2d.cu ├── 3A-func-eval-2D ├── .cproject ├── .project ├── Makefile └── func_eval.cu ├── Makefile.inc ├── extra ├── 04_image_filtering │ ├── Makefile │ ├── filter.cu │ ├── image.jpg │ ├── original.png │ ├── reference.png │ └── visualize.m ├── 05_reduction │ ├── Makefile │ └── reduction.cu ├── 06_unsharp_mark │ ├── Makefile │ ├── filter.m │ ├── original.png │ ├── reference.png │ ├── result.png │ ├── unsharp_mask.cu │ └── visualize.m ├── N0_event_timing │ ├── Makefile │ └── timing.cu ├── N1_root_finding │ ├── .cproject │ ├── .project │ ├── Makefile │ ├── polynomial.nb │ └── root_finding.cu ├── N2_64bit_shuffle │ ├── Makefile │ └── shuffle.cu ├── N3_func_eval_template │ ├── .cproject │ ├── .project │ ├── Makefile │ └── func_eval.cu └── N4_templates │ ├── Makefile │ └── templates.cu └── util ├── .cproject ├── .project ├── bmp.h.orig ├── math.h ├── png.h └── util.h /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | {one line to give the program's name and a brief idea of what it does.} 635 | Copyright (C) {year} {name of author} 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | {project} Copyright (C) {year} {fullname} 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # CUDA-tutorial -------------------------------------------------------------------------------- /exercise/00-hello-world/hello.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | 5 | __global__ void hello_gpu() 6 | { 7 | printf( "\"Hello, world!\", says the GPU.\n" ); 8 | } 9 | 10 | void hello_cpu() 11 | { 12 | printf( "\"Hello, world!\", says the CPU.\n" ); 13 | } 14 | 15 | // host code entrance 16 | int main( int argc, char **argv ) 17 | { 18 | hello_cpu(); 19 | hello_gpu <<< 16, 512>>>(); 20 | cudaDeviceSynchronize(); 21 | } 22 | -------------------------------------------------------------------------------- /exercise/1A-func-eval-1D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /exercise/1A-func-eval-1D/func_eval.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __inline__ __host__ __device__ double f( double x ) 13 | { 14 | return sin( 2.0 * x ) * cos( 7.0 * x ) * exp( x ); 15 | } 16 | 17 | __global__ void evaluate( double *y, const int n ) 18 | { 19 | int i = global_thread_id(); 20 | y[i] = f( ( double )i / ( double )n ); 21 | } 22 | 23 | // host code entrance 24 | int main( int argc, char **argv ) 25 | { 26 | int N = 128 * 1024 * 1024; 27 | 28 | // timing register 29 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1, t_GPU_2; 30 | // allocate host memory 31 | double *hst_y, *ref_y; 32 | hst_y = new double[N]; 33 | ref_y = new double[N]; 34 | // allocate device memory 35 | double *dev_y; 36 | cudaMalloc( &dev_y, N * sizeof( double ) ); 37 | 38 | t_GPU_0 = get_time(); 39 | 40 | // do computation on GPU 41 | evaluate <<< N / 1024, 1024 >>> ( dev_y, N ); 42 | cudaDeviceSynchronize(); 43 | 44 | t_GPU_1 = get_time(); 45 | 46 | // copy result back to CPU 47 | cudaMemcpy( hst_y, dev_y, N * sizeof( double ), cudaMemcpyDefault ); 48 | 49 | t_GPU_2 = get_time(); 50 | 51 | t_CPU_0 = get_time(); 52 | 53 | // calculate reference value 54 | #pragma omp parallel for 55 | for( int i = 0; i < N; i++ ) ref_y[i] = f( ( double )i / ( double )N ); 56 | 57 | t_CPU_1 = get_time(); 58 | 59 | // compare 60 | bool match = true; 61 | for( int i = 0; i < N; i++ ) { 62 | match = match && 63 | ( fabs( ref_y[i] - hst_y[i] ) < 8 * std::numeric_limits::epsilon() ); 64 | } 65 | 66 | // output 67 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 68 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 69 | std::cout << "Data transfer from GPU took " << t_GPU_2 - t_GPU_1 << " secs." << std::endl; 70 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 71 | 72 | // free up resources 73 | delete [] hst_y; 74 | delete [] ref_y; 75 | cudaDeviceReset(); 76 | } 77 | -------------------------------------------------------------------------------- /exercise/1B-daxpy/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /exercise/1B-daxpy/daxpy.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | struct Timer { 7 | double t0; 8 | Timer() : t0( omp_get_wtime() ) {} 9 | double stop() { return omp_get_wtime() - t0; } 10 | }; 11 | 12 | __global__ void daxpy( double *x, double *y, double *z, const int N ) { 13 | int n_threads = gridDim.x * blockDim.x; 14 | int my_tid = blockIdx.x * blockDim.x + threadIdx.x; 15 | for(int i = my_tid; i < N; i += n_threads ) z[i] = 2. * x[i] + y[i]; 16 | } 17 | 18 | int main() { 19 | int N = 128 * 1024 * 1024; 20 | std::vector x(N), y(N), z(N); 21 | 22 | // initialize the array with random contents 23 | printf("initializing..."); 24 | fflush(stdout); 25 | for(int i = 0; i < N; i++) { 26 | x[i] = rand() / double(RAND_MAX); 27 | y[i] = rand() / double(RAND_MAX); 28 | } 29 | printf("done\n"); 30 | 31 | // calculate on CPU, serial 32 | Timer t_serial; 33 | for(int i=0;i hst_z(N); 45 | cudaMalloc( &dev_x, N * sizeof(double) ); 46 | cudaMalloc( &dev_y, N * sizeof(double) ); 47 | cudaMalloc( &dev_z, N * sizeof(double) ); 48 | Timer t_upload; 49 | cudaMemcpy( dev_x, x.data(), N * sizeof(double), cudaMemcpyDefault ); 50 | cudaMemcpy( dev_y, y.data(), N * sizeof(double), cudaMemcpyDefault ); 51 | printf("GPU, data upload: %.3lf s\n", t_upload.stop() ); 52 | Timer t_gpu; 53 | daxpy<<< 30, 1024 >>>( dev_x, dev_y, dev_z, N ); 54 | cudaDeviceSynchronize(); 55 | printf("GPU, computation: %.3lf s\n", t_gpu.stop() ); 56 | Timer t_download; 57 | cudaMemcpy( hst_z.data(), dev_z, N * sizeof(double), cudaMemcpyDefault ); 58 | printf("GPU, data download: %.3lf s\n", t_download.stop() ); 59 | 60 | // correctness check 61 | bool pass = true; 62 | for(int i=0;i 1e-10 ) pass = false; 63 | printf("Correctness check: %s\n", pass ? "PASS" : "FAIL" ); 64 | 65 | } 66 | 67 | -------------------------------------------------------------------------------- /exercise/20-hello-world-2D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /exercise/20-hello-world-2D/hello2D.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | 5 | __global__ void hello_gpu() 6 | { 7 | printf( "\"Hello, world!\", says GPU block (%d,%d) thread (%d,%d).\n", blockIdx.x, blockIdx.y, threadIdx.x, threadIdx.y ); 8 | } 9 | 10 | void hello_cpu() 11 | { 12 | printf( "\"Hello, world!\", says the CPU.\n" ); 13 | } 14 | 15 | // host code entrance 16 | int main( int argc, char **argv ) 17 | { 18 | hello_cpu(); 19 | printf( "launching 2x2 blocks each containing 4 threads\n" ); 20 | hello_gpu <<< dim3( 2, 2, 1 ), dim3( 4, 1, 1 ) >>>(); 21 | cudaDeviceSynchronize(); 22 | printf( "launching 2x2 blocks each containing 2x2 threads\n" ); 23 | hello_gpu <<< dim3( 2, 2, 1 ), dim3( 2, 2, 1 ) >>>(); 24 | cudaDeviceSynchronize(); 25 | cudaDeviceSynchronize(); 26 | } 27 | -------------------------------------------------------------------------------- /exercise/3A-func-eval-2D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /exercise/3A-func-eval-2D/func_eval.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __inline__ __host__ __device__ double f( double x ) 13 | { 14 | // TO-DO 15 | } 16 | 17 | __global__ void evaluate( double *y, const int n ) 18 | { 19 | // TO-DO 20 | } 21 | 22 | // host code entrance 23 | int main( int argc, char **argv ) 24 | { 25 | int N = 1024 /*TO-DO: adjust N*/; 26 | 27 | // timing register 28 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1, t_GPU_2; 29 | // allocate host memory 30 | double *hst_y, *ref_y; 31 | hst_y = new double[N]; 32 | ref_y = new double[N]; 33 | // allocate device memory 34 | double * dev_y; 35 | cudaMalloc( /*TO-DO*/ ); 36 | 37 | t_GPU_0 = get_time(); 38 | 39 | // do computation on GPU 40 | evaluate <<< /*TO-DO*/ >>> ( /*TO-DO*/ ); 41 | cudaDeviceSynchronize(); 42 | 43 | t_GPU_1 = get_time(); 44 | 45 | // copy result back to CPU 46 | cudaMemcpy( /*TO-DO*/ ); 47 | 48 | t_GPU_2 = get_time(); 49 | 50 | t_CPU_0 = get_time(); 51 | 52 | // calculate reference value 53 | #pragma omp parallel for 54 | for( int i = 0; i < N; i++ ) ref_y[i] = f( i * 1.0 / N ); 55 | 56 | t_CPU_1 = get_time(); 57 | 58 | // compare 59 | bool match = true; 60 | for( int i = 0; i < N; i++ ) { 61 | match = match && 62 | ( fabs( ref_y[i] - hst_y[i] ) < 16 * std::numeric_limits::epsilon() ); 63 | } 64 | 65 | // output 66 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 67 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 68 | std::cout << "Data transfer from GPU took " << t_GPU_2 - t_GPU_1 << " secs." << std::endl; 69 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 70 | 71 | // free up resources 72 | delete [] hst_y; 73 | delete [] ref_y; 74 | cudaDeviceReset(); 75 | } 76 | -------------------------------------------------------------------------------- /exercise/3B-image-filtering/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) result.png 15 | -------------------------------------------------------------------------------- /exercise/3B-image-filtering/caonima.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/exercise/3B-image-filtering/caonima.jpg -------------------------------------------------------------------------------- /exercise/3B-image-filtering/filter-cpu.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | #include "../util/util.h" 13 | #include "../util/png.h" 14 | 15 | struct png_image { 16 | uint width, height; 17 | std::vector pixel; 18 | 19 | png_image( const char fn[] ) { 20 | load( fn ); 21 | } 22 | uint load( const char fn[] ) { 23 | return lodepng::decode( pixel, width, height, fn, LCT_GREY, 8 ); 24 | } 25 | uint save( const char fn[] ) { 26 | return lodepng::encode( fn, &pixel[0], width, height, LCT_GREY, 8 ); 27 | } 28 | unsigned char *data() { 29 | return &pixel[0]; 30 | } 31 | }; 32 | 33 | // host code entrance 34 | int main( int argc, char **argv ) 35 | { 36 | // the grayscale image is stored in image.data(), with one value for each pixel 37 | png_image image( argc > 1 ? argv[1] : "original.png" ); 38 | png_image new_image = image; 39 | 40 | // image.data() returns pointer to the pixel data 41 | 42 | #pragma omp parallel for 43 | for( int i = 0; i < image.height; i++ ) { 44 | for( int j = 0; j < image.width; j++ ) { 45 | float sum = 0.f; 46 | for( int k = 0; k < 3; k++ ) { 47 | for( int l = 0; l < 3; l++ ) { 48 | int K = i - 1 + k; 49 | int L = j - 1 + l; 50 | // using 'symmetric' boundary here 51 | if( K < 0 ) K = -K; 52 | if( L < 0 ) L = -L; 53 | if( K >= image.height ) K = 2 * image.height - K; 54 | if( L >= image.width ) L = 2 * image.width - L; 55 | // adding the contributions 56 | sum += image.pixel[K * image.width + L] * 0.111111; 57 | } 58 | } 59 | new_image.pixel[i * new_image.width + j] = sum; 60 | } 61 | } 62 | 63 | new_image.save( "result.png" ); 64 | } 65 | -------------------------------------------------------------------------------- /exercise/3B-image-filtering/filter.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | #include "../util/util.h" 13 | #include "../util/png.h" 14 | 15 | struct png_image { 16 | uint width, height; 17 | std::vector pixel; 18 | 19 | png_image( const char fn[] ) { 20 | load( fn ); 21 | } 22 | uint load( const char fn[] ) { 23 | return lodepng::decode( pixel, width, height, fn, LCT_GREY, 8 ); 24 | } 25 | uint save( const char fn[] ) { 26 | return lodepng::encode( fn, &pixel[0], width, height, LCT_GREY, 8 ); 27 | } 28 | unsigned char *data() { 29 | return &pixel[0]; 30 | } 31 | }; 32 | 33 | __global__ void filter( const unsigned char * image, unsigned char * new_image, const int width, const int height ) 34 | { 35 | int i = threadIdx.y + blockIdx.y * blockDim.y; 36 | int j = threadIdx.x + blockIdx.x * blockDim.x; 37 | if( i < height && j < width ) { 38 | float sum = 0.f; 39 | for( int k = 0; k < 3; k++ ) { 40 | for( int l = 0; l < 3; l++ ) { 41 | int K = i - 1 + k; 42 | int L = j - 1 + l; 43 | // using 'symmetric' boundary here 44 | if( K < 0 ) K = -K; 45 | if( L < 0 ) L = -L; 46 | if( K >= height ) K = 2 * height - K; 47 | if( L >= width ) L = 2 * width - L; 48 | // adding the contributions 49 | sum += image[K * width + L] * 0.111111; 50 | } 51 | } 52 | new_image[i * width + j] = sum; 53 | } 54 | } 55 | 56 | // host code entrance 57 | int main( int argc, char **argv ) 58 | { 59 | // the grayscale image is stored in image.data(), with one value for each pixel 60 | png_image image( argc > 1 ? argv[1] : "original.png" ); 61 | png_image new_image = image; 62 | 63 | unsigned char *dev_image, *dev_new_image; 64 | cudaMalloc( &dev_image, sizeof( unsigned char ) * image.width * image.height ); 65 | cudaMalloc( &dev_new_image, sizeof( unsigned char ) * image.width * image.height ); 66 | cudaMemcpy( dev_image, image.data(), sizeof( unsigned char ) * image.width * image.height, cudaMemcpyDefault ); 67 | 68 | // image.data() returns pointer to the pixel data 69 | dim3 n_threads( 16, 16 ); 70 | dim3 n_blocks( ( image.width + n_threads.x - 1 ) / n_threads.x, ( image.height + n_threads.y - 1 ) / n_threads.y ); 71 | filter <<< n_blocks, n_threads>>>( dev_image, dev_new_image, image.width, image.height ); 72 | 73 | cudaMemcpy( new_image.data(), dev_new_image, sizeof( unsigned char ) * image.width * image.height, cudaMemcpyDefault ); 74 | 75 | new_image.save( "result.png" ); 76 | } 77 | -------------------------------------------------------------------------------- /exercise/3B-image-filtering/original.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/exercise/3B-image-filtering/original.png -------------------------------------------------------------------------------- /exercise/3B-image-filtering/reference.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/exercise/3B-image-filtering/reference.png -------------------------------------------------------------------------------- /exercise/3B-image-filtering/visualize.m: -------------------------------------------------------------------------------- 1 | hold on; 2 | 3 | filter = ones( 3 ) / 9; 4 | ori = imread('original.png'); 5 | ref = imfilter( ori, filter, 'symmetric' ); 6 | imwrite( ref, 'reference.png' ); 7 | try 8 | res = imread('result.png'); 9 | catch 10 | res = uint8( zeros( size(ori) ) ); 11 | end 12 | 13 | subplot(1,3,1); 14 | imshow(ori); 15 | title('original'); 16 | subplot(1,3,2); 17 | imshow(ref); 18 | title('reference'); 19 | subplot(1,3,3); 20 | imshow(res); 21 | title('result'); 22 | 23 | -------------------------------------------------------------------------------- /exercise/40-reduction/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /exercise/40-reduction/reduction.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __global__ void mean( float *value, float *sum, const int n ) 13 | { 14 | // TO-DO 15 | } 16 | 17 | // host code entrance 18 | int main( int argc, char **argv ) 19 | { 20 | int N = 128 * 1024 * 1024; 21 | 22 | // allocate host memory 23 | float hst_sum = 0.f, ref_sum = 0.f; 24 | float *hst_value; 25 | hst_value = new float[N]; 26 | 27 | srand( 0 ); // deterministic result can be helpful for debugging 28 | for( int i = 0; i < N; i++ ) hst_value[i] = float( rand() ) / float( RAND_MAX ); 29 | 30 | // allocate device memory 31 | /*TO-DO*/ 32 | 33 | // do computation on GPU 34 | mean <<< /*TO-DO*/ >>> ( /*TO-DO*/ ); 35 | 36 | // copy result back to CPU 37 | /*TO-DO*/ 38 | 39 | // calculate reference value 40 | for( int i = 0; i < N; i++ ) ref_sum += hst_value[i]; 41 | 42 | hst_sum /= N; 43 | ref_sum /= N; 44 | 45 | // compare & output 46 | std::cout << "CPU result: " << ref_sum << std::endl; 47 | std::cout << "GPU result: " << hst_sum << std::endl; 48 | std::cout << "CPU/GPU result match: " << ( fabs( hst_sum - ref_sum ) < 10 * std::numeric_limits::epsilon() ? "YES" : "NO" ) << std::endl; 49 | 50 | // free up resources 51 | delete [] hst_value; 52 | cudaDeviceReset(); 53 | } 54 | -------------------------------------------------------------------------------- /exercise/Makefile.inc: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | CC = nvcc 4 | CFLAGS = -O3 -arch=sm_35 -Xcompiler -fopenmp 5 | DPFLAGS = -rdc=true -lcudadevrt 6 | SHELL = /bin/bash 7 | RM = rm 8 | -------------------------------------------------------------------------------- /exercise/matlab/fourier.m: -------------------------------------------------------------------------------- 1 | %% load image 2 | img = imread( 'iron-man.jpg' ); 3 | img_gray = rgb2gray( img ); 4 | 5 | %% do FFT on CPU 6 | fprintf('performing FFT on CPU\n'); 7 | tic; fft_cpu = fft2( img_gray ); toc 8 | spectral_cpu = log( abs( fftshift( fft_cpu ) ) ); 9 | 10 | %% do FFT on GPU 11 | fprintf('performing FFT on GPU\n'); 12 | try 13 | img_gpu = gpuArray( img_gray ); 14 | tic; fft_gpu = fft2( img_gpu ); toc 15 | fft_gpu = gather( fft_gpu ); 16 | spectral_gpu = log( abs( fftshift( fft_gpu ) ) ); 17 | end 18 | 19 | %% visualize result 20 | close all 21 | figure(1); imshow(img) 22 | figure(2); imagesc( spectral_cpu ); colormap jet 23 | try 24 | figure(3); imagesc( spectral_gpu ); colormap jet 25 | end -------------------------------------------------------------------------------- /exercise/matlab/iron-man.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/exercise/matlab/iron-man.jpg -------------------------------------------------------------------------------- /exercise/util/util.h: -------------------------------------------------------------------------------- 1 | #ifndef UTIL_H_ 2 | #define UTIL_H_ 3 | 4 | #ifdef __CUDACC__ 5 | 6 | __device__ int block_id() { 7 | return blockIdx.x + blockIdx.y * gridDim.x + blockIdx.z * gridDim.x * gridDim.y; 8 | } 9 | 10 | __device__ int block_num() { 11 | return gridDim.x * gridDim.y * gridDim.z; 12 | } 13 | 14 | __device__ int block_size() { 15 | return blockDim.x * blockDim.y * blockDim.z; 16 | } 17 | 18 | __device__ int local_thread_id() { 19 | return threadIdx.x + threadIdx.y * blockDim.x + threadIdx.z * blockDim.x * blockDim.y; 20 | } 21 | 22 | __device__ int global_thread_id() { 23 | return block_id() * block_size() + local_thread_id(); 24 | } 25 | 26 | __device__ int global_thread_num() { 27 | return block_num() * block_size(); 28 | } 29 | 30 | #endif 31 | 32 | #include 33 | #include 34 | 35 | double get_time() 36 | { 37 | struct timespec time; 38 | clock_gettime( CLOCK_REALTIME, &time ); 39 | return (double)time.tv_sec + (double)time.tv_nsec * 1.0e-9 ; 40 | } 41 | 42 | #endif 43 | -------------------------------------------------------------------------------- /solution/00-hello-world/hello.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | 5 | __global__ void hello_gpu() 6 | { 7 | printf( "\"Hello, world!\", says the GPU.\n" ); 8 | } 9 | 10 | void hello_cpu() 11 | { 12 | printf( "\"Hello, world!\", says the CPU.\n" ); 13 | } 14 | 15 | // host code entrance 16 | int main( int argc, char **argv ) 17 | { 18 | hello_cpu(); 19 | hello_gpu <<< 1, 1>>>(); 20 | cudaDeviceSynchronize(); 21 | } 22 | -------------------------------------------------------------------------------- /solution/1A-func-eval-1D/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | -------------------------------------------------------------------------------- /solution/1A-func-eval-1D/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | 0_0_hello_world 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /solution/1A-func-eval-1D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/1A-func-eval-1D/func_eval.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __inline__ __host__ __device__ double f( double x ) { 13 | return sin( 2.0*x ) * cos( 7.0*x ) * exp( x ); 14 | } 15 | 16 | __global__ void evaluate( double *y, const int n ) 17 | { 18 | int i = global_thread_id(); 19 | y[i] = f( (double)i / (double)n ); 20 | } 21 | 22 | // host code entrance 23 | int main( int argc, char **argv ) 24 | { 25 | int N = 128 * 1024 * 1024; 26 | 27 | // timing register 28 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1, t_GPU_2; 29 | // allocate host memory 30 | double *hst_y, *ref_y; 31 | hst_y = new double[N]; 32 | ref_y = new double[N]; 33 | // allocate device memory 34 | double *dev_y; 35 | cudaMalloc( &dev_y, N * sizeof( double ) ); 36 | 37 | t_GPU_0 = get_time(); 38 | 39 | // do computation on GPU 40 | evaluate <<< N / 1024, 1024 >>> ( dev_y, N ); 41 | cudaDeviceSynchronize(); 42 | 43 | t_GPU_1 = get_time(); 44 | 45 | // copy result back to CPU 46 | cudaMemcpy( hst_y, dev_y, N * sizeof( double ), cudaMemcpyDefault ); 47 | 48 | t_GPU_2 = get_time(); 49 | 50 | t_CPU_0 = get_time(); 51 | 52 | // calculate reference value 53 | #pragma omp parallel for 54 | for( int i = 0; i < N; i++ ) ref_y[i] = f( (double)i / (double)N ); 55 | 56 | t_CPU_1 = get_time(); 57 | 58 | // compare 59 | bool match = true; 60 | for( int i = 0; i < N; i++ ) { 61 | match = match && 62 | ( fabs( ref_y[i] - hst_y[i] ) < 8 * std::numeric_limits::epsilon() ); 63 | } 64 | 65 | // output 66 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 67 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 68 | std::cout << "Data transfer from GPU took " << t_GPU_2 - t_GPU_1 << " secs." << std::endl; 69 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 70 | 71 | // free up resources 72 | delete [] hst_y; 73 | delete [] ref_y; 74 | cudaDeviceReset(); 75 | } 76 | -------------------------------------------------------------------------------- /solution/1A-func-eval-1D/func_eval_0copy.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __inline__ __host__ __device__ double f( double x ) { 13 | return sin( 2.0*x ) * cos( 7.0*x ) * exp( x ); 14 | } 15 | 16 | __global__ void evaluate( double *y, const int n ) 17 | { 18 | int i = global_thread_id(); 19 | y[i] = f( (double)i / (double)n ); 20 | } 21 | 22 | // host code entrance 23 | int main( int argc, char **argv ) 24 | { 25 | int N = 128 * 1024 * 1024; 26 | 27 | // timing register 28 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1, t_GPU_2; 29 | // allocate host memory 30 | double *hst_y, *ref_y; 31 | #if 0 32 | hst_y = new double[N]; 33 | #else 34 | double t_hal_0 = omp_get_wtime(); 35 | cudaMallocHost( &hst_y, sizeof(double)*N ); 36 | double t_hal_1 = omp_get_wtime(); 37 | printf("cudaMallocHost took %lf ms\n", (t_hal_1-t_hal_0)*1000); 38 | #endif 39 | double t_cnew_0 = omp_get_wtime(); 40 | ref_y = new double[N]; 41 | double t_cnew_1 = omp_get_wtime(); 42 | printf("C++ operator new took %lf ms\n", (t_cnew_1-t_cnew_0)*1000); 43 | // allocate device memory 44 | double *dev_y; 45 | cudaSetDevice(0); 46 | double t0 = omp_get_wtime(); 47 | //cudaMalloc( &dev_y, N * sizeof( double ) ); 48 | dev_y = hst_y; 49 | double t1 = omp_get_wtime(); 50 | printf("cudaMalloc took %lf ms\n", (t1-t0)*1000); 51 | 52 | t_GPU_0 = get_time(); 53 | 54 | // do computation on GPU 55 | evaluate <<< N / 1024, 1024 >>> ( dev_y, N ); 56 | cudaDeviceSynchronize(); 57 | 58 | t_GPU_1 = get_time(); 59 | 60 | // copy result back to CPU 61 | // cudaMemcpy( hst_y, dev_y, N * sizeof( double ), cudaMemcpyDefault ); 62 | 63 | t_GPU_2 = get_time(); 64 | 65 | t_CPU_0 = get_time(); 66 | 67 | // calculate reference value 68 | #pragma omp parallel for 69 | for( int i = 0; i < N; i++ ) ref_y[i] = f( (double)i / (double)N ); 70 | 71 | t_CPU_1 = get_time(); 72 | 73 | // compare 74 | bool match = true; 75 | for( int i = 0; i < N; i++ ) { 76 | match = match && 77 | ( fabs( ref_y[i] - hst_y[i] ) < 8 * std::numeric_limits::epsilon() ); 78 | } 79 | 80 | // output 81 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 82 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 83 | std::cout << "Data transfer from GPU took " << t_GPU_2 - t_GPU_1 << " secs. Bandwidth = " << double(N)*sizeof(double)/(t_GPU_2-t_GPU_1)/1e9 << " GB/s" << std::endl; 84 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 85 | 86 | // free up resources 87 | //delete [] hst_y; 88 | delete [] ref_y; 89 | cudaDeviceReset(); 90 | } 91 | -------------------------------------------------------------------------------- /solution/1B-daxpy/daxpy.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | struct Timer { 7 | double t0; 8 | Timer() : t0( omp_get_wtime() ) {} 9 | double stop() { return omp_get_wtime() - t0; } 10 | }; 11 | 12 | __global__ void daxpy( double *x, double *y, double *z, const int N ) { 13 | int n_threads = gridDim.x * blockDim.x; 14 | int my_tid = blockIdx.x * blockDim.x + threadIdx.x; 15 | for(int i = my_tid; i < N; i += n_threads ) z[i] = 2. * x[i] + y[i]; 16 | } 17 | 18 | int main() { 19 | int N = 128 * 1024 * 1024; 20 | std::vector x(N), y(N), z(N); 21 | 22 | // initialize the array with random contents 23 | printf("initializing..."); 24 | fflush(stdout); 25 | for(int i = 0; i < N; i++) { 26 | x[i] = rand() / double(RAND_MAX); 27 | y[i] = rand() / double(RAND_MAX); 28 | } 29 | printf("done\n"); 30 | 31 | // calculate on CPU, serial 32 | Timer t_serial; 33 | for(int i=0;i hst_z(N); 45 | cudaMalloc( &dev_x, N * sizeof(double) ); 46 | cudaMalloc( &dev_y, N * sizeof(double) ); 47 | cudaMalloc( &dev_z, N * sizeof(double) ); 48 | Timer t_upload; 49 | cudaMemcpy( dev_x, x.data(), N * sizeof(double), cudaMemcpyDefault ); 50 | cudaMemcpy( dev_y, y.data(), N * sizeof(double), cudaMemcpyDefault ); 51 | printf("GPU, data upload: %.3lf s\n", t_upload.stop() ); 52 | Timer t_gpu; 53 | daxpy<<< 30, 1024 >>>( dev_x, dev_y, dev_z, N ); 54 | cudaDeviceSynchronize(); 55 | printf("GPU, computation: %.3lf s\n", t_gpu.stop() ); 56 | Timer t_download; 57 | cudaMemcpy( hst_z.data(), dev_z, N * sizeof(double), cudaMemcpyDefault ); 58 | printf("GPU, data download: %.3lf s\n", t_download.stop() ); 59 | 60 | // correctness check 61 | bool pass = true; 62 | for(int i=0;i 1e-10 ) pass = false; 63 | printf("Correctness check: %s\n", pass ? "PASS" : "FAIL" ); 64 | 65 | } 66 | 67 | -------------------------------------------------------------------------------- /solution/20-hello-world-2D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/20-hello-world-2D/hello2d.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | 5 | __global__ void hello_gpu() 6 | { 7 | printf( "\"Hello, world!\", says GPU block (%d,%d) thread (%d,%d).\n", blockIdx.x, blockIdx.y, threadIdx.x, threadIdx.y ); 8 | } 9 | 10 | void hello_cpu() 11 | { 12 | printf( "\"Hello, world!\", says the CPU.\n" ); 13 | } 14 | 15 | // host code entrance 16 | int main( int argc, char **argv ) 17 | { 18 | hello_cpu(); 19 | printf("launching 2x2 blocks each containing 4 threads\n"); 20 | hello_gpu <<< dim3(2,2,1), dim3(4,1,1) >>>(); 21 | cudaDeviceSynchronize(); 22 | printf("launching 2x2 blocks each containing 2x2 threads\n"); 23 | hello_gpu <<< dim3(2,2,1), dim3(2,2,1) >>>(); 24 | cudaDeviceSynchronize(); 25 | cudaDeviceSynchronize(); 26 | } 27 | -------------------------------------------------------------------------------- /solution/3A-func-eval-2D/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | -------------------------------------------------------------------------------- /solution/3A-func-eval-2D/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | 0_0_hello_world 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /solution/3A-func-eval-2D/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/3A-func-eval-2D/func_eval.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | 9 | #include "../util/util.h" 10 | 11 | __inline__ __host__ __device__ double f( double x, double y ) 12 | { 13 | return sin( 5.0 * x ) * cos( 16.0 * y ) * exp( x ); 14 | } 15 | 16 | __global__ void evaluate( double *z, const int n ) 17 | { 18 | int i = threadIdx.x + blockIdx.x * blockDim.x; 19 | int j = threadIdx.y + blockIdx.y * blockDim.y; 20 | z[i + j * n] = f( ( double )i / ( double )n, ( double )j / ( double )n ); 21 | } 22 | 23 | // host code entrance 24 | int main( int argc, char **argv ) 25 | { 26 | int N = 4 * 1024; 27 | 28 | // timing register 29 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1; 30 | // allocate host memory 31 | double *hst_z, *ref_z; 32 | hst_z = new double[N * N]; 33 | ref_z = new double[N * N]; 34 | // allocate device memory 35 | double *dev_z; 36 | cudaMalloc( &dev_z, N * N * sizeof( double ) ); 37 | 38 | t_GPU_0 = get_time(); 39 | 40 | // do computation on GPU 41 | dim3 n_blocks( N / 32, N / 32, 1 ), n_threads( 32, 32, 1 ); 42 | 43 | evaluate <<< n_blocks, n_threads >>> ( dev_z, N ); 44 | // copy result back to CPU 45 | cudaMemcpy( hst_z, dev_z, N * N * sizeof( double ), cudaMemcpyDefault ); 46 | 47 | t_GPU_1 = get_time(); 48 | 49 | t_CPU_0 = get_time(); 50 | 51 | // calculate reference value 52 | for( int i = 0; i < N; i++ ) { 53 | for( int j = 0 ; j < N ; j++ ) { 54 | ref_z[i + j * N] = f( ( double )i / ( double )N, ( double )j / ( double )N ); 55 | } 56 | } 57 | 58 | t_CPU_1 = get_time(); 59 | 60 | // compare 61 | bool match = true; 62 | for( int i = 0; i < N; i++ ) { 63 | for( int j = 0 ; j < N ; j++ ) { 64 | match = match && 65 | ( fabs( ref_z[i] - hst_z[i] ) < 8 * std::numeric_limits::epsilon() ); 66 | } 67 | } 68 | 69 | // output 70 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 71 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 72 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 73 | 74 | // free up resources 75 | delete [] hst_z; 76 | delete [] ref_z; 77 | cudaDeviceReset(); 78 | } 79 | -------------------------------------------------------------------------------- /solution/Makefile.inc: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | CC = nvcc 4 | CFLAGS = -O3 -arch=sm_35 -Xcompiler -fopenmp 5 | DPFLAGS = -rdc=true -lcudadevrt 6 | SHELL = /bin/bash 7 | RM = rm 8 | -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/filter.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | #include "../util/util.h" 13 | #include "../util/png.h" 14 | 15 | struct png_image { 16 | uint width, height; 17 | std::vector pixel; 18 | 19 | png_image( const char fn[] ) { 20 | load( fn ); 21 | } 22 | uint load( const char fn[] ) { 23 | return lodepng::decode( pixel, width, height, fn, LCT_GREY, 8 ); 24 | } 25 | uint save( const char fn[] ) { 26 | return lodepng::encode( fn, &pixel[0], width, height, LCT_GREY, 8 ); 27 | } 28 | unsigned char *data() { 29 | return &pixel[0]; 30 | } 31 | }; 32 | 33 | // host code entrance 34 | int main( int argc, char **argv ) 35 | { 36 | // the grayscale image is stored in image.data(), with one value for each pixel 37 | png_image image( argc > 1 ? argv[1] : "original.png" ); 38 | 39 | // the gaussian filter 40 | const static int r = 3; 41 | const static int d = 7; 42 | const static float gauss_filter [] = { 43 | 0.0049, 0.0092, 0.0134, 0.0152, 0.0134, 0.0092, 0.0049, \ 44 | 0.0092, 0.0172, 0.0250, 0.0283, 0.0250, 0.0172, 0.0092, \ 45 | 0.0134, 0.0250, 0.0364, 0.0412, 0.0364, 0.0250, 0.0134, \ 46 | 0.0152, 0.0283, 0.0412, 0.0467, 0.0412, 0.0283, 0.0152, \ 47 | 0.0134, 0.0250, 0.0364, 0.0412, 0.0364, 0.0250, 0.0134, \ 48 | 0.0092, 0.0172, 0.0250, 0.0283, 0.0250, 0.0172, 0.0092, \ 49 | 0.0049, 0.0092, 0.0134, 0.0152, 0.0134, 0.0092, 0.0049 50 | }; 51 | 52 | png_image new_image = image; 53 | 54 | // image.data() returns pointer to the pixel data 55 | 56 | #pragma omp parallel for 57 | for( int i = 0; i < image.height; i++ ) { 58 | for( int j = 0; j < image.width; j++ ) { 59 | float sum = 0.f; 60 | for( int k = 0; k < d; k++ ) { 61 | for( int l = 0; l < d; l++ ) { 62 | int K = i - r + k; 63 | int L = j - r + l; 64 | // using 'symmetric' boundary here 65 | if( K < 0 ) K = -K; 66 | if( L < 0 ) L = -L; 67 | if( K >= image.height ) K = 2 * image.height - K; 68 | if( L >= image.width ) L = 2 * image.width - L; 69 | // adding the contributions 70 | sum += image.pixel[K * image.width + L] * gauss_filter[k * d + l]; 71 | } 72 | } 73 | new_image.pixel[i * new_image.width + j] = sum; 74 | } 75 | } 76 | 77 | new_image.save( "result.png" ); 78 | } 79 | -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/image.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/04_image_filtering/image.jpg -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/original.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/04_image_filtering/original.png -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/reference.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/04_image_filtering/reference.png -------------------------------------------------------------------------------- /solution/extra/04_image_filtering/visualize.m: -------------------------------------------------------------------------------- 1 | hold on; 2 | 3 | ori = imread('original.png'); 4 | ref = imread('reference.png'); 5 | try 6 | res = imread('result.png'); 7 | catch 8 | res = uint8( zeros( size(ori) ) ); 9 | end 10 | 11 | subplot(2,3,1); 12 | imshow(ori); 13 | title('original'); 14 | subplot(2,3,2); 15 | imshow(ref); 16 | title('reference'); 17 | subplot(2,3,3); 18 | imshow(res); 19 | title('result'); 20 | 21 | rangex = 500:590; 22 | rangey = 600:760; 23 | subplot(2,3,4); 24 | imshow(ori(rangex,rangey)); 25 | title('original'); 26 | subplot(2,3,5); 27 | imshow(ref(rangex,rangey)); 28 | title('reference'); 29 | subplot(2,3,6); 30 | imshow(res(rangex,rangey)); 31 | title('result'); 32 | -------------------------------------------------------------------------------- /solution/extra/05_reduction/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/05_reduction/reduction.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __global__ void mean( float *value, float *sum, const int n ) 13 | { 14 | float my_sum = 0; 15 | for( int i = global_thread_id(); i < n; i += global_thread_num() ) { 16 | my_sum += value[i]; 17 | } 18 | atomicAdd( sum, my_sum ); 19 | } 20 | 21 | // host code entrance 22 | int main( int argc, char **argv ) 23 | { 24 | int N = 128 * 1024 * 1024; 25 | 26 | // allocate host memory 27 | float hst_sum = 0.f, ref_sum = 0.f; 28 | float *hst_value; 29 | hst_value = new float[N]; 30 | // allocate device memory 31 | float *dev_value, *dev_sum; 32 | cudaMalloc( &dev_value, N * sizeof( float ) ); 33 | cudaMalloc( &dev_sum, sizeof( float ) ); 34 | 35 | srand( 0 ); // deterministic result can be helpful for debugging 36 | for( int i = 0; i < N; i++ ) hst_value[i] = float( rand() ) / float( RAND_MAX ); 37 | 38 | cudaMemset( dev_sum, 0, sizeof( float ) ); 39 | cudaMemcpy( dev_value, hst_value, N * sizeof( float ), cudaMemcpyDefault ); 40 | 41 | // do computation on GPU 42 | mean <<< 16, 1024 >>> ( dev_value, dev_sum, N ); 43 | 44 | // copy result back to CPU 45 | cudaMemcpy( &hst_sum, dev_sum, sizeof( float ), cudaMemcpyDefault ); 46 | 47 | // calculate reference value 48 | for( int i = 0; i < N; i++ ) ref_sum += hst_value[i]; 49 | 50 | hst_sum /= N; 51 | ref_sum /= N; 52 | 53 | // compare & output 54 | std::cout << "CPU result: " << ref_sum << std::endl; 55 | std::cout << "GPU result: " << hst_sum << std::endl; 56 | std::cout << "CPU/GPU result match: " << ( fabs( hst_sum - ref_sum ) < 10 * std::numeric_limits::epsilon() ? "YES" : "NO" ) << std::endl; 57 | 58 | // free up resources 59 | delete [] hst_value; 60 | cudaDeviceReset(); 61 | } 62 | -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/filter.m: -------------------------------------------------------------------------------- 1 | gauss=fspecial('gaussian',5,1); 2 | lap=[-1,-1,-1;-1,9,-1;-1,-1,-1]/2; 3 | sharp=conv2( gauss, lap, 'full' ); 4 | [x,y]=meshgrid([0:1/6:1]); 5 | surf(x,y,sharp); -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/original.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/06_unsharp_mark/original.png -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/reference.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/06_unsharp_mark/reference.png -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/result.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/06_unsharp_mark/result.png -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/unsharp_mask.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | #include "../util/util.h" 13 | #include "../util/png.h" 14 | 15 | struct png_image { 16 | uint width, height; 17 | std::vector pixel; 18 | 19 | png_image( const char fn[] ) { 20 | load( fn ); 21 | } 22 | uint load( const char fn[] ) { 23 | return lodepng::decode( pixel, width, height, fn, LCT_GREY, 8 ); 24 | } 25 | uint save( const char fn[] ) { 26 | return lodepng::encode( fn, &pixel[0], width, height, LCT_GREY, 8 ); 27 | } 28 | unsigned char *data() { 29 | return &pixel[0]; 30 | } 31 | }; 32 | 33 | float point_filter( png_image &image, const float filter[], int i, int j, int d ) 34 | { 35 | float sum = 0.f; 36 | int r = ( d - 1 ) / 2; 37 | for( int k = 0; k < d; k++ ) { 38 | for( int l = 0; l < d; l++ ) { 39 | int K = i - r + k; 40 | int L = j - r + l; 41 | // using 'symmetric' boundary here 42 | if( K < 0 ) K = -K; 43 | if( L < 0 ) L = -L; 44 | if( K >= image.height ) K = 2 * image.height - K; 45 | if( L >= image.width ) L = 2 * image.width - L; 46 | // adding the contributions 47 | sum += image.pixel[K * image.width + L] * filter[k * d + l]; 48 | } 49 | } 50 | return sum; 51 | } 52 | 53 | // host code entrance 54 | int main( int argc, char **argv ) 55 | { 56 | // the grayscale image is stored in image.data(), with one value for each pixel 57 | png_image image( argc > 1 ? argv[1] : "original.png" ); 58 | 59 | const static float threshold = 125.0; 60 | // filters 61 | const static float gauss_filter [] = { 62 | 0.0049, 0.0092, 0.0134, 0.0152, 0.0134, 0.0092, 0.0049, \ 63 | 0.0092, 0.0172, 0.0250, 0.0283, 0.0250, 0.0172, 0.0092, \ 64 | 0.0134, 0.0250, 0.0364, 0.0412, 0.0364, 0.0250, 0.0134, \ 65 | 0.0152, 0.0283, 0.0412, 0.0467, 0.0412, 0.0283, 0.0152, \ 66 | 0.0134, 0.0250, 0.0364, 0.0412, 0.0364, 0.0250, 0.0134, \ 67 | 0.0092, 0.0172, 0.0250, 0.0283, 0.0250, 0.0172, 0.0092, \ 68 | 0.0049, 0.0092, 0.0134, 0.0152, 0.0134, 0.0092, 0.0049 69 | }; 70 | const static float sharp_filter [] = { 71 | -0.0015, -0.0081, -0.0191, -0.0243, -0.0191, -0.0081, -0.0015, \ 72 | -0.0081, -0.0298, -0.0382, -0.0234, -0.0382, -0.0298, -0.0081, \ 73 | -0.0191, -0.0382, 0.0523, 0.1792, 0.0523, -0.0382, -0.0191, \ 74 | -0.0243, -0.0234, 0.1792, 0.4136, 0.1792, -0.0234, -0.0243, \ 75 | -0.0191, -0.0382, 0.0523, 0.1792, 0.0523, -0.0382, -0.0191, \ 76 | -0.0081, -0.0298, -0.0382, -0.0234, -0.0382, -0.0298, -0.0081, \ 77 | -0.0015, -0.0081, -0.0191, -0.0243, -0.0191, -0.0081, -0.0015, 78 | }; 79 | const static float dx_filter [] = { 80 | -1.0f, 0.0f, 1.0f, \ 81 | -1.0f, 0.0f, 1.0f, \ 82 | -1.0f, 0.0f, 1.0f 83 | }; 84 | const static float dy_filter [] = { 85 | -1.0f, -1.0f, -1.0f, \ 86 | 0.0f, 0.0f, 0.0f, \ 87 | 1.0f, 1.0f, 1.0f 88 | }; 89 | 90 | png_image new_image = image; 91 | 92 | // image.data() returns pointer to the pixel data 93 | 94 | #pragma omp parallel for 95 | for( int i = 0; i < image.height; i++ ) { 96 | for( int j = 0; j < image.width; j++ ) { 97 | // calculate local contrast 98 | float dx = point_filter( image, dx_filter, i, j, 3 ); 99 | float dy = point_filter( image, dy_filter, i, j, 3 ); 100 | float contrast = std::sqrt( dx * dx + dy * dy ); 101 | // sharpen the edges and blur everything else 102 | if( contrast > threshold ) { 103 | new_image.pixel[i * new_image.width + j] = min( max( 0.0f, point_filter( image, sharp_filter, i, j, 7 ) ), 255.0 ); 104 | } else { 105 | new_image.pixel[i * new_image.width + j] = point_filter( image, gauss_filter, i, j, 7 ); 106 | } 107 | } 108 | } 109 | 110 | new_image.save( "result.png" ); 111 | } 112 | -------------------------------------------------------------------------------- /solution/extra/06_unsharp_mark/visualize.m: -------------------------------------------------------------------------------- 1 | hold on; 2 | 3 | ori = imread('original.png'); 4 | ref = imread('reference.png'); 5 | try 6 | res = imread('result.png'); 7 | catch 8 | res = uint8( zeros( size(ori) ) ); 9 | end 10 | 11 | subplot(2,3,1); 12 | imshow(ori); 13 | title('original'); 14 | subplot(2,3,2); 15 | imshow(ref); 16 | title('reference'); 17 | subplot(2,3,3); 18 | imshow(res); 19 | title('result'); 20 | 21 | rangex = 500:590; 22 | rangey = 600:760; 23 | subplot(2,3,4); 24 | imshow(ori(rangex,rangey)); 25 | title('original'); 26 | subplot(2,3,5); 27 | imshow(ref(rangex,rangey)); 28 | title('reference'); 29 | subplot(2,3,6); 30 | imshow(res(rangex,rangey)); 31 | title('result'); 32 | -------------------------------------------------------------------------------- /solution/extra/N0_event_timing/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/N0_event_timing/timing.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | #include 9 | 10 | #include "../util/util.h" 11 | 12 | __inline__ __host__ __device__ double f( double x ) 13 | { 14 | return sin( 2.0 * x ) * cos( 7.0 * x ) * exp( x ); 15 | } 16 | 17 | __global__ void evaluate( double *y, const int n ) 18 | { 19 | int i = global_thread_id(); 20 | y[i] = f( ( double )i / ( double )n ); 21 | } 22 | 23 | // host code entrance 24 | int main( int argc, char **argv ) 25 | { 26 | int N = 128 * 1024 * 1024; 27 | 28 | // timing register 29 | double t_CPU_0, t_CPU_1; 30 | // allocate host memory 31 | double *hst_y, *ref_y; 32 | hst_y = new double[N]; 33 | ref_y = new double[N]; 34 | // allocate device memory 35 | double *dev_y; 36 | cudaMalloc( &dev_y, N * sizeof( double ) ); 37 | 38 | // TO-DO: insert event 0 here 39 | insert_zeroth_event; 40 | 41 | // do computation on GPU 42 | evaluate <<< N / 1024, 1024 >>> ( dev_y, N ); 43 | 44 | // TO-DO: insert event 1 here 45 | insert_first_event; 46 | 47 | // copy result back to CPU 48 | cudaMemcpyAsync( hst_y, dev_y, N * sizeof( double ), cudaMemcpyDefault ); 49 | 50 | // TO-DO: insert event 2 here 51 | insert_second_event; 52 | 53 | cudaDeviceSynchronize(); 54 | 55 | t_CPU_0 = get_time(); 56 | 57 | // calculate reference value 58 | #pragma omp parallel for 59 | for( int i = 0; i < N; i++ ) ref_y[i] = f( ( double )i / ( double )N ); 60 | 61 | t_CPU_1 = get_time(); 62 | 63 | // compare 64 | bool match = true; 65 | for( int i = 0; i < N; i++ ) { 66 | match = match && 67 | ( fabs( ref_y[i] - hst_y[i] ) < 8 * std::numeric_limits::epsilon() ); 68 | } 69 | 70 | // output 71 | // TO-DO query time elapsed between events 72 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 73 | std::cout << "Computation on GPU took " << get_time_between_event_0_and_1 << " secs." << std::endl; 74 | std::cout << "Data transfer from GPU took " << get_time_between_event_1_and_2 << " secs." << std::endl; 75 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 76 | 77 | // free up resources 78 | delete [] hst_y; 79 | delete [] ref_y; 80 | cudaDeviceReset(); 81 | } 82 | -------------------------------------------------------------------------------- /solution/extra/N1_root_finding/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | -------------------------------------------------------------------------------- /solution/extra/N1_root_finding/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | 0_1_root_finding 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /solution/extra/N1_root_finding/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ $(DPFLAGS) 12 | 13 | clean: 14 | $(RM) $(EXE) 15 | -------------------------------------------------------------------------------- /solution/extra/N1_root_finding/polynomial.nb: -------------------------------------------------------------------------------- 1 | (* Content-type: application/vnd.wolfram.mathematica *) 2 | 3 | (*** Wolfram Notebook File ***) 4 | (* http://www.wolfram.com/nb *) 5 | 6 | (* CreatedBy='Mathematica 8.0' *) 7 | 8 | (*CacheID: 234*) 9 | (* Internal cache information: 10 | NotebookFileLineBreakTest 11 | NotebookFileLineBreakTest 12 | NotebookDataPosition[ 157, 7] 13 | NotebookDataLength[ 4103, 132] 14 | NotebookOptionsPosition[ 3648, 111] 15 | NotebookOutlinePosition[ 3985, 126] 16 | CellTagsIndexPosition[ 3942, 123] 17 | WindowFrame->Normal*) 18 | 19 | (* Beginning of Notebook Content *) 20 | Notebook[{ 21 | 22 | Cell[CellGroupData[{ 23 | Cell[BoxData[{ 24 | RowBox[{ 25 | RowBox[{"x0", "=", 26 | RowBox[{"2345", "/", 27 | RowBox[{"(", 28 | RowBox[{"2", "^", "12"}], ")"}]}]}], ";"}], "\[IndentingNewLine]", 29 | RowBox[{ 30 | RowBox[{"x1", "=", "2"}], ";"}], "\[IndentingNewLine]", 31 | RowBox[{ 32 | RowBox[{"x2", "=", "4"}], ";"}], "\[IndentingNewLine]", 33 | RowBox[{ 34 | RowBox[{"x3", "=", "8"}], ";"}], "\[IndentingNewLine]", 35 | RowBox[{"SetPrecision", "[", 36 | RowBox[{ 37 | RowBox[{"Expand", "[", 38 | RowBox[{ 39 | RowBox[{"(", 40 | RowBox[{"x", "-", "x0"}], ")"}], 41 | RowBox[{"(", 42 | RowBox[{"x", "-", "x1"}], ")"}], 43 | RowBox[{"(", 44 | RowBox[{"x", "-", "x2"}], ")"}], 45 | RowBox[{"(", 46 | RowBox[{"x", "-", "x3"}], ")"}]}], "]"}], ",", "15"}], "]"}]}], "Input",\ 47 | 48 | CellChangeTimes->{{3.6072738149482737`*^9, 3.607273907970641*^9}, { 49 | 3.607274446337237*^9, 3.607274494047564*^9}, {3.607274732165848*^9, 50 | 3.607274752804946*^9}, {3.607861312418264*^9, 3.607861314366634*^9}}], 51 | 52 | Cell[BoxData[ 53 | RowBox[{"36.640625`15.", "-", 54 | RowBox[{"96.060546875`15.", " ", "x"}], "+", 55 | RowBox[{"64.01513671875`15.", " ", 56 | SuperscriptBox["x", "2"]}], "-", 57 | RowBox[{"14.572509765625`15.", " ", 58 | SuperscriptBox["x", "3"]}], "+", 59 | SuperscriptBox["x", "4"]}]], "Output", 60 | CellChangeTimes->{{3.607273842288871*^9, 3.607273854919466*^9}, { 61 | 3.60727388626904*^9, 3.607273908221857*^9}, {3.607274476909288*^9, 62 | 3.607274495157909*^9}, {3.6072747345306463`*^9, 3.607274753242082*^9}, 63 | 3.607861315219453*^9}] 64 | }, Open ]], 65 | 66 | Cell[CellGroupData[{ 67 | 68 | Cell[BoxData[{ 69 | RowBox[{"x0", "=", 70 | RowBox[{ 71 | RowBox[{ 72 | RowBox[{"Sqrt", "[", "73", "]"}], "/", 73 | RowBox[{"Sqrt", "[", "79", "]"}]}], "*", "1.0"}]}], "\[IndentingNewLine]", 74 | RowBox[{ 75 | RowBox[{"x1", "=", "2"}], ";"}], "\[IndentingNewLine]", 76 | RowBox[{ 77 | RowBox[{"x2", "=", "3"}], ";"}], "\[IndentingNewLine]", 78 | RowBox[{ 79 | RowBox[{"x3", "=", "4"}], ";"}], "\[IndentingNewLine]", 80 | RowBox[{"SetPrecision", "[", 81 | RowBox[{ 82 | RowBox[{"Expand", "[", 83 | RowBox[{ 84 | RowBox[{"(", 85 | RowBox[{"x", "-", "x0"}], ")"}], 86 | RowBox[{"(", 87 | RowBox[{"x", "-", "x1"}], ")"}], 88 | RowBox[{"(", 89 | RowBox[{"x", "-", "x2"}], ")"}], 90 | RowBox[{"(", 91 | RowBox[{"x", "-", "x3"}], ")"}]}], "]"}], ",", "15"}], "]"}]}], "Input",\ 92 | 93 | CellChangeTimes->{{3.607861331336321*^9, 3.607861380676744*^9}, { 94 | 3.6078614264231243`*^9, 3.607861429933855*^9}}], 95 | 96 | Cell[BoxData["0.9612755239323388`"], "Output", 97 | CellChangeTimes->{{3.6078613506259203`*^9, 3.607861381090825*^9}, { 98 | 3.60786142674866*^9, 3.6078614302293787`*^9}}], 99 | 100 | Cell[BoxData[ 101 | RowBox[{"23.07061257437613122078801097813993692398`15.", "-", 102 | RowBox[{"48.99316362224081444765033666044473648071`15.", " ", "x"}], "+", 103 | RowBox[{"34.65147971539104787552787456661462783813`15.", " ", 104 | SuperscriptBox["x", "2"]}], "-", 105 | RowBox[{"9.96127552393233806071748404065147042274`15.", " ", 106 | SuperscriptBox["x", "3"]}], "+", 107 | SuperscriptBox["x", "4"]}]], "Output", 108 | CellChangeTimes->{{3.6078613506259203`*^9, 3.607861381090825*^9}, { 109 | 3.60786142674866*^9, 3.607861430230616*^9}}] 110 | }, Open ]] 111 | }, 112 | WindowSize->{740, 867}, 113 | WindowMargins->{{523, Automatic}, {-9, Automatic}}, 114 | FrontEndVersion->"8.0 for Linux x86 (64-bit) (October 10, 2011)", 115 | StyleDefinitions->"Default.nb" 116 | ] 117 | (* End of Notebook Content *) 118 | 119 | (* Internal cache information *) 120 | (*CellTagsOutline 121 | CellTagsIndex->{} 122 | *) 123 | (*CellTagsIndex 124 | CellTagsIndex->{} 125 | *) 126 | (*NotebookFileOutline 127 | Notebook[{ 128 | Cell[CellGroupData[{ 129 | Cell[579, 22, 940, 27, 104, "Input"], 130 | Cell[1522, 51, 524, 11, 51, "Output"] 131 | }, Open ]], 132 | Cell[CellGroupData[{ 133 | Cell[2083, 67, 865, 26, 104, "Input"], 134 | Cell[2951, 95, 164, 2, 29, "Output"], 135 | Cell[3118, 99, 514, 9, 51, "Output"] 136 | }, Open ]] 137 | } 138 | ] 139 | *) 140 | 141 | (* End of internal cache information *) 142 | -------------------------------------------------------------------------------- /solution/extra/N1_root_finding/root_finding.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | 9 | #include "../util/util.h" 10 | 11 | #if 0 12 | __device__ double f( double x ) 13 | { 14 | return 36.640625 + x * ( -96.060546875 + x * ( 64.01513671875 + x * ( -14.572509765625 + x ) ) ); 15 | } 16 | #endif 17 | 18 | #if 1 19 | __device__ double f( double x ) 20 | { 21 | return 23.0706125743761 + x * ( -48.9931636222408 + x * ( 34.6514797153910 + x * ( -9.9612755239323 + x ) ) ); 22 | } 23 | #endif 24 | 25 | __global__ void find_root( const double left, const double right, const double tolerance, const int n ) 26 | { 27 | if( blockIdx.x * blockDim.x + threadIdx.x == 0 ) printf( "searching within [%.15lf,%.15lf)\n", left, right ); 28 | __syncthreads(); 29 | 30 | double delta = ( right - left ) / n; 31 | for( int i = blockIdx.x * blockDim.x + threadIdx.x; i < n; i += gridDim.x * blockDim.x ) { 32 | double a = i * delta + left; 33 | double b = a + delta; 34 | if( signbit( f(a) * f(b) ) ) { 35 | if( delta < tolerance ) { 36 | printf( "root found as %.15lf\n", ( b + a ) * 0.5 ); 37 | } else { 38 | find_root <<< 1, 1024>>>( a, b, tolerance, n ); 39 | cudaDeviceSynchronize(); 40 | } 41 | } 42 | } 43 | } 44 | 45 | // host code entrance 46 | int main( int argc, char **argv ) 47 | { 48 | double tolerance = argc > 1 ? atof( argv[1] ) : 1E-7; 49 | int n = argc > 2 ? atof( argv[2] ) : 1024; 50 | 51 | find_root <<< 1, 1024>>>( 0.0, 1.0, tolerance, n ); 52 | cudaDeviceSynchronize(); 53 | } 54 | -------------------------------------------------------------------------------- /solution/extra/N2_64bit_shuffle/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/N2_64bit_shuffle/shuffle.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | 9 | __device__ __inline__ long long __shfl( long long var, int srcLane, int width = warpSize ) 10 | { 11 | int lo, hi; 12 | asm volatile( "mov.b64 {%0, %1}, %2; " : "=r"( lo ), "=r"( hi ) : "l"( var ) ); 13 | lo = __shfl( lo, srcLane, width ); 14 | hi = __shfl( hi, srcLane, width ); 15 | long long r; 16 | asm volatile( "mov.b64 %0, {%1, %2}; " : "=l"( r ) : "r"( lo ), "r"( hi ) ); 17 | return r; 18 | } 19 | 20 | __device__ __inline__ double __shfl( double var, int srcLane, int width = warpSize ) 21 | { 22 | int lo, hi; 23 | asm volatile( "mov.b64 {%0, %1}, %2; " : "=r"( lo ), "=r"( hi ) : "d"( var ) ); 24 | lo = __shfl( lo, srcLane, width ); 25 | hi = __shfl( hi, srcLane, width ); 26 | double r; 27 | asm volatile( "mov.b64 %0, {%1, %2}; " : "=d"( r ) : "r"( lo ), "r"( hi ) ); 28 | return r; 29 | } 30 | 31 | __global__ void shuffle_int64_test() 32 | { 33 | long long i = threadIdx.x; 34 | long long p = __shfl( i, warpSize - 1 - threadIdx.x ); 35 | printf( "thread %d got value %ld\n", threadIdx.x, p ); 36 | } 37 | 38 | __global__ void shuffle_float64_test() 39 | { 40 | double i = threadIdx.x; 41 | double p = __shfl( i, warpSize - 1 - threadIdx.x ); 42 | printf( "thread %d got value %lf\n", threadIdx.x, p ); 43 | } 44 | 45 | int main() 46 | { 47 | shuffle_int64_test <<< 1, 32>>>(); 48 | cudaDeviceSynchronize(); 49 | shuffle_float64_test <<< 1, 32>>>(); 50 | cudaDeviceSynchronize(); 51 | } 52 | -------------------------------------------------------------------------------- /solution/extra/N3_func_eval_template/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | -------------------------------------------------------------------------------- /solution/extra/N3_func_eval_template/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | 0_0_hello_world 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /solution/extra/N3_func_eval_template/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/N3_func_eval_template/func_eval.cu: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | 6 | #include 7 | #include 8 | 9 | #include "../util/util.h" 10 | #include "../util/math.h" 11 | 12 | template 13 | __inline__ __host__ __device__ REAL f( REAL x ) 14 | { 15 | #ifdef __CUDA_ARCH__ 16 | return cuda::sin( REAL( 2.0 ) * x ) * cuda::cos( REAL( 7.0 ) * x ) * cuda::exp( x ); 17 | #else 18 | return std::sin( REAL( 2.0 ) * x ) * std::cos( REAL( 7.0 ) * x ) * std::exp( x ); 19 | #endif 20 | } 21 | 22 | template 23 | __global__ void evaluate( REAL *y, const int n ) 24 | { 25 | int i = global_thread_id(); 26 | y[i] = f( ( REAL )i / ( REAL )n ); 27 | } 28 | 29 | // host code entrance 30 | template 31 | void evaluate() 32 | { 33 | int N = 128 * 1024 * 1024; 34 | 35 | // timing register 36 | double t_CPU_0, t_CPU_1, t_GPU_0, t_GPU_1, t_GPU_2; 37 | // allocate host memory 38 | REAL *hst_y, *ref_y; 39 | hst_y = new REAL[N]; 40 | ref_y = new REAL[N]; 41 | // allocate device memory 42 | REAL *dev_y; 43 | cudaMalloc( &dev_y, N * sizeof( REAL ) ); 44 | 45 | t_GPU_0 = get_time(); 46 | 47 | // do computation on GPU 48 | evaluate <<< N / 1024, 1024 >>> ( dev_y, N ); 49 | cudaDeviceSynchronize(); 50 | 51 | t_GPU_1 = get_time(); 52 | 53 | // copy result back to CPU 54 | cudaMemcpy( hst_y, dev_y, N * sizeof( REAL ), cudaMemcpyDefault ); 55 | 56 | t_GPU_2 = get_time(); 57 | 58 | t_CPU_0 = get_time(); 59 | 60 | // calculate reference value 61 | #pragma omp parallel for 62 | for( int i = 0; i < N; i++ ) ref_y[i] = f( ( REAL )i / ( REAL )N ); 63 | 64 | t_CPU_1 = get_time(); 65 | 66 | // compare 67 | bool match = true; 68 | for( int i = 0; i < N; i++ ) { 69 | match = match && 70 | ( fabs( ref_y[i] - hst_y[i] ) < 8 * std::numeric_limits::epsilon() ); 71 | } 72 | 73 | // output 74 | std::cout << "Computation on CPU took " << t_CPU_1 - t_CPU_0 << " secs." << std::endl; 75 | std::cout << "Computation on GPU took " << t_GPU_1 - t_GPU_0 << " secs." << std::endl; 76 | std::cout << "Data transfer from GPU took " << t_GPU_2 - t_GPU_1 << " secs." << std::endl; 77 | std::cout << "CPU/GPU result match: " << ( match ? "YES" : "NO" ) << std::endl; 78 | std::cout << "Precision mode: " << ( sizeof( REAL ) == 4 ? "single" : "double" ) << std::endl; 79 | 80 | // free up resources 81 | delete [] hst_y; 82 | delete [] ref_y; 83 | cudaFree( dev_y ); 84 | } 85 | 86 | int main() 87 | { 88 | evaluate(); 89 | evaluate(); 90 | } 91 | -------------------------------------------------------------------------------- /solution/extra/N4_templates/Makefile: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | include ../Makefile.inc 4 | 5 | SCR = $(wildcard *.cu) 6 | EXE = $(SCR:.cu=.x) 7 | 8 | default: $(EXE) 9 | 10 | %.x : %.cu 11 | $(CC) $(CFLAGS) $< -o $@ 12 | 13 | clean: 14 | $(RM) $(EXE) -------------------------------------------------------------------------------- /solution/extra/N4_templates/templates.cu: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/yhtang/CUDA-tutorial/21a1273d6fbfa659c1b90a4b84d98b11b5bffecd/solution/extra/N4_templates/templates.cu -------------------------------------------------------------------------------- /solution/util/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | -------------------------------------------------------------------------------- /solution/util/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | util 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /solution/util/bmp.h.orig: -------------------------------------------------------------------------------- 1 | /* 2 | * bmp.h 3 | * 4 | * Created on: Apr 29, 2014 5 | * Author: ytang 6 | */ 7 | 8 | #ifndef BMP_H_ 9 | #define BMP_H_ 10 | 11 | #include 12 | #include 13 | #include 14 | 15 | #if sizeof(int) == 2 16 | typedef int WORD; 17 | #else 18 | typedef short int WORD; 19 | #endif 20 | 21 | #if sizeof(int) == 4 22 | typedef int DWORD; 23 | #else 24 | typedef long DWORD; 25 | #endif 26 | 27 | struct bitmap 28 | { 29 | char *pixel; 30 | int width, height; 31 | 32 | bitmap() { 33 | pixel = NULL; 34 | width = height = 0; 35 | } 36 | ~bitmap() { 37 | if (pixel) delete [] pixel; 38 | } 39 | }; 40 | 41 | typedef struct tag_bitmapFILEHEADER 42 | { 43 | WORD bfType; 44 | DWORD bfSize; 45 | WORD bfReserved1; 46 | WORD bfReserved2; 47 | DWORD bfOffBits; 48 | 49 | } _bitmapFILEHEADER; 50 | 51 | typedef struct tag_bitmapINFOHEADER 52 | { 53 | DWORD biSize; 54 | DWORD biWidth; 55 | DWORD biHeight; 56 | WORD biPlanes; 57 | WORD biBitCount; 58 | DWORD biCompression; 59 | DWORD biSizeImage; 60 | DWORD biXPelsPerMeter; 61 | DWORD biYPelsPerMeter; 62 | DWORD biClrUsed; 63 | DWORD biClrImportant; 64 | } _bitmapINFOHEADER; 65 | 66 | bitmap * LoadBMP(char *fn) 67 | { 68 | bitmap * image; 69 | _bitmapFILEHEADER FileHeader; 70 | _bitmapINFOHEADER bmiHeader; 71 | int width, height, linebytes; 72 | 73 | image = new bitmap; 74 | if (image==NULL) return NULL; 75 | std::ifstream fp(fn); 76 | if (fp.fail()) { delete image; return NULL; } 77 | 78 | if (fread((char *)&FileHeader, sizeof(FileHeader), 1, fp) != 1) 79 | { 80 | farfree(image); 81 | fclose(fp); 82 | return NULL; 83 | } 84 | 85 | if (FileHeader.bfType != 0X4D42) 86 | { 87 | farfree(image); 88 | fclose(fp); 89 | return NULL; 90 | } 91 | 92 | if (fread((char *)&bmiHeader, sizeof(bmiHeader), 1, fp) != 1) 93 | { 94 | farfree(image); 95 | fclose(fp); 96 | return NULL; 97 | } 98 | 99 | if (bmiHeader.biBitCount > 8) 100 | { 101 | farfree(image); 102 | fclose(fp); 103 | return NULL; 104 | } 105 | 106 | if (bmiHeader.biCompression != 0) 107 | { 108 | farfree(image); 109 | fclose(fp); 110 | return NULL; 111 | } 112 | 113 | image->width = width = (unsigned int)bmiHeader.biWidth; 114 | image->height = height = (unsigned int)bmiHeader.biHeight; 115 | linebytes = ((width*(long)bmiHeader.biBitCount+31)>>5)<<2; 116 | if ( ( image->pixel = (char *)malloc(linebytes*height) ) ==NULL ) 117 | { 118 | farfree(image); 119 | fclose(fp); 120 | return NULL; 121 | } 122 | fseek(fp,FileHeader.bfOffBits,SEEK_SET); 123 | for(i=0;ipixel+linebytes*(height-i-1),linebytes,1,fp)!=1) 126 | { 127 | farfree(image->pixel); 128 | farfree(image); 129 | fclose(fp); 130 | return NULL; 131 | } 132 | } 133 | fclose(fp); 134 | return image; 135 | } 136 | 137 | 138 | #endif /* BMP_H_ */ 139 | -------------------------------------------------------------------------------- /solution/util/math.h: -------------------------------------------------------------------------------- 1 | #ifndef MATH_H_ 2 | #define MATH_H_ 3 | 4 | #include 5 | 6 | namespace cuda { 7 | 8 | // SIN 9 | template 10 | __inline__ __device__ T sin( T x ) { return ::sin(x); } 11 | 12 | template<> 13 | __inline__ __device__ float sin( float x ) { return ::sinf(x); } 14 | 15 | #define Overload(fun) \ 16 | template \ 17 | __inline__ __device__ T fun( T x ) { return ::fun(x); } \ 18 | template<> \ 19 | __inline__ __device__ float fun( float x ) { return ::fun##f(x); } 20 | 21 | Overload(cos); 22 | Overload(exp); 23 | Overload(log); 24 | Overload(tan); 25 | 26 | #undef Overload 27 | 28 | template 29 | struct literal { 30 | T value; 31 | literal( T v = 0 ) : value(v) {} 32 | __inline__ __host__ __device__ operator T& () { 33 | return value; 34 | } 35 | }; 36 | 37 | } 38 | 39 | 40 | #endif 41 | -------------------------------------------------------------------------------- /solution/util/util.h: -------------------------------------------------------------------------------- 1 | #ifndef UTIL_H_ 2 | #define UTIL_H_ 3 | 4 | #ifdef __CUDACC__ 5 | 6 | __device__ int block_id() { 7 | return blockIdx.x + blockIdx.y * gridDim.x + blockIdx.z * gridDim.x * gridDim.y; 8 | } 9 | 10 | __device__ int block_num() { 11 | return gridDim.x * gridDim.y * gridDim.z; 12 | } 13 | 14 | __device__ int block_size() { 15 | return blockDim.x * blockDim.y * blockDim.z; 16 | } 17 | 18 | __device__ int local_thread_id() { 19 | return threadIdx.x + threadIdx.y * blockDim.x + threadIdx.z * blockDim.x * blockDim.y; 20 | } 21 | 22 | __device__ int global_thread_id() { 23 | return block_id() * block_size() + local_thread_id(); 24 | } 25 | 26 | __device__ int global_thread_num() { 27 | return block_num() * block_size(); 28 | } 29 | 30 | #endif 31 | 32 | #include 33 | #include 34 | 35 | double get_time() 36 | { 37 | struct timespec time; 38 | clock_gettime( CLOCK_REALTIME, &time ); 39 | return (double)time.tv_sec + (double)time.tv_nsec * 1.0e-9 ; 40 | } 41 | 42 | #endif 43 | --------------------------------------------------------------------------------