Blame view
tools/cub-1.8.0/test/test_warp_scan.cu
19.9 KB
8dcb6dfcb
first commit
|
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 27 28 29 30 31 32 33 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 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 |
/******************************************************************************
* Copyright (c) 2011, Duane Merrill. All rights reserved.
* Copyright (c) 2011-2018, NVIDIA CORPORATION. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* * Neither the name of the NVIDIA CORPORATION nor the
* names of its contributors may be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
* DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
******************************************************************************/
/******************************************************************************
* Test of WarpScan utilities
******************************************************************************/
// Ensure printing of CUDA runtime errors to console
#define CUB_STDERR
#include <stdio.h>
#include <typeinfo>
#include <cub/warp/warp_scan.cuh>
#include <cub/util_allocator.cuh>
#include "test_util.h"
using namespace cub;
//---------------------------------------------------------------------
// Globals, constants and typedefs
//---------------------------------------------------------------------
static const int NUM_WARPS = 2;
bool g_verbose = false;
int g_repeat = 0;
CachingDeviceAllocator g_allocator(true);
/**
* Primitive variant to test
*/
enum TestMode
{
BASIC,
AGGREGATE,
};
/**
* \brief WrapperFunctor (for precluding test-specialized dispatch to *Sum variants)
*/
template<typename OpT>
struct WrapperFunctor
{
OpT op;
WrapperFunctor(OpT op) : op(op) {}
template <typename T>
__host__ __device__ __forceinline__ T operator()(const T &a, const T &b) const
{
return op(a, b);
}
};
//---------------------------------------------------------------------
// Test kernels
//---------------------------------------------------------------------
/// Exclusive scan basic
template <typename WarpScanT, typename T, typename ScanOpT, typename IsPrimitiveT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
T &initial_value,
ScanOpT &scan_op,
T &aggregate,
Int2Type<BASIC> test_mode,
IsPrimitiveT is_primitive)
{
// Test basic warp scan
warp_scan.ExclusiveScan(data, data, initial_value, scan_op);
}
/// Exclusive scan aggregate
template <
typename WarpScanT,
typename T,
typename ScanOpT,
typename IsPrimitiveT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
T &initial_value,
ScanOpT &scan_op,
T &aggregate,
Int2Type<AGGREGATE> test_mode,
IsPrimitiveT is_primitive)
{
// Test with cumulative aggregate
warp_scan.ExclusiveScan(data, data, initial_value, scan_op, aggregate);
}
/// Exclusive sum basic
template <
typename WarpScanT,
typename T>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
T &initial_value,
Sum &scan_op,
T &aggregate,
Int2Type<BASIC> test_mode,
Int2Type<true> is_primitive)
{
// Test basic warp scan
warp_scan.ExclusiveSum(data, data);
}
/// Exclusive sum aggregate
template <
typename WarpScanT,
typename T>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
T &initial_value,
Sum &scan_op,
T &aggregate,
Int2Type<AGGREGATE> test_mode,
Int2Type<true> is_primitive)
{
// Test with cumulative aggregate
warp_scan.ExclusiveSum(data, data, aggregate);
}
/// Inclusive scan basic
template <
typename WarpScanT,
typename T,
typename ScanOpT,
typename IsPrimitiveT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
NullType &initial_value,
ScanOpT &scan_op,
T &aggregate,
Int2Type<BASIC> test_mode,
IsPrimitiveT is_primitive)
{
// Test basic warp scan
warp_scan.InclusiveScan(data, data, scan_op);
}
/// Inclusive scan aggregate
template <
typename WarpScanT,
typename T,
typename ScanOpT,
typename IsPrimitiveT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
NullType &initial_value,
ScanOpT &scan_op,
T &aggregate,
Int2Type<AGGREGATE> test_mode,
IsPrimitiveT is_primitive)
{
// Test with cumulative aggregate
warp_scan.InclusiveScan(data, data, scan_op, aggregate);
}
/// Inclusive sum basic
template <
typename WarpScanT,
typename T,
typename InitialValueT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
NullType &initial_value,
Sum &scan_op,
T &aggregate,
Int2Type<BASIC> test_mode,
Int2Type<true> is_primitive)
{
// Test basic warp scan
warp_scan.InclusiveSum(data, data);
}
/// Inclusive sum aggregate
template <
typename WarpScanT,
typename T,
typename InitialValueT>
__device__ __forceinline__ void DeviceTest(
WarpScanT &warp_scan,
T &data,
NullType &initial_value,
Sum &scan_op,
T &aggregate,
Int2Type<AGGREGATE> test_mode,
Int2Type<true> is_primitive)
{
// Test with cumulative aggregate
warp_scan.InclusiveSum(data, data, aggregate);
}
/**
* WarpScan test kernel
*/
template <
int LOGICAL_WARP_THREADS,
TestMode TEST_MODE,
typename T,
typename ScanOpT,
typename InitialValueT>
__global__ void WarpScanKernel(
T *d_in,
T *d_out,
T *d_aggregate,
ScanOpT scan_op,
InitialValueT initial_value,
clock_t *d_elapsed)
{
// Cooperative warp-scan utility type (1 warp)
typedef WarpScan<T, LOGICAL_WARP_THREADS> WarpScanT;
// Allocate temp storage in shared memory
__shared__ typename WarpScanT::TempStorage temp_storage[NUM_WARPS];
// Get warp index
int warp_id = threadIdx.x / LOGICAL_WARP_THREADS;
// Per-thread tile data
T data = d_in[threadIdx.x];
// Start cycle timer
__threadfence_block(); // workaround to prevent clock hoisting
clock_t start = clock();
__threadfence_block(); // workaround to prevent clock hoisting
T aggregate;
// Test scan
WarpScanT warp_scan(temp_storage[warp_id]);
DeviceTest(
warp_scan,
data,
initial_value,
scan_op,
aggregate,
Int2Type<TEST_MODE>(),
Int2Type<Traits<T>::PRIMITIVE>());
// Stop cycle timer
__threadfence_block(); // workaround to prevent clock hoisting
clock_t stop = clock();
__threadfence_block(); // workaround to prevent clock hoisting
// Store data
d_out[threadIdx.x] = data;
if (TEST_MODE != BASIC)
{
// Store aggregate
d_aggregate[threadIdx.x] = aggregate;
}
// Store time
if (threadIdx.x == 0)
{
*d_elapsed = (start > stop) ? start - stop : stop - start;
}
}
//---------------------------------------------------------------------
// Host utility subroutines
//---------------------------------------------------------------------
/**
* Initialize exclusive-scan problem (and solution)
*/
template <
typename T,
typename ScanOpT>
void Initialize(
GenMode gen_mode,
T *h_in,
T *h_reference,
int logical_warp_items,
ScanOpT scan_op,
T initial_value,
T warp_aggregates[NUM_WARPS])
{
for (int w = 0; w < NUM_WARPS; ++w)
{
int base_idx = (w * logical_warp_items);
int i = base_idx;
InitValue(gen_mode, h_in[i], i);
T warp_aggregate = h_in[i];
h_reference[i] = initial_value;
T inclusive = scan_op(initial_value, h_in[i]);
for (i = i + 1; i < base_idx + logical_warp_items; ++i)
{
InitValue(gen_mode, h_in[i], i);
h_reference[i] = inclusive;
inclusive = scan_op(inclusive, h_in[i]);
warp_aggregate = scan_op(warp_aggregate, h_in[i]);
}
warp_aggregates[w] = warp_aggregate;
}
}
/**
* Initialize inclusive-scan problem (and solution)
*/
template <
typename T,
typename ScanOpT>
void Initialize(
GenMode gen_mode,
T *h_in,
T *h_reference,
int logical_warp_items,
ScanOpT scan_op,
NullType,
T warp_aggregates[NUM_WARPS])
{
for (int w = 0; w < NUM_WARPS; ++w)
{
int base_idx = (w * logical_warp_items);
int i = base_idx;
InitValue(gen_mode, h_in[i], i);
T warp_aggregate = h_in[i];
T inclusive = h_in[i];
h_reference[i] = inclusive;
for (i = i + 1; i < base_idx + logical_warp_items; ++i)
{
InitValue(gen_mode, h_in[i], i);
inclusive = scan_op(inclusive, h_in[i]);
warp_aggregate = scan_op(warp_aggregate, h_in[i]);
h_reference[i] = inclusive;
}
warp_aggregates[w] = warp_aggregate;
}
}
/**
* Test warp scan
*/
template <
int LOGICAL_WARP_THREADS,
TestMode TEST_MODE,
typename T,
typename ScanOpT,
typename InitialValueT> // NullType implies inclusive-scan, otherwise inclusive scan
void Test(
GenMode gen_mode,
ScanOpT scan_op,
InitialValueT initial_value)
{
enum {
TOTAL_ITEMS = LOGICAL_WARP_THREADS * NUM_WARPS,
};
// Allocate host arrays
T *h_in = new T[TOTAL_ITEMS];
T *h_reference = new T[TOTAL_ITEMS];
T *h_aggregate = new T[TOTAL_ITEMS];
// Initialize problem
T aggregates[NUM_WARPS];
Initialize(
gen_mode,
h_in,
h_reference,
LOGICAL_WARP_THREADS,
scan_op,
initial_value,
aggregates);
if (g_verbose)
{
printf("Input:
");
DisplayResults(h_in, TOTAL_ITEMS);
printf("
");
}
for (int w = 0; w < NUM_WARPS; ++w)
{
for (int i = 0; i < LOGICAL_WARP_THREADS; ++i)
{
h_aggregate[(w * LOGICAL_WARP_THREADS) + i] = aggregates[w];
}
}
// Initialize/clear device arrays
T *d_in = NULL;
T *d_out = NULL;
T *d_aggregate = NULL;
clock_t *d_elapsed = NULL;
CubDebugExit(g_allocator.DeviceAllocate((void**)&d_in, sizeof(T) * TOTAL_ITEMS));
CubDebugExit(g_allocator.DeviceAllocate((void**)&d_out, sizeof(T) * (TOTAL_ITEMS + 1)));
CubDebugExit(g_allocator.DeviceAllocate((void**)&d_aggregate, sizeof(T) * TOTAL_ITEMS));
CubDebugExit(g_allocator.DeviceAllocate((void**)&d_elapsed, sizeof(clock_t)));
CubDebugExit(cudaMemcpy(d_in, h_in, sizeof(T) * TOTAL_ITEMS, cudaMemcpyHostToDevice));
CubDebugExit(cudaMemset(d_out, 0, sizeof(T) * (TOTAL_ITEMS + 1)));
CubDebugExit(cudaMemset(d_aggregate, 0, sizeof(T) * TOTAL_ITEMS));
// Run kernel
printf("Test-mode %d (%s), gen-mode %d (%s), %s warpscan, %d warp threads, %s (%d bytes) elements:
",
TEST_MODE, typeid(TEST_MODE).name(),
gen_mode, typeid(gen_mode).name(),
(Equals<InitialValueT, NullType>::VALUE) ? "Inclusive" : "Exclusive",
LOGICAL_WARP_THREADS,
typeid(T).name(),
(int) sizeof(T));
fflush(stdout);
// Run aggregate/prefix kernel
WarpScanKernel<LOGICAL_WARP_THREADS, TEST_MODE><<<1, TOTAL_ITEMS>>>(
d_in,
d_out,
d_aggregate,
scan_op,
initial_value,
d_elapsed);
printf("\tElapsed clocks: ");
DisplayDeviceResults(d_elapsed, 1);
CubDebugExit(cudaPeekAtLastError());
CubDebugExit(cudaDeviceSynchronize());
// Copy out and display results
printf("\tScan results: ");
int compare = CompareDeviceResults(h_reference, d_out, TOTAL_ITEMS, g_verbose, g_verbose);
printf("%s
", compare ? "FAIL" : "PASS");
AssertEquals(0, compare);
// Copy out and display aggregate
if (TEST_MODE == AGGREGATE)
{
printf("\tScan aggregate: ");
compare = CompareDeviceResults(h_aggregate, d_aggregate, TOTAL_ITEMS, g_verbose, g_verbose);
printf("%s
", compare ? "FAIL" : "PASS");
AssertEquals(0, compare);
}
// Cleanup
if (h_in) delete[] h_in;
if (h_reference) delete[] h_reference;
if (h_aggregate) delete[] h_aggregate;
if (d_in) CubDebugExit(g_allocator.DeviceFree(d_in));
if (d_out) CubDebugExit(g_allocator.DeviceFree(d_out));
if (d_aggregate) CubDebugExit(g_allocator.DeviceFree(d_aggregate));
if (d_elapsed) CubDebugExit(g_allocator.DeviceFree(d_elapsed));
}
/**
* Run battery of tests for different primitive variants
*/
template <
int LOGICAL_WARP_THREADS,
typename ScanOpT,
typename T>
void Test(
GenMode gen_mode,
ScanOpT scan_op,
T initial_value)
{
// Exclusive
Test<LOGICAL_WARP_THREADS, BASIC, T>(gen_mode, scan_op, T());
Test<LOGICAL_WARP_THREADS, AGGREGATE, T>(gen_mode, scan_op, T());
// Exclusive (non-specialized, so we can use initial-value)
Test<LOGICAL_WARP_THREADS, BASIC, T>(gen_mode, WrapperFunctor<ScanOpT>(scan_op), initial_value);
Test<LOGICAL_WARP_THREADS, AGGREGATE, T>(gen_mode, WrapperFunctor<ScanOpT>(scan_op), initial_value);
// Inclusive
Test<LOGICAL_WARP_THREADS, BASIC, T>(gen_mode, scan_op, NullType());
Test<LOGICAL_WARP_THREADS, AGGREGATE, T>(gen_mode, scan_op, NullType());
}
/**
* Run battery of tests for different data types and scan ops
*/
template <int LOGICAL_WARP_THREADS>
void Test(GenMode gen_mode)
{
// Get device ordinal
int device_ordinal;
CubDebugExit(cudaGetDevice(&device_ordinal));
// Get ptx version
int ptx_version;
CubDebugExit(PtxVersion(ptx_version));
// primitive
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (char) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (short) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (int) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (long) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (long long) 99);
if (gen_mode != RANDOM) {
// Only test numerically stable inputs
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (float) 99);
if (ptx_version > 100)
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), (double) 99);
}
// primitive (alternative scan op)
Test<LOGICAL_WARP_THREADS>(gen_mode, Max(), (unsigned char) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Max(), (unsigned short) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Max(), (unsigned int) 99);
Test<LOGICAL_WARP_THREADS>(gen_mode, Max(), (unsigned long long) 99);
// vec-2
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_uchar2(17, 21));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_ushort2(17, 21));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_uint2(17, 21));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_ulong2(17, 21));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_ulonglong2(17, 21));
if (gen_mode != RANDOM) {
// Only test numerically stable inputs
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_float2(17, 21));
if (ptx_version > 100)
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_double2(17, 21));
}
// vec-4
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_char4(17, 21, 32, 85));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_short4(17, 21, 32, 85));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_int4(17, 21, 32, 85));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_long4(17, 21, 32, 85));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_longlong4(17, 21, 32, 85));
if (gen_mode != RANDOM) {
// Only test numerically stable inputs
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_float4(17, 21, 32, 85));
if (ptx_version > 100)
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), make_double4(17, 21, 32, 85));
}
// complex
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), TestFoo::MakeTestFoo(17, 21, 32, 85));
Test<LOGICAL_WARP_THREADS>(gen_mode, Sum(), TestBar(17, 21));
}
/**
* Run battery of tests for different problem generation options
*/
template <int LOGICAL_WARP_THREADS>
void Test()
{
Test<LOGICAL_WARP_THREADS>(UNIFORM);
Test<LOGICAL_WARP_THREADS>(INTEGER_SEED);
Test<LOGICAL_WARP_THREADS>(RANDOM);
}
/**
* Main
*/
int main(int argc, char** argv)
{
// Initialize command line
CommandLineArgs args(argc, argv);
g_verbose = args.CheckCmdLineFlag("v");
args.GetCmdLineArgument("repeat", g_repeat);
// Print usage
if (args.CheckCmdLineFlag("help"))
{
printf("%s "
"[--device=<device-id>] "
"[--repeat=<repetitions of entire test suite>]"
"[--v] "
"
", argv[0]);
exit(0);
}
// Initialize device
CubDebugExit(args.DeviceInit());
#ifdef QUICK_TEST
// Compile/run quick tests
Test<32, AGGREGATE, int>(UNIFORM, Sum(), (int) 0);
Test<32, AGGREGATE, float>(UNIFORM, Sum(), (float) 0);
Test<32, AGGREGATE, long long>(UNIFORM, Sum(), (long long) 0);
Test<32, AGGREGATE, double>(UNIFORM, Sum(), (double) 0);
typedef KeyValuePair<int, float> T;
cub::Sum sum_op;
Test<32, AGGREGATE, T>(UNIFORM, ReduceBySegmentOp<cub::Sum>(sum_op), T());
#else
// Compile/run thorough tests
for (int i = 0; i <= g_repeat; ++i)
{
// Test logical warp sizes
Test<32>();
Test<16>();
Test<9>();
Test<2>();
}
#endif
return 0;
}
|