intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-13 10:14:36 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
1fa5710dffcac4d22b23c74982b586ec4792c81b
compute-runtime/shared/test/unit_test/helpers/flattened_id_tests.cpp
Mateusz Jablonski 04a66d2b61 Move helpers tests to shared 
ray tracing tests
engine node helper xehp and later tests
hw helper tests

move shared ult from command queue to helpers

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2022-11-07 15:36:22 +01:00

258 lines
12 KiB
C++
Raw Blame History

/*
* Copyright (C) 2018-2022 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "shared/source/helpers/basic_math.h"
#include "shared/source/helpers/string.h"
#include "shared/test/common/helpers/unit_test_helper.h"
#include "shared/test/common/test_macros/hw_test.h"
#include <algorithm>
#include <cstdint>
using namespace NEO;
union GRF {
float fRegs[8];
uint32_t dwRegs[8];
uint16_t wRegs[16];
};
// clang-format off
uint16_t gFlattenedIDsSimD8[] =
{
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
};
uint16_t gFlattenedIDsSimD1632[] =
{
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47,
0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f,
0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57,
0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f,
0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f,
0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f,
0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f,
0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97,
0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f,
0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf,
0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7,
0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf,
0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7,
0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf,
0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7,
0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf,
0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7,
0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef,
0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff,
};
// clang-format on
size_t getSizeFlattenedIDs(uint32_t simd, uint32_t lwsX, uint32_t lwsY, uint32_t lwsZ) {
auto lws = lwsX * lwsY * lwsZ;
lws = std::min(lws, 256u);
auto iDsPerGrf = std::min(simd, 16u);
auto numGRFs = Math::divideAndRoundUp(lws, iDsPerGrf);
return numGRFs * sizeof(GRF);
}
void generateFlattenedIDs(void *buffer, uint32_t simd, uint32_t lwsX, uint32_t lwsY, uint32_t lwsZ) {
auto pSrc = simd == 8 ? gFlattenedIDsSimD8 : gFlattenedIDsSimD1632;
auto copySize = getSizeFlattenedIDs(simd, lwsX, lwsY, lwsZ);
memcpy_s(buffer, copySize, pSrc, copySize);
}
struct FlattenedIDFixture : ::testing::TestWithParam<std::tuple<int, int, int, int>> {
void SetUp() override {
simd = std::get<0>(GetParam());
localWorkSizeX = std::get<1>(GetParam());
localWorkSizeY = std::get<2>(GetParam());
localWorkSizeZ = std::get<3>(GetParam());
localWorkSize = localWorkSizeX * localWorkSizeY * localWorkSizeZ;
if (localWorkSize > 256) {
localWorkSizeY = std::min(256 / localWorkSizeX, localWorkSizeY);
localWorkSizeZ = std::min(256 / (localWorkSizeX * localWorkSizeY), localWorkSizeZ);
localWorkSize = localWorkSizeX * localWorkSizeY * localWorkSizeZ;
}
memset(buffer, 0xff, sizeof(buffer));
}
void validateIDWithinLimits(uint32_t simd, uint32_t lwsX, uint32_t lwsY, uint32_t lwsZ, bool useFullRow) {
auto numWorkItems = lwsX * lwsY * lwsZ;
auto idsPerThread = simd;
// As per BackEnd HLD, SIMD32 has 32 flattenedIDs per channel. SIMD8/16 has up to 16 flattenedIDs.
auto skipPerThread = (simd == 32 || useFullRow) ? 32 : 16;
auto pBuffer = buffer;
size_t itemIndex = 0;
auto numWorkItemsRemaining = numWorkItems;
while (numWorkItemsRemaining > 0) {
EXPECT_LT(pBuffer[itemIndex], numWorkItems);
++itemIndex;
if (idsPerThread == itemIndex) {
pBuffer += skipPerThread;
itemIndex = 0;
}
--numWorkItemsRemaining;
}
}
void validateAllWorkItemsCovered(uint32_t simd, uint32_t lwsX, uint32_t lwsY, uint32_t lwsZ, bool useFullRow) {
auto numWorkItems = lwsX * lwsY * lwsZ;
auto idsPerThread = simd;
// As per BackEnd HLD, SIMD32 has 32 localIDs per channel. SIMD8/16 has up to 16 localIDs.
auto skipPerThread = (simd == 32 || useFullRow) ? 32 : 16;
// Initialize local ID hit table
uint32_t localIDHitTable[8];
memset(localIDHitTable, 0, sizeof(localIDHitTable));
auto pBuffer = buffer;
size_t itemIndex = 0;
auto numWorkItemsRemaining = numWorkItems;
while (numWorkItemsRemaining > 0) {
// Flatten out the IDs
auto workItem = pBuffer[itemIndex];
ASSERT_LT(workItem, 256u);
// Look up in the hit table
auto &hitItem = localIDHitTable[workItem / 32];
auto hitBit = 1 << (workItem % 32);
// No double-hits
EXPECT_EQ(0u, hitItem & hitBit);
// Set that work item as hit
hitItem |= hitBit;
++itemIndex;
if (idsPerThread == itemIndex) {
pBuffer += skipPerThread;
itemIndex = 0;
}
--numWorkItemsRemaining;
}
// All entries in hit table should be in form of n^2 - 1
for (unsigned int i : localIDHitTable) {
EXPECT_EQ(0u, i & (i + 1));
}
}
// Test parameters
uint32_t localWorkSize;
uint32_t localWorkSizeX;
uint32_t localWorkSizeY;
uint32_t localWorkSizeZ;
uint32_t simd;
// Provide support for a max LWS of 256
// 32 threads @ SIMD8
// 16 lanes per thread (SIMD8 - only 8 used)
uint16_t buffer[32 * 16];
};
HWTEST_P(FlattenedIDFixture, WhenGeneratingFlattenedIdsThenAllIdsWithinLimits) {
generateFlattenedIDs(buffer, simd, localWorkSizeX, localWorkSizeY, localWorkSizeZ);
validateIDWithinLimits(simd, localWorkSizeX, localWorkSizeY, localWorkSizeZ, UnitTestHelper<FamilyType>::useFullRowForLocalIdsGeneration);
}
HWTEST_P(FlattenedIDFixture, WhenGeneratingFlattenedIdsThenAllWorkItemsCovered) {
generateFlattenedIDs(buffer, simd, localWorkSizeX, localWorkSizeY, localWorkSizeZ);
validateAllWorkItemsCovered(simd, localWorkSizeX, localWorkSizeY, localWorkSizeZ, UnitTestHelper<FamilyType>::useFullRowForLocalIdsGeneration);
}
TEST_P(FlattenedIDFixture, WhenGeneratingFlattenedIdsThenLocalIdsAreCorrect) {
auto sizePerThreadData = getSizeFlattenedIDs(
simd,
localWorkSizeX,
localWorkSizeY,
localWorkSizeZ);
// Should be multiple of GRFs
auto sizeGRF = sizeof(GRF);
EXPECT_EQ(0u, sizePerThreadData % sizeGRF);
auto workItems = localWorkSizeX * localWorkSizeY * localWorkSizeZ;
auto idsPerGRF = (simd == 8) ? 8 : 16;
auto numGRFsExpected = Math::divideAndRoundUp(workItems, idsPerGRF);
EXPECT_EQ(numGRFsExpected, sizePerThreadData / sizeGRF);
}
#define SIMDParams ::testing::Values(8, 16, 32)
#if HEAVY_DUTY_TESTING
#define LWSXParams ::testing::Values(1, 7, 8, 9, 15, 16, 17, 31, 32, 33, 64, 128, 256)
#define LWSYParams ::testing::Values(1, 2, 3, 4, 5, 6, 7, 8)
#define LWSZParams ::testing::Values(1, 2, 3, 4)
#else
#define LWSXParams ::testing::Values(1, 7, 8, 9, 15, 16, 17, 31, 32, 33, 64, 128, 256)
#define LWSYParams ::testing::Values(1, 2, 4, 8)
#define LWSZParams ::testing::Values(1)
#endif
INSTANTIATE_TEST_CASE_P(AllCombinations, FlattenedIDFixture, ::testing::Combine(SIMDParams, LWSXParams, LWSYParams, LWSZParams));
// To debug a specific configuration replace the list of Values with specific values.
// NOTE: You'll need a unique test prefix
INSTANTIATE_TEST_CASE_P(SingleTest, FlattenedIDFixture,
::testing::Combine(
::testing::Values(32), //SIMD
::testing::Values(5), //LWSX
::testing::Values(6), //LWSY
::testing::Values(7))); //LWSZ
Reference in New Issue View Git Blame Copy Permalink