From 6df6ef778dbace8d32a3c6ac0f35c35584cf7f06 Mon Sep 17 00:00:00 2001 From: rohit pathania Date: Thu, 24 May 2018 16:55:46 +0530 Subject: [PATCH] Kernel group memory dynamic allocation, basic allocation and free test Change-Id: I17fdb77f17567ac1b429d9a571cac70ac1e64dd4 --- .../suites/functional/memory_allocation.cc | 504 ++++++++++++++++++ rocrtst/suites/functional/memory_allocation.h | 89 ++++ rocrtst/suites/test_common/CMakeLists.txt | 5 + .../kernels/groupMemoryDynamic_kernels.cl | 57 ++ rocrtst/suites/test_common/main.cc | 15 + 5 files changed, 670 insertions(+) create mode 100755 rocrtst/suites/functional/memory_allocation.cc create mode 100755 rocrtst/suites/functional/memory_allocation.h create mode 100755 rocrtst/suites/test_common/kernels/groupMemoryDynamic_kernels.cl diff --git a/rocrtst/suites/functional/memory_allocation.cc b/rocrtst/suites/functional/memory_allocation.cc new file mode 100755 index 0000000000..098593c000 --- /dev/null +++ b/rocrtst/suites/functional/memory_allocation.cc @@ -0,0 +1,504 @@ +/* + * ============================================================================= + * ROC Runtime Conformance Release License + * ============================================================================= + * The University of Illinois/NCSA + * Open Source License (NCSA) + * + * Copyright (c) 2018, Advanced Micro Devices, Inc. + * All rights reserved. + * + * Developed by: + * + * AMD Research and AMD ROC Software Development + * + * Advanced Micro Devices, Inc. + * + * www.amd.com + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal with the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * - Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimers. + * - Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimers in + * the documentation and/or other materials provided with the distribution. + * - Neither the names of , + * nor the names of its contributors may be used to endorse or promote + * products derived from this Software without specific prior written + * permission. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS WITH THE SOFTWARE. + * + */ + + +#include +#include +#include +#include +#include +#include + +#include "suites/functional/memory_allocation.h" +#include "common/base_rocr_utils.h" +#include "common/common.h" +#include "common/helper_funcs.h" +#include "common/hsatimer.h" +#include "gtest/gtest.h" +#include "hsa/hsa.h" +#include "hsa/hsa_ext_finalize.h" + +static const uint32_t kNumBufferElements = 256; +static const int kValue = 5; + + +// This wrapper atomically writes the provided header and setup to the +// provided AQL packet. The provided AQL packet address should be in the +// queue memory space. +static inline void AtomicSetPacketHeader(uint16_t header, uint16_t setup, + hsa_kernel_dispatch_packet_t* queue_packet) { + __atomic_store_n(reinterpret_cast(queue_packet), + header | (setup << 16), __ATOMIC_RELEASE); +} + + + +MemoryAllocationTest::MemoryAllocationTest(bool launch_GroupMemory, + bool launch_BasicAllocateFree) : TestBase() { + set_num_iteration(10); // Number of iterations to execute of the main test; + // This is a default value which can be overridden + // on the command line. + std::string name; + std::string desc; + + name = "RocR Memory Test "; + if (launch_GroupMemory) { + name += " For Kernel Dynamic Memory Alocation"; + desc += " This test Allocate group memory in kernel dynamically."; + } else if (launch_BasicAllocateFree) { + name += " For BasicAllocateFree"; + desc += " This test Allocate And free Memory on all the availble pool " + " on which allocation is allowed on RocR Agents."; + } + set_title(name); + set_description(desc); + + memset(&aql(), 0, sizeof(hsa_kernel_dispatch_packet_t)); +} + +MemoryAllocationTest::~MemoryAllocationTest(void) { +} + +// Any 1-time setup involving member variables used in the rest of the test +// should be done here. +void MemoryAllocationTest::SetUp(void) { + hsa_status_t err; + + TestBase::SetUp(); + + err = rocrtst::SetDefaultAgents(this); + ASSERT_EQ(HSA_STATUS_SUCCESS, err); + + err = rocrtst::SetPoolsTypical(this); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // Fill up the kernel packet except header + err = rocrtst::InitializeAQLPacket(this, &aql()); + ASSERT_EQ(HSA_STATUS_SUCCESS, err); + return; +} + +void MemoryAllocationTest::Run(void) { + // Compare required profile for this test case with what we're actually + // running on + if (!rocrtst::CheckProfile(this)) { + return; + } + + TestBase::Run(); +} + +void MemoryAllocationTest::DisplayTestInfo(void) { + TestBase::DisplayTestInfo(); +} + +void MemoryAllocationTest::DisplayResults(void) const { + // Compare required profile for this test case with what we're actually + // running on + if (!rocrtst::CheckProfile(this)) { + return; + } + + return; +} + +void MemoryAllocationTest::Close() { + // This will close handles opened within rocrtst utility calls and call + // hsa_shut_down(), so it should be done after other hsa cleanup + TestBase::Close(); +} + +void MemoryAllocationTest::WriteAQLPktToQueue(hsa_queue_t* q) { + void* queue_base = q->base_address; + const uint32_t queue_mask = q->size - 1; + uint64_t index = hsa_queue_add_write_index_relaxed(q, 1); + + reinterpret_cast( + queue_base)[index & queue_mask] = aql(); +} + + + + + +typedef struct __attribute__ ((aligned(16))) args_t { + uint32_t *a; + uint32_t *b; + uint32_t grp_offset; + uint32_t count; + } args; + + +static const char kSubTestSeparator[] = " **************************"; + +static void PrintMemorySubtestHeader(const char *header) { + std::cout << " *** Memory Allocation Test: " << header << " ***" << std::endl; +} + +static const int kMemoryAllocSize = 1024; + + + + +void MemoryAllocationTest::GroupMemoryDynamicAllocation(hsa_agent_t cpuAgent, + hsa_agent_t gpuAgent) { + hsa_status_t err; + + // Get Global Memory Pool on the gpuAgent to allocate gpu buffers + hsa_amd_memory_pool_t gpu_pool; + err = hsa_amd_agent_iterate_memory_pools(gpuAgent, + rocrtst::GetGlobalMemoryPool, + &gpu_pool); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + hsa_amd_memory_pool_access_t access; + hsa_amd_agent_memory_pool_get_info(cpuAgent, gpu_pool, + HSA_AMD_AGENT_MEMORY_POOL_INFO_ACCESS, + &access); + if (access != HSA_AMD_MEMORY_POOL_ACCESS_NEVER_ALLOWED) { + // hsa objects + hsa_queue_t *queue = NULL; // command queue + + // get queue size + uint32_t queue_size = 0; + err = hsa_agent_get_info(gpuAgent, + HSA_AGENT_INFO_QUEUE_MAX_SIZE, &queue_size); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // create queue + err = hsa_queue_create(gpuAgent, + queue_size, HSA_QUEUE_TYPE_MULTI, + NULL, NULL, 0, 0, &queue); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // Get System Memory Pool on the cpuAgent to allocate host side buffers + hsa_amd_memory_pool_t global_pool; + err = hsa_amd_agent_iterate_memory_pools(cpuAgent, + rocrtst::GetGlobalMemoryPool, + &global_pool); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // Find a memory pool that supports kernel arguments. + hsa_amd_memory_pool_t kernarg_pool; + err = hsa_amd_agent_iterate_memory_pools(cpuAgent, + rocrtst::GetKernArgMemoryPool, + &kernarg_pool); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // Allocate the host side buffers + // (Indata,kernArg) on system memory + uint32_t *Indata = NULL; + args *kernArgs = NULL; + + err = hsa_amd_memory_pool_allocate(global_pool, + kMemoryAllocSize, 0, + reinterpret_cast(&Indata)); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + + // Allocate the kernel argument buffer from the kernarg_pool. + err = hsa_amd_memory_pool_allocate(kernarg_pool, sizeof(args_t), 0, + reinterpret_cast(&kernArgs)); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // initialize the host buffers + for (int i = 0; i < kMemoryAllocSize; ++i) { + // unsigned int seed = time(NULL); + Indata[i] = i; + } + + // for the dGPU, we have coarse grained local memory, + // so allocate memory for it on the GPU's GLOBAL segment . + + // Get local memory of GPU to allocate device side buffers + uint32_t *OutData = NULL; + err = hsa_amd_memory_pool_allocate(gpu_pool, kMemoryAllocSize, 0, + reinterpret_cast(&OutData)); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + + // Allow cpuAgent access to all allocated GPU memory. + err = hsa_amd_agents_allow_access(1, &cpuAgent, NULL, OutData); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + memset(OutData, 0, kMemoryAllocSize * sizeof(int)); + + // Allow gpuAgent access to all allocated system memory. + err = hsa_amd_agents_allow_access(1, &gpuAgent, NULL, Indata); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + err = hsa_amd_agents_allow_access(1, &gpuAgent, NULL, kernArgs); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + uint32_t grp_offset = group_segment_size(); + kernArgs->a = Indata; + // gpu memory where data will be copied from dynamically group memory + kernArgs->b = OutData; + kernArgs->grp_offset = grp_offset; + kernArgs->count = kMemoryAllocSize; + + // Fill up the kernel packet except header + err = rocrtst::InitializeAQLPacket(this, &aql()); + ASSERT_EQ(HSA_STATUS_SUCCESS, err); + + // Create the executable, get symbol by name and load the code object + set_kernel_file_name("groupMemoryDynamic_kernels.hsaco"); + set_kernel_name("group_memory_dynamic"); + err = rocrtst::LoadKernelFromObjFile(this, &gpuAgent); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // The total byte size of group memory, static + dynamic + uint32_t total_grp_byte_size = group_segment_size() + kMemoryAllocSize * sizeof(uint32_t); + if (verbosity() > 0) { + std::cout << "aql.total_grp_byte_size" << total_grp_byte_size << std::endl; + } + + // Fill up the kernel packet except header + err = rocrtst::InitializeAQLPacket(this, &aql()); + ASSERT_EQ(HSA_STATUS_SUCCESS, err); + + aql().workgroup_size_x = 256; + aql().workgroup_size_y = 1; + aql().workgroup_size_z = 1; + aql().grid_size_y = 1; + aql().grid_size_z = 1; + aql().private_segment_size = 0; + aql().grid_size_x = kMemoryAllocSize; + aql().group_segment_size = total_grp_byte_size; + aql().kernel_object = kernel_object(); + aql().kernarg_address = kernArgs; + + const uint32_t queue_mask = queue->size - 1; + + // Load index for writing header later to command queue at same index + uint64_t index = hsa_queue_load_write_index_relaxed(queue); + + // This function simply copies the data we've collected so far into our + // local AQL packet, except the the setup and header fields. + WriteAQLPktToQueue(queue); + + + aql().header = HSA_PACKET_TYPE_KERNEL_DISPATCH; + aql().header |= HSA_FENCE_SCOPE_SYSTEM << + HSA_PACKET_HEADER_ACQUIRE_FENCE_SCOPE; + aql().header |= HSA_FENCE_SCOPE_SYSTEM << + HSA_PACKET_HEADER_RELEASE_FENCE_SCOPE; + + void* q_base = queue->base_address; + // Set the Aql packet header + AtomicSetPacketHeader(aql().header, aql().setup, + &(reinterpret_cast + (q_base))[index & queue_mask]); + + + // ringdoor bell + hsa_signal_store_relaxed(queue->doorbell_signal, index); + + // wait for the signal and reset it for future use + while (hsa_signal_wait_scacquire(aql().completion_signal, HSA_SIGNAL_CONDITION_LT, 1, + (uint64_t)-1, HSA_WAIT_STATE_ACTIVE)) { } + + hsa_signal_store_relaxed(aql().completion_signal, 1); + + // compare Results + for (int i = 0; i < kMemoryAllocSize; ++i) { + if (verbosity() > 0) { + // std::cout<< i << "OutData[i]" << OutData[i] << "Indata[i]" << Indata[i] < 0) { + std::cout<< "Test not applicable as system is not large bar." + "Skipping."<< std::endl; + std::cout << kSubTestSeparator << std::endl; + } + return; + } +} + + + +void MemoryAllocationTest::GroupMemoryDynamicAllocation(void) { + hsa_status_t err; + if (verbosity() > 0) { + PrintMemorySubtestHeader("Memory Group dynamic allocation"); + } + // find all cpu agents + std::vector cpus; + err = hsa_iterate_agents(rocrtst::IterateCPUAgents, &cpus); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + // find all gpu agents + std::vector gpus; + err = hsa_iterate_agents(rocrtst::IterateGPUAgents, &gpus); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + for (unsigned int i = 0 ; i< gpus.size(); ++i) { + GroupMemoryDynamicAllocation(cpus[0], gpus[i]); + } + + if (verbosity() > 0) { + std::cout << "subtest Passed" << std::endl; + std::cout << kSubTestSeparator << std::endl; + } +} + + +static void PrintAgentNameAndType(hsa_agent_t agent) { + hsa_status_t err; + + char ag_name[64]; + hsa_device_type_t ag_type; + + err = hsa_agent_get_info(agent, HSA_AGENT_INFO_NAME, ag_name); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + err = hsa_agent_get_info(agent, HSA_AGENT_INFO_DEVICE, &ag_type); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + std::cout << " Agent: " << ag_name << " ("; + switch (ag_type) { + case HSA_DEVICE_TYPE_CPU: + std::cout << "CPU)"; + break; + case HSA_DEVICE_TYPE_GPU: + std::cout << "GPU)"; + break; + case HSA_DEVICE_TYPE_DSP: + std::cout << "DSP)"; + break; + } + std::cout << std::endl; + return; +} + +static void PrintSegmentNameAndType(uint32_t segment) { + switch (segment) { + case HSA_AMD_SEGMENT_GLOBAL: + std::cout << " GLOBAL SEGMENT"; + break; + case HSA_AMD_SEGMENT_GROUP: + std::cout << " GROUP SEGMENT"; + break; + case HSA_AMD_SEGMENT_PRIVATE: + std::cout << " PRIVATE SEGMENT"; + break; + case HSA_AMD_SEGMENT_READONLY: + std::cout << " READONLY SEGMENT"; + break; + default: + std::cout << " no segment"; + break; + } + std::cout << std::endl; + return; +} + +void MemoryAllocationTest::MemoryBasicAllocationAndFree(hsa_agent_t agent, + hsa_amd_memory_pool_t pool) { + hsa_status_t err; + + rocrtst::pool_info_t pool_i; + err = rocrtst::AcquirePoolInfo(pool, &pool_i); + ASSERT_EQ(HSA_STATUS_SUCCESS, err); + + if (verbosity() > 0) { + PrintAgentNameAndType(agent); + } + + // if allocation is allowed in this pool allocate the memory + // and then free it + if (pool_i.alloc_allowed) { + if (verbosity() > 0) { + PrintSegmentNameAndType(pool_i.segment); + } + size_t max_size; + err = hsa_amd_memory_pool_get_info(pool, HSA_AMD_MEMORY_POOL_INFO_SIZE, + &max_size); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + max_size = (max_size > kMemoryAllocSize) ? kMemoryAllocSize : max_size; + + char *memoryPtr; + err = hsa_amd_memory_pool_allocate(pool, max_size , 0, + reinterpret_cast(&memoryPtr)); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + if (memoryPtr) { + err = hsa_amd_memory_pool_free(memoryPtr); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + } + } + return; +} + + +void MemoryAllocationTest::MemoryBasicAllocationAndFree(void) { + hsa_status_t err; + std::vector> agent_pools; + if (verbosity() > 0) { + PrintMemorySubtestHeader("MemoryBasicAllocationAndFree"); + } + + err = rocrtst::GetAgentPools(&agent_pools); + ASSERT_EQ(err, HSA_STATUS_SUCCESS); + + auto pool_idx = 0; + for (auto a : agent_pools) { + for (auto p : a->pools) { + if (verbosity() > 0) { + std::cout << " Pool " << pool_idx++ << ":" << std::endl; + } + MemoryBasicAllocationAndFree(a->agent, p); + } + } + + if (verbosity() > 0) { + std::cout << "subtest Passed" << std::endl; + std::cout << kSubTestSeparator << std::endl; + } +} + diff --git a/rocrtst/suites/functional/memory_allocation.h b/rocrtst/suites/functional/memory_allocation.h new file mode 100755 index 0000000000..3ca04114d7 --- /dev/null +++ b/rocrtst/suites/functional/memory_allocation.h @@ -0,0 +1,89 @@ +/* + * ============================================================================= + * ROC Runtime Conformance Release License + * ============================================================================= + * The University of Illinois/NCSA + * Open Source License (NCSA) + * + * Copyright (c) 2018, Advanced Micro Devices, Inc. + * All rights reserved. + * + * Developed by: + * + * AMD Research and AMD ROC Software Development + * + * Advanced Micro Devices, Inc. + * + * www.amd.com + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal with the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * - Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimers. + * - Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimers in + * the documentation and/or other materials provided with the distribution. + * - Neither the names of , + * nor the names of its contributors may be used to endorse or promote + * products derived from this Software without specific prior written + * permission. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS WITH THE SOFTWARE. + * + */ +#ifndef ROCRTST_SUITES_FUNCTIONAL_MEMORY_ALLOCATION_H_ +#define ROCRTST_SUITES_FUNCTIONAL_MEMORY_ALLOCATION_H_ + + +#include "common/base_rocr.h" +#include "hsa/hsa.h" +#include "suites/test_common/test_base.h" + +class MemoryAllocationTest : public TestBase { + public: + MemoryAllocationTest(bool launch_GroupMemory, + bool launch_BasicAllocateFree); + + // @Brief: Destructor for test case of MemoryTest + virtual ~MemoryAllocationTest(); + + // @Brief: Setup the environment for measurement + virtual void SetUp(); + + // @Brief: Core measurement execution + virtual void Run(); + + // @Brief: Clean up and retrive the resource + virtual void Close(); + + // @Brief: Display results + virtual void DisplayResults() const; + + // @Brief: Display information about what this test does + virtual void DisplayTestInfo(void); + + void GroupMemoryDynamicAllocation(void); + + void MemoryBasicAllocationAndFree(void); + + + private: + void GroupMemoryDynamicAllocation(hsa_agent_t cpuAgent, hsa_agent_t gpuAgent); + void MemoryBasicAllocationAndFree(hsa_agent_t agent, + hsa_amd_memory_pool_t pool); + + void WriteAQLPktToQueue(hsa_queue_t* q); +}; + +#endif // ROCRTST_SUITES_FUNCTIONAL_MEMORY_ALLOCATION_H_ diff --git a/rocrtst/suites/test_common/CMakeLists.txt b/rocrtst/suites/test_common/CMakeLists.txt index 14658547cd..7061514e99 100755 --- a/rocrtst/suites/test_common/CMakeLists.txt +++ b/rocrtst/suites/test_common/CMakeLists.txt @@ -342,6 +342,11 @@ set(BITCODE_LIBS "${COMMON_BITCODE_LIBS}") set(CL_FILE_LIST "${KERNELS_DIR}/signal_operations.cl") build_sample_for_devices("signal_operations") +# groupMemoryDynamic +set(BITCODE_LIBS "${COMMON_BITCODE_LIBS}") +set(CL_FILE_LIST "${KERNELS_DIR}/groupMemoryDynamic_kernels.cl") +build_sample_for_devices("groupMemoryDynamic") + # Build rules add_executable(${ROCRTST} ${performanceSources} ${functionalSources} ${negativeSources} ${stressSources} ${common_srcs} ${testCommonSources}) diff --git a/rocrtst/suites/test_common/kernels/groupMemoryDynamic_kernels.cl b/rocrtst/suites/test_common/kernels/groupMemoryDynamic_kernels.cl new file mode 100755 index 0000000000..94d331d802 --- /dev/null +++ b/rocrtst/suites/test_common/kernels/groupMemoryDynamic_kernels.cl @@ -0,0 +1,57 @@ +/* + * ============================================================================= + * ROC Runtime Conformance Release License + * ============================================================================= + * The University of Illinois/NCSA + * Open Source License (NCSA) + * + * Copyright (c) 2018, Advanced Micro Devices, Inc. + * All rights reserved. + * + * Developed by: + * + * AMD Research and AMD ROC Software Development + * + * Advanced Micro Devices, Inc. + * + * www.amd.com + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal with the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * - Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimers. + * - Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimers in + * the documentation and/or other materials provided with the distribution. + * - Neither the names of , + * nor the names of its contributors may be used to endorse or promote + * products derived from this Software without specific prior written + * permission. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS WITH THE SOFTWARE. + * + */ + + + +__kernel void group_memory_dynamic(__global uint * in, + __global uint * out,__local uint *grp_offset,__private uint count) { + __local uint grp[256]; + __private int gid = get_global_id(0); + if( gid > count) + return; + grp_offset[gid] = in[gid]; + out[gid] = grp_offset[gid]; + return; +} diff --git a/rocrtst/suites/test_common/main.cc b/rocrtst/suites/test_common/main.cc index 9327a1338d..c4a142272e 100755 --- a/rocrtst/suites/test_common/main.cc +++ b/rocrtst/suites/test_common/main.cc @@ -54,6 +54,7 @@ #include "suites/functional/ipc.h" #include "suites/functional/memory_alignment.h" #include "suites/functional/memory_atomics.h" +#include "suites/functional/memory_allocation.h" #include "suites/performance/dispatch_time.h" #include "suites/performance/memory_async_copy.h" #include "suites/performance/memory_async_copy_numa.h" @@ -138,6 +139,20 @@ TEST(rocrtstFunc, MemoryAccessTests) { RunCustomTestEpilog(&mt); } +TEST(rocrtstFunc, GroupMemoryAllocationTest) { + MemoryAllocationTest ma(true, false); + RunCustomTestProlog(&ma); + ma.GroupMemoryDynamicAllocation(); + RunCustomTestEpilog(&ma); +} + +TEST(rocrtstFunc, MemoryAllocateAndFreeTest) { + MemoryAllocationTest ma(false, true); + RunCustomTestProlog(&ma); + ma.MemoryBasicAllocationAndFree(); + RunCustomTestEpilog(&ma); +} + TEST(rocrtstFunc, Concurrent_Init_Test) { ConcurrentInitTest ci; RunCustomTestProlog(&ci);