AI/rocm-systems
2
0
Forka 0
Codice Problemi Pull Request Actions Pacchetti Progetti Rilasci Wiki Attività

SWDEV-555178 - Calculate phys mem offset for remap range (#1879)

Sfoglia il codice sorgente
This commit is contained in:
marandje
2025-12-23 10:27:42 +01:00
committato da GitHub
parent 719556fbba
commit 3e49440495
5 ha cambiato i file con 11 aggiunte e 102 eliminazioni
Scarica il file Patch Scarica il file Diff Expand all files Collapse all files
projects/clr/rocclr/device/pal/palvirtual.cpp
+8 -5
Vedi File
@@ -2319,12 +2319,18 @@ void VirtualGPU::submitVirtualMap(amd::VirtualMapCommand& vcmd) {
// Create a view, since original base obj will map the whole memory and multimap cases wont work.
amd::Memory* vaddr_sub_obj = nullptr;
Pal::IGpuMemory* phymem_igpu_mem = nullptr;
size_t vaddr_offset = 0;
size_t phys_offset = 0;
if (phys_mem_obj != nullptr) {
constexpr bool kParent = false;
vaddr_sub_obj = phys_mem_obj->getContext().devices()[0]->CreateVirtualBuffer(
phys_mem_obj->getContext(), const_cast<void*>(vcmd.ptr()), vcmd.size(),
phys_mem_obj->getUserData().deviceId, phys_mem_obj->getUserData().locationType, kParent);
pal::Memory* phys_pal_mem = dev().getGpuMemory(phys_mem_obj);
phymem_igpu_mem = phys_pal_mem->iMem();
phys_offset = phys_pal_mem->offset();
} else {
vaddr_sub_obj = amd::MemObjMap::FindMemObj(vcmd.ptr());
}
@@ -2335,11 +2341,8 @@ void VirtualGPU::submitVirtualMap(amd::VirtualMapCommand& vcmd) {
// The imem() in the backend is shared between base and sub/view object.
pal::Memory* vaddr_pal_mem = dev().getGpuMemory(vaddr_base_obj);
Pal::IGpuMemory* phymem_igpu_mem =
(phys_mem_obj == nullptr) ? nullptr : dev().getGpuMemory(phys_mem_obj)->iMem();
Pal::VirtualMemoryRemapRange range{vaddr_pal_mem->iMem(), vaddr_offset,
phymem_igpu_mem, 0,
phymem_igpu_mem, phys_offset,
vcmd.size(), Pal::VirtualGpuMemAccessMode::NoAccess};
// Wait for previous operations before unmap
@@ -2364,7 +2367,7 @@ void VirtualGPU::submitVirtualMap(amd::VirtualMapCommand& vcmd) {
phys_mem_obj->getUserData().vaddr_mem_obj = vaddr_sub_obj;
} else {
// assert the vaddr_mem_obj is mapped and needs to be removed
amd::Memory* vaddr_sub_obj = amd::MemObjMap::FindMemObj(vcmd.ptr());
vaddr_sub_obj = amd::MemObjMap::FindMemObj(vcmd.ptr());
assert(vaddr_sub_obj != nullptr);
assert(vcmd.ptr() == vaddr_sub_obj->getSvmPtr());
projects/hip-tests/catch/hipTestMain/config/config_amd_windows
-2
Vedi File
@@ -177,7 +177,6 @@
"Unit_hipStreamCreateWithPriority_MulthreadNonblockingflag",
"SWDEV-396617 ExecMemcpyNodeSetParamsFromSymbol fails in direction",
"SWDEV-396616 hipMemMap returns invalid error",
"Unit_hipMemVmm_Basic",
"SWDEV-396615 mGPUs not considered correctly",
"Unit_hipManagedKeyword_MultiGpu",
"Disabling test tracked SWDEV-391555",
@@ -699,7 +698,6 @@
"====================================================",
"Unit_hipEventRecord",
"========Rock_Window_Failures_on_gfx1151===========================================",
"Unit_hipMemVmm_Uncached",
"Unit_Uuid_FntlTstsFor_SetEnv_HIP_VISIBLE_DEVICES",
"Unit_UUID_setEnv_Thread",
"Unit_hipTexRefSetArray_Positive",
projects/hip-tests/catch/unit/memory/hipMemVmm.cc
+2 -2
Vedi File
@@ -58,9 +58,9 @@ TEST_CASE("Unit_hipMemVmm_Basic") {
HIP_CHECK(hipMemGetAllocationGranularity(&granularity, &memAllocationProp,
hipMemAllocationGranularityRecommended));
size_t size = 4 * 1024;
size_t size = granularity * 4;
void* reservedAddress{nullptr};
HIP_CHECK(hipMemAddressReserve(&reservedAddress, size, granularity, nullptr, 0));
HIP_CHECK(hipMemAddressReserve(&reservedAddress, size, 0, nullptr, 0));
hipMemGenericAllocationHandle_t gaHandle{nullptr};
HIP_CHECK(hipMemCreate(&gaHandle, size, &memAllocationProp, 0));
projects/hip-tests/catch/unit/virtualMemoryManagement/CMakeLists.txt
+1 -2
Vedi File
@@ -55,8 +55,7 @@ set(TEST_SRC
hipMemGetAllocationPropertiesFromHandle.cc
hipMemMap.cc
hipMemRelease.cc
hipMemUnmap.cc
hipMemVmm_old.cc)
hipMemUnmap.cc)
hip_add_exe_to_target(NAME VirtualMemoryManagementTest
TEST_SRC ${TEST_SRC}
projects/hip-tests/catch/unit/virtualMemoryManagement/hipMemVmm_old.cc
-91
Vedi File
@@ -1,91 +0,0 @@
/*
Copyright (c) 2022 Advanced Micro Devices, Inc. All rights reserved.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in 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:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANNTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER INN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR INN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
*/
/* Test Case Description:
1) This testcase verifies the basic scenario - supported on
all devices
*/
#include <cstdio>
#include <hip_test_common.hh>
#include <hip_test_kernels.hh>
#include <hip_test_checkers.hh>
/*
This testcase verifies HIP Mem VMM API basic scenario - supported on all devices
*/
TEST_CASE("Unit_hipMemVmm_Basic") {
CTX_CREATE();
int vmm = 0;
HIP_CHECK(hipDeviceGetAttribute(&vmm, hipDeviceAttributeVirtualMemoryManagementSupported, 0));
INFO("hipDeviceAttributeVirtualMemoryManagementSupported: " << vmm);
if (vmm == 0) {
SUCCEED(
"GPU 0 doesn't support hipDeviceAttributeVirtualMemoryManagement "
"attribute. Hence skipping the testing with Pass result.\n");
return;
}
size_t granularity = 0;
hipMemAllocationProp memAllocationProp{};
memAllocationProp.type = hipMemAllocationTypePinned;
memAllocationProp.location.id = 0;
memAllocationProp.location.type = hipMemLocationTypeDevice;
HIP_CHECK(hipMemGetAllocationGranularity(&granularity, &memAllocationProp,
hipMemAllocationGranularityRecommended));
size_t size = granularity;
void* reservedAddress{nullptr};
HIP_CHECK(hipMemAddressReserve(&reservedAddress, size, 0, nullptr, 0));
hipMemGenericAllocationHandle_t gaHandle;
HIP_CHECK(hipMemCreate(&gaHandle, size, &memAllocationProp, 0));
HIP_CHECK(hipMemMap(reservedAddress, size, 0, gaHandle, 0));
hipDevice_t device;
HIP_CHECK(hipDeviceGet(&device, 0));
hipMemAccessDesc desc;
desc.location.type = hipMemLocationTypeDevice;
desc.location.id = device;
desc.flags = hipMemAccessFlagsProtReadWrite;
std::vector<char> values(size);
const char value = 1;
HIP_CHECK(hipMemSetAccess(reservedAddress, size, &desc, 1));
HIP_CHECK(hipMemset(reservedAddress, value, size));
HIP_CHECK(hipMemcpy(&values[0], reservedAddress, size, hipMemcpyDeviceToHost));
for (size_t i = 0; i < size; ++i) {
REQUIRE(values[i] == value);
}
HIP_CHECK(hipMemUnmap(reservedAddress, size));
HIP_CHECK(hipMemRelease(gaHandle));
HIP_CHECK(hipMemAddressFree(reservedAddress, size));
CTX_DESTROY();
}