AI/rocm-systems
2
0
フォーク 0
コード イシュー プルリクエスト Actions パッケージ プロジェクト リリース Wiki アクティビティ

kfdtest: Clean up comments

ソースを参照 
Consolidate style (use /* */ for multi-line), fix typos,
use dword instad of DWORD/DWord

Change-Id: I620e45c1687550db41127e45641b7d79d28223a1
このコミットが含まれているのは:
Kent Russell
2018-08-14 09:52:31 -04:00
親 db92d5af23
コミット 414042abf7
51個のファイルの変更、427行の追加、414行の削除
Patchファイルをダウンロード Diffファイルをダウンロード すべてのファイルを展開 すべてのファイルを折り畳む
tests/kfdtest/src/KFDMemoryTest.cpp
+75 -66
ファイルの表示
@@ -130,10 +130,10 @@ void KFDMemoryTest::TearDown() {
#define GB(x) ((x) << 30)
/*
* try to map as much as possible system memory to gpu.
* lets see if kfd support 1TB memory correctly or not.
* And after this test case, we can observe if there is any sideeffect.
* NOTICE: there are memory usage limit checks in hsa/kfd according to the total
* Try to map as much as possible system memory to gpu
* to see if KFD supports 1TB memory correctly or not.
* After this test case, we can observe if there are any side effects.
* NOTICE: There are memory usage limit checks in hsa/kfd according to the total
* physical system memory.
*/
TEST_F(KFDMemoryTest, MMapLarge) {
@@ -187,19 +187,19 @@ TEST_F(KFDMemoryTest, MMapLarge) {
TEST_END
}
/* keep memory mapped to default node
/* Keep memory mapped to default node
* Keep mapping/unmapping memory to/from non-default node
* A shader running on default node consistantly access
* memory - make sure memory is always accessible on default,
* i.e., there is no gpu vm fault.
* A shader running on default node consistantly accesses
* memory - make sure memory is always accessible by default,
* i.e. there is no gpu vm fault.
* Synchronization b/t host program and shader:
* 1. host initialize src and dst buffer to 0
* 2. shader keep reading src buffer and check value
* 3. host write src buffer to 0x5678 to indicate quit, polling dst until it becomes 0x5678
* 4. shader write dst buffer to 0x5678 after src changed to 0x5678, quit
* 5. host program quit after dst becomes 0x5678
* Need at least two gpu nodes to run the test. The defaut node has to be a gfx9 node.
* Otherwise, test is skipped. Use kfdtest --node=$$ to specify the defaut node
* 1. Host initializes src and dst buffer to 0
* 2. Shader keeps reading src buffer and check value
* 3. Host writes src buffer to 0x5678 to indicate quit, polling dst until it becomes 0x5678
* 4. Shader write dst buffer to 0x5678 after src changes to 0x5678, then quits
* 5. Host program quits after dst becomes 0x5678
* Need at least two gpu nodes to run the test. The default node has to be a gfx9 node,
* otherwise, test is skipped. Use kfdtest --node=$$ to specify the default node
* This test case is introduced as a side-result of investigation of SWDEV-134798, which
* is a gpu vm fault while running rocr conformance test. Here we try to simulate the
* same test behaviour.
@@ -250,7 +250,7 @@ TEST_F(KFDMemoryTest, MapUnmapToNodes) {
hsaKmtMapMemoryToGPUNodes(srcBuffer.As<void*>(), PAGE_SIZE, NULL, memFlags, (i>>5)&1+1, mapNodes);
}
/* fill src buffer so shader quits */
/* Fill src buffer so shader quits */
srcBuffer.Fill(0x5678);
WaitOnValue(dstBuffer.As<uint32_t *>(), 0x5678);
ASSERT_EQ(*dstBuffer.As<uint32_t *>(), 0x5678);
@@ -258,7 +258,7 @@ TEST_F(KFDMemoryTest, MapUnmapToNodes) {
TEST_END
}
// basic test of hsaKmtMapMemoryToGPU and hsaKmtUnmapMemoryToGPU
// Basic test of hsaKmtMapMemoryToGPU and hsaKmtUnmapMemoryToGPU
TEST_F(KFDMemoryTest , MapMemoryToGPU) {
TEST_START(TESTPROFILE_RUNALL)
@@ -280,7 +280,7 @@ TEST_F(KFDMemoryTest , MapMemoryToGPU) {
TEST_END
}
// following tests are for hsaKmtAllocMemory with invalid params
// Following tests are for hsaKmtAllocMemory with invalid params
TEST_F(KFDMemoryTest, InvalidMemoryPointerAlloc) {
TEST_START(TESTPROFILE_RUNALL)
@@ -299,7 +299,7 @@ TEST_F(KFDMemoryTest, ZeroMemorySizeAlloc) {
TEST_END
}
// basic test for hsaKmtAllocMemory
// Basic test for hsaKmtAllocMemory
TEST_F(KFDMemoryTest, MemoryAlloc) {
TEST_START(TESTPROFILE_RUNALL)
@@ -381,7 +381,8 @@ TEST_F(KFDMemoryTest, MemoryRegister) {
HsaMemoryBuffer sdmaBuffer((void *)&stackData[sdmaOffset], sizeof(HSAuint32));
/* Create PM4 and SDMA queues before fork+COW to test queue
* eviction and restore */
* eviction and restore
*/
PM4Queue pm4Queue;
SDMAQueue sdmaQueue;
ASSERT_SUCCESS(pm4Queue.Create(defaultGPUNode));
@@ -392,7 +393,8 @@ TEST_F(KFDMemoryTest, MemoryRegister) {
/* First submit just so the queues are not empty, and to get the
* TLB populated (in case we need to flush TLBs somewhere after
* updating the page tables) */
* updating the page tables)
*/
Dispatch dispatch0(isaBuffer);
dispatch0.SetArgs(srcBuffer.As<void*>(), dstBuffer.As<void*>());
dispatch0.Submit(pm4Queue);
@@ -410,7 +412,8 @@ TEST_F(KFDMemoryTest, MemoryRegister) {
* make any write access to the stack because we want the
* parent to make the first write access and get a new copy. A
* busy loop is the safest way to do that, since any function
* call (e.g. sleep) would write to the stack. */
* call (e.g. sleep) would write to the stack.
*/
while (1)
{}
WARN() << "Shouldn't get here!" << std::endl;
@@ -419,13 +422,15 @@ TEST_F(KFDMemoryTest, MemoryRegister) {
/* Parent process writes to COW page(s) and gets a new copy. MMU
* notifier needs to update the GPU mapping(s) for the test to
* pass. */
* pass.
*/
globalData = 0xD00BED00;
stackData[dstOffset] = 0xdeadbeef;
stackData[sdmaOffset] = 0xdeadbeef;
/* Terminate the child process before a possible test failure that
* would leave it spinning in the background indefinitely. */
* would leave it spinning in the background indefinitely.
*/
int status;
EXPECT_EQ(0, kill(pid, SIGTERM));
EXPECT_EQ(pid, waitpid(pid, &status, 0));
@@ -516,10 +521,11 @@ TEST_F(KFDMemoryTest, MemoryRegisterSamePtr) {
TEST_END
}
// FlatScratchAccess
// Since HsaMemoryBuffer has to be associated with a specific GPU node, this function in the current form
// will not work for multiple GPU nodes. For now test only one default GPU node.
// TODO: Generalize it to support multiple nodes
/* FlatScratchAccess
* Since HsaMemoryBuffer has to be associated with a specific GPU node, this function in the current form
* will not work for multiple GPU nodes. For now test only one default GPU node.
* TODO: Generalize it to support multiple nodes
*/
#define SCRATCH_SLICE_SIZE 0x10000
#define SCRATCH_SLICE_NUM 3
@@ -558,24 +564,23 @@ TEST_F(KFDMemoryTest, FlatScratchAccess) {
// Map everything for test below
ASSERT_SUCCESS(hsaKmtMapMemoryToGPU(scratchBuffer.As<char*>(), SCRATCH_SIZE, NULL));
// source & destination memory buffers
HsaMemoryBuffer srcMemBuffer(PAGE_SIZE, defaultGPUNode);
HsaMemoryBuffer dstMemBuffer(PAGE_SIZE, defaultGPUNode);
// Initialize the srcBuffer to some fixed value
srcMemBuffer.Fill(0x01010101);
// Initialize a buffer with a DWORD copy ISA
// Initialize a buffer with a dword copy ISA
m_pIsaGen->CompileShader((m_FamilyId >= FAMILY_AI) ? gfx9_ScratchCopyDword : gfx8_ScratchCopyDword,
"ScratchCopyDword", isaBuffer);
const HsaNodeProperties *pNodeProperties = m_NodeInfo.GetNodeProperties(defaultGPUNode);
// TODO: Add support to all GPU Nodes.
// The loop over the system nodes is removed as the test can be executed only on GPU nodes. This
// also requires changes to be made to all the HsaMemoryBuffer variables defined above, as
// HsaMemoryBuffer is now associated with a Node.
/* TODO: Add support to all GPU Nodes.
* The loop over the system nodes is removed as the test can be executed only on GPU nodes. This
* also requires changes to be made to all the HsaMemoryBuffer variables defined above, as
* HsaMemoryBuffer is now associated with a Node.
*/
if (pNodeProperties != NULL) {
// Get the aperture of the scratch buffer
HsaMemoryProperties *memoryProperties = new HsaMemoryProperties[pNodeProperties->NumMemoryBanks];
@@ -585,7 +590,7 @@ TEST_F(KFDMemoryTest, FlatScratchAccess) {
for (unsigned int bank = 0; bank < pNodeProperties->NumMemoryBanks; bank++) {
if (memoryProperties[bank].HeapType == HSA_HEAPTYPE_GPU_SCRATCH) {
int numWaves = 4; // WAVES must be >= # SE
int waveSize = 1; // amount of space used by each wave in units of 256 dwords...
int waveSize = 1; // Amount of space used by each wave in units of 256 dwords
PM4Queue queue;
ASSERT_SUCCESS(queue.Create(defaultGPUNode));
@@ -595,25 +600,24 @@ TEST_F(KFDMemoryTest, FlatScratchAccess) {
// Create a dispatch packet to copy
Dispatch dispatchSrcToScratch(isaBuffer);
// setup the dispatch packet
// Setup the dispatch packet
// Copying from the source Memory Buffer to the scratch buffer
dispatchSrcToScratch.SetArgs(srcMemBuffer.As<void*>(), reinterpret_cast<void*>(scratchApertureAddr));
dispatchSrcToScratch.SetDim(1, 1, 1);
dispatchSrcToScratch.SetScratch(numWaves, waveSize, scratchBuffer.As<uint64_t>());
// submit the packet
// Submit the packet
dispatchSrcToScratch.Submit(queue);
dispatchSrcToScratch.Sync();
// Create another dispatch packet to copy scratch buffer contents to destination buffer.
Dispatch dispatchScratchToDst(isaBuffer);
// set the arguments to copy from the scratch buffer
// to the destination buffer
// Set the arguments to copy from the scratch buffer to the destination buffer
dispatchScratchToDst.SetArgs(reinterpret_cast<void*>(scratchApertureAddr), dstMemBuffer.As<void*>());
dispatchScratchToDst.SetDim(1, 1, 1);
dispatchScratchToDst.SetScratch(numWaves, waveSize, scratchBuffer.As<uint64_t>());
// submit the packet
// Submit the packet
dispatchScratchToDst.Submit(queue);
dispatchScratchToDst.Sync();
@@ -708,7 +712,7 @@ void KFDMemoryTest::BigBufferSystemMemory(int defaultGPUNode, HSAuint64 granular
lastTestedSize = sizeMB;
}
/* Save the biggest allocated system buffer forsignal handling test */
/* Save the biggest allocated system buffer for signal handling test */
LOG() << "The biggest allocated system buffer is " << std::dec
<< lastTestedSize << "MB" << std::endl;
if (lastSize)
@@ -781,7 +785,8 @@ void KFDMemoryTest::BigBufferVRAM(int defaultGPUNode, HSAuint64 granularityMB,
* is small. For example, on a typical Carrizo platform, the biggest allocated
* system buffer could be more than 14G even though it only has 4G memory.
* In that situation, it will take too much time to finish the test, because of
* the onerous memory swap operation. So we limit the buffer size that way.*/
* the onerous memory swap operation. So we limit the buffer size that way.
*/
TEST_F(KFDMemoryTest, BigBufferStressTest) {
if (!is_dgpu()) {
LOG() << "Skipping test: Running on APU fails and locks the system." << std::endl;
@@ -804,7 +809,8 @@ TEST_F(KFDMemoryTest, BigBufferStressTest) {
BigBufferVRAM(defaultGPUNode, granularityMB, NULL);
/* Repeatedly allocate and map big buffers in system memory until it fails,
* then unmap and free them. */
* then unmap and free them.
*/
#define ARRAY_ENTRIES 2048
int i = 0;
@@ -875,7 +881,8 @@ TEST_F(KFDMemoryTest, MMBench) {
/* Two SDMA queues to interleave user mode SDMA with memory
* management on either SDMA engine. Make the queues long enough
* to buffer at least nBufs x WriteData packets (7 dwords per
* packet). */
* packet).
*/
SDMAQueue sdmaQueue[2];
ASSERT_SUCCESS(sdmaQueue[0].Create(defaultGPUNode, PAGE_SIZE*8));
ASSERT_SUCCESS(sdmaQueue[1].Create(defaultGPUNode, PAGE_SIZE*8));
@@ -1094,7 +1101,8 @@ TEST_F(KFDMemoryTest, QueryPointerInfo) {
* to access its memory like a debugger would. Child copies data in
* the parent process using PTRACE_PEEKDATA and PTRACE_POKEDATA. After
* the child terminates, the parent checks that the copy was
* successful. */
* successful.
*/
TEST_F(KFDMemoryTest, PtraceAccess) {
TEST_START(TESTPROFILE_RUNALL)
@@ -1108,13 +1116,14 @@ TEST_F(KFDMemoryTest, PtraceAccess) {
void *mem[2];
unsigned i;
// Offset in the VRAM buffer to test crossing non-contiguous
// buffer boundaries. The second access starting from offset
// sizeof(HSAint64)+1 will cross a node boundary in a single access,
// for node sizes of 4MB or smaller.
/* Offset in the VRAM buffer to test crossing non-contiguous
* buffer boundaries. The second access starting from offset
* sizeof(HSAint64)+1 will cross a node boundary in a single access,
* for node sizes of 4MB or smaller.
*/
const HSAuint64 VRAM_OFFSET = (4 << 20) - 2 * sizeof(HSAint64);
// alloc system memory from node 0 and initialize it
// Alloc system memory from node 0 and initialize it
memFlags.ui32.NonPaged = 0;
ASSERT_SUCCESS(hsaKmtAllocMemory(0, PAGE_SIZE*2, memFlags, &mem[0]));
for (i = 0; i < 4*sizeof(HSAint64) + 4; i++) {
@@ -1122,7 +1131,7 @@ TEST_F(KFDMemoryTest, PtraceAccess) {
(reinterpret_cast<HSAuint8 *>(mem[0]))[PAGE_SIZE+i] = 0; // destination
}
// try to alloc local memory from GPU node
// Try to alloc local memory from GPU node
memFlags.ui32.NonPaged = 1;
if (m_NodeInfo.IsGPUNodeLargeBar(defaultGPUNode)) {
EXPECT_SUCCESS(hsaKmtAllocMemory(defaultGPUNode, PAGE_SIZE*2 + (4 << 20),
@@ -1137,13 +1146,14 @@ TEST_F(KFDMemoryTest, PtraceAccess) {
mem[1] = NULL;
}
// Allow any process to trace this one. If kernel is built without
// Yama, this is not needed, and this call will fail.
/* Allow any process to trace this one. If kernel is built without
* Yama, this is not needed, and this call will fail.
*/
#ifdef PR_SET_PTRACER
prctl(PR_SET_PTRACER, PR_SET_PTRACER_ANY, 0, 0, 0);
#endif
// Find out my pid so the child can trace it
// Find current pid so the child can trace it
pid_t tracePid = getpid();
// Fork the child
@@ -1168,8 +1178,7 @@ TEST_F(KFDMemoryTest, PtraceAccess) {
} while (!WIFSTOPPED(traceStatus));
for (i = 0; i < 4; i++) {
// Test 4 different (mis-)alignments, leaving 1-byte
// gaps between longs
// Test 4 different (mis-)alignments, leaving 1-byte gaps between longs
HSAuint8 *addr = reinterpret_cast<HSAuint8 *>(reinterpret_cast<long *>(mem[0]) + i) + i;
errno = 0;
long data = ptrace(PTRACE_PEEKDATA, tracePid, addr, NULL);
@@ -1264,7 +1273,7 @@ TEST_F(KFDMemoryTest, PtraceAccessInvisibleVram) {
ASSERT_SUCCESS(hsaKmtAllocMemory(defaultGPUNode, size, memFlags, &mem));
ASSERT_SUCCESS(hsaKmtMapMemoryToGPU(mem, size, NULL));
/* set the word before 4M boundary to 0xdeadbeefdeadbeef
/* Set the word before 4M boundary to 0xdeadbeefdeadbeef
* and the word after 4M boundary to 0xcafebabecafebabe
*/
mem0 = reinterpret_cast<void *>(reinterpret_cast<HSAuint8 *>(mem) + VRAM_OFFSET);
@@ -1309,7 +1318,7 @@ TEST_F(KFDMemoryTest, PtraceAccessInvisibleVram) {
waitpid(tracePid, &traceStatus, 0);
} while (!WIFSTOPPED(traceStatus));
/* peek the memory */
/* Peek the memory */
errno = 0;
HSAint64 data0 = ptrace(PTRACE_PEEKDATA, tracePid, mem0, NULL);
ASSERT_EQ(0, errno);
@@ -1318,7 +1327,7 @@ TEST_F(KFDMemoryTest, PtraceAccessInvisibleVram) {
ASSERT_EQ(0, errno);
ASSERT_EQ(data[1], data1);
/* swap mem0 and mem1 by poking */
/* Swap mem0 and mem1 by poking */
ASSERT_EQ(0, ptrace(PTRACE_POKEDATA, tracePid, mem0, reinterpret_cast<void *>(data[1])));
ASSERT_EQ(0, errno);
ASSERT_EQ(0, ptrace(PTRACE_POKEDATA, tracePid, mem1, reinterpret_cast<void *>(data[0])));
@@ -1404,7 +1413,7 @@ TEST_F(KFDMemoryTest, SignalHandling) {
size = (sysMemSize >> 2) & ~(HSAuint64)(PAGE_SIZE - 1);
ASSERT_SUCCESS(hsaKmtAllocMemory(0 /* system */, size, m_MemoryFlags, reinterpret_cast<void**>(&pDb)));
// verify that pDb is not null before it's being used
// Verify that pDb is not null before it's being used
ASSERT_NE(nullPtr, pDb) << "hsaKmtAllocMemory returned a null pointer";
pid_t childPid = fork();
@@ -1473,7 +1482,7 @@ TEST_F(KFDMemoryTest, CheckZeroInitializationSysMem) {
return;
}
/* check the first 64 bit */
/* Check the first 64 bits */
EXPECT_EQ(0, pDb[0]);
pDb[0] = 1;
@@ -1495,7 +1504,7 @@ TEST_F(KFDMemoryTest, CheckZeroInitializationSysMem) {
}
static inline void access(volatile void *sd, int size, int rw) {
/* Most like sit in cache*/
/* Most likely sitting in cache*/
static struct DUMMY {
char dummy[1024];
} dummy;
@@ -1509,8 +1518,8 @@ static inline void access(volatile void *sd, int size, int rw) {
}
/*
* on large-ber system, test the visible vram access speed.
* kfd is not allowd to alloc visible vram on non-largebar system.
* On large-bar system, test the visible vram access speed.
* KFD is not allowed to alloc visible vram on non-largebar system.
*/
TEST_F(KFDMemoryTest, MMBandWidth) {
TEST_REQUIRE_ENV_CAPABILITIES(ENVCAPS_64BITLINUX);
@@ -1571,7 +1580,7 @@ TEST_F(KFDMemoryTest, MMBandWidth) {
memFlags.ui32.HostAccess = 1;
memFlags.ui32.NonPaged = 0;
} else {
/* alloc visible vram*/
/* Alloc visible vram*/
allocNode = defaultGPUNode;
memFlags.ui32.PageSize = HSA_PAGE_SIZE_4KB;
memFlags.ui32.HostAccess = 1;