AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'rocmsmi/amd-staging' into HEAD

Browse Source 
Change-Id: I65ed7f3a0d1b6e58bc8377932d7c39db21d1b422
This commit is contained in:
Galantsev, Dmitrii
2023-09-21 19:10:41 -05:00
parent def17accbe e0483f2ee2
commit 5c41319c83
38 changed files with 1475 additions and 1055 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/amd_smi_test/functional/frequencies_read.cc
+13 -9
View File
@@ -126,16 +126,20 @@ void TestFrequenciesRead::Run(void) {
} else if (err == AMDSMI_STATUS_NOT_YET_IMPLEMENTED) {
std::cout << "\t**Get " << name <<
": Not implemented on this machine" << std::endl;
// special driver issue, shouldn't normally occur
} else if (err == AMDSMI_STATUS_UNEXPECTED_DATA) {
std::cerr << "WARN: Clock file [" << FreqEnumToStr(t) << "] exists on device [" << i << "] but empty!" << std::endl;
std::cerr << " Likely a driver issue!" << std::endl;
} else {
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported " << name << " clock frequencies: ";
std::cout << f.num_supported << std::endl;
print_frequencies(&f);
// Verify api support checking functionality is working
err = amdsmi_get_clk_freq(processor_handles_[i], t, nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
CHK_ERR_ASRT(err)
IF_VERB(STANDARD) {
std::cout << "\t**Supported " << name << " clock frequencies: ";
std::cout << f.num_supported << std::endl;
print_frequencies(&f);
// Verify api support checking functionality is working
err = amdsmi_get_clk_freq(processor_handles_[i], t, nullptr);
ASSERT_EQ(err, AMDSMI_STATUS_INVAL);
}
}
};
tests/amd_smi_test/functional/frequencies_read_write.cc
+24 -56
View File
@@ -104,8 +104,7 @@ void TestFrequenciesReadWrite::Run(void) {
for (uint32_t dv_ind = 0; dv_ind < num_monitor_devs(); ++dv_ind) {
PrintDeviceHeader(processor_handles_[dv_ind]);
for (uint32_t clk = (uint32_t)CLK_TYPE_FIRST;
clk <= CLK_TYPE__MAX; ++clk) {
for (uint32_t clk = CLK_TYPE_FIRST; clk <= CLK_TYPE__MAX; ++clk) {
amdsmi_clk = (amdsmi_clk_type_t)clk;
auto freq_read = [&]() -> bool {
@@ -121,14 +120,20 @@ void TestFrequenciesReadWrite::Run(void) {
std::cout << "\t**Set " << FreqEnumToStr(amdsmi_clk) <<
": Not supported on this machine" << std::endl;
return false;
} else {
// CHK_ERR_ASRT(ret)
IF_VERB(STANDARD) {
std::cout << "Initial frequency for clock " <<
FreqEnumToStr(amdsmi_clk) << " is " << f.current << std::endl;
}
return true;
}
// special driver issue, shouldn't normally occur
if (ret == AMDSMI_STATUS_UNEXPECTED_DATA) {
std::cerr << "WARN: Clock file [" << FreqEnumToStr(amdsmi_clk) << "] exists on device [" << dv_ind << "] but empty!" << std::endl;
std::cerr << " Likely a driver issue!" << std::endl;
}
// CHK_ERR_ASRT(ret)
IF_VERB(STANDARD) {
std::cout << "Initial frequency for clock " <<
FreqEnumToStr(amdsmi_clk) << " is " << f.current << std::endl;
}
return true;
};
auto freq_write = [&]() {
@@ -152,19 +157,18 @@ void TestFrequenciesReadWrite::Run(void) {
std::endl;
}
ret = amdsmi_set_clk_freq(processor_handles_[dv_ind], amdsmi_clk, freq_bitmask);
//Certain ASICs does not allow to set particular clocks. If set function for a clock returns
//permission error despite root access, manually set ret value to success and return
if (ret == AMDSMI_STATUS_NO_PERM && geteuid() == 0) {
std::cout << "\t**Set " << FreqEnumToStr(amdsmi_clk) <<
": Not supported on this machine. Skipping..." << std::endl;
ret = AMDSMI_STATUS_SUCCESS;
return;
} else if (ret == AMDSMI_STATUS_NOT_SUPPORTED) {
// Certain ASICs does not allow to set particular clocks. If set function for a clock returns
// permission error despite root access, manually set ret value to success and return
//
// Sometimes setting clock frequencies is completely not supported
if ((ret == AMDSMI_STATUS_NO_PERM && geteuid() == 0) ||
(ret == AMDSMI_STATUS_NOT_SUPPORTED)) {
std::cout << "\t**Set " << FreqEnumToStr(amdsmi_clk) <<
": Not supported on this machine. Skipping..." << std::endl;
ret = AMDSMI_STATUS_SUCCESS;
return;
}
CHK_ERR_ASRT(ret)
ret = amdsmi_get_clk_freq(processor_handles_[dv_ind], amdsmi_clk, &f);
if (ret != AMDSMI_STATUS_SUCCESS) {
@@ -187,7 +191,9 @@ void TestFrequenciesReadWrite::Run(void) {
}
ret = amdsmi_set_gpu_perf_level(processor_handles_[dv_ind], AMDSMI_DEV_PERF_LEVEL_AUTO);
if (ret != AMDSMI_STATUS_SUCCESS) {
if (ret == AMDSMI_STATUS_NOT_SUPPORTED) {
std::cout << "\t**Setting performance level is not supported on this machine. Skipping..." << std::endl;
ret = AMDSMI_STATUS_SUCCESS;
return;
}
};
@@ -199,44 +205,6 @@ void TestFrequenciesReadWrite::Run(void) {
}
freq_write();
CHK_ERR_ASRT(ret)
#if 0
ret = amdsmi_get_clk_freq(dv_ind, amdsmi_clk, &f);
CHK_ERR_ASRT(ret)
IF_VERB(STANDARD) {
std::cout << "Initial frequency for clock " << amdsmi_clk << " is " <<
f.current << std::endl;
}
// Set clocks to something other than the usual default of the lowest
// frequency.
freq_bitmask = 0b01100; // Try the 3rd and 4th clocks
std::string freq_bm_str =
std::bitset<AMDSMI_MAX_NUM_FREQUENCIES>(freq_bitmask).to_string();
freq_bm_str.erase(0, std::min(freq_bm_str.find_first_not_of('0'),
freq_bm_str.size()-1));
IF_VERB(STANDARD) {
std::cout << "Setting frequency mask for clock " << amdsmi_clk <<
" to 0b" << freq_bm_str << " ..." << std::endl;
}
ret = amdsmi_set_clk_freq(dv_ind, amdsmi_clk, freq_bitmask);
CHK_ERR_ASRT(ret)
ret = amdsmi_get_clk_freq(dv_ind, amdsmi_clk, &f);
CHK_ERR_ASRT(ret)
IF_VERB(STANDARD) {
std::cout << "Frequency is now index " << f.current << std::endl;
std::cout << "Resetting mask to all frequencies." << std::endl;
}
ret = amdsmi_set_clk_freq(dv_ind, amdsmi_clk, 0xFFFFFFFF);
CHK_ERR_ASRT(ret)
ret = amdsmi_set_gpu_perf_level(dv_ind, AMDSMI_DEV_PERF_LEVEL_AUTO);
CHK_ERR_ASRT(ret)
#endif
}
}
}
tests/amd_smi_test/functional/gpu_metrics_read.cc
+2 -2
View File
@@ -177,8 +177,8 @@ void TestGpuMetricsRead::Run(void) {
<< std::to_string(smu.pcie_link_speed) << '\n';
std::cout << "gfx_activity_acc="
<< std::dec << smu.gfx_activity_acc << '\n';
std::cout << "mem_actvity_acc="
<< std::dec << smu.mem_actvity_acc << '\n';
std::cout << "mem_activity_acc="
<< std::dec << smu.mem_activity_acc << '\n';
for (int i = 0; i < AMDSMI_NUM_HBM_INSTANCES; ++i) {
std::cout << "temperature_hbm[" << i << "]=" << std::dec <<
tests/amd_smi_test/functional/power_cap_read_write.cc
+7
View File
@@ -112,6 +112,13 @@ void TestPowerCapReadWrite::Run(void) {
max = info.max_power_cap;
orig = info.default_power_cap;
// Check if power cap is within the range
// skip the test otherwise
if (orig < min || orig > max) {
std::cout << "Power cap is not within the range. Skipping test for " << dv_ind << std::endl;
continue;
}
new_cap = (max + min)/2;
IF_VERB(STANDARD) {
tests/amd_smi_test/functional/power_read_write.cc
+4
View File
@@ -126,6 +126,10 @@ void TestPowerReadWrite::Run(void) {
if (ret == AMDSMI_STATUS_NOT_SUPPORTED) {
std::cout << "The power profile presets settings is not supported. "
<< std::endl;
// Verify api support checking functionality is working
ret = amdsmi_get_gpu_power_profile_presets(processor_handles_[dv_ind], 0, nullptr);
ASSERT_EQ(ret, AMDSMI_STATUS_NOT_SUPPORTED);
continue;
}
CHK_ERR_ASRT(ret)
tests/amd_smi_test/main.cc
-3
View File
@@ -106,14 +106,12 @@ static void RunCustomTestProlog(TestBase *test) {
}
test->SetUp();
test->Run();
return;
}
static void RunCustomTestEpilog(TestBase *tst) {
if (sRSMIGlvalues->verbosity >= TestBase::VERBOSE_STANDARD) {
tst->DisplayResults();
}
tst->Close();
return;
}
// If the test case one big test, you should use RunGenericTest()
@@ -125,7 +123,6 @@ static void RunCustomTestEpilog(TestBase *tst) {
static void RunGenericTest(TestBase *test) {
RunCustomTestProlog(test);
RunCustomTestEpilog(test);
return;
}
tests/amd_smi_test/rsmitst.exclude
+1 -4
View File
@@ -63,10 +63,7 @@ $BLACKLIST_ALL_ASICS\
# /sys/class/kfd/kfd/topology/nodes/*/properties
FILTER[90400]=\
$BLACKLIST_ALL_ASICS\
"rsmitstReadOnly.TestVoltCurvRead:"\
"rsmitstReadOnly.TestFrequenciesRead:"\
"rsmitstReadWrite.TestFrequenciesReadWrite:"\
"rsmitstReadWrite.TestPowerReadWrite"
"rsmitstReadOnly.TestVoltCurvRead"
FILTER[90401]=${FILTER[90400]}
FILTER[90402]=${FILTER[90400]}
tests/amd_smi_test/test_base.cc
+4 -7
View File
@@ -43,7 +43,7 @@
*
*/
#include <assert.h>
#include <cassert>
#include "amd_smi/amdsmi.h"
#include "test_base.h"
@@ -61,10 +61,9 @@ static const char kResultsLabel[] = "TEST RESULTS";
// This one is used outside this file
const char kSetupLabel[] = "TEST SETUP";
TestBase::TestBase() : setup_failed_(false), description_("") {
}
TestBase::~TestBase() {
TestBase::TestBase() : setup_failed_(false) {
}
TestBase::~TestBase() = default;
void TestBase::MakeHeaderStr(const char *inStr,
std::string *outStr) const {
@@ -155,8 +154,6 @@ void TestBase::SetUp(uint64_t init_flags) {
std::cout << "No AMD SMI tests can be run." << std::endl;
}
}
return;
}
void TestBase::PrintDeviceHeader(amdsmi_processor_handle dv_ind) {
@@ -254,7 +251,7 @@ void TestBase::set_description(std::string d) {
size_t endlptr;
for (size_t i = le; i < description_.size(); i += le) {
endlptr = description_.find_last_of(" ", i);
endlptr = description_.find_last_of(' ', i);
description_.replace(endlptr, 1, "\n");
i = endlptr;
}
tests/amd_smi_test/test_base.h
+2 -2
View File
@@ -45,6 +45,7 @@
#ifndef TESTS_AMD_SMI_TEST_TEST_BASE_H_
#define TESTS_AMD_SMI_TEST_TEST_BASE_H_
#include <cstdint>
#include <string>
#include <vector>
#include "amd_smi/amdsmi.h"
@@ -150,9 +151,8 @@ class TestBase {
"\t===> Abort is over-ridden due to dont_fail command line option." \
<< std::endl; \
return; \
} else { \
ASSERT_EQ(AMDSMI_STATUS_SUCCESS, (RET)); \
} \
ASSERT_EQ(AMDSMI_STATUS_SUCCESS, (RET)); \
}
void MakeHeaderStr(const char *inStr, std::string *outStr);
tests/amd_smi_test/test_common.cc
+3 -3
View File
@@ -43,13 +43,13 @@
*
*/
#include <assert.h>
#include <stdint.h>
#include <getopt.h>
#include <cassert>
#include <cstdint>
#include <iostream>
#include <string>
#include <map>
#include <string>
#include "test_base.h"
#include "test_common.h"
tests/amd_smi_test/test_common.h
+2 -2
View File
@@ -74,7 +74,7 @@ void DumpMonitorInfo(const TestBase *test);
#endif
#define DISPLAY_AMDSMI_ERR(RET) { \
if (RET != AMDSMI_STATUS_SUCCESS) { \
if ((RET) != AMDSMI_STATUS_SUCCESS) { \
const char *err_str; \
std::cout << "\t===> ERROR: AMDSMI call returned " << (RET) << std::endl; \
amdsmi_status_code_to_string((RET), &err_str); \
@@ -91,7 +91,7 @@ void DumpMonitorInfo(const TestBase *test);
} \
}
#define CHK_AMDSMI_PERM_ERR(RET) { \
if (RET == AMDSMI_STATUS_NO_PERM) { \
if ((RET) == AMDSMI_STATUS_NO_PERM) { \
std::cout << "This command requires root access." << std::endl; \
} else { \
DISPLAY_AMDSMI_ERR(RET) \