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

ROCm SMI LIB: Add Missing GPU Blocks

Browse Source 
This patch adds the following 4 missing GPU blocks to the SMI LIB:
-RSMI_GPU_BLOCK_MMHUB
-RSMI_GPU_BLOCK_PCIE_BIF
-RSMI_GPU_BLOCK_HDP
-RSMI_GPU_BLOCK_XGMI_WAFL

Signed-off-by: Ori Messinger <Ori.Messinger@amd.com>
Change-Id: Ia1ec6f53e195f4bf7b8f073d6bed4fdb6572e546


[ROCm/amdsmi commit: 9d6403bb17]
This commit is contained in:
Ori Messinger
2022-05-03 05:33:04 -04:00
parent 65841a8fd0
commit 821ffaa5b9
4 changed files with 56 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/amdsmi/include/rocm_smi/rocm_smi_device.h
+4
View File
@@ -118,6 +118,10 @@ enum DevInfoTypes {
kDevErrCntSDMA,
kDevErrCntUMC,
kDevErrCntGFX,
kDevErrCntMMHUB,
kDevErrCntPCIEBIF,
kDevErrCntHDP,
kDevErrCntXGMIWAFL,
kDevErrCntFeatures,
kDevMemTotGTT,
kDevMemTotVisVRAM,
projects/amdsmi/oam/src/amd_oam.cc
+16
View File
@@ -341,6 +341,22 @@ get_device_error_count(oam_dev_handle_t *handle,
type = amd::smi::kDevErrCntGFX;
break;
case RSMI_GPU_BLOCK_MMHUB:
type = amd::smi::kDevErrCntMMHUB;
break;
case RSMI_GPU_BLOCK_PCIE_BIF:
type = amd::smi::kDevErrCntPCIEBIF;
break;
case RSMI_GPU_BLOCK_HDP:
type = amd::smi::kDevErrCntHDP;
break;
case RSMI_GPU_BLOCK_XGMI_WAFL:
type = amd::smi::kDevErrCntXGMIWAFL;
break;
default:
return RSMI_STATUS_NOT_SUPPORTED;
}
projects/amdsmi/src/rocm_smi.cc
+16
View File
@@ -604,6 +604,22 @@ rsmi_dev_ecc_count_get(uint32_t dv_ind, rsmi_gpu_block_t block,
type = amd::smi::kDevErrCntGFX;
break;
case RSMI_GPU_BLOCK_MMHUB:
type = amd::smi::kDevErrCntMMHUB;
break;
case RSMI_GPU_BLOCK_PCIE_BIF:
type = amd::smi::kDevErrCntPCIEBIF;
break;
case RSMI_GPU_BLOCK_HDP:
type = amd::smi::kDevErrCntHDP;
break;
case RSMI_GPU_BLOCK_XGMI_WAFL:
type = amd::smi::kDevErrCntXGMIWAFL;
break;
default:
return RSMI_STATUS_NOT_SUPPORTED;
}
projects/amdsmi/src/rocm_smi_device.cc
+20
View File
@@ -99,6 +99,10 @@ static const char *kDevPCIEThruPutFName = "pcie_bw";
static const char *kDevErrCntSDMAFName = "ras/sdma_err_count";
static const char *kDevErrCntUMCFName = "ras/umc_err_count";
static const char *kDevErrCntGFXFName = "ras/gfx_err_count";
static const char *kDevErrCntMMHUBFName = "ras/mmhub_err_count";
static const char *kDevErrCntPCIEBIFFName = "ras/pcie_bif_err_count";
static const char *kDevErrCntHDPFName = "ras/hdp_err_count";
static const char *kDevErrCntXGMIWAFLFName = "ras/xgmi_wafl_err_count";
static const char *kDevErrCntFeaturesFName = "ras/features";
static const char *kDevMemPageBadFName = "ras/gpu_vram_bad_pages";
static const char *kDevMemTotGTTFName = "mem_info_gtt_total";
@@ -241,6 +245,10 @@ static const std::map<DevInfoTypes, const char *> kDevAttribNameMap = {
{kDevErrCntSDMA, kDevErrCntSDMAFName},
{kDevErrCntUMC, kDevErrCntUMCFName},
{kDevErrCntGFX, kDevErrCntGFXFName},
{kDevErrCntMMHUB, kDevErrCntMMHUBFName},
{kDevErrCntPCIEBIF, kDevErrCntPCIEBIFFName},
{kDevErrCntHDP, kDevErrCntHDPFName},
{kDevErrCntXGMIWAFL, kDevErrCntXGMIWAFLFName},
{kDevErrCntFeatures, kDevErrCntFeaturesFName},
{kDevMemTotGTT, kDevMemTotGTTFName},
{kDevMemTotVisVRAM, kDevMemTotVisVRAMFName},
@@ -339,6 +347,10 @@ static std::map<DevInfoTypes, uint8_t> kDevInfoVarTypeToRSMIVariant = {
{kDevErrCntUMC, RSMI_GPU_BLOCK_UMC},
{kDevErrCntSDMA, RSMI_GPU_BLOCK_SDMA},
{kDevErrCntGFX, RSMI_GPU_BLOCK_GFX},
{kDevErrCntMMHUB, RSMI_GPU_BLOCK_MMHUB},
{kDevErrCntPCIEBIF, RSMI_GPU_BLOCK_PCIE_BIF},
{kDevErrCntHDP, RSMI_GPU_BLOCK_HDP},
{kDevErrCntXGMIWAFL, RSMI_GPU_BLOCK_XGMI_WAFL},
// rsmi_event_group_t
{kDevDFCountersAvailable, RSMI_EVNT_GRP_XGMI}
@@ -467,6 +479,10 @@ static const std::map<const char *, dev_depends_t> kDevFuncDependsMap = {
kDevErrCntUMC,
kDevErrCntSDMA,
kDevErrCntGFX,
kDevErrCntMMHUB,
kDevErrCntPCIEBIF,
kDevErrCntHDP,
kDevErrCntXGMIWAFL,
}
}
},
@@ -852,6 +868,10 @@ int Device::readDevInfo(DevInfoTypes type, std::vector<std::string> *val) {
case kDevErrCntSDMA:
case kDevErrCntUMC:
case kDevErrCntGFX:
case kDevErrCntMMHUB:
case kDevErrCntPCIEBIF:
case kDevErrCntHDP:
case kDevErrCntXGMIWAFL:
case kDevMemPageBad:
return readDevInfoMultiLineStr(type, val);
break;