Skip to content

Commit

Permalink
copy metrics descriptions to analysis_configs YAML files
Browse files Browse the repository at this point in the history 
Signed-off-by: Peter Park <[email protected]>
  • Loading branch information
@peterjunpark
peterjunpark committed Jan 13, 2025
1 parent 7e7d99d commit 92a3f11
Show file tree
Hide file tree
Showing 11 changed files with 898 additions and 126 deletions.
126 changes: 125 additions & 1 deletion src/rocprof_compute_soc/analysis_configs/gfx90a/0200_system-speed-of-light.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,131 @@
# Add description/tips for each metric in this section.
# So it could be shown in hover.
Metric Description:
SALU: &SALU_anchor Scalar Arithmetic Logic Unit
Speed-of-Light:
VALU FLOPs: &VALU_FLOPs_desc >-
The total floating-point operations executed per second on the VALU. This is also presented as
a percent of the peak theoretical FLOPs achievable on the specific accelerator. Note: this
does not include any floating-point operations from MFMA instructions.
VALU IOPs: &VALU_IOPs_desc >-
The total integer operations executed per second on the VALU. This is also presented as a
percent of the peak theoretical IOPs achievable on the specific accelerator. Note: this does
not include any integer operations from MFMA instructions.
MFMA FLOPs (BF16): &MFMA_FLOPs_BF16_desc >-
The total number of 16-bit brain floating point MFMA operations executed per second. Note:
this does not include any 16-bit brain floating point operations from VALU instructions. This
is also presented as a percent of the peak theoretical BF16 MFMA operations achievable on the
specific accelerator.
MFMA FLOPs (F16): &MFMA_FLOPs_F16_desc >-
The total number of 16-bit floating point MFMA operations executed per second. Note: this does
not include any 16-bit floating point operations from VALU instructions. This is also
presented as a percent of the peak theoretical F16 MFMA operations achievable on the specific
accelerator.
MFMA FLOPs (F32): &MFMA_FLOPs_F32_desc >-
The total number of 32-bit floating point MFMA operations executed per second. Note: this does
not include any 32-bit floating point operations from VALU instructions. This is also
presented as a percent of the peak theoretical F32 MFMA operations achievable on the specific
accelerator.
MFMA FLOPs (F64): &MFMA_FLOPs_F64_desc >-
The total number of 64-bit floating point MFMA operations executed per second. Note: this does
not include any 64-bit floating point operations from VALU instructions. This is also
presented as a percent of the peak theoretical F64 MFMA operations achievable on the specific
accelerator.
MFMA IOPs (INT8): &MFMA_IOPs_INT8_desc >-
The total number of 8-bit integer MFMA operations executed per second. Note: this does not
include any 8-bit integer operations from VALU instructions. This is also presented as a
percent of the peak theoretical INT8 MFMA operations achievable on the specific accelerator.
SALU Utilization: &SALU_Utilization_desc >-
Indicates what percent of the kernel's duration the SALU was busy executing instructions.
Computed as the ratio of the total number of cycles spent by the scheduler issuing SALU or
SMEM instructions over the total CU cycles.
VALU Utilization: &VALU_Utilization_desc >-
Indicates what percent of the kernel's duration the VALU was busy executing instructions. Does
not include VMEM operations. Computed as the ratio of the total number of cycles spent by the
scheduler issuing VALU instructions over the total CU cycles.
MFMA Utilization: &MFMA_Utilization_desc >-
Indicates what percent of the kernel's duration the MFMA unit was busy executing instructions.
Computed as the ratio of the total number of cycles the MFMA was busy over the total CU
cycles.
VMEM Utilization: &VMEM_Utilization_desc >-
Indicates what percent of the kernel's duration the VMEM unit was busy executing instructions,
including both global/generic and spill/scratch operations (see the VMEM instruction count
metrics) for more detail). Does not include VALU operations. Computed as the ratio of the
total number of cycles spent by the scheduler issuing VMEM instructions over the total CU
cycles.
Branch Utilization: &Branch_Utilization_desc >-
Indicates what percent of the kernel's duration the branch unit was busy executing
instructions. Computed as the ratio of the total number of cycles spent by the scheduler
issuing branch instructions over the total CU cycles.
VALU Active Threads: &VALU_Active_Threads_desc >-
Indicates the average level of divergence within a wavefront over the lifetime of the kernel.
The number of work-items that were active in a wavefront during execution of each VALU
instruction, time-averaged over all VALU instructions run on all wavefronts in the kernel.
IPC: &IPC_desc >-
The ratio of the total number of instructions executed on the CU over the total active CU
cycles. This is also presented as a percent of the peak theoretical bandwidth achievable on
the specific accelerator.
Wavefront Occupancy: &Wavefront_Occupancy_desc >-
The time-averaged number of wavefronts resident on the accelerator over the lifetime of the
kernel. Note: this metric may be inaccurate for short-running kernels (less than 1ms). This is
also presented as a percent of the peak theoretical occupancy achievable on the specific
accelerator.
Theoretical LDS Bandwidth: &Theoretical_LDS_Bandwidth_desc >-
Indicates the maximum amount of bytes that could have been loaded from, stored to, or
atomically updated in the LDS per unit time (see LDS Bandwidth example for more detail). This
is also presented as a percent of the peak theoretical F64 MFMA operations achievable on the
specific accelerator.
LDS Bank Conflicts/Access: &LDS_Bank_Conflicts_Access_desc >-
The ratio of the number of cycles spent in the LDS scheduler due to bank conflicts (as
determined by the conflict resolution hardware) to the base number of cycles that would be
spent in the LDS scheduler in a completely uncontended case. This is also presented in
normalized form (i.e., the Bank Conflict Rate).
vL1D Cache Hit Rate: &vL1D_Cache_Hit_Rate_desc >-
The ratio of the number of vL1D cache line requests that hit in vL1D cache over the total
number of cache line requests to the vL1D cache RAM.
vL1D Cache BW: &vL1D_Cache_BW_desc >-
The number of bytes looked up in the vL1D cache as a result of VMEM instructions per unit
time. The number of bytes is calculated as the number of cache lines requested multiplied by
the cache line size. This value does not consider partial requests; so, for example, if only a
single value is requested in a cache line, the data movement will still be counted as a full
cache line. This is also presented as a percent of the peak theoretical bandwidth achievable
on the specific accelerator.
L2 Cache Hit Rate: &L2_Cache_Hit_Rate_desc >-
The ratio of the number of L2 cache line requests that hit in the L2 cache over the total
number of incoming cache line requests to the L2 cache.
L2 Cache BW: &L2_Cache_BW_desc >-
The number of bytes looked up in the L2 cache per unit time. The number of bytes is calculated
as the number of cache lines requested multiplied by the cache line size. This value does not
consider partial requests; so, for example, if only a single value is requested in a cache
line, the data movement will still be counted as a full cache line. This is also presented as
a percent of the peak theoretical bandwidth achievable on the specific accelerator.
L2-Fabric Read BW: &L2-Fabric_Read_BW_desc >-
The number of bytes read by the L2 over the Infinity Fabric™ interface per unit time. This is
also presented as a percent of the peak theoretical bandwidth achievable on the specific
accelerator.
L2-Fabric Write BW: &L2-Fabric_Write_BW_desc >-
The number of bytes sent by the L2 over the Infinity Fabric interface by write and atomic
operations per unit time. This is also presented as a percent of the peak theoretical
bandwidth achievable on the specific accelerator.
L2-Fabric Read Latency: &L2-Fabric_Read_Latency_desc >-
The time-averaged number of cycles read requests spent in Infinity Fabric before data was
returned to the L2.
L2-Fabric Write Latency: &L2-Fabric_Write_Latency_desc >-
The time-averaged number of cycles write requests spent in Infinity Fabric before a completion
acknowledgement was returned to the L2.
sL1D Cache Hit Rate: &sL1D_Cache_Hit_Rate_desc >-
The percent of sL1D requests that hit on a previously loaded line the cache. Calculated as the
ratio of the number of sL1D requests that hit over the number of all sL1D requests.
sL1D Cache BW: &sL1D_Cache_BW_desc >-
The number of bytes looked up in the sL1D cache per unit time. This is also presented as a
percent of the peak theoretical bandwidth achievable on the specific accelerator.
L1I Hit Rate: &L1I_Hit_Rate_desc >-
The percent of L1I requests that hit on a previously loaded line the cache. Calculated as the
ratio of the number of L1I requests that hit over the number of all L1I requests.
L1I BW: &L1I_BW_desc >-
The number of bytes looked up in the L1I cache per unit time. This is also presented as a
percent of the peak theoretical bandwidth achievable on the specific accelerator.
L1I Fetch Latency: &L1I_Fetch_Latency_desc >-
The average number of cycles spent to fetch instructions to a CU.

# Define the panel properties and properties of each metric in the panel.
Panel Config:
Expand Down
55 changes: 43 additions & 12 deletions src/rocprof_compute_soc/analysis_configs/gfx90a/0500_command-processor.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,37 @@
# Add description/tips for each metric in this section.
# So it could be shown in hover.
Metric Description:
Command Processor Fetcher:
CPF Utilization: &CPF_Utilization_desc >-
Percent of total cycles where the CPF was busy actively doing any work. The ratio of CPF busy
cycles over total cycles counted by the CPF.
CPF Stall: &CPF_Stall_desc >-
Percent of CPF busy cycles where the CPF was stalled for any reason.
CPF-L2 Utilization: &CPF-L2_Utilization_desc >-
Percent of total cycles counted by the CPF-L2 interface where the CPF-L2 interface was active
doing any work. The ratio of CPF-L2 busy cycles over total cycles counted by the CPF-L2.
CPF-L2 Stall: &CPF-L2_Stall_desc >-
Percent of CPF-L2 busy cycles where the CPF-L2 interface was stalled for any reason.
CPF-UTCL1 Stall: &CPF-UTCL1_Stall_desc >-
Percent of CPF busy cycles where the CPF was stalled by address translation.
Packet Processor:
CPC Utilization: &CPC_Utilization_desc >-
Percent of total cycles where the CPC was busy actively doing any work. The ratio of CPC busy
cycles over total cycles counted by the CPC.
CPC Stall Rate: &CPC_Stall_Rate_desc >-
Percent of CPC busy cycles where the CPC was stalled for any reason.
CPC Packet Decoding Utilization: &CPC_Packet_Decoding_Utilization_desc >-
Percent of CPC busy cycles spent decoding commands for processing.
CPC-Workgroup Manager Utilization: &CPC-Workgroup_Manager_Utilization_desc >-
Percent of CPC busy cycles spent dispatching workgroups to the workgroup manager.
CPC-L2 Utilization: &CPC-L2_Utilization_desc >-
Percent of total cycles counted by the CPC-L2 interface where the CPC-L2 interface was active
doing any work.
CPC-UTCL1 Stall: &CPC-UTCL1_Stall_desc >-
Percent of CPC busy cycles where the CPC was stalled by address translation.
CPC-UTCL2 Utilization: &CPC-UTCL2_Utilization_desc >-
Percent of total cycles counted by the CPC's L2 address translation interface where the CPC
was busy doing address translation work.

# Define the panel properties and properties of each metric in the panel.
Panel Config:
Expand All @@ -27,7 +58,7 @@ Panel Config:
max: MAX((((100 * CPF_CPF_STAT_BUSY) / (CPF_CPF_STAT_BUSY + CPF_CPF_STAT_IDLE))
if ((CPF_CPF_STAT_BUSY + CPF_CPF_STAT_IDLE) != 0) else None))
unit: pct
tips:
tips: *CPF_Utilization_desc
CPF Stall:
avg: AVG((((100 * CPF_CPF_STAT_STALL) / CPF_CPF_STAT_BUSY) if (CPF_CPF_STAT_BUSY
!= 0) else None))
Expand All @@ -36,7 +67,7 @@ Panel Config:
max: MAX((((100 * CPF_CPF_STAT_STALL) / CPF_CPF_STAT_BUSY) if (CPF_CPF_STAT_BUSY
!= 0) else None))
unit: pct
tips:
tips: *CPF_Stall_desc
CPF-L2 Utilization:
avg: AVG((((100 * CPF_CPF_TCIU_BUSY) / (CPF_CPF_TCIU_BUSY + CPF_CPF_TCIU_IDLE))
if ((CPF_CPF_TCIU_BUSY + CPF_CPF_TCIU_IDLE) != 0) else None))
Expand All @@ -45,7 +76,7 @@ Panel Config:
max: MAX((((100 * CPF_CPF_TCIU_BUSY) / (CPF_CPF_TCIU_BUSY + CPF_CPF_TCIU_IDLE))
if ((CPF_CPF_TCIU_BUSY + CPF_CPF_TCIU_IDLE) != 0) else None))
unit: pct
tips:
tips: *CPF-L2_Utilization_desc
CPF-L2 Stall:
avg: AVG((((100 * CPF_CPF_TCIU_STALL) / CPF_CPF_TCIU_BUSY) if (CPF_CPF_TCIU_BUSY
!= 0) else None))
Expand All @@ -54,7 +85,7 @@ Panel Config:
max: MAX((((100 * CPF_CPF_TCIU_STALL) / CPF_CPF_TCIU_BUSY) if (CPF_CPF_TCIU_BUSY
!= 0) else None))
unit: pct
tips:
tips: *CPF-L2_Stall_desc
CPF-UTCL1 Stall:
avg: AVG(((100 * CPF_CMP_UTCL1_STALL_ON_TRANSLATION) / CPF_CPF_STAT_BUSY) if (CPF_CPF_STAT_BUSY
!= 0) else None)
Expand All @@ -63,7 +94,7 @@ Panel Config:
max: MAX(((100 * CPF_CMP_UTCL1_STALL_ON_TRANSLATION) / CPF_CPF_STAT_BUSY) if (CPF_CPF_STAT_BUSY
!= 0) else None)
unit: pct
tips:
tips: *CPF-UTCL1_Stall_desc

- metric_table:
id: 502
Expand All @@ -84,7 +115,7 @@ Panel Config:
max: MAX((((100 * CPC_CPC_STAT_BUSY) / (CPC_CPC_STAT_BUSY + CPC_CPC_STAT_IDLE))
if ((CPC_CPC_STAT_BUSY + CPC_CPC_STAT_IDLE) != 0) else None))
unit: pct
tips:
tips: *CPC_Utilization_desc
CPC Stall Rate:
avg: AVG((((100 * CPC_CPC_STAT_STALL) / CPC_CPC_STAT_BUSY) if (CPC_CPC_STAT_BUSY
!= 0) else None))
Expand All @@ -93,19 +124,19 @@ Panel Config:
max: MAX((((100 * CPC_CPC_STAT_STALL) / CPC_CPC_STAT_BUSY) if (CPC_CPC_STAT_BUSY
!= 0) else None))
unit: pct
tips:
tips: *CPC_Stall_Rate_desc
CPC Packet Decoding Utilization:
avg: AVG((100 * CPC_ME1_BUSY_FOR_PACKET_DECODE) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
min: MIN((100 * CPC_ME1_BUSY_FOR_PACKET_DECODE) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
max: MAX((100 * CPC_ME1_BUSY_FOR_PACKET_DECODE) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
unit: pct
tips:
tips: *CPC_Packet_Decoding_Utilization_desc
CPC-Workgroup Manager Utilization:
avg: AVG((100 * CPC_ME1_DC0_SPI_BUSY) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
min: MIN((100 * CPC_ME1_DC0_SPI_BUSY) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
max: MAX((100 * CPC_ME1_DC0_SPI_BUSY) / CPC_CPC_STAT_BUSY if (CPC_CPC_STAT_BUSY != 0) else None)
unit: Pct
tips:
tips: *CPC-Workgroup_Manager_Utilization_desc
CPC-L2 Utilization:
avg: AVG((((100 * CPC_CPC_TCIU_BUSY) / (CPC_CPC_TCIU_BUSY + CPC_CPC_TCIU_IDLE))
if ((CPC_CPC_TCIU_BUSY + CPC_CPC_TCIU_IDLE) != 0) else None))
Expand All @@ -114,7 +145,7 @@ Panel Config:
max: MAX((((100 * CPC_CPC_TCIU_BUSY) / (CPC_CPC_TCIU_BUSY + CPC_CPC_TCIU_IDLE))
if ((CPC_CPC_TCIU_BUSY + CPC_CPC_TCIU_IDLE) != 0) else None))
unit: pct
tips:
tips: *CPC-L2_Utilization_desc
CPC-UTCL1 Stall:
avg: AVG(((100 * CPC_UTCL1_STALL_ON_TRANSLATION) / CPC_CPC_STAT_BUSY) if (CPC_CPC_STAT_BUSY
!= 0) else None)
Expand All @@ -123,7 +154,7 @@ Panel Config:
max: MAX(((100 * CPC_UTCL1_STALL_ON_TRANSLATION) / CPC_CPC_STAT_BUSY) if (CPC_CPC_STAT_BUSY
!= 0) else None)
unit: pct
tips:
tips: *CPC-UTCL1_Stall_desc
CPC-UTCL2 Utilization:
avg: AVG((((100 * CPC_CPC_UTCL2IU_BUSY) / (CPC_CPC_UTCL2IU_BUSY + CPC_CPC_UTCL2IU_IDLE))
if ((CPC_CPC_UTCL2IU_BUSY + CPC_CPC_UTCL2IU_IDLE) != 0) else None))
Expand All @@ -132,4 +163,4 @@ Panel Config:
max: MAX((((100 * CPC_CPC_UTCL2IU_BUSY) / (CPC_CPC_UTCL2IU_BUSY + CPC_CPC_UTCL2IU_IDLE))
if ((CPC_CPC_UTCL2IU_BUSY + CPC_CPC_UTCL2IU_IDLE) != 0) else None))
unit: pct
tips:
tips: *CPC-UTCL2_Utilization_desc
Loading

0 comments on commit 92a3f11

Please sign in to comment.