OV - - Function _ZN6Kripke9SweepComm13workRemainingEv

/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm.cpp: 191 - 191

--------------------------------------------------------------------------------

191:   return (recv_requests.size() > 0 || queue_sdom_ids.size() > 0);

/usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_vector.h: 919 - 919

--------------------------------------------------------------------------------

919:       { return size_type(this->_M_impl._M_finish - this->_M_impl._M_start); }

/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm/SweepComm.cpp: 43 - 54

--------------------------------------------------------------------------------

43: bool SweepComm::workRemaining(void){

44:   // If there are outstanding subdomains to process, return true

45:   if(ParallelComm::workRemaining()){

[...]

51:   waitAllSends();

52:

53:   return false;

54: }

0x9c0c0 PUSH	%RBP

0x9c0c1 MOV	%RSP,%RBP

0x9c0c4 PUSH	%RBX

0x9c0c5 PUSH	%RAX

0x9c0c6 MOV	0x30(%RDI),%RAX

0x9c0ca MOV	0x60(%RDI),%RCX

0x9c0ce XOR	0x28(%RDI),%RAX

0x9c0d2 XOR	0x58(%RDI),%RCX

0x9c0d6 OR	%RAX,%RCX

0x9c0d9 SETNE	%BL

0x9c0dc JNE	9c0e3

0x9c0de CALL	a8fc0 <@plt_start@+0x4f0>

0x9c0e3 MOV	%EBX,%EAX

0x9c0e5 ADD	$0x8,%RSP

0x9c0e9 POP	%RBX

0x9c0ea POP	%RBP

0x9c0eb RET

0x9c0ec INT	$0x3

0x9c0ed INT	$0x3

0x9c0ee INT	$0x3

0x9c0ef INT	$0x3

Path /

Average path: Display a virtual path defined by average values of all real paths

Source file and lines	SweepComm.cpp:43-54
Module	libkripke.so

The function is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm.cpp:191
/usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_vector.h:919
/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm/SweepComm.cpp:43-45,51-54

Warnings:

Ignoring paths for analysis
Failed to get the number of paths

gain
potential
hint
expert

Vectorization

Your function is not vectorized. Only 11% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your function, you can lower the cost of an iteration from 3.00 to 0.35 cycles (8.47x speedup).

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

No data for this section

CALL instructions

Detected function call instructions.

Details

Calling (and then returning from) a function prevents many compiler optimizations (like vectorization), breaks control flow (which reduces pipeline performance) and executes extra instructions to save/restore the registers used inside it, which is very expensive (dozens of cycles). Consider to inline small functions.

unknown: 1 occurrences

Type of elements and instruction set

No instructions are processing arithmetic or math operations on FP elements. This function is probably writing/copying data or processing integer elements.

Matching between your function (in the source code) and the binary function

The binary function does not contain any FP arithmetical operations. The binary function is loading 32 bytes.

General properties

nb instructions	21
nb uops	18
loop length	48
used x86 registers	6
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	2.00	1.50	1.50	1.50	1.50	1.67	1.67	1.67	0.00	0.00	0.00	0.00	0.00	0.00
cycles	2.00	1.50	1.50	1.50	1.50	1.67	1.67	1.67	0.00	0.00	0.00	0.00	0.00	0.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	3.00
Dispatch	2.00
Overall L1	3.00

Vectorization ratios

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Vector efficiency ratios

all	11%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	11%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 3.00 cycles. At this rate:

11% of peak load performance is reached (10.67 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

ASM code

In the binary file, the address of the function is: 9c0c0

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
PUSH %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOV 0x30(%RDI),%RAX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV 0x60(%RDI),%RCX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
XOR 0x28(%RDI),%RAX	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33
XOR 0x58(%RDI),%RCX	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33
OR %RAX,%RCX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
SETNE %BL	1	0.50	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 9c0e3 <_ZN6Kripke9SweepComm13workRemainingEv+0x23>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
CALL a8fc0 <@plt_start@+0x4f0>	2	0.50	0	0	0	0.50	0.33	0.33	0.33	0	0	0	0	0	0	0	0.50
MOV %EBX,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
ADD $0x8,%RSP	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33
RET	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50
INT $0x3
INT $0x3
INT $0x3
INT $0x3

Source file and lines	SweepComm.cpp:43-54
Module	libkripke.so

The function is defined in:

/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm.cpp:191
/usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/stl_vector.h:919
/beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/build/Kripke/src/Kripke/ParallelComm/SweepComm.cpp:43-45,51-54

Warnings:

Ignoring paths for analysis
Failed to get the number of paths

gain
potential
hint
expert

Vectorization

Your function is not vectorized. Only 11% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your function, you can lower the cost of an iteration from 3.00 to 0.35 cycles (8.47x speedup).

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

No data for this section

CALL instructions

Detected function call instructions.

Details

Calling (and then returning from) a function prevents many compiler optimizations (like vectorization), breaks control flow (which reduces pipeline performance) and executes extra instructions to save/restore the registers used inside it, which is very expensive (dozens of cycles). Consider to inline small functions.

unknown: 1 occurrences

Type of elements and instruction set

No instructions are processing arithmetic or math operations on FP elements. This function is probably writing/copying data or processing integer elements.

Matching between your function (in the source code) and the binary function

The binary function does not contain any FP arithmetical operations. The binary function is loading 32 bytes.

General properties

nb instructions	21
nb uops	18
loop length	48
used x86 registers	6
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 cycles

Back-end

	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5
uops	2.00	1.50	1.50	1.50	1.50	1.67	1.67	1.67	0.00	0.00	0.00	0.00	0.00	0.00
cycles	2.00	1.50	1.50	1.50	1.50	1.67	1.67	1.67	0.00	0.00	0.00	0.00	0.00	0.00

Execution ports to units layout:

ALU0/BRU0: ALU
ALU1: ALU
ALU2: ALU
ALU3: ALU
BRU1:
AGU0 (256 bits): store address, load
AGU1 (256 bits): store address, load
AGU2 (256 bits): store address, load
FP0 (256 bits): VPU, DIV/SQRT
FP1 (256 bits): VPU, DIV/SQRT
FP2 (256 bits): VPU
FP3 (256 bits): VPU
FP4 (256 bits): FP store data
FP5 (256 bits): FP store data

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	3.00
Dispatch	2.00
Overall L1	3.00

Vectorization ratios

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

Vector efficiency ratios

all	11%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	11%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 3.00 cycles. At this rate:

11% of peak load performance is reached (10.67 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

ASM code

In the binary file, the address of the function is: 9c0c0

Instruction	Nb FU	ALU0/BRU0	ALU1	ALU2	ALU3	BRU1	AGU0	AGU1	AGU2	FP0	FP1	FP2	FP3	FP4	FP5	Latency	Recip. throughput
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
PUSH %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50
MOV 0x30(%RDI),%RAX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
MOV 0x60(%RDI),%RCX	1	0	0	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	3	0.33
XOR 0x28(%RDI),%RAX	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33
XOR 0x58(%RDI),%RCX	1	0.25	0.25	0.25	0.25	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33
OR %RAX,%RCX	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
SETNE %BL	1	0.50	0	0	0.50	0	0	0	0	0	0	0	0	0	0	1	0.50
JNE 9c0e3 <_ZN6Kripke9SweepComm13workRemainingEv+0x23>	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50-1
CALL a8fc0 <@plt_start@+0x4f0>	2	0.50	0	0	0	0.50	0.33	0.33	0.33	0	0	0	0	0	0	0	0.50
MOV %EBX,%EAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17
ADD $0x8,%RSP	1	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0	0	0	1	0.25
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.33
RET	1	0.50	0	0	0	0.50	0	0	0	0	0	0	0	0	0	1	0.50
INT $0x3
INT $0x3
INT $0x3
INT $0x3

▶Scalability Runs Description

Run 2x1	Number processes: 2Number nodes: 1Number processes per node: 2Run Command: <executable> --groups 1024 --procs 2,1,1MPI Command: mpirun -np <number_processes>Dataset: Run Directory: /beegfs/hackathon/users/eoseret/qaas_runs/170-850-6313/intel/Kripke/run/oneview_runs/compilers/aocc_10/oneview_run_1708510900I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spreadOMP_NUM_THREADS: 1
Run 2x2	OMP_NUM_THREADS: 2I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x4	OMP_NUM_THREADS: 4I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x8	OMP_NUM_THREADS: 8I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x16	OMP_NUM_THREADS: 16I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x32	OMP_NUM_THREADS: 32I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x64	OMP_NUM_THREADS: 64I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread
Run 2x96	OMP_NUM_THREADS: 96I_MPI_PIN_DOMAIN: auto:scatterOMP_PLACES: threadsOMP_PROC_BIND: spread

(2x1) Efficiency	(2x1) Potential Speed-Up (%)	(2x2) Efficiency	(2x2) Potential Speed-Up (%)	(2x4) Efficiency	(2x4) Potential Speed-Up (%)	(2x8) Efficiency	(2x8) Potential Speed-Up (%)	(2x16) Efficiency	(2x16) Potential Speed-Up (%)	(2x32) Efficiency	(2x32) Potential Speed-Up (%)	(2x64) Efficiency	(2x64) Potential Speed-Up (%)	(2x96) Efficiency	(2x96) Potential Speed-Up (%)
1	0	1.5	-0	1	0	1	0	1	0	1	0	1	0	1	0

Speed-Up and Efficiency

▶Detailed Speed-Up and Efficiency

Run	Number of threads	Efficiency (ideal is 1)	Speedup	Ideal Speedup	Time (s)	Coverage (%)
2x1	1	1	1	1	0.11	0.01
2x2	1	1.5	3	2	0.06	0
2x4	1	1	1	4	0.03	0
2x8	1	1	1	8	0	0
2x16	1	1	1	16	0	0
2x32	1	1	1	32	0	0
2x64	1	1	1	64	0.06	0
2x96	1	1	1	96	0.06	0

Name	Coverage (%)	Time (s)
○_ZN6Kripke9SweepComm13workRemainingEv	0.01	0.06

Report Configuration

Vectorization

Details

Workaround

Execution units bottlenecks

CALL instructions

Details

Type of elements and instruction set

Matching between your function (in the source code) and the binary function

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Vectorization

Details

Workaround

Execution units bottlenecks

CALL instructions

Details

Type of elements and instruction set

Matching between your function (in the source code) and the binary function

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

▶Scalability Runs Description

▼Colums Filter

Speed-Up and Efficiency

▶Detailed Speed-Up and Efficiency