OV - exec - Function hypre_IJMatrixAssembleParCSR._omp

	Function: hypre_IJMatrixAssembleParCSR._omp_fn.1	Module: exec	Source: IJMatrix_parcsr.c:2798-2812	Coverage (incl. loops): 0.74% \| (excl. loops): 0.00%

/home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c: 2798 - 2812

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

2798: #pragma omp parallel for private (i,j,j0,temp)

2799: #endif

2800:          for (i=0; i < num_rows; i++)

2801:          {

2802:    	    j0 = diag_i[i];

2803:    	    for (j=j0; j < diag_i[i+1]; j++)

2804:    	    {

2805:    	       diag_j[j] -= col_0;

2806:    	       if (diag_j[j] == i)

2807:    	       {

2808:    	          temp = diag_data[j0];

2809:    	          diag_data[j0] = diag_data[j];

2810:    	          diag_data[j] = temp;

2811:    	          diag_j[j] = diag_j[j0];

2812:    	          diag_j[j0] = i;

0x504d80 STP	X29, X30, [SP, #976]!

0x504d84 ADD	X29, SP, #0

0x504d88 STP	X19, X20, [SP, #16]

0x504d8c ORR	X20, XZR, X0

0x504d90 STR	X21, [SP, #32]

0x504d94 LDR	X21, [X0, #24]

0x504d98 BL	410d60

0x504d9c SBFM	X19, X0, #0, #31

0x504da0 BL	410c90

0x504da4 SBFM	X2, X0, #0, #31

0x504da8 SDIV	X9, X21, X19

0x504dac MSUB	X1, X9, X19, X21

0x504db0 CMP	X2, X1

0x504db4 B.LT	504e7c

0x504db8 MADD	X18, X9, X2, X1

0x504dbc ADD	X11, X9, X18

0x504dc0 CMP	X18, X11

0x504dc4 B.GE	504e30

0x504dc8 LDP	X10, X3, [X20]

0x504dcc LDR	X6, [X20, #16]

0x504dd0 ADD	X4, X10, #8

0x504dd4 LDR	X5, [X20, #32]

0x504dd8 HINT	#0

0x504ddc HINT	#0

      (3537) 0x504de0 LDR	X0, [X10, X18,LSL #3]

      (3537) 0x504de4 LDR	X12, [X4, X18,LSL #3]

      (3537) 0x504de8 ADD	X8, X6, X0,LSL #3

      (3537) 0x504dec ADD	X7, X3, X0,LSL #3

      (3537) 0x504df0 CMP	X0, X12

      (3537) 0x504df4 B.GE	504e24

    (3536) 0x504df8 HINT	#0

    (3536) 0x504dfc HINT	#0

    (3536) 0x504e00 LDR	X13, [X3, X0,LSL #3]

    (3536) 0x504e04 SUB	X14, X13, X5

    (3536) 0x504e08 STR	X14, [X3, X0,LSL #3]

    (3536) 0x504e0c CMP	X18, X14

    (3536) 0x504e10 B.EQ	504e40

    (3536) 0x504e14 LDR	X15, [X4, X18,LSL #3]

    (3536) 0x504e18 ADD	X0, X0, #1

    (3536) 0x504e1c CMP	X15, X0

    (3536) 0x504e20 B.GT	504e00

      (3537) 0x504e24 ADD	X18, X18, #1

      (3537) 0x504e28 CMP	X11, X18

      (3537) 0x504e2c B.NE	504de0

(3538) 0x504e30 LDR	X21, [SP, #32]

(3538) 0x504e34 LDP	X19, X20, [SP, #16]

(3538) 0x504e38 LDP	X29, X30, [SP], #48

(3538) 0x504e3c RET

  (3535) 0x504e40 LDR	D30, [X6, X0,LSL #3]

  (3535) 0x504e44 LDR	D31, [X8]

  (3535) 0x504e48 LDR	X16, [X7]

  (3535) 0x504e4c STR	D30, [X8]

  (3535) 0x504e50 STR	D31, [X6, X0,LSL #3]

  (3535) 0x504e54 STR	X16, [X3, X0,LSL #3]

  (3535) 0x504e58 ADD	X0, X0, #1

  (3535) 0x504e5c STR	X18, [X7]

  (3535) 0x504e60 LDR	X17, [X4, X18,LSL #3]

  (3535) 0x504e64 CMP	X17, X0

  (3535) 0x504e68 B.GT	504e00

  (3535) 0x504e6c ADD	X18, X18, #1

  (3535) 0x504e70 CMP	X11, X18

  (3535) 0x504e74 B.NE	504de0

(3538) 0x504e78 B	504e30

0x504e7c ADD	X9, X9, #1

0x504e80 MOVZ	X1, #0

0x504e84 B	504db8

Coverage (%)	Name	Source Location	Module
►98.98+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	thread_start		libc.so.6
►1.02+	GOMP_parallel	libgomp.h:980	libgomp.so.1.0.0
○	hypre_IJMatrixAssembleParCSR	IJMatrix_parcsr.c:2798	exec
○	BuildIJLaplacian27pt	amg.c:2272	exec
○	main	amg.c:274	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start	amg.c:253	exec

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

min	med	avg	max

Percentile Index	10	20	30	40	50	60	70	80	90	100
Value

Path /

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

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run gcc_0

Source file and lines	IJMatrix_parcsr.c:2798-2812
Module	exec

The function is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c:2798-2812.

Warnings:
Ignoring paths for analysis

gain
potential
hint
expert

Vectorization

Your function is not vectorized. Only 64% of vector register length is used (average across all VPU instructions).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers).

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 20.00 to 1.00 cycles (20.00x speedup).

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 40 bytes. The binary function is storing 40 bytes.

General properties

nb instructions	27
loop length	108
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	2.50	2.50	2.67	2.58	2.67	2.50	3.00	2.58	0.00	0.00	0.00	0.00	2.50	2.17	2.33	1.50	1.50
cycles	2.50	2.50	2.67	2.58	2.67	2.50	3.00	2.58	0.00	0.00	0.00	0.00	2.50	2.17	2.33	1.50	1.50

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions

5.00-20.00

Cycles summary

Front-end	3.13
Overall L1	5.00-20.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 20.00 cycles. At this rate:

2% of peak load performance is reached (2.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
4% of peak store performance is reached (2.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 504d80

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
STP X29, X30, [SP, #976]!	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ADD X29, SP, #0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR X20, XZR, X0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
STR X21, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (50.0%)
LDR X21, [X0, #24]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
BL 410d60 <@plt_start@+0x5a0>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SBFM X19, X0, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
BL 410c90 <@plt_start@+0x4d0>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SBFM X2, X0, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
SDIV X9, X21, X19	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	5-20	5-20	N/A
MSUB X1, X9, X19, X21	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
CMP X2, X1	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.LT 504e7c <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xfc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MADD X18, X9, X2, X1	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
ADD X11, X9, X18	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
CMP X18, X11	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.GE 504e30 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDP X10, X3, [X20]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.50	scal (100.0%)
LDR X6, [X20, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
ADD X4, X10, #8	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
LDR X5, [X20, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
HINT #0																					N/A
HINT #0																					N/A
ADD X9, X9, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVZ X1, #0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
B 504db8 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0x38>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run gcc_0

Source file and lines	IJMatrix_parcsr.c:2798-2812
Module	exec

The function is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c:2798-2812.

Warnings:
Ignoring paths for analysis

gain
potential
hint
expert

Vectorization

Your function is not vectorized. Only 64% of vector register length is used (average across all VPU instructions).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers).

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 20.00 to 1.00 cycles (20.00x speedup).

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 40 bytes. The binary function is storing 40 bytes.

General properties

nb instructions	27
loop length	108
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	2.50	2.50	2.67	2.58	2.67	2.50	3.00	2.58	0.00	0.00	0.00	0.00	2.50	2.17	2.33	1.50	1.50
cycles	2.50	2.50	2.67	2.58	2.67	2.50	3.00	2.58	0.00	0.00	0.00	0.00	2.50	2.17	2.33	1.50	1.50

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions

5.00-20.00

Cycles summary

Front-end	3.13
Overall L1	5.00-20.00

Vectorization ratios

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 20.00 cycles. At this rate:

2% of peak load performance is reached (2.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
4% of peak store performance is reached (2.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the function is: 504d80

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
STP X29, X30, [SP, #976]!	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ADD X29, SP, #0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
STP X19, X20, [SP, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (100.0%)
ORR X20, XZR, X0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
STR X21, [SP, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0.50	0.50	1	0.50	scal (50.0%)
LDR X21, [X0, #24]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
BL 410d60 <@plt_start@+0x5a0>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SBFM X19, X0, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
BL 410c90 <@plt_start@+0x4d0>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SBFM X2, X0, #0, #31	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (100.0%)
SDIV X9, X21, X19	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	5-20	5-20	N/A
MSUB X1, X9, X19, X21	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
CMP X2, X1	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.LT 504e7c <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xfc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MADD X18, X9, X2, X1	1	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
ADD X11, X9, X18	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
CMP X18, X11	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.GE 504e30 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xb0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDP X10, X3, [X20]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.50	scal (100.0%)
LDR X6, [X20, #16]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
ADD X4, X10, #8	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
LDR X5, [X20, #32]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	scal (50.0%)
HINT #0																					N/A
HINT #0																					N/A
ADD X9, X9, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVZ X1, #0	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	1	0.17	scal (50.0%)
B 504db8 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0x38>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Name	Coverage (%)	Time (s)
▼hypre_IJMatrixAssembleParCSR._omp_fn.1–	0.74	0.12
▼Loop 3538 - IJMatrix_parcsr.c:2798-2812 - exec–	0.00	0.00
▼Loop 3535 - IJMatrix_parcsr.c:2802-2812 - exec–	0.17	0.02
▼Loop 3536 - IJMatrix_parcsr.c:2802-2811 - exec–	0.57	0.07
○Loop 3537 - IJMatrix_parcsr.c:2802-2811 - exec	0.00	0.00

Report Configuration

Vectorization

Details

Expensive FP math instructions/calls

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Expensive FP math instructions/calls

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code