OV - exec - Loop 3409

0x4e1760 LDR	X0, [X10, X30,LSL #3]    [1]

0x4e1764 LDR	X12, [X5, X30,LSL #3]    [2]

0x4e1768 UBFM	X4, X0, #61, #60

0x4e176c ADD	X8, X7, X4

0x4e1770 ADD	X13, X3, X4

0x4e1774 CMP	X0, X12

0x4e1778 B.GE	4e17a4

0x4e17a4 ADD	X30, X30, #1

0x4e17a8 CMP	X11, X30

0x4e17ac B.NE	4e1760

/home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c: 2802 - 2811

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

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];

Coverage (%)	Name	Source Location	Module
►100.00+	GOMP_parallel		libomp.so
○	hypre_IJMatrixAssembleParCSR	IJMatrix_parcsr.c:2817	exec
○	BuildIJLaplacian27pt	amg.c:2272	exec
○	main	amg.c:274	exec
○	__libc_start_main		libc-2.31.so
○	_start	amg.c:599	exec

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.20
Bottlenecks	P2, P3, P4, P5,
Function	hypre_IJMatrixAssembleParCSR._omp_fn.1
Source	IJMatrix_parcsr.c:2802-2811
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	1.50
CQA cycles if no scalar integer	1.50
CQA cycles if FP arith vectorized	1.50
CQA cycles if fully vectorized	0.38
Front-end cycles	1.25
DIV/SQRT cycles	1.00
P0 cycles	1.00
P1 cycles	1.50
P2 cycles	1.50
P3 cycles	1.50
P4 cycles	1.50
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.67
P10 cycles	0.67
P11 cycles	0.67
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	10.00
Nb uops	10.00
Nb loads	NA
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	10.67
Bytes prefetched	0.00
Bytes loaded	16.00
Bytes stored	0.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	25.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.20
Bottlenecks	P2, P3, P4, P5,
Function	hypre_IJMatrixAssembleParCSR._omp_fn.1
Source	IJMatrix_parcsr.c:2802-2811
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	1.50
CQA cycles if no scalar integer	1.50
CQA cycles if FP arith vectorized	1.50
CQA cycles if fully vectorized	0.38
Front-end cycles	1.25
DIV/SQRT cycles	1.00
P0 cycles	1.00
P1 cycles	1.50
P2 cycles	1.50
P3 cycles	1.50
P4 cycles	1.50
P5 cycles	0.00
P6 cycles	0.00
P7 cycles	0.00
P8 cycles	0.00
P9 cycles	0.67
P10 cycles	0.67
P11 cycles	0.67
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	10.00
Nb uops	10.00
Nb loads	NA
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	10.67
Bytes prefetched	0.00
Bytes loaded	16.00
Bytes stored	0.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	NA
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	25.00

Path /

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

Function	hypre_IJMatrixAssembleParCSR._omp_fn.1
Source file and lines	IJMatrix_parcsr.c:2802-2811
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c:2802-2811.

The related source loop is not unrolled or unrolled with no peel/tail loop.
Warnings:

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 1.50 to 0.37 cycles (4.00x speedup).

Details

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

Workaround

Try another compiler or update/tune your current one
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop 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)

No data for this section

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

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 16 bytes.

General properties

nb instructions	10
loop length	40
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

1.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.00	1.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00
cycles	1.00	1.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	1.25
Overall L1	1.50

Vectorization ratios

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
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 iteration of the binary loop takes 1.50 cycles. At this rate:

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

ASM code

In the binary file, the address of the loop is: 4e17a4

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	Latency	Recip. throughput
LDR X0, [X10, X30,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X12, [X5, X30,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
UBFM X4, X0, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
ADD X8, X7, X4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
ADD X13, X3, X4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X0, X12	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.GE 4e17a4 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xa0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
ADD X30, X30, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X11, X30	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.NE 4e1760 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0x5c>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50

Function	hypre_IJMatrixAssembleParCSR._omp_fn.1
Source file and lines	IJMatrix_parcsr.c:2802-2811
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/IJ_mv/IJMatrix_parcsr.c:2802-2811.

The related source loop is not unrolled or unrolled with no peel/tail loop.
Warnings:

Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 1.50 to 0.37 cycles (4.00x speedup).

Details

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

Workaround

Try another compiler or update/tune your current one
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop 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)

No data for this section

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

The binary loop does not contain any FP arithmetical operations. The binary loop is loading 16 bytes.

General properties

nb instructions	10
loop length	40
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

1.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	1.00	1.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00
cycles	1.00	1.00	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.67	0.67	0.67	0.00	0.00

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	1.25
Overall L1	1.50

Vectorization ratios

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
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 iteration of the binary loop takes 1.50 cycles. At this rate:

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

ASM code

In the binary file, the address of the loop is: 4e17a4

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P10	P11	P12	Latency	Recip. throughput
LDR X0, [X10, X30,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X12, [X5, X30,LSL #3]	1	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
UBFM X4, X0, #61, #60	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
ADD X8, X7, X4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
ADD X13, X3, X4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X0, X12	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.GE 4e17a4 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0xa0>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50
ADD X30, X30, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.25
CMP X11, X30	1	0	0	0.25	0.25	0.25	0.25	0	0	0	1	0.33
B.NE 4e1760 <hypre_IJMatrixAssembleParCSR._omp_fn.1+0x5c>	1	0.50	0.50	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code