OV - exec - Loop 164

0x43a640 ADD	X18, X18, #8

0x43a644 SUBS	X13, X13, #1

0x43a648 ADD	X2, X2, #8

0x43a64c B.EQ	43a6a8

0x43a650 LDR	X0, [X2]    [3]

0x43a654 LDR	X3, [X22, X0,LSL #3]    [1]

0x43a658 CMP	X3, X5

0x43a65c B.LT	43a67c

0x43a660 LDR	D5, [X18]    [2]

0x43a664 FMUL	D6, D3, D5

0x43a668 FCMP	D6, #0

0x43a66c B.PL	43a67c

0x43a670 LDR	D6, [X15, X3,LSL #3]    [4]

0x43a674 FMADD	D5, D4, D5, D6

0x43a678 STR	D5, [X15, X3,LSL #3]    [4]

0x43a67c CMP	X0, X26

0x43a680 B.NE	43a640

0x43a684 LDR	D5, [X18]    [2]

0x43a688 FMUL	D6, D3, D5

0x43a68c FCMP	D6, #0

0x43a690 B.PL	43a640

0x43a694 FMADD	D2, D4, D5, D2

0x43a698 B	43a640

/home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/parcsr_ls/par_lr_interp.c: 1644 - 1651

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

1644:                  for(jj1 = A_diag_i[i1]+1; jj1 < A_diag_i[i1+1]; jj1++)

1645:                  {

1646:                    i2 = A_diag_j[jj1];

1647:                    if(P_marker[i2] >= jj_begin_row && (sgn*A_diag_data[jj1]) < 0)

1648:                       P_diag_data[P_marker[i2]] +=

1649:                       distribute*A_diag_data[jj1];

1650:                        if(i2 == i && (sgn*A_diag_data[jj1]) < 0)

1651:                          diagonal += distribute*A_diag_data[jj1];

Coverage (%)	Name	Source Location	Module

Path /

Metric	Value
CQA speedup if no scalar integer	1.50
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	1.04
Bottlenecks	P0, P1,
Function	.omp_outlined.#0x439400
Source	par_lr_interp.c:1644-1651
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	3.00
CQA cycles if no scalar integer	2.00
CQA cycles if FP arith vectorized	3.00
CQA cycles if fully vectorized	0.75
Front-end cycles	2.88
DIV/SQRT cycles	3.00
P0 cycles	3.00
P1 cycles	1.25
P2 cycles	1.25
P3 cycles	1.25
P4 cycles	1.25
P5 cycles	2.00
P6 cycles	1.75
P7 cycles	1.75
P8 cycles	1.50
P9 cycles	2.17
P10 cycles	1.83
P11 cycles	2.00
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	23.00
Nb uops	23.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	2.00
Nb FLOP add-sub	0.00
Nb FLOP mul	2.00
Nb FLOP fma	2.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	16.00
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	8.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
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	25.00
Vector-efficiency ratio store	25.00
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	1.50
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	1.04
Bottlenecks	P0, P1,
Function	.omp_outlined.#0x439400
Source	par_lr_interp.c:1644-1651
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	3.00
CQA cycles if no scalar integer	2.00
CQA cycles if FP arith vectorized	3.00
CQA cycles if fully vectorized	0.75
Front-end cycles	2.88
DIV/SQRT cycles	3.00
P0 cycles	3.00
P1 cycles	1.25
P2 cycles	1.25
P3 cycles	1.25
P4 cycles	1.25
P5 cycles	2.00
P6 cycles	1.75
P7 cycles	1.75
P8 cycles	1.50
P9 cycles	2.17
P10 cycles	1.83
P11 cycles	2.00
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	23.00
Nb uops	23.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	2.00
Nb FLOP add-sub	0.00
Nb FLOP mul	2.00
Nb FLOP fma	2.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	16.00
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	8.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
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	25.00
Vector-efficiency ratio store	25.00
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	.omp_outlined.#0x439400
Source file and lines	par_lr_interp.c:1644-1651
Module	exec

The loop is defined in /home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/parcsr_ls/par_lr_interp.c:1644-1651.

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

Ignoring paths for analysis
Too many paths. Rerun with max-paths=9
RecMII not computed since number of paths is unknown or > max_paths

Try to simplify control and/or increase the maximum number of paths per function/loop through the 'max-paths-nb' option.

This loop has 9 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

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 3.00 to 0.75 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)

FMA

Detected 2 FMA (fused multiply-add) operations.

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

The binary loop is composed of 6 FP arithmetical operations:

2: addition or subtraction (all inside FMA instructions)
4: multiply (2 inside FMA instructions)

The binary loop is loading 40 bytes. The binary loop is storing 8 bytes.

Arithmetic intensity

Arithmetic intensity is 0.12 FP operations per loaded or stored byte.

General properties

nb instructions	23
loop length	92
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	3.00	3.00	1.25	1.25	1.25	1.25	2.00	1.75	1.75	1.50	2.17	1.83	2.00	0.00	0.00
cycles	3.00	3.00	1.25	1.25	1.25	1.25	2.00	1.75	1.75	1.50	2.17	1.83	2.00	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	2.88
Overall L1	3.00

Vectorization ratios

all	0%
load	0%
store	0%
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 3.00 cycles. At this rate:

13% of peak load performance is reached (13.33 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
5% of peak store performance is reached (2.67 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 43a650

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
ADD X18, X18, #8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SUBS X13, X13, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
ADD X2, X2, #8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.EQ 43a6a8 <.omp_outlined.+0x12a8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X0, [X2]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X3, [X22, X0,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CMP X3, X5	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.LT 43a67c <.omp_outlined.+0x127c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D5, [X18]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D6, D3, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D6, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.PL 43a67c <.omp_outlined.+0x127c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D6, [X15, X3,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMADD D5, D4, D5, D6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
STR D5, [X15, X3,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
CMP X0, X26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.NE 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D5, [X18]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D6, D3, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D6, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.PL 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMADD D2, D4, D5, D2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
B 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	.omp_outlined.#0x439400
Source file and lines	par_lr_interp.c:1644-1651
Module	exec

The loop is defined in /home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/parcsr_ls/par_lr_interp.c:1644-1651.

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

Ignoring paths for analysis
Too many paths. Rerun with max-paths=9
RecMII not computed since number of paths is unknown or > max_paths

Try to simplify control and/or increase the maximum number of paths per function/loop through the 'max-paths-nb' option.

This loop has 9 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 3.00 to 2.00 cycles (1.50x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

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 3.00 to 0.75 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)

FMA

Detected 2 FMA (fused multiply-add) operations.

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

The binary loop is composed of 6 FP arithmetical operations:

2: addition or subtraction (all inside FMA instructions)
4: multiply (2 inside FMA instructions)

The binary loop is loading 40 bytes. The binary loop is storing 8 bytes.

Arithmetic intensity

Arithmetic intensity is 0.12 FP operations per loaded or stored byte.

General properties

nb instructions	23
loop length	92
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.88 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	3.00	3.00	1.25	1.25	1.25	1.25	2.00	1.75	1.75	1.50	2.17	1.83	2.00	0.00	0.00
cycles	3.00	3.00	1.25	1.25	1.25	1.25	2.00	1.75	1.75	1.50	2.17	1.83	2.00	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	2.88
Overall L1	3.00

Vectorization ratios

all	0%
load	0%
store	0%
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 3.00 cycles. At this rate:

13% of peak load performance is reached (13.33 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
5% of peak store performance is reached (2.67 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 43a650

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
ADD X18, X18, #8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
SUBS X13, X13, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
ADD X2, X2, #8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.EQ 43a6a8 <.omp_outlined.+0x12a8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X0, [X2]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X3, [X22, X0,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CMP X3, X5	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.LT 43a67c <.omp_outlined.+0x127c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D5, [X18]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D6, D3, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D6, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.PL 43a67c <.omp_outlined.+0x127c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D6, [X15, X3,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMADD D5, D4, D5, D6	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
STR D5, [X15, X3,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
CMP X0, X26	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.NE 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR D5, [X18]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FMUL D6, D3, D5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	3	0.25
FCMP D6, #0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.PL 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
FMADD D2, D4, D5, D2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	4	0.25
B 43a640 <.omp_outlined.+0x1240>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

FMA

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

FMA

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code