OV - exec - Loop 2624

	Loop Id: 2624	Module: exec	Source: ams.c:3664-3682	Coverage: 2.87%

0x47956c LDR	D1, [X8]

0x479570 FMUL	D0, D0, D1

0x479574 LDR	D1, [X0, X9,LSL #3]

0x479578 FDIV	D0, D0, D1

0x47957c LDR	D1, [X1, X9,LSL #3]

0x479580 FADD	D0, D0, D1

0x479584 STR	D0, [X1, X9,LSL #3]

0x479588 ORR	X9, XZR, X7

0x47958c CMP	X7, X2

0x479590 B.EQ	4796f0

0x479594 LDR	X21, [X11, X9,LSL #3]

0x479598 LDR	D1, [X19]

0x47959c LDR	D0, [X10, X21,LSL #3]

0x4795a0 FCMP	D0, D1

0x4795a4 B.EQ	4796e0

0x4795a8 LDR	D0, [X12, X9,LSL #3]

0x4795ac ADD	X7, X9, #1

0x4795b0 LDR	X22, [X11, X7,LSL #3]

0x4795b4 SUBS	X23, X21, X22

0x4795b8 B.GE	479638

0x4795bc SUB	W24, W22, W21

0x4795c0 ANDS	X24, X24, #4160

0x4795c4 B.EQ	4795e4

  (2628) 0x4795c8 LDR	X25, [X13, X21,LSL #3]

  (2628) 0x4795cc LDR	D1, [X10, X21,LSL #3]

  (2628) 0x4795d0 ADD	X21, X21, #1

  (2628) 0x4795d4 SUBS	X24, X24, #1

  (2628) 0x4795d8 LDR	D2, [X14, X25,LSL #3]

  (2628) 0x4795dc FMSUB	D0, D1, D2, D0

  (2628) 0x4795e0 B.NE	4795c8

0x4795e4 CMN	X23, #4

0x4795e8 B.HI	479638

0x4795ec SUB	X22, X22, X21

0x4795f0 ADD	X23, X3, X21,LSL #3

0x4795f4 ADD	X21, X4, X21,LSL #3

0x4795f8 HINT	#0

0x4795fc HINT	#0

  (2627) 0x479600 LDP	X24, X25, [X21, #1008]

  (2627) 0x479604 LDP	D2, D3, [X23, #1008]

  (2627) 0x479608 SUBS	X22, X22, #4

  (2627) 0x47960c LDR	D1, [X14, X24,LSL #3]

  (2627) 0x479610 FMSUB	D0, D2, D1, D0

  (2627) 0x479614 LDR	D1, [X14, X25,LSL #3]

  (2627) 0x479618 LDP	X24, X25, [X21], #32

  (2627) 0x47961c LDR	D2, [X14, X24,LSL #3]

  (2627) 0x479620 FMSUB	D0, D3, D1, D0

  (2627) 0x479624 LDP	D1, D3, [X23], #32

  (2627) 0x479628 FMSUB	D0, D1, D2, D0

  (2627) 0x47962c LDR	D1, [X14, X25,LSL #3]

  (2627) 0x479630 FMSUB	D0, D3, D1, D0

  (2627) 0x479634 B.NE	479600

0x479638 LDR	X21, [X15, X9,LSL #3]

0x47963c LDR	X22, [X15, X7,LSL #3]

0x479640 SUBS	X23, X21, X22

0x479644 B.GE	47956c

0x479648 SUB	W24, W22, W21

0x47964c ANDS	X24, X24, #4160

0x479650 B.EQ	47967c

0x479654 HINT	#0

0x479658 HINT	#0

0x47965c HINT	#0

  (2626) 0x479660 LDR	X25, [X16, X21,LSL #3]

  (2626) 0x479664 LDR	D1, [X17, X21,LSL #3]

  (2626) 0x479668 ADD	X21, X21, #1

  (2626) 0x47966c SUBS	X24, X24, #1

  (2626) 0x479670 LDR	D2, [X18, X25,LSL #3]

  (2626) 0x479674 FMSUB	D0, D1, D2, D0

  (2626) 0x479678 B.NE	479660

0x47967c CMN	X23, #4

0x479680 B.HI	47956c

0x479684 SUB	X22, X22, X21

0x479688 ADD	X23, X5, X21,LSL #3

0x47968c ADD	X21, X6, X21,LSL #3

0x479690 HINT	#0

0x479694 HINT	#0

0x479698 HINT	#0

0x47969c HINT	#0

  (2625) 0x4796a0 LDP	X24, X25, [X21, #1008]

  (2625) 0x4796a4 LDP	D2, D3, [X23, #1008]

  (2625) 0x4796a8 SUBS	X22, X22, #4

  (2625) 0x4796ac LDR	D1, [X18, X24,LSL #3]

  (2625) 0x4796b0 FMSUB	D0, D2, D1, D0

  (2625) 0x4796b4 LDR	D1, [X18, X25,LSL #3]

  (2625) 0x4796b8 LDP	X24, X25, [X21], #32

  (2625) 0x4796bc LDR	D2, [X18, X24,LSL #3]

  (2625) 0x4796c0 FMSUB	D0, D3, D1, D0

  (2625) 0x4796c4 LDP	D1, D3, [X23], #32

  (2625) 0x4796c8 FMSUB	D0, D1, D2, D0

  (2625) 0x4796cc LDR	D1, [X18, X25,LSL #3]

  (2625) 0x4796d0 FMSUB	D0, D3, D1, D0

  (2625) 0x4796d4 B.NE	4796a0

0x4796d8 B	47956c

/home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c: 3664 - 3682

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

3664:       for (i = 0; i < n; i++)

3665:       {

3666:          /*-----------------------------------------------------------

3667:           * If diagonal is nonzero, relax point i; otherwise, skip it.

3668:           *-----------------------------------------------------------*/

3669:          if (A_diag_data[A_diag_i[i]] != zero)

3670:          {

3671:             res = f_data[i];

3672:             for (jj = A_diag_i[i]; jj < A_diag_i[i+1]; jj++)

3673:             {

3674:                ii = A_diag_j[jj];

3675:                res -= A_diag_data[jj] * Vtemp_data[ii];

3676:             }

3677:             for (jj = A_offd_i[i]; jj < A_offd_i[i+1]; jj++)

3678:             {

3679:                ii = A_offd_j[jj];

3680:                res -= A_offd_data[jj] * Vext_data[ii];

3681:             }

3682:             u_data[i] += (relax_weight*res)/l1_norms[i];

Coverage (%)	Name	Source Location	Module
○100.00	__kmp_invoke_microtask		libomp.so

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 /

Metric	Value
CQA speedup if no scalar integer	1.62
CQA speedup if FP arith vectorized	1.37
CQA speedup if fully vectorized	2.70
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.14
Bottlenecks	micro-operation queue,
Function	hypre_ParCSRRelaxThreads.omp_outlined.28
Source	ams.c:3664-3664,ams.c:3669-3672,ams.c:3677-3677,ams.c:3682-3682
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.13
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	3.74
CQA cycles if fully vectorized	1.90
Front-end cycles	5.13
P0 cycles	4.50
P1 cycles	4.50
P2 cycles	2.92
P3 cycles	2.92
P4 cycles	2.83
P5 cycles	2.83
P6 cycles	2.75
P7 cycles	2.75
P8 cycles	1.25
P9 cycles	1.25
P10 cycles	1.25
P11 cycles	1.25
P12 cycles	3.83
P13 cycles	3.50
P14 cycles	3.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.50
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	50.00
Nb uops	41.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	0.59
Nb FLOP add-sub	1.00
Nb FLOP mul	1.00
Nb FLOP fma	0.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	17.17
Bytes prefetched	0.00
Bytes loaded	80.00
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	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	48.08
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
Vector-efficiency ratio mul	50.00
Vector-efficiency ratio add_sub	44.44
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	50.00

Metric	Value
CQA speedup if no scalar integer	1.62
CQA speedup if FP arith vectorized	1.37
CQA speedup if fully vectorized	2.70
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.14
Bottlenecks	micro-operation queue,
Function	hypre_ParCSRRelaxThreads.omp_outlined.28
Source	ams.c:3664-3664,ams.c:3669-3672,ams.c:3677-3677,ams.c:3682-3682
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	5.13
CQA cycles if no scalar integer	3.17
CQA cycles if FP arith vectorized	3.74
CQA cycles if fully vectorized	1.90
Front-end cycles	5.13
P0 cycles	4.50
P1 cycles	4.50
P2 cycles	2.92
P3 cycles	2.92
P4 cycles	2.83
P5 cycles	2.83
P6 cycles	2.75
P7 cycles	2.75
P8 cycles	1.25
P9 cycles	1.25
P10 cycles	1.25
P11 cycles	1.25
P12 cycles	3.83
P13 cycles	3.50
P14 cycles	3.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	1.75 - 3.50
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	50.00
Nb uops	41.00
Nb loads	NA
Nb stores	1.00
Nb stack references	0.00
FLOP/cycle	0.59
Nb FLOP add-sub	1.00
Nb FLOP mul	1.00
Nb FLOP fma	0.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	17.17
Bytes prefetched	0.00
Bytes loaded	80.00
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	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	48.08
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
Vector-efficiency ratio mul	50.00
Vector-efficiency ratio add_sub	44.44
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	50.00

Path /

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

Function	hypre_ParCSRRelaxThreads.omp_outlined.28
Source file and lines	ams.c:3664-3682
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c:3664-3682.

Analyzed code is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c:3664,3669-3672,3677,3682.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=25
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed 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 25 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 5.13 to 3.17 cycles (1.62x 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. Only 48% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 5.13 to 1.90 cycles (2.70x 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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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

Presence of both ADD/SUB and MUL operations.

Workaround

Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

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

The binary loop is composed of 3 FP arithmetical operations:

1: addition or subtraction
1: multiply
1: divide

The binary loop is loading 80 bytes. The binary loop is storing 8 bytes.

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	50
loop length	200
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

5.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	4.50	4.50	2.92	2.92	2.83	2.83	2.75	2.75	1.25	1.25	1.25	1.25	3.83	3.50	3.67	0.00	0.00
cycles	4.50	4.50	2.92	2.92	2.83	2.83	2.75	2.75	1.25	1.25	1.25	1.25	3.83	3.50	3.67	0.00	0.00

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

1.75-3.50

Cycles summary

Front-end	5.13
Overall L1	5.13

Vectorization ratios

INT

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

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

INT+FP

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 5.13 cycles. At this rate:

16% of peak load performance is reached (15.61 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (1.56 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 479594

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
LDR D1, [X8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FMUL D0, D0, D1	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	3	0.25	scal (50.0%)
LDR D1, [X0, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FDIV D0, D0, D1	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	0	0	7-15	1.75-3.50	scal (50.0%)
LDR D1, [X1, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FADD D0, D0, D1	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25	scal (50.0%)
STR D0, [X1, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
ORR X9, XZR, X7	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
CMP X7, X2	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.EQ 4796f0 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x270>	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
LDR X21, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR D1, [X19]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
LDR D0, [X10, X21,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FCMP D0, D1	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
B.EQ 4796e0 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x260>	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
LDR D0, [X12, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
ADD X7, 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
LDR X22, [X11, X7,LSL #3]	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%)
SUBS X23, X21, X22	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 479638 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1b8>	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
SUB W24, W22, W21	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 (25.0%)
ANDS X24, X24, #4160	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.EQ 4795e4 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x164>	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
CMN X23, #4	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.HI 479638 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1b8>	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
SUB X22, X22, X21	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%)
ADD X23, X3, X21,LSL #3	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%)
ADD X21, X4, X21,LSL #3	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
HINT #0																					N/A
HINT #0																					N/A
LDR X21, [X15, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X22, [X15, X7,LSL #3]	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%)
SUBS X23, X21, X22	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 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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
SUB W24, W22, W21	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 (25.0%)
ANDS X24, X24, #4160	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.EQ 47967c <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1fc>	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
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
CMN X23, #4	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.HI 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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
SUB X22, X22, X21	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%)
ADD X23, X5, X21,LSL #3	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%)
ADD X21, X6, X21,LSL #3	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
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
B 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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

Function	hypre_ParCSRRelaxThreads.omp_outlined.28
Source file and lines	ams.c:3664-3682
Module	exec

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=25
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

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 5.13 to 3.17 cycles (1.62x speedup).

Workaround

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

Vectorization

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:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
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

Presence of both ADD/SUB and MUL operations.

Workaround

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

The binary loop is composed of 3 FP arithmetical operations:

1: addition or subtraction
1: multiply
1: divide

The binary loop is loading 80 bytes. The binary loop is storing 8 bytes.

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	50
loop length	200
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

5.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	4.50	4.50	2.92	2.92	2.83	2.83	2.75	2.75	1.25	1.25	1.25	1.25	3.83	3.50	3.67	0.00	0.00
cycles	4.50	4.50	2.92	2.92	2.83	2.83	2.75	2.75	1.25	1.25	1.25	1.25	3.83	3.50	3.67	0.00	0.00

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

1.75-3.50

Cycles summary

Front-end	5.13
Overall L1	5.13

Vectorization ratios

INT

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

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

INT+FP

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 5.13 cycles. At this rate:

16% of peak load performance is reached (15.61 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (1.56 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 479594

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16	Latency	Recip. throughput	Vectorization
LDR D1, [X8]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FMUL D0, D0, D1	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	3	0.25	scal (50.0%)
LDR D1, [X0, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FDIV D0, D0, D1	1	0	0	0	0	0	0	0	0	0.50	0	0.50	0	0	0	0	0	0	7-15	1.75-3.50	scal (50.0%)
LDR D1, [X1, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FADD D0, D0, D1	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	2	0.25	scal (50.0%)
STR D0, [X1, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	0	0	2	0.50	scal (50.0%)
ORR X9, XZR, X7	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
CMP X7, X2	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.EQ 4796f0 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x270>	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
LDR X21, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR D1, [X19]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
LDR D0, [X10, X21,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
FCMP D0, D1	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	2	1	scal (50.0%)
B.EQ 4796e0 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x260>	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
LDR D0, [X12, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	6	0.33	scal (50.0%)
ADD X7, 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
LDR X22, [X11, X7,LSL #3]	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%)
SUBS X23, X21, X22	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 479638 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1b8>	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
SUB W24, W22, W21	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 (25.0%)
ANDS X24, X24, #4160	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.EQ 4795e4 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x164>	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
CMN X23, #4	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.HI 479638 <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1b8>	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
SUB X22, X22, X21	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%)
ADD X23, X3, X21,LSL #3	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%)
ADD X21, X4, X21,LSL #3	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
HINT #0																					N/A
HINT #0																					N/A
LDR X21, [X15, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	0	0	4	0.33	N/A
LDR X22, [X15, X7,LSL #3]	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%)
SUBS X23, X21, X22	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 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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
SUB W24, W22, W21	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 (25.0%)
ANDS X24, X24, #4160	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.EQ 47967c <hypre_ParCSRRelaxThreads.omp_outlined.28+0x1fc>	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
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
CMN X23, #4	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.HI 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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
SUB X22, X22, X21	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%)
ADD X23, X5, X21,LSL #3	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%)
ADD X21, X6, X21,LSL #3	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
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
HINT #0																					N/A
B 47956c <hypre_ParCSRRelaxThreads.omp_outlined.28+0xec>	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

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

FMA

Workaround

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

Workaround

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