OV - exec - Loop 520

0x421708 ORR	W1, WZR, W19

0x42170c ADRP	X0,

0x421710 ADD	X0, X0, #792

0x421714 BL	40f770

0x421718 LDP	X20, X19, [SP, #128]

0x42171c LDP	X22, X21, [SP, #112]

0x421720 LDP	X24, X23, [SP, #96]

0x421724 LDP	X26, X25, [SP, #80]

0x421728 LDP	X28, X27, [SP, #64]

0x42172c LDP	X29, X30, [SP, #48]

0x421730 ADD	SP, SP, #144

0x421734 RET

0x421738 LDR	X10, [X29, #112]

0x42173c FMOV	D0, #1.0000000

0x421740 B	421750

0x421744 CMP	X9, X8

0x421748 ADD	X9, X9, #1

0x42174c B.EQ	421708

0x421750 LDR	X12, [X20, X9,LSL #3]

0x421754 LDR	D1, [X21, X12,LSL #3]

0x421758 FCMP	D1, D0

0x42175c B.LE	421744

0x421760 ADD	X11, X12, #1

0x421764 LDR	X13, [X22, X12,LSL #3]

0x421768 LDR	X14, [X22, X11,LSL #3]

0x42176c B	42178c

  (524) 0x421780 STR	XZR, [X24, X15,LSL #3]

  (524) 0x421784 LDR	X14, [X22, X11,LSL #3]

  (524) 0x421788 ADD	X13, X13, #1

  (524) 0x42178c CMP	X13, X14

  (524) 0x421790 B.GE	4217c0

  (524) 0x421794 LDR	X15, [X23, X13,LSL #3]

  (524) 0x421798 LDR	D2, [X21, X15,LSL #3]

  (524) 0x42179c FCMP	D2, D0

  (524) 0x4217a0 B.LE	421788

  (524) 0x4217a4 FCMP	D1, D2

  (524) 0x4217a8 B.GT	421780

  (524) 0x4217ac FCMP	D2, D1

  (524) 0x4217b0 ORR	X15, XZR, X12

  (524) 0x4217b4 B.GT	421780

  (524) 0x4217b8 B	421788

0x4217c0 LDR	X13, [X25, X12,LSL #3]

0x4217c4 LDR	X14, [X25, X11,LSL #3]

0x4217c8 CMP	X13, X14

0x4217cc B.GE	421744

/home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c: 2354 - 2372

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

2354: #pragma omp parallel for private(ig, i, jS, j, jj) HYPRE_SMP_SCHEDULE

2355: #endif

2356:         for (ig = 0; ig < graph_size; ig++)

2357:         {

2358:            i = graph_array[ig];

2359:            if (measure_array[i] > 1)

2360:            {

2361:               for (jS = S_diag_i[i]; jS < S_diag_i[i+1]; jS++)

2362:               {

2363:                  j = S_diag_j[jS];

2364:                  if (measure_array[j] > 1)

2365:                  {

2366:                     if (measure_array[i] > measure_array[j])

2367:                        CF_marker[j] = 0;

2368:                     else if (measure_array[j] > measure_array[i])

2369:                        CF_marker[i] = 0;

2370:                  }

2371:               } /* for each local neighbor j of i */

2372:               for (jS = S_offd_i[i]; jS < S_offd_i[i+1]; jS++)

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

Path /

Metric	Value
CQA speedup if no scalar integer	2.60
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.43
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.16
Bottlenecks	P10, P11, P12,
Function	.omp_outlined..19
Source	par_coarsen.c:2354-2361,par_coarsen.c:2372-2372
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.33
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	4.33
CQA cycles if fully vectorized	0.58
Front-end cycles	3.75
DIV/SQRT cycles	3.50
P0 cycles	3.50
P1 cycles	2.25
P2 cycles	2.25
P3 cycles	2.25
P4 cycles	2.25
P5 cycles	1.00
P6 cycles	0.50
P7 cycles	0.25
P8 cycles	0.25
P9 cycles	4.33
P10 cycles	4.33
P11 cycles	4.33
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	30.00
Nb uops	30.00
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	35.08
Bytes prefetched	0.00
Bytes loaded	152.00
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	0.00
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	22.50
Vector-efficiency ratio load	25.00
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	18.75

Metric	Value
CQA speedup if no scalar integer	2.60
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	7.43
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.16
Bottlenecks	P10, P11, P12,
Function	.omp_outlined..19
Source	par_coarsen.c:2354-2361,par_coarsen.c:2372-2372
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.33
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	4.33
CQA cycles if fully vectorized	0.58
Front-end cycles	3.75
DIV/SQRT cycles	3.50
P0 cycles	3.50
P1 cycles	2.25
P2 cycles	2.25
P3 cycles	2.25
P4 cycles	2.25
P5 cycles	1.00
P6 cycles	0.50
P7 cycles	0.25
P8 cycles	0.25
P9 cycles	4.33
P10 cycles	4.33
P11 cycles	4.33
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	30.00
Nb uops	30.00
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	35.08
Bytes prefetched	0.00
Bytes loaded	152.00
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	0.00
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	22.50
Vector-efficiency ratio load	25.00
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	18.75

Path /

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

Function	.omp_outlined..19
Source file and lines	par_coarsen.c:2354-2372
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2354-2372.

Analyzed code is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2354-2361,2372.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
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

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 4.33 to 1.67 cycles (2.60x 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 22% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 4.33 to 0.58 cycles (7.43x 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
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA)

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 4.33 to 0.50 cycles (8.67x speedup).

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 152 bytes.

General properties

nb instructions	30
loop length	120
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	3.50	3.50	2.25	2.25	2.25	2.25	1.00	0.50	0.25	0.25	4.33	4.33	4.33	0.00	0.00
cycles	3.50	3.50	2.25	2.25	2.25	2.25	1.00	0.50	0.25	0.25	4.33	4.33	4.33	0.00	0.00

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	3.75
Overall L1	4.33

Vectorization ratios

all	0%
load	0%
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 4.33 cycles. At this rate:

36% of peak load performance is reached (35.08 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 421744

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
ORR W1, WZR, W19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADRP X0, <4eb70c>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADD X0, X0, #792	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
BL 40f770 <@plt_start@+0x520>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50
LDP X20, X19, [SP, #128]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X22, X21, [SP, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X24, X23, [SP, #96]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X26, X25, [SP, #80]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X28, X27, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X29, X30, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
ADD SP, SP, #144	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
RET	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X10, [X29, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
FMOV D0, #1.0000000	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
B 421750 <.omp_outlined..19+0xf0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
CMP X9, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
ADD X9, X9, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.EQ 421708 <.omp_outlined..19+0xa8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X12, [X20, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR D1, [X21, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FCMP D1, D0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.LE 421744 <.omp_outlined..19+0xe4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADD X11, X12, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
LDR X13, [X22, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X14, [X22, X11,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
B 42178c <.omp_outlined..19+0x12c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X13, [X25, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X14, [X25, X11,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CMP X13, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.GE 421744 <.omp_outlined..19+0xe4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	.omp_outlined..19
Source file and lines	par_coarsen.c:2354-2372
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2354-2372.

Analyzed code is defined in /home/hbollore/qaas/qaas-runs/169-817-3176/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2354-2361,2372.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
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

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 4.33 to 1.67 cycles (2.60x 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 22% of vector register length is used (average across all VPU instructions). By vectorizing your loop, you can lower the cost of an iteration from 4.33 to 0.58 cycles (7.43x 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
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA)

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 4.33 to 0.50 cycles (8.67x speedup).

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 152 bytes.

General properties

nb instructions	30
loop length	120
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

3.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	3.50	3.50	2.25	2.25	2.25	2.25	1.00	0.50	0.25	0.25	4.33	4.33	4.33	0.00	0.00
cycles	3.50	3.50	2.25	2.25	2.25	2.25	1.00	0.50	0.25	0.25	4.33	4.33	4.33	0.00	0.00

Cycles executing div or sqrt instructions

NA

Cycles summary

Front-end	3.75
Overall L1	4.33

Vectorization ratios

all	0%
load	0%
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 4.33 cycles. At this rate:

36% of peak load performance is reached (35.08 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 421744

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
ORR W1, WZR, W19	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADRP X0, <4eb70c>	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
ADD X0, X0, #792	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
BL 40f770 <@plt_start@+0x520>	1	0.50	0.50	0.50	0.50	0	0	0	0	0	0	0	0	0	1	0.50
LDP X20, X19, [SP, #128]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X22, X21, [SP, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X24, X23, [SP, #96]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X26, X25, [SP, #80]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X28, X27, [SP, #64]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
LDP X29, X30, [SP, #48]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	1
ADD SP, SP, #144	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
RET	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X10, [X29, #112]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
FMOV D0, #1.0000000	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
B 421750 <.omp_outlined..19+0xf0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
CMP X9, X8	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
ADD X9, X9, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
B.EQ 421708 <.omp_outlined..19+0xa8>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X12, [X20, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR D1, [X21, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FCMP D1, D0	1	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1
B.LE 421744 <.omp_outlined..19+0xe4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
ADD X11, X12, #1	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
LDR X13, [X22, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X14, [X22, X11,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
B 42178c <.omp_outlined..19+0x12c>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50
LDR X13, [X25, X12,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
LDR X14, [X25, X11,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33
CMP X13, X14	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
B.GE 421744 <.omp_outlined..19+0xe4>	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

Expensive FP math instructions/calls

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

Code clean check

Workaround

Vectorization

Details

Workaround

Expensive FP math instructions/calls

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