OV - exec - Loop 103

	Loop Id: 103	Module: exec	Source: par_coarsen.c:2446-2477 [...]	Coverage: 0.06%

0x42fd30 CMPQ	$0,(%RCX)

0x42fd34 JLE	42fd68

0x42fd36 MOVQ	$0x1,(%RCX)

0x42fd3d ADD	$0x8,%RSI

0x42fd41 CMP	%RSI,%R12

0x42fd44 JE	42fe1c

0x42fd4a MOV	(%RSI),%R8

0x42fd4d LEA	(,%R8,8),%RDX

0x42fd55 LEA	(%R9,%RDX,1),%RCX

0x42fd59 VCOMISD	(%R15,%RDX,1),%XMM0

0x42fd5f JBE	42fd30

0x42fd61 MOVQ	$-0x1,(%RCX)

0x42fd68 VMOVQ	%XMM1,%RDI

0x42fd6d VMOVQ	%XMM1,%R14

0x42fd72 LEA	0x8(%RDI,%RDX,1),%R13

0x42fd77 MOV	(%R14,%RDX,1),%RAX

0x42fd7b LEA	0x8(%RDX),%R14

0x42fd7f MOV	(%R13),%R8

0x42fd83 CMP	%R8,%RAX

0x42fd86 JGE	42fdc2

0x42fd88 NOPW	%CS:(%RAX,%RAX,1)

0x42fd93 NOPW	%CS:(%RAX,%RAX,1)

0x42fd9e XCHG	%AX,%AX

  (106) 0x42fda0 MOV	(%R10,%RAX,8),%RDI

  (106) 0x42fda4 CMPQ	$0,(%R9,%RDI,8)

  (106) 0x42fda9 JLE	42fe30

  (106) 0x42fdaf MOVQ	$-0x1,(%RCX)

  (106) 0x42fdb6 MOV	(%R13),%R8

  (106) 0x42fdba INC	%RAX

  (106) 0x42fdbd CMP	%RAX,%R8

  (106) 0x42fdc0 JG	42fda0

0x42fdc2 VMOVQ	%XMM2,%RAX

0x42fdc7 LEA	(%RAX,%R14,1),%R14

0x42fdcb MOV	(%RAX,%RDX,1),%R13

0x42fdcf MOV	(%R14),%RDI

0x42fdd2 CMP	%RDI,%R13

0x42fdd5 JGE	42fd3d

  (104) 0x42fddb MOV	(%R11,%R13,8),%R8

  (104) 0x42fddf CMPQ	$0,(%RBX,%R8,8)

  (104) 0x42fde4 JLE	42fe07

  (105) 0x42fde6 MOVQ	$-0x1,(%RCX)

  (105) 0x42fded MOV	(%R14),%RDI

  (105) 0x42fdf0 INC	%R13

  (105) 0x42fdf3 CMP	%RDI,%R13

  (105) 0x42fdf6 JGE	42fd3d

  (105) 0x42fdfc MOV	(%R11,%R13,8),%R8

  (105) 0x42fe00 CMPQ	$0,(%RBX,%R8,8)

  (105) 0x42fe05 JG	42fde6

  (104) 0x42fe07 INC	%R13

  (104) 0x42fe0a CMP	%RDI,%R13

  (104) 0x42fe0d JL	42fddb

0x42fe0f ADD	$0x8,%RSI

0x42fe13 CMP	%RSI,%R12

0x42fe16 JNE	42fd4a

  (106) 0x42fe30 INC	%RAX

  (106) 0x42fe33 CMP	%R8,%RAX

  (106) 0x42fe36 JL	42fda0

0x42fe3c JMP	42fdc2

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c: 2446 - 2477

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

2446:        if(measure_array[i]<1.) CF_marker[i]= F_PT;

[...]

2453:        if (CF_marker[i] > 0) CF_marker[i] = C_PT;

[...]

2465: 	 for (jS = S_diag_i[i]; jS < S_diag_i[i+1]; jS++)

2466:          {

2467: 	   /* j is the column number, or the local number of the point influencing i */

2468: 	   j = S_diag_j[jS];

2469: 	   if (CF_marker[j] > 0) /* j is a C-point */

2470: 	     CF_marker[i] = F_PT;

2471: 	 }

2472: 	 /* now the external part */

2473: 	 for (jS = S_offd_i[i]; jS < S_offd_i[i+1]; jS++)

2474: 	 {

2475: 	   j = S_offd_j[jS];

2476: 	   if (CF_marker_offd[j] > 0) /* j is a C-point */

2477: 	     CF_marker[i] = F_PT;

Coverage (%)	Name	Source Location	Module
►96.54+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►2.87+	GOMP_parallel	libgomp.h:982	libgomp.so.1.0.0
○	hypre_BoomerAMGCoarsenPMIS	par_coarsen.c:2492	exec
○	hypre_BoomerAMGSetup	par_amg_setup.c:612	exec
○	hypre_PCGSetup	pcg.c:234	exec
○	main	amg.c:398	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 /

Metric	Value
CQA speedup if no scalar integer	2.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	13.89
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.26
Bottlenecks	micro-operation queue,
Function	hypre_BoomerAMGCoarsenPMIS._omp_fn.7
Source	par_coarsen.c:2446-2446,par_coarsen.c:2453-2453,par_coarsen.c:2465-2465,par_coarsen.c:2473-2473
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.00
CQA cycles if no scalar integer	2.00
CQA cycles if FP arith vectorized	4.00
CQA cycles if fully vectorized	0.29
Front-end cycles	4.00
P0 cycles	3.17
P1 cycles	3.17
P2 cycles	3.17
P3 cycles	3.17
P4 cycles	3.17
P5 cycles	3.17
P6 cycles	2.50
P7 cycles	2.50
P8 cycles	2.50
P9 cycles	2.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	2.00
P15 cycles	2.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	33.00
Nb uops	32.00
Nb loads	7.00
Nb stores	2.00
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	18.00
Bytes prefetched	0.00
Bytes loaded	56.00
Bytes stored	16.00
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	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	11.25
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	6.25
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	2.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	13.89
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.26
Bottlenecks	micro-operation queue,
Function	hypre_BoomerAMGCoarsenPMIS._omp_fn.7
Source	par_coarsen.c:2446-2446,par_coarsen.c:2453-2453,par_coarsen.c:2465-2465,par_coarsen.c:2473-2473
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.00
CQA cycles if no scalar integer	2.00
CQA cycles if FP arith vectorized	4.00
CQA cycles if fully vectorized	0.29
Front-end cycles	4.00
P0 cycles	3.17
P1 cycles	3.17
P2 cycles	3.17
P3 cycles	3.17
P4 cycles	3.17
P5 cycles	3.17
P6 cycles	2.50
P7 cycles	2.50
P8 cycles	2.50
P9 cycles	2.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	2.00
P15 cycles	2.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	33.00
Nb uops	32.00
Nb loads	7.00
Nb stores	2.00
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	18.00
Bytes prefetched	0.00
Bytes loaded	56.00
Bytes stored	16.00
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	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	11.25
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	6.25
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	12.50

Path /

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

Function	hypre_BoomerAMGCoarsenPMIS._omp_fn.7
Source file and lines	par_coarsen.c:2446-2477
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2446,2453,2465-2477.

Analyzed code is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2446,2453,2465,2473.

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 4.00 to 2.00 cycles (2.00x 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 11% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 4.00 to 0.29 cycles (13.89x speedup).

Details

All SSE/AVX 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)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

No data for this section

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

LEA: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2465
VCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2446

Type of elements and instruction set

1 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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 56 bytes (7 double precision FP elements). The binary loop is storing 16 bytes (2 double precision FP elements).

General properties

nb instructions	33
nb uops	32
loop length	152
used x86 registers	11
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.00 cycles
front end	4.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	3.17	3.17	3.17	3.17	3.17	3.17	2.50	2.50	2.50	2.50	0.00	0.00	0.50	0.50	2.00	2.00
cycles	3.17	3.17	3.17	3.17	3.17	3.17	2.50	2.50	2.50	2.50	0.00	0.00	0.50	0.50	2.00	2.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions

Cycles summary

Front-end	4.00
Dispatch	3.17
Overall L1	4.00

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	11%
load	12%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	12%

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	11%
load	12%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 4.00 cycles. At this rate:

5% of peak load performance is reached (14.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (4.00 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 4.00 to 3.17 cycles (1.26x speedup).

ASM code

In the binary file, the address of the loop is: 42fd4a

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
CMPQ $0,(%RCX)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	scal (12.5%)
JLE 42fd68 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0xc8>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOVQ $0x1,(%RCX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	4	0.50	scal (6.3%)
ADD $0x8,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JE 42fe1c <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x17c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOV (%RSI),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
LEA (,%R8,8),%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
LEA (%R9,%RDX,1),%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
VCOMISD (%R15,%RDX,1),%XMM0	2	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0.50	0.50	6	0.50	scal (12.5%)
JBE 42fd30 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x90>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOVQ $-0x1,(%RCX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	4	0.50	scal (6.3%)
VMOVQ %XMM1,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMOVQ %XMM1,%R14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
LEA 0x8(%RDI,%RDX,1),%R13	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	0	0	0	0	2	0.50	N/A
MOV (%R14,%RDX,1),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
LEA 0x8(%RDX),%R14	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
MOV (%R13),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
CMP %R8,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 42fdc2 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x122>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
XCHG %AX,%AX	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
VMOVQ %XMM2,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
LEA (%RAX,%R14,1),%R14	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
MOV (%RAX,%RDX,1),%R13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOV (%R14),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
CMP %RDI,%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 42fd3d <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x9d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
ADD $0x8,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 42fd4a <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0xaa>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
JMP 42fdc2 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x122>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	1	N/A

Function	hypre_BoomerAMGCoarsenPMIS._omp_fn.7
Source file and lines	par_coarsen.c:2446-2477
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 4.00 to 2.00 cycles (2.00x speedup).

Workaround

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

Vectorization

Details

All SSE/AVX 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)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

No data for this section

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

LEA: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2465
VCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2446

Type of elements and instruction set

1 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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

General properties

nb instructions	33
nb uops	32
loop length	152
used x86 registers	11
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.00 cycles
front end	4.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	3.17	3.17	3.17	3.17	3.17	3.17	2.50	2.50	2.50	2.50	0.00	0.00	0.50	0.50	2.00	2.00
cycles	3.17	3.17	3.17	3.17	3.17	3.17	2.50	2.50	2.50	2.50	0.00	0.00	0.50	0.50	2.00	2.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

Cycles executing div or sqrt instructions

Cycles summary

Front-end	4.00
Dispatch	3.17
Overall L1	4.00

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	11%
load	12%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
other	12%

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	11%
load	12%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 4.00 cycles. At this rate:

5% of peak load performance is reached (14.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
3% of peak store performance is reached (4.00 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 42fd4a

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
CMPQ $0,(%RCX)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	1	0.25	scal (12.5%)
JLE 42fd68 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0xc8>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOVQ $0x1,(%RCX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	4	0.50	scal (6.3%)
ADD $0x8,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JE 42fe1c <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x17c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOV (%RSI),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
LEA (,%R8,8),%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
LEA (%R9,%RDX,1),%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
VCOMISD (%R15,%RDX,1),%XMM0	2	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0.50	0.50	6	0.50	scal (12.5%)
JBE 42fd30 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x90>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOVQ $-0x1,(%RCX)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	4	0.50	scal (6.3%)
VMOVQ %XMM1,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMOVQ %XMM1,%R14	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
LEA 0x8(%RDI,%RDX,1),%R13	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	0	0	0	0	2	0.50	N/A
MOV (%R14,%RDX,1),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
LEA 0x8(%RDX),%R14	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
MOV (%R13),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
CMP %R8,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 42fdc2 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x122>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
XCHG %AX,%AX	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
VMOVQ %XMM2,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
LEA (%RAX,%R14,1),%R14	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.25	N/A
MOV (%RAX,%RDX,1),%R13	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOV (%R14),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
CMP %RDI,%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 42fd3d <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x9d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
ADD $0x8,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 42fd4a <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0xaa>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A
JMP 42fdc2 <hypre_BoomerAMGCoarsenPMIS._omp_fn.7+0x122>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	1	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Complex instructions

Details

Type of elements and instruction set

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Complex instructions

Details

Type of elements and instruction set

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code