OV - exec - Loop 300

	Loop Id: 300	Module: exec	Source: par_coarsen.c:2354-2381	Coverage: 0.06%

0x43d850 LEA	0x1(%RCX),%RDX

0x43d854 CMP	%RAX,%RCX

0x43d857 MOV	%RDX,%RCX

0x43d85a JE	43d82b

0x43d85c MOV	(%R11,%RCX,8),%RDX

0x43d860 MOVSD	(%R13,%RDX,8),%XMM1

0x43d867 UCOMISD	%XMM1,%XMM0

0x43d86b JAE	43d850

0x43d86d MOV	(%R12,%RDX,8),%RSI

0x43d871 CMP	0x8(%R12,%RDX,8),%RSI

0x43d876 MOV	-0x38(%RBP),%R8

0x43d87a JL	43d8b2

0x43d87c MOV	(%R14,%RDX,8),%RSI

0x43d880 CMP	0x8(%R14,%RDX,8),%RSI

0x43d885 JGE	43d850

0x43d887 MOV	-0x38(%RBP),%RDI

0x43d88b LEA	(%RDI,%RDX,8),%RDI

0x43d88f JMP	43d8fd

  (302) 0x43d8a0 MOVQ	$0,(%R8,%RDI,8)

  (302) 0x43d8a8 INC	%RSI

  (302) 0x43d8ab CMP	0x8(%R12,%RDX,8),%RSI

  (302) 0x43d8b0 JGE	43d87c

  (302) 0x43d8b2 MOV	(%R15,%RSI,8),%RDI

  (302) 0x43d8b6 MOVSD	(%R13,%RDI,8),%XMM2

  (302) 0x43d8bd UCOMISD	%XMM2,%XMM0

  (302) 0x43d8c1 JAE	43d8a8

  (302) 0x43d8c3 UCOMISD	%XMM2,%XMM1

  (302) 0x43d8c7 JA	43d8a0

  (302) 0x43d8c9 UCOMISD	%XMM1,%XMM2

  (302) 0x43d8cd MOV	%RDX,%RDI

  (302) 0x43d8d0 JA	43d8a0

  (302) 0x43d8d2 JMP	43d8a8

  (301) 0x43d8e0 MOV	-0x58(%RBP),%R9

  (301) 0x43d8e4 LEA	(%R9,%R8,8),%R8

  (301) 0x43d8e8 MOVQ	$0,(%R8)

  (301) 0x43d8ef INC	%RSI

  (301) 0x43d8f2 CMP	0x8(%R14,%RDX,8),%RSI

  (301) 0x43d8f7 JGE	43d850

  (301) 0x43d8fd MOV	(%R10,%RSI,8),%R8

  (301) 0x43d901 LEA	(%R8,%RBX,1),%R9

  (301) 0x43d905 MOVSD	(%R13,%R9,8),%XMM2

  (301) 0x43d90c UCOMISD	%XMM2,%XMM0

  (301) 0x43d910 JAE	43d8ef

  (301) 0x43d912 UCOMISD	%XMM2,%XMM1

  (301) 0x43d916 JA	43d8e0

  (301) 0x43d918 UCOMISD	%XMM1,%XMM2

  (301) 0x43d91c MOV	%RDI,%R8

  (301) 0x43d91f JA	43d8e8

  (301) 0x43d921 JMP	43d8ef

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c: 2354 - 2381

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

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++)

2373:               {

2374:                  jj = S_offd_j[jS];

2375:                  j = num_variables+jj;

2376:                  if (measure_array[j] > 1)

2377:                  {

2378:                     if (measure_array[i] > measure_array[j])

2379:                        CF_marker_offd[jj] = 0;

2380:                     else if (measure_array[j] > measure_array[i])

2381:                        CF_marker[i] = 0;

Coverage (%)	Name	Source Location	Module
►96.78+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so
○	_INTERNALea951ae6::__kmp_launc[...]		libiomp5.so
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►2.37+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_fork_call		libiomp5.so
○	__kmpc_fork_call		libiomp5.so
○	hypre_BoomerAMGCoarsenPMIS	par_coarsen.c:2354	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		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	1.78
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.19
Bottlenecks	micro-operation queue,
Function	hypre_BoomerAMGCoarsenPMIS.extracted.15
Source	par_coarsen.c:2354-2361,par_coarsen.c:2372-2372,par_coarsen.c:2381-2381
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	2.38
CQA cycles if no scalar integer	1.33
CQA cycles if FP arith vectorized	2.38
CQA cycles if fully vectorized	0.30
Front-end cycles	2.38
P0 cycles	1.67
P1 cycles	1.67
P2 cycles	1.67
P3 cycles	1.67
P4 cycles	1.67
P5 cycles	1.67
P6 cycles	2.00
P7 cycles	2.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.50
P15 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	18.00
Nb uops	19.00
Nb loads	8.00
Nb stores	0.00
Nb stack references	1.00
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	26.95
Bytes prefetched	0.00
Bytes loaded	64.00
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	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
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	1.78
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.19
Bottlenecks	micro-operation queue,
Function	hypre_BoomerAMGCoarsenPMIS.extracted.15
Source	par_coarsen.c:2354-2361,par_coarsen.c:2372-2372,par_coarsen.c:2381-2381
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	2.38
CQA cycles if no scalar integer	1.33
CQA cycles if FP arith vectorized	2.38
CQA cycles if fully vectorized	0.30
Front-end cycles	2.38
P0 cycles	1.67
P1 cycles	1.67
P2 cycles	1.67
P3 cycles	1.67
P4 cycles	1.67
P5 cycles	1.67
P6 cycles	2.00
P7 cycles	2.00
P8 cycles	2.00
P9 cycles	2.00
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.50
P15 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	18.00
Nb uops	19.00
Nb loads	8.00
Nb stores	0.00
Nb stack references	1.00
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	26.95
Bytes prefetched	0.00
Bytes loaded	64.00
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	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
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.extracted.15
Source file and lines	par_coarsen.c:2354-2381
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:2354-2381.

Analyzed code is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2354-2361,2372,2381.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=26
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 26 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 2.38 to 1.33 cycles (1.78x 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. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 2.38 to 0.30 cycles (8.00x 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:
- 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)

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.

UCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2359

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 64 bytes (8 double precision FP elements).

General properties

nb instructions	18
nb uops	19
loop length	65
used x86 registers	11
used mmx registers	0
used xmm registers	2
used ymm registers	0
used zmm registers	0
nb stack references	1

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	2.38 cycles
front end	2.38 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	1.67	1.67	1.67	1.67	1.67	1.67	2.00	2.00	2.00	2.00	0.00	0.00	0.50	0.50	0.50	0.50
cycles	1.67	1.67	1.67	1.67	1.67	1.67	2.00	2.00	2.00	2.00	0.00	0.00	0.50	0.50	0.50	0.50

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	2.38
Dispatch	2.00
Overall L1	2.38

Vectorization ratios

INT

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)
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	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%

Vector efficiency ratios

INT

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)
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	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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 2.38 cycles. At this rate:

10% of peak load performance is reached (26.95 out of 256.00 bytes loaded 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 2.38 to 2.00 cycles (1.19x speedup).

ASM code

In the binary file, the address of the loop is: 43d85c

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
LEA 0x1(%RCX),%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
CMP %RAX,%RCX	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%)
MOV %RDX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JE 43d82b <hypre_BoomerAMGCoarsenPMIS.extracted.15+0x9b>	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 (%R11,%RCX,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%R13,%RDX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
UCOMISD %XMM1,%XMM0	2	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0.50	0.50	7	1	scal (12.5%)
JAE 43d850 <hypre_BoomerAMGCoarsenPMIS.extracted.15+0xc0>	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 (%R12,%RDX,8),%RSI	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 0x8(%R12,%RDX,8),%RSI	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%)
MOV -0x38(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	scal (12.5%)
JL 43d8b2 <hypre_BoomerAMGCoarsenPMIS.extracted.15+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
MOV (%R14,%RDX,8),%RSI	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 0x8(%R14,%RDX,8),%RSI	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	N/A
JGE 43d850 <hypre_BoomerAMGCoarsenPMIS.extracted.15+0xc0>	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 -0x38(%RBP),%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
LEA (%RDI,%RDX,8),%RDI	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
JMP 43d8fd <hypre_BoomerAMGCoarsenPMIS.extracted.15+0x16d>	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.extracted.15
Source file and lines	par_coarsen.c:2354-2381
Module	exec

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=26
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 2.38 to 1.33 cycles (1.78x 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. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 2.38 to 0.30 cycles (8.00x 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:
- 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)

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.

UCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_coarsen.c:2359

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 64 bytes (8 double precision FP elements).

General properties

nb instructions	18
nb uops	19
loop length	65
used x86 registers	11
used mmx registers	0
used xmm registers	2
used ymm registers	0
used zmm registers	0
nb stack references	1

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	2.38 cycles
front end	2.38 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	1.67	1.67	1.67	1.67	1.67	1.67	2.00	2.00	2.00	2.00	0.00	0.00	0.50	0.50	0.50	0.50
cycles	1.67	1.67	1.67	1.67	1.67	1.67	2.00	2.00	2.00	2.00	0.00	0.00	0.50	0.50	0.50	0.50

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	2.38
Dispatch	2.00
Overall L1	2.38

Vectorization ratios

INT

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)
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	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%

Vector efficiency ratios

INT

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)
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	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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 2.38 cycles. At this rate:

10% of peak load performance is reached (26.95 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 43d85c

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
LEA 0x1(%RCX),%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
CMP %RAX,%RCX	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%)
MOV %RDX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JE 43d82b <hypre_BoomerAMGCoarsenPMIS.extracted.15+0x9b>	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 (%R11,%RCX,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%R13,%RDX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
UCOMISD %XMM1,%XMM0	2	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0.50	0.50	7	1	scal (12.5%)
JAE 43d850 <hypre_BoomerAMGCoarsenPMIS.extracted.15+0xc0>	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 (%R12,%RDX,8),%RSI	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 0x8(%R12,%RDX,8),%RSI	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%)
MOV -0x38(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	scal (12.5%)
JL 43d8b2 <hypre_BoomerAMGCoarsenPMIS.extracted.15+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
MOV (%R14,%RDX,8),%RSI	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 0x8(%R14,%RDX,8),%RSI	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	N/A
JGE 43d850 <hypre_BoomerAMGCoarsenPMIS.extracted.15+0xc0>	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 -0x38(%RBP),%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
LEA (%RDI,%RDX,8),%RDI	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
JMP 43d8fd <hypre_BoomerAMGCoarsenPMIS.extracted.15+0x16d>	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