OV - exec - Loop 1059

	Loop Id: 1059	Module: exec	Source: par_multi_interp.c:1811-1837	Coverage: 0.42%

0x26e8d0 INC	%RCX

0x26e8d3 CMP	%RCX,-0x30(%RBP)

0x26e8d7 JE	26e9c4

0x26e8dd MOV	-0xd0(%RBP),%RAX

0x26e8e4 MOV	-0xb0(%RBP),%R9

0x26e8eb MOV	(%RAX,%RCX,8),%R10

0x26e8ef CMP	%R8,(%R9,%R10,8)

0x26e8f3 JNE	26e9a0

0x26e8f9 MOV	-0xa8(%RBP),%RAX

0x26e900 MOV	(%RAX,%R10,8),%R9

0x26e904 MOV	0x8(%RAX,%R10,8),%R11

0x26e909 CMP	%R11,%R9

0x26e90c JGE	26e93d

0x26e90e XCHG	%AX,%AX

  (1061) 0x26e910 MOV	(%R14,%R9,8),%RAX

  (1061) 0x26e914 VMOVSD	(%RSI,%RCX,8),%XMM2

  (1061) 0x26e919 VMULSD	(%RDX,%R9,8),%XMM2,%XMM2

  (1061) 0x26e91f INC	%R9

  (1061) 0x26e922 MOV	(%R15,%RAX,8),%RAX

  (1061) 0x26e926 VADDSD	(%RDX,%RAX,8),%XMM2,%XMM5

  (1061) 0x26e92b VADDSD	%XMM2,%XMM4,%XMM4

  (1061) 0x26e92f VADDSD	%XMM2,%XMM3,%XMM3

  (1061) 0x26e933 VMOVSD	%XMM5,(%RDX,%RAX,8)

  (1061) 0x26e938 CMP	%R9,%R11

  (1061) 0x26e93b JNE	26e910

0x26e93d MOV	-0xa0(%RBP),%RAX

0x26e944 MOV	(%RAX,%R10,8),%R9

0x26e948 MOV	0x8(%RAX,%R10,8),%R10

0x26e94d CMP	%R10,%R9

0x26e950 JGE	26e8d0

0x26e956 NOPW	%CS:(%RAX,%RAX,1)

  (1060) 0x26e960 MOV	(%RBX,%R9,8),%R11

  (1060) 0x26e964 VMOVSD	(%RSI,%RCX,8),%XMM2

  (1060) 0x26e969 VMULSD	(%RDI,%R9,8),%XMM2,%XMM2

  (1060) 0x26e96f INC	%R9

  (1060) 0x26e972 MOV	(%R12,%R11,8),%R11

  (1060) 0x26e976 VADDSD	(%RDI,%R11,8),%XMM2,%XMM5

  (1060) 0x26e97c VADDSD	%XMM2,%XMM4,%XMM4

  (1060) 0x26e980 VADDSD	%XMM2,%XMM3,%XMM3

  (1060) 0x26e984 VMOVSD	%XMM5,(%RDI,%R11,8)

  (1060) 0x26e98a CMP	%R9,%R10

  (1060) 0x26e98d JNE	26e960

0x26e98f JMP	26e8d0

0x26e9a0 MOV	-0x38(%RBP),%RAX

0x26e9a4 CMPQ	$-0x3,(%RAX,%R10,8)

0x26e9a9 JE	26e8d0

0x26e9af VADDSD	(%RSI,%RCX,8),%XMM3,%XMM3

0x26e9b4 JMP	26e8d0

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_multi_interp.c: 1811 - 1837

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

1811:                 for (j=A_diag_i[i1]+1; j < A_diag_i[i1+1]; j++)

1812:                 {

1813:                j1 = A_diag_j[j];

1814:                if (tmp_marker[j1] == i1)

1815:                {

1816:                   for (k=P_diag_i[j1]; k < P_diag_i[j1+1]; k++)

1817:                   {

1818:                  k1 = P_diag_j[k];

1819:                      alfa = A_diag_data[j]*P_diag_data[k];

1820:                      P_diag_data[tmp_array[k1]] += alfa;

1821:                      sum_C += alfa;

1822:                      sum_N += alfa;

1823:                       }

1824:                   for (k=P_offd_i[j1]; k < P_offd_i[j1+1]; k++)

1825:                   {

1826:                  k1 = P_offd_j[k];

1827:                      alfa = A_diag_data[j]*P_offd_data[k];

1828:                      P_offd_data[tmp_array_offd[k1]] += alfa;

1829:                      sum_C += alfa;

1830:                      sum_N += alfa;

1831:                       }

1832:                    }

1833:                    else

1834:                    {

1835:                       if (CF_marker[j1] != -3 &&

1836:                 (num_functions == 1 || dof_func[i1] == dof_func[j1]))

1837:                  sum_N += A_diag_data[j];

Coverage (%)	Name	Source Location	Module
►96.34+	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_launch_thread		libomp.so
○	__kmp_launch_worker(void*)		libomp.so
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►2.48+	__kmp_invoke_microtask		libomp.so
○	__kmp_invoke_task_func		libomp.so
○	__kmp_fork_call		libomp.so
○	__kmpc_fork_call		libomp.so
○	hypre_BoomerAMGBuildMultipass	par_multi_interp.c:1737	exec
○	hypre_BoomerAMGSetup	par_amg_setup.c:737	exec
○	hypre_PCGSetup	pcg.c:234	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	2.10
CQA speedup if FP arith vectorized	2.67
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.17
Bottlenecks	P6, P7, P8, P9,
Function	hypre_BoomerAMGBuildMultipass.omp_outlined.20
Source	par_multi_interp.c:1811-1816,par_multi_interp.c:1824-1824,par_multi_interp.c:1835-1837
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	3.50
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	1.31
CQA cycles if fully vectorized	0.44
Front-end cycles	3.00
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.00
P3 cycles	2.33
P4 cycles	2.33
P5 cycles	2.33
P6 cycles	3.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	3.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.00
P15 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	26.00
Nb uops	24.00
Nb loads	14.00
Nb stores	0.00
Nb stack references	6.00
FLOP/cycle	0.29
Nb FLOP add-sub	1.00
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	32.00
Bytes prefetched	0.00
Bytes loaded	112.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
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	2.10
CQA speedup if FP arith vectorized	2.67
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.17
Bottlenecks	P6, P7, P8, P9,
Function	hypre_BoomerAMGBuildMultipass.omp_outlined.20
Source	par_multi_interp.c:1811-1816,par_multi_interp.c:1824-1824,par_multi_interp.c:1835-1837
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	3.50
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	1.31
CQA cycles if fully vectorized	0.44
Front-end cycles	3.00
P0 cycles	2.00
P1 cycles	2.00
P2 cycles	2.00
P3 cycles	2.33
P4 cycles	2.33
P5 cycles	2.33
P6 cycles	3.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	3.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.50
P13 cycles	0.50
P14 cycles	0.00
P15 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	26.00
Nb uops	24.00
Nb loads	14.00
Nb stores	0.00
Nb stack references	6.00
FLOP/cycle	0.29
Nb FLOP add-sub	1.00
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	32.00
Bytes prefetched	0.00
Bytes loaded	112.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
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_BoomerAMGBuildMultipass.omp_outlined.20
Source file and lines	par_multi_interp.c:1811-1837
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_multi_interp.c:1811-1837.

Analyzed code is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_multi_interp.c:1811-1816,1824,1835-1837.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=6
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 6 execution paths.

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

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

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

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 3.50 to 1.67 cycles (2.10x 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 3.50 to 0.44 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

Performance is limited by reading data from caches/RAM (load units are a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 3.50 to 3.00 cycles (1.17x speedup).

Workaround

Read less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop
- use the 'restrict' C99 keyword

No data for this section

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 is composed of 1 FP arithmetical operations:

1: addition or subtraction

The binary loop is loading 112 bytes (14 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.01 FP operations per loaded or stored byte.

General properties

nb instructions	26
nb uops	24
loop length	129
used x86 registers	8
used mmx registers	0
used xmm registers	1
used ymm registers	0
used zmm registers	0
nb stack references	6

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.00	2.00	2.00	2.33	2.33	2.33	3.50	3.50	3.50	3.50	0.00	0.00	0.50	0.50	0.00	0.00
cycles	2.00	2.00	2.00	2.33	2.33	2.33	3.50	3.50	3.50	3.50	0.00	0.00	0.50	0.50	0.00	0.00

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	3.00
Dispatch	3.50
Overall L1	3.50

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	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

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%

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	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
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 3.50 cycles. At this rate:

12% of peak load performance is reached (32.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 26e8dd

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	Latency	Recip. throughput	Vectorization
INC %RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP %RCX,-0x30(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (12.5%)
JE 26e9c4 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x794>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV -0xd0(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV -0xb0(%RBP),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%RCX,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMP %R8,(%R9,%R10,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (12.5%)
JNE 26e9a0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x770>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV -0xa8(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%R10,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RAX,%R10,8),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
CMP %R11,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 26e93d <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x70d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
XCHG %AX,%AX	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOV -0xa0(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%R10,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RAX,%R10,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMP %R10,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	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	0.09	N/A
JMP 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A
MOV -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMPQ $-0x3,(%RAX,%R10,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	N/A
JE 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
VADDSD (%RSI,%RCX,8),%XMM3,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50	scal (12.5%)
JMP 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A

Function	hypre_BoomerAMGBuildMultipass.omp_outlined.20
Source file and lines	par_multi_interp.c:1811-1837
Module	exec

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=6
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 3.50 to 1.67 cycles (2.10x 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 3.50 to 0.44 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

Workaround

Read less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop
- use the 'restrict' C99 keyword

No data for this section

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 is composed of 1 FP arithmetical operations:

1: addition or subtraction

The binary loop is loading 112 bytes (14 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.01 FP operations per loaded or stored byte.

General properties

nb instructions	26
nb uops	24
loop length	129
used x86 registers	8
used mmx registers	0
used xmm registers	1
used ymm registers	0
used zmm registers	0
nb stack references	6

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	3.00 cycles
front end	3.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.00	2.00	2.00	2.33	2.33	2.33	3.50	3.50	3.50	3.50	0.00	0.00	0.50	0.50	0.00	0.00
cycles	2.00	2.00	2.00	2.33	2.33	2.33	3.50	3.50	3.50	3.50	0.00	0.00	0.50	0.50	0.00	0.00

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	3.00
Dispatch	3.50
Overall L1	3.50

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	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

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%

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	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
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 3.50 cycles. At this rate:

12% of peak load performance is reached (32.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 26e8dd

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P12	P13	Latency	Recip. throughput	Vectorization
INC %RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP %RCX,-0x30(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (12.5%)
JE 26e9c4 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x794>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV -0xd0(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV -0xb0(%RBP),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%RCX,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMP %R8,(%R9,%R10,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (12.5%)
JNE 26e9a0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x770>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV -0xa8(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%R10,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RAX,%R10,8),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
CMP %R11,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 26e93d <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x70d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
XCHG %AX,%AX	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOV -0xa0(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV (%RAX,%R10,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RAX,%R10,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMP %R10,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JGE 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	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	0.09	N/A
JMP 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A
MOV -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
CMPQ $-0x3,(%RAX,%R10,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	1	0.25	N/A
JE 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
VADDSD (%RSI,%RCX,8),%XMM3,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	2	0.50	scal (12.5%)
JMP 26e8d0 <hypre_BoomerAMGBuildMultipass.omp_outlined.20+0x6a0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	1	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Type of elements and instruction set

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

Arithmetic intensity

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

Workaround

Type of elements and instruction set

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code