OV - exec - Loop 1058

	Loop Id: 1058	Module: exec	Source: par_multi_interp.c:1072-1125	Coverage: 0.04%

0x272ee0 MOV	-0x70(%RBP),%RDX

0x272ee4 INC	%RDX

0x272ee7 CMP	%RDX,-0xe0(%RBP)

0x272eee JLE	27316a

0x272ef4 MOV	%RDX,-0x70(%RBP)

0x272ef8 MOV	-0xd8(%RBP),%RSI

0x272eff MOV	-0x98(%RBP),%R8

0x272f06 MOV	(%RSI,%RDX,8),%RDI

0x272f0a MOV	(%R8,%RDI,8),%RDX

0x272f0e MOV	%RDI,%R13

0x272f11 NOT	%R13

0x272f14 MOV	%RDI,-0x30(%RBP)

0x272f18 CMP	%RDX,0x8(%R8,%RDI,8)

0x272f1d JLE	273070

0x272f23 MOV	-0xc8(%RBP),%RSI

0x272f2a MOV	(%RSI),%RSI

0x272f2d MOV	%RSI,-0x88(%RBP)

0x272f34 MOV	-0xf0(%RBP),%RSI

0x272f3b MOV	(%RSI),%RSI

0x272f3e MOV	%RSI,-0x48(%RBP)

0x272f42 MOV	-0x90(%RBP),%RSI

0x272f49 MOV	-0xd0(%RBP),%R8

0x272f50 MOV	(%RSI),%RDI

0x272f53 MOV	(%R8),%RSI

0x272f56 MOV	%RSI,-0xe8(%RBP)

0x272f5d MOV	-0xf8(%RBP),%RSI

0x272f64 MOV	(%RSI),%RSI

0x272f67 MOV	%RSI,-0x40(%RBP)

0x272f6b MOV	-0x80(%RBP),%RSI

0x272f6f MOV	(%RSI),%R10

0x272f72 JMP	272f99

  (1061) 0x272f80 MOV	-0x98(%RBP),%RSI

  (1061) 0x272f87 MOV	-0x30(%RBP),%R8

  (1061) 0x272f8b INC	%RDX

  (1061) 0x272f8e CMP	%RDX,0x8(%RSI,%R8,8)

  (1061) 0x272f93 JLE	273070

  (1061) 0x272f99 MOV	-0x110(%RBP),%RSI

  (1061) 0x272fa0 MOV	(%RBX),%R9

  (1061) 0x272fa3 MOV	-0x108(%RBP),%R8

  (1061) 0x272faa MOV	(%RSI,%RDX,8),%R11

  (1061) 0x272fae DEC	%R9

  (1061) 0x272fb1 CMP	%R9,(%R8,%R11,8)

  (1061) 0x272fb5 JNE	272f80

  (1061) 0x272fb7 MOV	-0x48(%RBP),%RSI

  (1061) 0x272fbb MOV	0x8(%RSI,%R11,8),%R9

  (1061) 0x272fc0 TEST	%R9,%R9

  (1061) 0x272fc3 JLE	273010

  (1061) 0x272fc5 MOV	-0x88(%RBP),%RSI

  (1061) 0x272fcc MOV	(%RSI,%R11,8),%R8

  (1061) 0x272fd0 ADD	%R8,%R9

  (1061) 0x272fd3 JMP	272fe8

    (1063) 0x272fe0 INC	%R8

    (1063) 0x272fe3 CMP	%R9,%R8

    (1063) 0x272fe6 JGE	273010

    (1063) 0x272fe8 MOV	(%RBX),%R12

    (1063) 0x272feb MOV	-0x8(%RDI,%R12,8),%RSI

    (1063) 0x272ff0 MOV	(%RSI,%R8,8),%RSI

    (1063) 0x272ff4 CMP	%R13,(%R14,%RSI,8)

    (1063) 0x272ff8 JE	272fe0

    (1063) 0x272ffa MOV	(%RDI,%R12,8),%R12

    (1063) 0x272ffe MOV	%RSI,(%R12,%RAX,8)

    (1063) 0x273002 INC	%RAX

    (1063) 0x273005 MOV	%R13,(%R14,%RSI,8)

    (1063) 0x273009 JMP	272fe0

  (1061) 0x273010 MOV	-0x40(%RBP),%RSI

  (1061) 0x273014 MOV	0x8(%RSI,%R11,8),%R9

  (1061) 0x273019 TEST	%R9,%R9

  (1061) 0x27301c JLE	272f80

  (1061) 0x273022 MOV	-0xe8(%RBP),%RSI

  (1061) 0x273029 MOV	(%RSI,%R11,8),%R8

  (1061) 0x27302d ADD	%R8,%R9

  (1061) 0x273030 JMP	27304c

    (1062) 0x273040 INC	%R8

    (1062) 0x273043 CMP	%R9,%R8

    (1062) 0x273046 JGE	272f80

    (1062) 0x27304c MOV	(%RBX),%R11

    (1062) 0x27304f MOV	-0x8(%R10,%R11,8),%RSI

    (1062) 0x273054 MOV	(%RSI,%R8,8),%RSI

    (1062) 0x273058 CMP	%R13,(%R15,%RSI,8)

    (1062) 0x27305c JE	273040

    (1062) 0x27305e MOV	(%R10,%R11,8),%R11

    (1062) 0x273062 MOV	%RSI,(%R11,%RCX,8)

    (1062) 0x273066 INC	%RCX

    (1062) 0x273069 MOV	%R13,(%R15,%RSI,8)

    (1062) 0x27306d JMP	273040

0x273070 MOV	-0x78(%RBP),%RSI

0x273074 MOV	-0x30(%RBP),%RDI

0x273078 MOV	(%RSI,%RDI,8),%RDX

0x27307c MOV	0x8(%RSI,%RDI,8),%RDI

0x273081 CMP	%RDI,%RDX

0x273084 JGE	272ee0

0x27308a MOV	-0x150(%RBP),%RSI

0x273091 MOV	(%RSI),%RSI

0x273094 MOV	%RSI,-0x40(%RBP)

0x273098 MOV	-0x158(%RBP),%RSI

0x27309f MOV	(%RSI),%RSI

0x2730a2 MOV	%RSI,-0x48(%RBP)

0x2730a6 MOV	-0x160(%RBP),%RSI

0x2730ad MOV	(%RSI),%R9

0x2730b0 JMP	2730d9

  (1059) 0x2730c0 MOV	-0x78(%RBP),%RSI

  (1059) 0x2730c4 MOV	-0x30(%RBP),%RDI

  (1059) 0x2730c8 MOV	0x8(%RSI,%RDI,8),%RDI

  (1059) 0x2730cd INC	%RDX

  (1059) 0x2730d0 CMP	%RDI,%RDX

  (1059) 0x2730d3 JGE	272ee0

  (1059) 0x2730d9 MOV	-0x100(%RBP),%RSI

  (1059) 0x2730e0 MOV	(%RBX),%R8

  (1059) 0x2730e3 MOV	-0x50(%RBP),%R10

  (1059) 0x2730e7 MOV	(%RSI,%RDX,8),%RSI

  (1059) 0x2730eb DEC	%R8

  (1059) 0x2730ee CMP	%R8,(%R10,%RSI,8)

  (1059) 0x2730f2 JNE	2730cd

  (1059) 0x2730f4 MOV	-0x48(%RBP),%R8

  (1059) 0x2730f8 MOV	0x8(%R8,%RSI,8),%R10

  (1059) 0x2730fd TEST	%R10,%R10

  (1059) 0x273100 JLE	2730cd

  (1059) 0x273102 MOV	-0x40(%RBP),%RDI

  (1059) 0x273106 MOV	-0x80(%RBP),%R11

  (1059) 0x27310a MOV	-0x90(%RBP),%R8

  (1059) 0x273111 MOV	(%RDI,%RSI,8),%RDI

  (1059) 0x273115 MOV	(%R11),%R11

  (1059) 0x273118 MOV	(%R8),%R8

  (1059) 0x27311b ADD	%RDI,%R10

  (1059) 0x27311e JMP	273128

    (1060) 0x273120 INC	%RDI

    (1060) 0x273123 CMP	%R10,%RDI

    (1060) 0x273126 JGE	2730c0

    (1060) 0x273128 MOV	(%RBX),%R12

    (1060) 0x27312b MOV	(%R9,%R12,8),%RSI

    (1060) 0x27312f MOV	(%RSI,%RDI,8),%RSI

    (1060) 0x273133 TEST	%RSI,%RSI

    (1060) 0x273136 JS	273150

    (1060) 0x273138 CMP	%R13,(%R15,%RSI,8)

    (1060) 0x27313c JE	273120

    (1060) 0x27313e MOV	(%R11,%R12,8),%R12

    (1060) 0x273142 MOV	%RSI,(%R12,%RCX,8)

    (1060) 0x273146 INC	%RCX

    (1060) 0x273149 MOV	%R13,(%R15,%RSI,8)

    (1060) 0x27314d JMP	273120

    (1060) 0x273150 NOT	%RSI

    (1060) 0x273153 CMP	%R13,(%R14,%RSI,8)

    (1060) 0x273157 JE	273120

    (1060) 0x273159 MOV	(%R8,%R12,8),%R12

    (1060) 0x27315d MOV	%RSI,(%R12,%RAX,8)

    (1060) 0x273161 INC	%RAX

    (1060) 0x273164 MOV	%R13,(%R14,%RSI,8)

    (1060) 0x273168 JMP	273120

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_multi_interp.c: 1072 - 1125

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

1072:           for (i=thread_start; i < thread_stop; i++)

1073:           {

1074:              i1 = pass_array[i];

1075:              for (j=S_diag_i[i1]; j < S_diag_i[i1+1]; j++)

1076:              {

1077:                 j1 = S_diag_j[j];

1078:                 if (assigned[j1] == pass-1)

1079:                 {

1080:                    j_start = P_diag_start[j1];

1081:                    j_end = j_start+P_diag_i[j1+1];

1082:                    for (k=j_start; k < j_end; k++)

1083:                    {

1084:                       k1 = P_diag_pass[pass-1][k];

1085:                       if (P_marker[k1] != -i1-1)

1086:                       {

1087:                           P_diag_pass[pass][cnt_nz++] = k1;

1088:                           P_marker[k1] = -i1-1;

1089:                       }

1090:                    }

1091:                    j_start = P_offd_start[j1];

1092:                    j_end = j_start+P_offd_i[j1+1];

1093:                    for (k=j_start; k < j_end; k++)

1094:                    {

1095:                       k1 = P_offd_pass[pass-1][k];

1096:                       if (P_marker_offd[k1] != -i1-1)

1097:                       {

1098:                          P_offd_pass[pass][cnt_nz_offd++] = k1;

1099:                          P_marker_offd[k1] = -i1-1;

1100:                       }

1101:                    }

1102:                 }

1103:              }

1104:              for (j=S_offd_i[i1]; j < S_offd_i[i1+1]; j++)

1105:              {

1106:                 j1 = S_offd_j[j];

1107:                 if (assigned_offd[j1] == pass-1)

1108:                 {

1109:                    j_start = Pext_start[j1];

1110:                    j_end = j_start+Pext_i[j1+1];

1111:                    for (k=j_start; k < j_end; k++)

1112:                    {

1113:                       k1 = Pext_pass[pass][k];

1114:                       if (k1 < 0)

1115:                       {

1116:                          if (P_marker[-k1-1] != -i1-1)

1117:                          {

1118:                             P_diag_pass[pass][cnt_nz++] = -k1-1;

1119:                             P_marker[-k1-1] = -i1-1;

1120:                          }

1121:                       }

1122:                       else if (P_marker_offd[k1] != -i1-1)

1123:                       {

1124:                          P_offd_pass[pass][cnt_nz_offd++] = k1;

1125:                          P_marker_offd[k1] = -i1-1;

Coverage (%)	Name	Source Location	Module
►96.13+	__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
►2.64+	__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:891	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	1.00
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	1.61
Bottlenecks	P6, P7, P8, P9,
Function	hypre_BoomerAMGBuildMultipass.omp_outlined.10
Source	par_multi_interp.c:1072-1075,par_multi_interp.c:1099-1099,par_multi_interp.c:1104-1104
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	9.25
CQA cycles if no scalar integer	9.25
CQA cycles if FP arith vectorized	9.25
CQA cycles if fully vectorized	1.16
Front-end cycles	5.75
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	9.25
P7 cycles	9.25
P8 cycles	9.25
P9 cycles	9.25
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
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	46.00
Nb uops	46.00
Nb loads	29.00
Nb stores	8.00
Nb stack references	19.00
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	32.00
Bytes prefetched	0.00
Bytes loaded	232.00
Bytes stored	64.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
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	1.00
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	1.61
Bottlenecks	P6, P7, P8, P9,
Function	hypre_BoomerAMGBuildMultipass.omp_outlined.10
Source	par_multi_interp.c:1072-1075,par_multi_interp.c:1099-1099,par_multi_interp.c:1104-1104
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	9.25
CQA cycles if no scalar integer	9.25
CQA cycles if FP arith vectorized	9.25
CQA cycles if fully vectorized	1.16
Front-end cycles	5.75
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	9.25
P7 cycles	9.25
P8 cycles	9.25
P9 cycles	9.25
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
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	46.00
Nb uops	46.00
Nb loads	29.00
Nb stores	8.00
Nb stack references	19.00
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	32.00
Bytes prefetched	0.00
Bytes loaded	232.00
Bytes stored	64.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
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_BoomerAMGBuildMultipass.omp_outlined.10
Source file and lines	par_multi_interp.c:1072-1125
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:1072-1125.

Analyzed code is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/parcsr_ls/par_multi_interp.c:1072-1075,1099,1104.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. If you really need to analyze all of the 77 paths individually, rerun with max-paths=77
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 77 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

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 9.25 to 1.16 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
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)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 9.25 to 5.75 cycles (1.61x speedup).

Workaround

Read less array elements
Write 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

No instructions are processing arithmetic or math operations on FP elements. This loop is probably writing/copying data or processing integer elements.

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 232 bytes. The binary loop is storing 64 bytes.

General properties

nb instructions	46
nb uops	46
loop length	214
used x86 registers	8
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	19

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	5.75 cycles
front end	5.75 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	9.25	9.25	9.25	9.25	0.00	0.00	0.00	0.00	0.00	0.00
cycles	1.67	1.67	1.67	1.67	1.67	1.67	9.25	9.25	9.25	9.25	0.00	0.00	0.00	0.00	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	5.75
Dispatch	9.25
Overall L1	9.25

Vectorization ratios

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

all	12%
load	12%
store	12%
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 9.25 cycles. At this rate:

9% of peak load performance is reached (25.08 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
5% of peak store performance is reached (6.92 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 272ef4

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	Latency	Recip. throughput	Vectorization
MOV -0x70(%RBP),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
INC %RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP %RDX,-0xe0(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	1	0.25	scal (12.5%)
JLE 27316a <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0xaba>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV %RDX,-0x70(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xd8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0x98(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI,%RDX,8),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%R8,%RDI,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RDI,%R13	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (12.5%)
NOT %R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
MOV %RDI,-0x30(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
CMP %RDX,0x8(%R8,%RDI,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	1	0.25	scal (12.5%)
JLE 273070 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x9c0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV -0xc8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x88(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xf0(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x48(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x90(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0xd0(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%R8),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0xe8(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xf8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x80(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (12.5%)
JMP 272f99 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x8e9>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
MOV -0x78(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0x30(%RBP),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI,%RDI,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV 0x8(%RSI,%RDI,8),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
CMP %RDI,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
JGE 272ee0 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x830>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV -0x150(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x158(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x48(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x160(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (12.5%)
JMP 2730d9 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0xa29>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A

Function	hypre_BoomerAMGBuildMultipass.omp_outlined.10
Source file and lines	par_multi_interp.c:1072-1125
Module	exec

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. If you really need to analyze all of the 77 paths individually, rerun with max-paths=77
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

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 9.25 to 1.16 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
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)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 9.25 to 5.75 cycles (1.61x speedup).

Workaround

Read less array elements
Write 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

No instructions are processing arithmetic or math operations on FP elements. This loop is probably writing/copying data or processing integer elements.

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 232 bytes. The binary loop is storing 64 bytes.

General properties

nb instructions	46
nb uops	46
loop length	214
used x86 registers	8
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	19

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	5.75 cycles
front end	5.75 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	9.25	9.25	9.25	9.25	0.00	0.00	0.00	0.00	0.00	0.00
cycles	1.67	1.67	1.67	1.67	1.67	1.67	9.25	9.25	9.25	9.25	0.00	0.00	0.00	0.00	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	5.75
Dispatch	9.25
Overall L1	9.25

Vectorization ratios

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

all	12%
load	12%
store	12%
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 9.25 cycles. At this rate:

9% of peak load performance is reached (25.08 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
5% of peak store performance is reached (6.92 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 272ef4

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	Latency	Recip. throughput	Vectorization
MOV -0x70(%RBP),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
INC %RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP %RDX,-0xe0(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	1	0.25	scal (12.5%)
JLE 27316a <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0xaba>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV %RDX,-0x70(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xd8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0x98(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI,%RDX,8),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%R8,%RDI,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RDI,%R13	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (12.5%)
NOT %R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
MOV %RDI,-0x30(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
CMP %RDX,0x8(%R8,%RDI,8)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	1	0.25	scal (12.5%)
JLE 273070 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x9c0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV -0xc8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x88(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xf0(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x48(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x90(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0xd0(%RBP),%R8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%R8),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0xe8(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0xf8(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x80(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (12.5%)
JMP 272f99 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x8e9>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
MOV -0x78(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV -0x30(%RBP),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI,%RDI,8),%RDX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV 0x8(%RSI,%RDI,8),%RDI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
CMP %RDI,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
JGE 272ee0 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0x830>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV -0x150(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x158(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %RSI,-0x48(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV -0x160(%RBP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV (%RSI),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (12.5%)
JMP 2730d9 <hypre_BoomerAMGBuildMultipass.omp_outlined.10+0xa29>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A

Report Configuration

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code