OV - exec - Loop 2655

	Loop Id: 2655	Module: exec	Source: ams.c:3532-3534	Coverage: 0.07%

0x4d8860 LDR	X14, [X28, X1,LSL #3]    [1]

0x4d8864 LDR	D2, [X21, X14,LSL #3]    [3]

0x4d8868 FCMPE	D2, #0

0x4d886c B.GE	4d887c

0x4d8870 LDR	D3, [X19, X1,LSL #3]    [2]

0x4d8874 FNEG	D4, D3

0x4d8878 STR	D4, [X19, X1,LSL #3]    [2]

0x4d887c ADD	X2, X1, #1

0x4d8880 CMP	X25, X2

0x4d8884 B.EQ	4d8960

0x4d8888 LDR	X17, [X28, X2,LSL #3]    [1]

0x4d888c LDR	D30, [X21, X17,LSL #3]    [3]

0x4d8890 FCMPE	D30, #0

0x4d8894 B.GE	4d88a4

0x4d8898 LDR	D6, [X19, X2,LSL #3]    [2]

0x4d889c FNEG	D28, D6

0x4d88a0 STR	D28, [X19, X2,LSL #3]    [2]

0x4d88a4 ADD	X30, X2, #1

0x4d88a8 LDR	X7, [X28, X30,LSL #3]    [1]

0x4d88ac LDR	D29, [X21, X7,LSL #3]    [3]

0x4d88b0 FCMPE	D29, #0

0x4d88b4 B.GE	4d88c4

0x4d88b8 LDR	D0, [X19, X30,LSL #3]    [2]

0x4d88bc FNEG	D1, D0

0x4d88c0 STR	D1, [X19, X30,LSL #3]    [2]

0x4d88c4 ADD	X8, X2, #2

0x4d88c8 LDR	X9, [X28, X8,LSL #3]    [1]

0x4d88cc LDR	D2, [X21, X9,LSL #3]    [3]

0x4d88d0 FCMPE	D2, #0

0x4d88d4 B.GE	4d88e4

0x4d88d8 LDR	D3, [X19, X8,LSL #3]    [2]

0x4d88dc FNEG	D4, D3

0x4d88e0 STR	D4, [X19, X8,LSL #3]    [2]

0x4d88e4 ADD	X16, X2, #3

0x4d88e8 LDR	X15, [X28, X16,LSL #3]    [1]

0x4d88ec LDR	D5, [X21, X15,LSL #3]    [3]

0x4d88f0 FCMPE	D5, #0

0x4d88f4 B.GE	4d8904

0x4d88f8 LDR	D7, [X19, X16,LSL #3]    [2]

0x4d88fc FNEG	D16, D7

0x4d8900 STR	D16, [X19, X16,LSL #3]    [2]

0x4d8904 ADD	X14, X2, #4

0x4d8908 LDR	X3, [X28, X14,LSL #3]    [1]

0x4d890c LDR	D17, [X21, X3,LSL #3]    [3]

0x4d8910 FCMPE	D17, #0

0x4d8914 B.GE	4d8924

0x4d8918 LDR	D18, [X19, X14,LSL #3]    [2]

0x4d891c FNEG	D19, D18

0x4d8920 STR	D19, [X19, X14,LSL #3]    [2]

0x4d8924 ADD	X1, X2, #5

0x4d8928 LDR	X18, [X28, X1,LSL #3]    [1]

0x4d892c LDR	D20, [X21, X18,LSL #3]    [3]

0x4d8930 FCMPE	D20, #0

0x4d8934 B.GE	4d8944

0x4d8938 LDR	D21, [X19, X1,LSL #3]    [2]

0x4d893c FNEG	D22, D21

0x4d8940 STR	D22, [X19, X1,LSL #3]    [2]

0x4d8944 ADD	X17, X2, #6

0x4d8948 LDR	X30, [X28, X17,LSL #3]    [1]

0x4d894c LDR	D31, [X21, X30,LSL #3]    [3]

0x4d8950 FCMPE	D31, #0

0x4d8954 B.MI	4d91e0

0x4d8958 ADD	X1, X2, #7

0x4d895c B	4d8860

0x4d91e0 LDR	D23, [X19, X17,LSL #3]    [2]

0x4d91e4 ADD	X1, X2, #7

0x4d91e8 FNEG	D24, D23

0x4d91ec STR	D24, [X19, X17,LSL #3]    [2]

0x4d91f0 B	4d8860

/home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c: 3532 - 3534

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

3532:       for (i = ns; i < ne; i++)

3533:          if (A_diag_data[A_diag_I[i]] < 0)

3534:             l1_norm[i] = -l1_norm[i];

Coverage (%)	Name	Source Location	Module
►98.11+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	thread_start		libc.so.6
►1.89+	GOMP_parallel	libgomp.h:980	libgomp.so.1.0.0
○	hypre_ParCSRComputeL1NormsThre[...]	ams.c:3546	exec
○	hypre_BoomerAMGSetup	par_amg_setup.c:1409	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	1.33
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	2.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.24
Bottlenecks	P12, P13, P14,
Function	hypre_ParCSRComputeL1NormsThreads._omp_fn.0
Source	ams.c:3532-3534
Source loop unroll info	multi-versionned
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	10.67
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	10.67
CQA cycles if fully vectorized	5.33
Front-end cycles	8.63
P0 cycles	5.50
P1 cycles	5.50
P2 cycles	1.75
P3 cycles	1.75
P4 cycles	1.67
P5 cycles	1.67
P6 cycles	1.58
P7 cycles	1.58
P8 cycles	8.00
P9 cycles	8.00
P10 cycles	4.00
P11 cycles	4.00
P12 cycles	10.67
P13 cycles	10.67
P14 cycles	10.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	69.00
Nb uops	69.00
Nb loads	NA
Nb stores	8.00
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	24.00
Bytes prefetched	0.00
Bytes loaded	192.00
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	50.00
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
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	50.00

Metric	Value
CQA speedup if no scalar integer	1.33
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	2.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.24
Bottlenecks	P12, P13, P14,
Function	hypre_ParCSRComputeL1NormsThreads._omp_fn.0
Source	ams.c:3532-3534
Source loop unroll info	multi-versionned
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	10.67
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	10.67
CQA cycles if fully vectorized	5.33
Front-end cycles	8.63
P0 cycles	5.50
P1 cycles	5.50
P2 cycles	1.75
P3 cycles	1.75
P4 cycles	1.67
P5 cycles	1.67
P6 cycles	1.58
P7 cycles	1.58
P8 cycles	8.00
P9 cycles	8.00
P10 cycles	4.00
P11 cycles	4.00
P12 cycles	10.67
P13 cycles	10.67
P14 cycles	10.67
P15 cycles	0.00
P16 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	69.00
Nb uops	69.00
Nb loads	NA
Nb stores	8.00
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	24.00
Bytes prefetched	0.00
Bytes loaded	192.00
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	50.00
Vector-efficiency ratio load	50.00
Vector-efficiency ratio store	50.00
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	50.00

Path /

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

Function	hypre_ParCSRComputeL1NormsThreads._omp_fn.0
Source file and lines	ams.c:3532-3534
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c:3532-3534.

The related source loop is multi-versionned.
Warnings:

Ignoring paths for analysis
Too many paths. If you really need to analyze all of the 256 paths individually, rerun with max-paths=256
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 256 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 10.67 to 8.00 cycles (1.33x 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. 2 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 10.67 to 5.33 cycles (2.00x speedup).

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: 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)

No data for this section

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 192 bytes. The binary loop is storing 64 bytes.

General properties

nb instructions	69
loop length	276
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

8.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	5.50	5.50	1.75	1.75	1.67	1.67	1.58	1.58	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00
cycles	5.50	5.50	1.75	1.75	1.67	1.67	1.58	1.58	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions

Cycles summary

Front-end	8.63
Overall L1	10.67

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	NA (no load vectorizable/vectorized instructions)
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.67 cycles. At this rate:

18% of peak load performance is reached (18.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
12% of peak store performance is reached (6.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4d8860

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X14, [X28, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D2, [X21, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D2, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d887c <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1fc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D3, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D4, D3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D4, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X2, X1, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
CMP X25, X2	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.EQ 4d8960 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR X17, [X28, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D30, [X21, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D30, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88a4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x224>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D6, [X19, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D28, D6	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D28, [X19, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X30, X2, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X7, [X28, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D29, [X21, X7,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D29, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88c4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x244>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D0, [X19, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D1, D0	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D1, [X19, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X8, X2, #2	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X9, [X28, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D2, [X21, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D2, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88e4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x264>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D3, [X19, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D4, D3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D4, [X19, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X16, X2, #3	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X15, [X28, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D5, [X21, X15,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D5, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8904 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x284>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D7, [X19, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D16, D7	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D16, [X19, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X14, X2, #4	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X3, [X28, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D17, [X21, X3,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D17, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8924 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2a4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D18, [X19, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D19, D18	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D19, [X19, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X1, X2, #5	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X18, [X28, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D20, [X21, X18,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D20, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8944 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2c4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D21, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D22, D21	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D22, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X17, X2, #6	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X30, [X28, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D31, [X21, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D31, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.MI 4d91e0 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0xb60>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
ADD X1, X2, #7	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
B 4d8860 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D23, [X19, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
ADD X1, X2, #7	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
FNEG D24, D23	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D24, [X19, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
B 4d8860 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Function	hypre_ParCSRComputeL1NormsThreads._omp_fn.0
Source file and lines	ams.c:3532-3534
Module	exec

The loop is defined in /home/hbollore/qaas/qaas-runs/174-161-6712/intel/AMG/build/AMG/AMG/parcsr_ls/ams.c:3532-3534.

The related source loop is multi-versionned.
Warnings:

Ignoring paths for analysis
Too many paths. If you really need to analyze all of the 256 paths individually, rerun with max-paths=256
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 10.67 to 8.00 cycles (1.33x speedup).

Workaround

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

Vectorization

Details

All VPU instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: 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)

No data for this section

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 192 bytes. The binary loop is storing 64 bytes.

General properties

nb instructions	69
loop length	276
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

8.63 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	5.50	5.50	1.75	1.75	1.67	1.67	1.58	1.58	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00
cycles	5.50	5.50	1.75	1.75	1.67	1.67	1.58	1.58	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00

Execution ports to units layout:

P0: BRU
P1: BRU
P2: ALU
P3: ALU
P4: ALU
P5: ALU
P6: ALU
P7: ALU
P8 (128 bits): VPU, FP store data, ALU, DIV/SQRT
P9 (128 bits): VPU, ALU, FP store data
P10 (128 bits): VPU, ALU, DIV/SQRT
P11 (128 bits): ALU, VPU
P12 (256 bits): store address, load
P13 (256 bits): store address, load
P14 (256 bits): load
P15 (64 bits): store data
P16 (64 bits): store data

Cycles executing div or sqrt instructions

Cycles summary

Front-end	8.63
Overall L1	10.67

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	NA (no load vectorizable/vectorized instructions)
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.67 cycles. At this rate:

18% of peak load performance is reached (18.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
12% of peak store performance is reached (6.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4d8860

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X14, [X28, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D2, [X21, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D2, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d887c <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1fc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D3, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D4, D3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D4, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X2, X1, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
CMP X25, X2	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	scal (50.0%)
B.EQ 4d8960 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR X17, [X28, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D30, [X21, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D30, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88a4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x224>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D6, [X19, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D28, D6	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D28, [X19, X2,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X30, X2, #1	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X7, [X28, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D29, [X21, X7,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D29, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88c4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x244>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D0, [X19, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D1, D0	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D1, [X19, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X8, X2, #2	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X9, [X28, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D2, [X21, X9,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D2, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d88e4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x264>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D3, [X19, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D4, D3	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D4, [X19, X8,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X16, X2, #3	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X15, [X28, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D5, [X21, X15,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D5, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8904 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x284>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D7, [X19, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D16, D7	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D16, [X19, X16,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X14, X2, #4	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X3, [X28, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D17, [X21, X3,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D17, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8924 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2a4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D18, [X19, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D19, D18	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D19, [X19, X14,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X1, X2, #5	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X18, [X28, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D20, [X21, X18,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D20, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8944 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x2c4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D21, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FNEG D22, D21	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D22, [X19, X1,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
ADD X17, X2, #6	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
LDR X30, [X28, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	4	0.33	N/A
LDR D31, [X21, X30,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
FCMPE D31, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.MI 4d91e0 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0xb60>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
ADD X1, X2, #7	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
B 4d8860 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D23, [X19, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33	scal (50.0%)
ADD X1, X2, #7	1	0	0	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	1	0.17	N/A
FNEG D24, D23	1	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25	scal (50.0%)
STR D24, [X19, X17,LSL #3]	1	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50	scal (50.0%)
B 4d8860 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x1e0>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code