OV - exec - Loop 2616

	Loop Id: 2616	Module: exec	Source: ams.c:3532-3534	Coverage: 0.03%

0x4d8c48 LDR	X14, [X24, X1,LSL #3]    [1]

0x4d8c4c LDR	D26, [X21, X14,LSL #3]    [3]

0x4d8c50 FCMPE	D26, #0

0x4d8c54 B.GE	4d8c64

0x4d8c58 LDR	D27, [X19, X1,LSL #3]    [2]

0x4d8c5c FNEG	D6, D27

0x4d8c60 STR	D6, [X19, X1,LSL #3]    [2]

0x4d8c64 ADD	X7, X1, #1

0x4d8c68 LDR	X0, [X24, X7,LSL #3]    [1]

0x4d8c6c LDR	D0, [X21, X0,LSL #3]    [3]

0x4d8c70 FCMPE	D0, #0

0x4d8c74 B.GE	4d8c84

0x4d8c78 LDR	D1, [X19, X7,LSL #3]    [2]

0x4d8c7c FNEG	D2, D1

0x4d8c80 STR	D2, [X19, X7,LSL #3]    [2]

0x4d8c84 ADD	X12, X7, #1

0x4d8c88 LDR	X3, [X24, X12,LSL #3]    [1]

0x4d8c8c LDR	D3, [X21, X3,LSL #3]    [3]

0x4d8c90 FCMPE	D3, #0

0x4d8c94 B.GE	4d8ca4

0x4d8c98 LDR	D4, [X19, X12,LSL #3]    [2]

0x4d8c9c FNEG	D5, D4

0x4d8ca0 STR	D5, [X19, X12,LSL #3]    [2]

0x4d8ca4 ADD	X1, X7, #2

0x4d8ca8 LDR	X16, [X24, X1,LSL #3]    [1]

0x4d8cac LDR	D7, [X21, X16,LSL #3]    [3]

0x4d8cb0 FCMPE	D7, #0

0x4d8cb4 B.GE	4d8cc4

0x4d8cb8 LDR	D16, [X19, X1,LSL #3]    [2]

0x4d8cbc FNEG	D17, D16

0x4d8cc0 STR	D17, [X19, X1,LSL #3]    [2]

0x4d8cc4 ADD	X11, X7, #3

0x4d8cc8 LDR	X2, [X24, X11,LSL #3]    [1]

0x4d8ccc LDR	D31, [X21, X2,LSL #3]    [3]

0x4d8cd0 FCMPE	D31, #0

0x4d8cd4 B.GE	4d8ce4

0x4d8cd8 LDR	D18, [X19, X11,LSL #3]    [2]

0x4d8cdc FNEG	D19, D18

0x4d8ce0 STR	D19, [X19, X11,LSL #3]    [2]

0x4d8ce4 ADD	X30, X7, #4

0x4d8ce8 LDR	X17, [X24, X30,LSL #3]    [1]

0x4d8cec LDR	D20, [X21, X17,LSL #3]    [3]

0x4d8cf0 FCMPE	D20, #0

0x4d8cf4 B.GE	4d8d04

0x4d8cf8 LDR	D28, [X19, X30,LSL #3]    [2]

0x4d8cfc FNEG	D29, D28

0x4d8d00 STR	D29, [X19, X30,LSL #3]    [2]

0x4d8d04 ADD	X18, X7, #5

0x4d8d08 LDR	X14, [X24, X18,LSL #3]    [1]

0x4d8d0c LDR	D24, [X21, X14,LSL #3]    [3]

0x4d8d10 FCMPE	D24, #0

0x4d8d14 B.GE	4d8d24

0x4d8d18 LDR	D21, [X19, X18,LSL #3]    [2]

0x4d8d1c FNEG	D22, D21

0x4d8d20 STR	D22, [X19, X18,LSL #3]    [2]

0x4d8d24 ADD	X0, X7, #6

0x4d8d28 LDR	X12, [X24, X0,LSL #3]    [1]

0x4d8d2c LDR	D30, [X21, X12,LSL #3]    [3]

0x4d8d30 FCMPE	D30, #0

0x4d8d34 B.GE	4d8d44

0x4d8d38 LDR	D23, [X19, X0,LSL #3]    [2]

0x4d8d3c FNEG	D25, D23

0x4d8d40 STR	D25, [X19, X0,LSL #3]    [2]

0x4d8d44 ADD	X1, X7, #7

0x4d8d48 CMP	X26, X1

0x4d8d4c B.NE	4d8c48

/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
►100.00+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	thread_start		libc.so.6

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	1.29
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	8.25
P0 cycles	4.50
P1 cycles	4.50
P2 cycles	1.58
P3 cycles	1.58
P4 cycles	1.50
P5 cycles	1.50
P6 cycles	1.42
P7 cycles	1.42
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	66.00
Nb uops	66.00
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	1.29
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	8.25
P0 cycles	4.50
P1 cycles	4.50
P2 cycles	1.58
P3 cycles	1.58
P4 cycles	1.50
P5 cycles	1.50
P6 cycles	1.42
P7 cycles	1.42
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	66.00
Nb uops	66.00
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	66
loop length	264
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

8.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	4.50	4.50	1.58	1.58	1.50	1.50	1.42	1.42	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00
cycles	4.50	4.50	1.58	1.58	1.50	1.50	1.42	1.42	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	8.25
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	NA (no other vectorizable/vectorized instructions)

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: 4d8c48

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X14, [X24, 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 D26, [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 D26, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8c64 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x5e4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D27, [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 D6, D27	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 D6, [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 X7, 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
LDR X0, [X24, X7,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 D0, [X21, X0,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 D0, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8c84 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x604>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D1, [X19, 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%)
FNEG D2, D1	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 D2, [X19, X7,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 X12, X7, #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 X3, [X24, X12,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 D3, [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 D3, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8ca4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x624>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D4, [X19, X12,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 D5, D4	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 D5, [X19, X12,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, X7, #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 X16, [X24, 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 D7, [X21, 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%)
FCMPE D7, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8cc4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x644>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D16, [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 D17, D16	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 D17, [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 X11, X7, #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 X2, [X24, X11,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, 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%)
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.GE 4d8ce4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x664>	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, X11,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, X11,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, X7, #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 X17, [X24, 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 D20, [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 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 4d8d04 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x684>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D28, [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 D29, D28	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 D29, [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 X18, X7, #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 X14, [X24, X18,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 D24, [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 D24, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8d24 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x6a4>	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, 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%)
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, X18,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 X0, X7, #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 X12, [X24, X0,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, X12,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 4d8d44 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x6c4>	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, X0,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 D25, 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 D25, [X19, X0,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, X7, #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
CMP X26, X1	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	N/A
B.NE 4d8c48 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x5c8>	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	66
loop length	264
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

8.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	P16
uops	4.50	4.50	1.58	1.58	1.50	1.50	1.42	1.42	8.00	8.00	4.00	4.00	10.67	10.67	10.67	0.00	0.00
cycles	4.50	4.50	1.58	1.58	1.50	1.50	1.42	1.42	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	8.25
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	NA (no other vectorizable/vectorized instructions)

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: 4d8c48

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	Latency	Recip. throughput	Vectorization
LDR X14, [X24, 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 D26, [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 D26, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8c64 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x5e4>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D27, [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 D6, D27	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 D6, [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 X7, 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
LDR X0, [X24, X7,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 D0, [X21, X0,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 D0, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8c84 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x604>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D1, [X19, 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%)
FNEG D2, D1	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 D2, [X19, X7,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 X12, X7, #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 X3, [X24, X12,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 D3, [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 D3, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8ca4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x624>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D4, [X19, X12,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 D5, D4	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 D5, [X19, X12,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, X7, #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 X16, [X24, 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 D7, [X21, 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%)
FCMPE D7, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8cc4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x644>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D16, [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 D17, D16	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 D17, [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 X11, X7, #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 X2, [X24, X11,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, 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%)
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.GE 4d8ce4 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x664>	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, X11,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, X11,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, X7, #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 X17, [X24, 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 D20, [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 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 4d8d04 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x684>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
LDR D28, [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 D29, D28	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 D29, [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 X18, X7, #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 X14, [X24, X18,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 D24, [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 D24, #0	1	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	2	1	scal (50.0%)
B.GE 4d8d24 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x6a4>	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, 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%)
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, X18,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 X0, X7, #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 X12, [X24, X0,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, X12,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 4d8d44 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x6c4>	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, X0,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 D25, 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 D25, [X19, X0,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, X7, #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
CMP X26, X1	1	0	0	0.25	0.25	0	0	0.25	0.25	0	0	0	0	0	0	0	1	0.33	N/A
B.NE 4d8c48 <hypre_ParCSRComputeL1NormsThreads._omp_fn.0+0x5c8>	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