OV - exec - Loop 1308

0x4a7c2c LDR	D1, [X1]    [5]

0x4a7c30 ADD	X0, X0, #4

0x4a7c34 CMP	X13, X0

0x4a7c38 FADD	D0, D0, D1

0x4a7c3c STR	D0, [X11, X9,LSL #3]    [1]

0x4a7c40 LDR	D1, [X1, X21,LSL #3]    [2]

0x4a7c44 FADD	D0, D0, D1

0x4a7c48 STR	D0, [X11, X9,LSL #3]    [1]

0x4a7c4c LDR	D1, [X1, X14]    [3]

0x4a7c50 FADD	D0, D0, D1

0x4a7c54 STR	D0, [X11, X9,LSL #3]    [1]

0x4a7c58 LDR	D1, [X1, X17]    [4]

0x4a7c5c ADD	X1, X1, X16

0x4a7c60 FADD	D0, D0, D1

0x4a7c64 STR	D0, [X11, X9,LSL #3]    [1]

0x4a7c68 B.NE	4a7c2c

/home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c: 577 - 579

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

577:                for (j = 0; j < num_threads; j++)

578:                {

579:                   y_data[i] += y_data_expand[j*y_size + i];

Coverage (%)	Name	Source Location	Module

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.00
Bottlenecks
Function	.omp_outlined..26#0x4a7a70
Source	csr_matvec.c:577-579
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	8.00
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	8.00
CQA cycles if fully vectorized	2.00
Front-end cycles	2.00
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.75
P2 cycles	0.75
P3 cycles	0.75
P4 cycles	0.75
P5 cycles	2.00
P6 cycles	2.00
P7 cycles	2.00
P8 cycles	2.00
P9 cycles	2.67
P10 cycles	2.67
P11 cycles	2.67
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	8
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	16.00
Nb uops	16.00
Nb loads	NA
Nb stores	4.00
Nb stack references	0.00
FLOP/cycle	0.50
Nb FLOP add-sub	4.00
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	8.00
Bytes prefetched	0.00
Bytes loaded	32.00
Bytes stored	32.00
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	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	25.00

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	4.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.00
Bottlenecks
Function	.omp_outlined..26#0x4a7a70
Source	csr_matvec.c:577-579
Source loop unroll info	unrolled by 4
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	4
CQA cycles	8.00
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	8.00
CQA cycles if fully vectorized	2.00
Front-end cycles	2.00
DIV/SQRT cycles	0.50
P0 cycles	0.50
P1 cycles	0.75
P2 cycles	0.75
P3 cycles	0.75
P4 cycles	0.75
P5 cycles	2.00
P6 cycles	2.00
P7 cycles	2.00
P8 cycles	2.00
P9 cycles	2.67
P10 cycles	2.67
P11 cycles	2.67
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.00
Inter-iter dependencies cycles	8
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	16.00
Nb uops	16.00
Nb loads	NA
Nb stores	4.00
Nb stack references	0.00
FLOP/cycle	0.50
Nb FLOP add-sub	4.00
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	8.00
Bytes prefetched	0.00
Bytes loaded	32.00
Bytes stored	32.00
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	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	25.00
Vector-efficiency ratio load	25.00
Vector-efficiency ratio store	25.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	25.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	25.00

Path /

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

Function	.omp_outlined..26#0x4a7a70
Source file and lines	csr_matvec.c:577-579
Module	exec

The loop is defined in /home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:577-579.

The related source loop is unrolled by 4 (including vectorization).

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 8.00 to 2.00 cycles (4.00x 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 is composed of 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 32 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.06 FP operations per loaded or stored byte.

General properties

nb instructions	16
loop length	64
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.75	0.75	0.75	0.75	2.00	2.00	2.00	2.00	2.67	2.67	2.67	0.00	0.00
cycles	0.50	0.50	0.75	0.75	0.75	0.75	2.00	2.00	2.00	2.00	2.67	2.67	2.67	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	8.00

Cycles summary

Front-end	2.00
Data deps.	8.00
Overall L1	8.00

Vectorization ratios

all	0%
load	0%
store	0%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 8.00 cycles. At this rate:

4% of peak load performance is reached (4.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
8% of peak store performance is reached (4.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4a7c2c

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
LDR D1, [X1]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X0, X0, #4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP X13, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X21,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X14]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X17]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X1, X1, X16	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
B.NE 4a7c2c <.omp_outlined..26+0x1bc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Function	.omp_outlined..26#0x4a7a70
Source file and lines	csr_matvec.c:577-579
Module	exec

The loop is defined in /home/hbollore/qaas-runs/171-233-5044/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:577-579.

The related source loop is unrolled by 4 (including vectorization).

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 4 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 8.00 to 2.00 cycles (4.00x 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 is composed of 4 FP arithmetical operations:

4: addition or subtraction

The binary loop is loading 32 bytes. The binary loop is storing 32 bytes.

Arithmetic intensity

Arithmetic intensity is 0.06 FP operations per loaded or stored byte.

General properties

nb instructions	16
loop length	64
nb stack references	0

Front-end

FIT IN UOP CACHE

front end

2.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14
uops	0.50	0.50	0.75	0.75	0.75	0.75	2.00	2.00	2.00	2.00	2.67	2.67	2.67	0.00	0.00
cycles	0.50	0.50	0.75	0.75	0.75	0.75	2.00	2.00	2.00	2.00	2.67	2.67	2.67	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions	NA
Longest recurrence chain latency (RecMII)	8.00

Cycles summary

Front-end	2.00
Data deps.	8.00
Overall L1	8.00

Vectorization ratios

all	0%
load	0%
store	0%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 8.00 cycles. At this rate:

4% of peak load performance is reached (4.00 out of 96.00 bytes loaded per cycle (GB/s @ 1GHz))
8% of peak store performance is reached (4.00 out of 48.00 bytes stored per cycle (GB/s @ 1GHz))

ASM code

In the binary file, the address of the loop is: 4a7c2c

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	Latency	Recip. throughput
LDR D1, [X1]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X0, X0, #4	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
CMP X13, X0	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X21,LSL #3]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X14]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
LDR D1, [X1, X17]	1	0	0	0	0	0	0	0	0	0	0	0.33	0.33	0.33	6	0.33
ADD X1, X1, X16	1	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	1	0.25
FADD D0, D0, D1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	2	0.25
STR D0, [X11, X9,LSL #3]	1	0	0	0	0	0	0	0.50	0.50	0	0	0.50	0.50	0	2	0.50
B.NE 4a7c2c <.omp_outlined..26+0x1bc>	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	0	1	0.50

Report Configuration

Vectorization

Details

Workaround

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code

Vectorization

Details

Workaround

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Cycles and memory resources usage

ASM code