OV - exec - Loop 5745

	Loop Id: 5745	Module: exec	Source: csr_matvec.c:248-341 [...]	Coverage: 0.66%

0x59731c MOV	-0x30(%RBP),%R11

0x597320 MOV	0x28(%RBP),%R15

0x597324 NOPW	%CS:(%RAX,%RAX,1)

0x597330 MOVSD	%XMM0,(%RAX,%R8,8)

0x597336 LEA	0x1(%RDI),%RSI

0x59733a CMP	%R12,%RDI

0x59733d MOV	%RSI,%RDI

0x597340 JE	598462

0x597346 LEA	(%R11,%RDI,1),%R8

0x59734a MOVSD	(%R15,%R8,8),%XMM0

0x597350 MOV	(%RBX,%R8,8),%R9

0x597354 MOV	0x8(%RBX,%R8,8),%R10

0x597359 SUB	%R9,%R10

0x59735c JLE	597330

0x59735e CMP	$0x8,%R10

0x597362 JB	597434

0x597368 MOV	%R10,%R11

0x59736b SHR	$0x3,%R11

0x59736f LEA	0x38(,%R9,8),%R13

0x597377 NOPW	(%RAX,%RAX,1)

  (5746) 0x597380 MOV	-0x38(%RDX,%R13,1),%RSI

  (5746) 0x597385 MOV	-0x30(%RDX,%R13,1),%R15

  (5746) 0x59738a MOVSD	(%RCX,%RSI,8),%XMM1

  (5746) 0x59738f MULSD	-0x38(%R14,%R13,1),%XMM1

  (5746) 0x597396 ADDSD	%XMM0,%XMM1

  (5746) 0x59739a MOVSD	(%RCX,%R15,8),%XMM0

  (5746) 0x5973a0 MULSD	-0x30(%R14,%R13,1),%XMM0

  (5746) 0x5973a7 MOV	-0x28(%RDX,%R13,1),%RSI

  (5746) 0x5973ac MOVSD	(%RCX,%RSI,8),%XMM2

  (5746) 0x5973b1 MULSD	-0x28(%R14,%R13,1),%XMM2

  (5746) 0x5973b8 ADDSD	%XMM0,%XMM2

  (5746) 0x5973bc MOV	-0x20(%RDX,%R13,1),%RSI

  (5746) 0x5973c1 MOVSD	(%RCX,%RSI,8),%XMM0

  (5746) 0x5973c6 MULSD	-0x20(%R14,%R13,1),%XMM0

  (5746) 0x5973cd MOV	-0x18(%RDX,%R13,1),%RSI

  (5746) 0x5973d2 MOVSD	(%RCX,%RSI,8),%XMM3

  (5746) 0x5973d7 MULSD	-0x18(%R14,%R13,1),%XMM3

  (5746) 0x5973de ADDSD	%XMM1,%XMM2

  (5746) 0x5973e2 ADDSD	%XMM0,%XMM3

  (5746) 0x5973e6 MOV	-0x10(%RDX,%R13,1),%RSI

  (5746) 0x5973eb MOVSD	(%RCX,%RSI,8),%XMM1

  (5746) 0x5973f0 MULSD	-0x10(%R14,%R13,1),%XMM1

  (5746) 0x5973f7 ADDSD	%XMM3,%XMM1

  (5746) 0x5973fb ADDSD	%XMM2,%XMM1

  (5746) 0x5973ff MOV	-0x8(%RDX,%R13,1),%RSI

  (5746) 0x597404 MOVSD	(%RCX,%RSI,8),%XMM2

  (5746) 0x597409 MULSD	-0x8(%R14,%R13,1),%XMM2

  (5746) 0x597410 MOV	(%RDX,%R13,1),%RSI

  (5746) 0x597414 MOVSD	(%RCX,%RSI,8),%XMM0

  (5746) 0x597419 MULSD	(%R14,%R13,1),%XMM0

  (5746) 0x59741f ADDSD	%XMM2,%XMM0

  (5746) 0x597423 ADDSD	%XMM1,%XMM0

  (5746) 0x597427 ADD	$0x40,%R13

  (5746) 0x59742b DEC	%R11

  (5746) 0x59742e JNE	597380

0x597434 MOV	%R10D,%R11D

0x597437 AND	$0x7,%R11D

0x59743b DEC	%R11

0x59743e CMP	$0x6,%R11

0x597442 JA	59731c

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c: 248 - 341

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

248:       hypre_assert(iBegin <= iEnd);

[...]

285:                   tempx += A_data[jj] * x_data[A_j[jj]];

[...]

334:             for (i = iBegin; i < iEnd; i++)

335:             {

336:                tempx = b_data[i];

337:                for (jj = A_i[i]; jj < A_i[i+1]; jj++)

338:                {

339:                   tempx += A_data[jj] * x_data[A_j[jj]];

340:                }

341:                y_data[i] = tempx;

Coverage (%)	Name	Source Location	Module
►96.46+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_launch_thread		libiomp5.so
○	_INTERNALea951ae6::__kmp_launc[...]		libiomp5.so
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►2.52+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so
○	__kmp_fork_call		libiomp5.so
○	__kmpc_fork_call		libiomp5.so
○	hypre_CSRMatrixMatvecOutOfPlac[...]	csr_matvec.c:243	exec
○	hypre_ParCSRMatrixMatvecOutOfP[...]	par_csr_matvec.c:237	exec
○	hypre_BoomerAMGCycle	par_cycle.c:456	exec
○	hypre_BoomerAMGSolve	par_amg_solve.c:272	exec
○	hypre_PCGSolve	pcg.c:545	exec
○	main	amg.c:419	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start		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	1.65
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	1.27
Bottlenecks	micro-operation queue,
Function	hypre_CSRMatrixMatvecOutOfPlace.extracted
Source	csr_matvec.c:248-248,csr_matvec.c:285-285,csr_matvec.c:334-337,csr_matvec.c:341-341
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	2.75
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	2.75
CQA cycles if fully vectorized	0.34
Front-end cycles	2.75
P0 cycles	2.17
P1 cycles	2.17
P2 cycles	2.17
P3 cycles	2.17
P4 cycles	2.17
P5 cycles	2.17
P6 cycles	1.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.50
P15 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	25.00
Nb uops	22.00
Nb loads	5.00
Nb stores	1.00
Nb stack references	2.00
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	17.45
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	8.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	12.50

Metric	Value
CQA speedup if no scalar integer	1.65
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	1.27
Bottlenecks	micro-operation queue,
Function	hypre_CSRMatrixMatvecOutOfPlace.extracted
Source	csr_matvec.c:248-248,csr_matvec.c:285-285,csr_matvec.c:334-337,csr_matvec.c:341-341
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	2.75
CQA cycles if no scalar integer	1.67
CQA cycles if FP arith vectorized	2.75
CQA cycles if fully vectorized	0.34
Front-end cycles	2.75
P0 cycles	2.17
P1 cycles	2.17
P2 cycles	2.17
P3 cycles	2.17
P4 cycles	2.17
P5 cycles	2.17
P6 cycles	1.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	0.00
P11 cycles	0.00
P12 cycles	0.00
P13 cycles	0.00
P14 cycles	0.50
P15 cycles	0.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	25.00
Nb uops	22.00
Nb loads	5.00
Nb stores	1.00
Nb stack references	2.00
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	17.45
Bytes prefetched	0.00
Bytes loaded	40.00
Bytes stored	8.00
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	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	12.50

Path /

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

Function	hypre_CSRMatrixMatvecOutOfPlace.extracted
Source file and lines	csr_matvec.c:248-341
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:248,285,334-341.

Analyzed code is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:248,285,334-337,341.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

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 2.75 to 1.67 cycles (1.65x speedup).

Workaround

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

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 2.75 to 0.34 cycles (8.00x speedup).

Details

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

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

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 40 bytes. The binary loop is storing 8 bytes.

General properties

nb instructions	25
nb uops	22
loop length	120
used x86 registers	12
used mmx registers	0
used xmm registers	1
used ymm registers	0
used zmm registers	0
nb stack references	2

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	2.75 cycles
front end	2.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.17	2.17	2.17	2.17	2.17	2.17	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.50	0.50
cycles	2.17	2.17	2.17	2.17	2.17	2.17	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.50	0.50

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	2.75
Dispatch	2.17
Overall L1	2.75

Vectorization ratios

INT

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

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	NA (no other vectorizable/vectorized instructions)

INT+FP

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%

Vector efficiency ratios

INT

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	12%

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	NA (no other vectorizable/vectorized instructions)

INT+FP

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 2.75 cycles. At this rate:

5% of peak load performance is reached (14.55 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (2.91 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 2.75 to 2.17 cycles (1.27x speedup).

ASM code

In the binary file, the address of the loop is: 597330

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P14	P15	Latency	Recip. throughput	Vectorization
MOV -0x30(%RBP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x28(%RBP),%R15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOVSD %XMM0,(%RAX,%R8,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
LEA 0x1(%RDI),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
CMP %R12,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
MOV %RSI,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JE 598462 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0x1b62>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
LEA (%R11,%RDI,1),%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
MOVSD (%R15,%R8,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV (%RBX,%R8,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RBX,%R8,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
SUB %R9,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JLE 597330 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xa30>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
CMP $0x8,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JB 597434 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xb34>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
SHR $0x3,%R11	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33	N/A
LEA 0x38(,%R9,8),%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOV %R10D,%R11D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
AND $0x7,%R11D	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
DEC %R11	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP $0x6,%R11	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JA 59731c <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xa1c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A

Function	hypre_CSRMatrixMatvecOutOfPlace.extracted
Source file and lines	csr_matvec.c:248-341
Module	exec

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Failed to get the number of paths
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

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 2.75 to 1.67 cycles (1.65x speedup).

Workaround

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

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers. By vectorizing your loop, you can lower the cost of an iteration from 2.75 to 0.34 cycles (8.00x speedup).

Details

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

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

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 40 bytes. The binary loop is storing 8 bytes.

General properties

nb instructions	25
nb uops	22
loop length	120
used x86 registers	12
used mmx registers	0
used xmm registers	1
used ymm registers	0
used zmm registers	0
nb stack references	2

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	2.75 cycles
front end	2.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	2.17	2.17	2.17	2.17	2.17	2.17	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.50	0.50
cycles	2.17	2.17	2.17	2.17	2.17	2.17	1.50	1.50	1.50	1.50	0.00	0.00	0.00	0.00	0.50	0.50

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	2.75
Dispatch	2.17
Overall L1	2.75

Vectorization ratios

INT

all	0%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

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	NA (no other vectorizable/vectorized instructions)

INT+FP

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%

Vector efficiency ratios

INT

all	12%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	12%

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	NA (no other vectorizable/vectorized instructions)

INT+FP

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 2.75 cycles. At this rate:

5% of peak load performance is reached (14.55 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (2.91 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 597330

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P14	P15	Latency	Recip. throughput	Vectorization
MOV -0x30(%RBP),%R11	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x28(%RBP),%R15	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOVSD %XMM0,(%RAX,%R8,8)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	1	0.50	scal (12.5%)
LEA 0x1(%RDI),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
CMP %R12,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
MOV %RSI,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JE 598462 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0x1b62>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
LEA (%R11,%RDI,1),%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
MOVSD (%R15,%R8,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.50	scal (12.5%)
MOV (%RBX,%R8,8),%R9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	N/A
MOV 0x8(%RBX,%R8,8),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	3	0.25	scal (12.5%)
SUB %R9,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JLE 597330 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xa30>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
CMP $0x8,%R10	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	scal (12.5%)
JB 597434 <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xb34>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A
MOV %R10,%R11	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
SHR $0x3,%R11	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.33	N/A
LEA 0x38(,%R9,8),%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.25	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
MOV %R10D,%R11D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
AND $0x7,%R11D	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
DEC %R11	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
CMP $0x6,%R11	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	1	0.17	N/A
JA 59731c <hypre_CSRMatrixMatvecOutOfPlace.extracted+0xa1c>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	1	0.50-1	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

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

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

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