OV - exec - Loop 5760

	Loop Id: 5760	Module: exec	Source: csr_matvec.c:194-195	Coverage: 0.04%

0x598b30 MOV	-0x38(%R14,%R9,1),%RCX    [2]

0x598b35 MOV	-0x30(%R14,%R9,1),%R10    [2]

0x598b3a MOVSD	(%RBX,%RCX,8),%XMM1    [3]

0x598b3f MULSD	-0x38(%R12,%R9,1),%XMM1    [1]

0x598b46 ADDSD	%XMM0,%XMM1

0x598b4a MOVSD	(%RBX,%R10,8),%XMM0    [3]

0x598b50 MULSD	-0x30(%R12,%R9,1),%XMM0    [1]

0x598b57 MOV	-0x28(%R14,%R9,1),%RCX    [2]

0x598b5c MOVSD	(%RBX,%RCX,8),%XMM2    [3]

0x598b61 MULSD	-0x28(%R12,%R9,1),%XMM2    [1]

0x598b68 ADDSD	%XMM0,%XMM2

0x598b6c MOV	-0x20(%R14,%R9,1),%RCX    [2]

0x598b71 MOVSD	(%RBX,%RCX,8),%XMM0    [3]

0x598b76 MULSD	-0x20(%R12,%R9,1),%XMM0    [1]

0x598b7d MOV	-0x18(%R14,%R9,1),%RCX    [2]

0x598b82 MOVSD	(%RBX,%RCX,8),%XMM3    [3]

0x598b87 MULSD	-0x18(%R12,%R9,1),%XMM3    [1]

0x598b8e ADDSD	%XMM1,%XMM2

0x598b92 ADDSD	%XMM0,%XMM3

0x598b96 MOV	-0x10(%R14,%R9,1),%RCX    [2]

0x598b9b MOVSD	(%RBX,%RCX,8),%XMM1    [3]

0x598ba0 MULSD	-0x10(%R12,%R9,1),%XMM1    [1]

0x598ba7 ADDSD	%XMM3,%XMM1

0x598bab ADDSD	%XMM2,%XMM1

0x598baf MOV	-0x8(%R14,%R9,1),%RCX    [2]

0x598bb4 MOVSD	(%RBX,%RCX,8),%XMM2    [3]

0x598bb9 MULSD	-0x8(%R12,%R9,1),%XMM2    [1]

0x598bc0 MOV	(%R14,%R9,1),%RCX    [2]

0x598bc4 MOVSD	(%RBX,%RCX,8),%XMM0    [3]

0x598bc9 MULSD	(%R12,%R9,1),%XMM0    [1]

0x598bcf ADDSD	%XMM2,%XMM0

0x598bd3 ADDSD	%XMM1,%XMM0

0x598bd7 ADD	$0x40,%R9

0x598bdb DEC	%R8

0x598bde JNE	598b30

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c: 194 - 195

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

194:                for (jj = A_i[m]; jj < A_i[m+1]; jj++)

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

Coverage (%)	Name	Source Location	Module
►96.28+	__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
►1.38+	__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:211	exec
○	hypre_ParCSRMatrixMatvecOutOfP[...]	par_csr_matvec.c:237	exec
○	hypre_BoomerAMGCycle	par_cycle.c:425	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
►1.06+	__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:211	exec
○	hypre_ParCSRMatrixMatvecOutOfP[...]	par_csr_matvec.c:237	exec
○	hypre_PCGSolve	pcg.c:496	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	1.00
CQA speedup if FP arith vectorized	1.88
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.33
Bottlenecks
Function	hypre_CSRMatrixMatvecOutOfPlace.extracted.19
Source	csr_matvec.c:194-195
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	8.00
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	4.25
CQA cycles if fully vectorized	1.00
Front-end cycles	4.25
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	0.33
P4 cycles	0.33
P5 cycles	0.33
P6 cycles	6.00
P7 cycles	6.00
P8 cycles	6.00
P9 cycles	6.00
P10 cycles	4.00
P11 cycles	4.00
P12 cycles	4.00
P13 cycles	4.00
P14 cycles	0.00
P15 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	8
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	35.00
Nb uops	34.00
Nb loads	24.00
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	2.00
Nb FLOP add-sub	8.00
Nb FLOP mul	8.00
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	0.00
Stride 0	0.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	1.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	12.50
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.88
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.33
Bottlenecks
Function	hypre_CSRMatrixMatvecOutOfPlace.extracted.19
Source	csr_matvec.c:194-195
Source loop unroll info	not unrolled or unrolled with no peel/tail loop
Source loop unroll confidence level	max
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	8.00
CQA cycles if no scalar integer	8.00
CQA cycles if FP arith vectorized	4.25
CQA cycles if fully vectorized	1.00
Front-end cycles	4.25
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	0.33
P4 cycles	0.33
P5 cycles	0.33
P6 cycles	6.00
P7 cycles	6.00
P8 cycles	6.00
P9 cycles	6.00
P10 cycles	4.00
P11 cycles	4.00
P12 cycles	4.00
P13 cycles	4.00
P14 cycles	0.00
P15 cycles	0.00
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	8
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	35.00
Nb uops	34.00
Nb loads	24.00
Nb stores	0.00
Nb stack references	0.00
FLOP/cycle	2.00
Nb FLOP add-sub	8.00
Nb FLOP mul	8.00
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	0.00
Stride 0	0.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	1.00
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	NA
Vectorization ratio other	NA
Vector-efficiency ratio all	12.50
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	NA
Vector-efficiency ratio mul	12.50
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	NA

Path /

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

Function	hypre_CSRMatrixMatvecOutOfPlace.extracted.19
Source file and lines	csr_matvec.c:194-195
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:194-195.

The related source loop is not unrolled or unrolled with no peel/tail loop.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 8.00 to 1.00 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.

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- use march=native
Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Irregular (variable stride) or indirect: 1 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

Try to remove indirect accesses. If applicable, precompute elements out of the innermost loop.

Type of elements and instruction set

16 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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

The binary loop is composed of 16 FP arithmetical operations:

8: addition or subtraction
8: multiply

The binary loop is loading 192 bytes (24 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.08 FP operations per loaded or stored byte.

Unroll opportunity

Loop is data access bound.

Workaround

Unroll your loop if trip count is significantly higher than target unroll factor and if some data references are common to consecutive iterations. This can be done manually. Or with the unroll (resp. unroll_and_jam) directive on top of the inner (resp. surrounding) loop. You can enforce an unroll factor: #pragma unroll_and_jam N, unroll_and_jam(N), unroll N or unroll(N)

General properties

nb instructions	35
nb uops	34
loop length	180
used x86 registers	7
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	0
ADD-SUB / MUL ratio	1.00

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	4.25 cycles
front end	4.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	0.33	0.33	0.33	6.00	6.00	6.00	6.00	4.00	4.00	4.00	4.00	0.00	0.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	6.00	6.00	6.00	6.00	4.00	4.00	4.00	4.00	0.00	0.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

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

Cycles summary

Front-end	4.25
Dispatch	6.00
Data deps.	8.00
Overall L1	8.00

Vectorization ratios

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

Vector efficiency ratios

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

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:

9% of peak load performance is reached (24.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 598b30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	Latency	Recip. throughput	Vectorization
MOV -0x38(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOV -0x30(%R14,%R9,1),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x38(%R12,%R9,1),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM0,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOVSD (%RBX,%R10,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x30(%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV -0x28(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x28(%R12,%R9,1),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM0,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x20(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x20(%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV -0x18(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x18(%R12,%R9,1),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM0,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x10(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x10(%R12,%R9,1),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM2,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x8(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x8(%R12,%R9,1),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV (%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD (%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM2,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADD $0x40,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
DEC %R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 598b30 <hypre_CSRMatrixMatvecOutOfPlace.extracted.19+0x1d0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A

Function	hypre_CSRMatrixMatvecOutOfPlace.extracted.19
Source file and lines	csr_matvec.c:194-195
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:194-195.

The related source loop is not unrolled or unrolled with no peel/tail loop.

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. 8 data elements could be processed at once in vector registers.

By vectorizing your loop, you can lower the cost of an iteration from 8.00 to 1.00 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.

FMA

Presence of both ADD/SUB and MUL operations.

Workaround

Pass to your compiler a micro-architecture specialization option:
- use march=native
Try to change order in which elements are evaluated (using parentheses) in arithmetic expressions containing both ADD/SUB and MUL operations to enable your compiler to generate FMA instructions wherever possible. For instance a + b*c is a valid FMA (MUL then ADD). However (a+b)* c cannot be translated into an FMA (ADD then MUL).

Slow data structures access

Detected data structures (typically arrays) that cannot be efficiently read/written

Details

Irregular (variable stride) or indirect: 1 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

Try to remove indirect accesses. If applicable, precompute elements out of the innermost loop.

Type of elements and instruction set

16 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

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

The binary loop is composed of 16 FP arithmetical operations:

8: addition or subtraction
8: multiply

The binary loop is loading 192 bytes (24 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.08 FP operations per loaded or stored byte.

Unroll opportunity

Loop is data access bound.

Workaround

General properties

nb instructions	35
nb uops	34
loop length	180
used x86 registers	7
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	0
ADD-SUB / MUL ratio	1.00

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	4.25 cycles
front end	4.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	0.33	0.33	0.33	6.00	6.00	6.00	6.00	4.00	4.00	4.00	4.00	0.00	0.00
cycles	0.33	0.33	0.33	0.33	0.33	0.33	6.00	6.00	6.00	6.00	4.00	4.00	4.00	4.00	0.00	0.00

Execution ports to units layout:

P0: ALU
P1: ALU
P2: ALU
P3: ALU, BRU
P4: ALU, BRU
P5: ALU, BRU
P6 (512 bits): store address, load
P7 (512 bits): store address, load
P8 (512 bits): store address, load
P9 (512 bits): store address, load
P10 (512 bits): VPU, DIV/SQRT
P11 (512 bits): VPU, DIV/SQRT
P12 (512 bits): VPU
P13 (512 bits): VPU
P14 (512 bits): FP store data
P15 (512 bits): FP store data

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

Cycles summary

Front-end	4.25
Dispatch	6.00
Data deps.	8.00
Overall L1	8.00

Vectorization ratios

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

Vector efficiency ratios

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

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:

9% of peak load performance is reached (24.00 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 598b30

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	Latency	Recip. throughput	Vectorization
MOV -0x38(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOV -0x30(%R14,%R9,1),%R10	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x38(%R12,%R9,1),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM0,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOVSD (%RBX,%R10,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x30(%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV -0x28(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x28(%R12,%R9,1),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM0,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x20(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x20(%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV -0x18(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x18(%R12,%R9,1),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM1,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM0,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x10(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x10(%R12,%R9,1),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM2,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
MOV -0x8(%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD -0x8(%R12,%R9,1),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
MOV (%R14,%R9,1),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	3	0.25	N/A
MOVSD (%RBX,%RCX,8),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	1	0.50	scal (12.5%)
MULSD (%R12,%R9,1),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	3	0.50	scal (12.5%)
ADDSD %XMM2,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADDSD %XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	2	0.50	scal (12.5%)
ADD $0x40,%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
DEC %R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	1	0.17	N/A
JNE 598b30 <hypre_CSRMatrixMatvecOutOfPlace.extracted.19+0x1d0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	1	0.50-1	N/A

Report Configuration

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Workaround

Slow data structures access

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

Unroll opportunity

Workaround

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Workaround

Slow data structures access

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

Unroll opportunity

Workaround

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code