OV - exec - Loop 154

	Loop Id: 154	Module: exec	Source: calc_dt_kernel.f90:99-129	Coverage: 3.40%

0x428e80 VMOVQ	%XMM23,%RDX

0x428e86 VMOVSD	(%RDX),%XMM3    [14]

0x428e8a VMINSD	%XMM7,%XMM3,%XMM7

0x428e8e LEA	0x1(%RAX),%RDX

0x428e92 VMINSD	%XMM1,%XMM7,%XMM1

0x428e96 VMINSD	%XMM1,%XMM12,%XMM12

0x428e9a CMP	%RAX,-0x38(%RBP)    [9]

0x428e9e JE	428fa0

0x428ea4 MOV	%RDX,%RAX

0x428ea7 VMOVSD	(%R13,%RAX,8),%XMM3    [10]

0x428eae VMOVSD	(%R14,%RAX,8),%XMM2    [11]

0x428eb4 VADDSD	%XMM3,%XMM3,%XMM5

0x428eb8 VDIVSD	(%R12,%RAX,8),%XMM5,%XMM0    [4]

0x428ebe VFMADD132SD	%XMM2,%XMM0,%XMM2

0x428ec3 VMOVAPD	%XMM4,%XMM0

0x428ec7 VMOVSD	(%RBX,%RAX,8),%XMM5    [13]

0x428ecc VMOVSD	0x8(%R8,%RAX,8),%XMM7    [12]

0x428ed3 VANDPD	%XMM6,%XMM0,%XMM3

0x428ed7 VADDSD	0x8(%R9,%RAX,8),%XMM7,%XMM1    [2]

0x428ede VMULSD	%XMM5,%XMM11,%XMM10

0x428ee2 VMULSD	0x8(%RDI,%RAX,8),%XMM1,%XMM4    [5]

0x428ee8 VMAXSD	%XMM10,%XMM3,%XMM1

0x428eed VANDPD	%XMM6,%XMM4,%XMM3

0x428ef1 VMULSD	%XMM5,%XMM15,%XMM7

0x428ef5 VMAXSD	%XMM3,%XMM1,%XMM1

0x428ef9 VMOVAPD	%XMM8,%XMM3

0x428efd VDIVSD	%XMM1,%XMM7,%XMM7

0x428f01 VMOVSD	0x8(%RSI,%RAX,8),%XMM8    [7]

0x428f07 VADDSD	%XMM8,%XMM3,%XMM1

0x428f0c VMULSD	(%R11,%RAX,8),%XMM1,%XMM3    [8]

0x428f12 VMOVAPD	%XMM9,%XMM1

0x428f16 VSUBSD	%XMM3,%XMM4,%XMM30

0x428f1c VMOVSD	0x8(%RCX,%RAX,8),%XMM9    [3]

0x428f22 VANDPD	%XMM6,%XMM3,%XMM3

0x428f26 VADDSD	%XMM9,%XMM1,%XMM1

0x428f2b VMULSD	(%R10,%RAX,8),%XMM1,%XMM1    [1]

0x428f31 VSUBSD	%XMM0,%XMM1,%XMM0

0x428f35 VANDPD	%XMM6,%XMM1,%XMM1

0x428f39 VADDSD	%XMM30,%XMM0,%XMM0

0x428f3f VSQRTSD	%XMM2,%XMM2,%XMM2

0x428f43 VMAXSD	%XMM3,%XMM1,%XMM1

0x428f47 VMULSD	%XMM5,%XMM14,%XMM30

0x428f4d VMAXSD	%XMM10,%XMM1,%XMM10

0x428f52 VMINSD	(%R15,%RAX,8),%XMM18,%XMM3    [6]

0x428f59 VDIVSD	%XMM10,%XMM30,%XMM1

0x428f5f VMAXSD	%XMM2,%XMM11,%XMM2

0x428f63 VMULSD	%XMM17,%XMM3,%XMM10

0x428f69 VADDSD	%XMM5,%XMM5,%XMM5

0x428f6d VDIVSD	%XMM2,%XMM10,%XMM3

0x428f71 VDIVSD	%XMM5,%XMM0,%XMM0

0x428f75 VMINSD	%XMM3,%XMM1,%XMM1

0x428f79 VCOMISD	%XMM0,%XMM16

0x428f7f JBE	428e80

0x428f85 VMOVQ	%XMM22,%RDX

0x428f8b VDIVSD	%XMM0,%XMM21,%XMM10

0x428f91 VXORPD	%XMM20,%XMM10,%XMM2

0x428f97 VMULSD	(%RDX),%XMM2,%XMM3    [15]

0x428f9b JMP	428e8a

/home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/calc_dt_kernel.f90: 99 - 129

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

99:         cc=soundspeed(j,k)*soundspeed(j,k)

100:         cc=cc+2.0_8*viscosity_a(j,k)/density0(j,k)

101:         cc=MAX(SQRT(cc),g_small)

102:

103:         dtct=dtc_safe*MIN(dsx,dsy)/cc

104:

105:         div=0.0

106:

107:         dv1=(xvel0(j  ,k)+xvel0(j  ,k+1))*xarea(j  ,k)

108:         dv2=(xvel0(j+1,k)+xvel0(j+1,k+1))*xarea(j+1,k)

109:

110:         div=div+dv2-dv1

111:

112:         dtut=dtu_safe*2.0_8*volume(j,k)/MAX(ABS(dv1),ABS(dv2),g_small*volume(j,k))

113:

114:         dv1=(yvel0(j,k  )+yvel0(j+1,k  ))*yarea(j,k  )

115:         dv2=(yvel0(j,k+1)+yvel0(j+1,k+1))*yarea(j,k+1)

116:

117:         div=div+dv2-dv1

118:

119:         dtvt=dtv_safe*2.0_8*volume(j,k)/MAX(ABS(dv1),ABS(dv2),g_small*volume(j,k))

120:

121:         div=div/(2.0_8*volume(j,k))

122:

123:         IF(div.LT.-g_small)THEN

124:           dtdivt=dtdiv_safe*(-1.0_8/div)

125:         ELSE

126:           dtdivt=g_big

127:         ENDIF

128:

129:         dt_min_val=MIN(dt_min_val,dtct,dtut,dtvt,dtdivt)

Coverage (%)	Name	Source Location	Module
►95.66+	gomp_thread_start	team.c:130	libgomp.so.1.0.0
○	start_thread		libc.so.6
○	__GI___clone3		libc.so.6
►3.53+	GOMP_parallel	libgomp.h:982	libgomp.so.1.0.0
○	calc_dt_kernel	calc_dt_kernel.f90:140	exec
○	calc_dt	calc_dt.f90:81	exec
○	timestep	timestep.f90:87	exec
○	hydro	hydro.f90:54	exec
○	main	clover_leaf.f90:41	exec
○	__libc_start_call_main		libc.so.6
○	__libc_start_main		libc.so.6
○	_start	clover_leaf.f90:41	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	5.08
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.13
Bottlenecks
Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source	calc_dt_kernel.f90:99-129
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	30.50
CQA cycles if no scalar integer	30.50
CQA cycles if FP arith vectorized	6.00
CQA cycles if fully vectorized	3.81
Front-end cycles	6.94
P0 cycles	0.67
P1 cycles	0.67
P2 cycles	0.67
P3 cycles	0.83
P4 cycles	0.83
P5 cycles	0.83
P6 cycles	3.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	3.50
P10 cycles	9.50
P11 cycles	9.50
P12 cycles	9.75
P13 cycles	9.75
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	30.50
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	54.50
Nb uops	55.50
Nb loads	14.00
Nb stores	0.00
Nb stack references	1.00
FLOP/cycle	0.79
Nb FLOP add-sub	8.00
Nb FLOP mul	7.50
Nb FLOP fma	1.00
Nb FLOP div	5.50
Nb FLOP rcp	0.00
Nb FLOP sqrt	1.00
Nb FLOP rsqrt	0.00
Bytes/cycle	3.69
Bytes prefetched	0.00
Bytes loaded	112.00
Bytes stored	0.00
Stride 0	2.00
Stride 1	0.00
Stride n	0.00
Stride unknown	12.00
Stride indirect	0.00
Vectorization ratio all	15.29
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	38.42
Vector-efficiency ratio all	14.41
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	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	17.30

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	5.42
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.25
Bottlenecks	P10, P11,
Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source	calc_dt_kernel.f90:99-129
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	32.50
CQA cycles if no scalar integer	32.50
CQA cycles if FP arith vectorized	6.00
CQA cycles if fully vectorized	4.06
Front-end cycles	7.13
P0 cycles	0.67
P1 cycles	0.67
P2 cycles	0.67
P3 cycles	1.00
P4 cycles	1.00
P5 cycles	1.00
P6 cycles	3.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	3.50
P10 cycles	10.00
P11 cycles	10.00
P12 cycles	10.00
P13 cycles	10.00
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	32.50
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	56.00
Nb uops	57.00
Nb loads	14.00
Nb stores	0.00
Nb stack references	1.00
FLOP/cycle	0.77
Nb FLOP add-sub	8.00
Nb FLOP mul	8.00
Nb FLOP fma	1.00
Nb FLOP div	6.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	1.00
Nb FLOP rsqrt	0.00
Bytes/cycle	3.45
Bytes prefetched	0.00
Bytes loaded	112.00
Bytes stored	0.00
Stride 0	2.00
Stride 1	0.00
Stride n	0.00
Stride unknown	12.00
Stride indirect	0.00
Vectorization ratio all	16.00
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	40.00
Vector-efficiency ratio all	14.50
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	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	17.50

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	4.75
CQA speedup if fully vectorized	8.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	3.00
Bottlenecks	P10, P11,
Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source	calc_dt_kernel.f90:99-129
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	28.50
CQA cycles if no scalar integer	28.50
CQA cycles if FP arith vectorized	6.00
CQA cycles if fully vectorized	3.56
Front-end cycles	6.75
P0 cycles	0.67
P1 cycles	0.67
P2 cycles	0.67
P3 cycles	0.67
P4 cycles	0.67
P5 cycles	0.67
P6 cycles	3.50
P7 cycles	3.50
P8 cycles	3.50
P9 cycles	3.50
P10 cycles	9.00
P11 cycles	9.00
P12 cycles	9.50
P13 cycles	9.50
P14 cycles	1.00
P15 cycles	1.00
DIV/SQRT cycles	28.50
Inter-iter dependencies cycles	2
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	53.00
Nb uops	54.00
Nb loads	14.00
Nb stores	0.00
Nb stack references	1.00
FLOP/cycle	0.81
Nb FLOP add-sub	8.00
Nb FLOP mul	7.00
Nb FLOP fma	1.00
Nb FLOP div	5.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	1.00
Nb FLOP rsqrt	0.00
Bytes/cycle	3.93
Bytes prefetched	0.00
Bytes loaded	112.00
Bytes stored	0.00
Stride 0	2.00
Stride 1	0.00
Stride n	0.00
Stride unknown	12.00
Stride indirect	0.00
Vectorization ratio all	14.58
Vectorization ratio load	0.00
Vectorization ratio store	NA
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	36.84
Vector-efficiency ratio all	14.32
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	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	17.11

Path /

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

Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source file and lines	calc_dt_kernel.f90:99-129
Module	exec

The loop is defined in /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/calc_dt_kernel.f90:99-129.

The related source loop is not unrolled or unrolled with no peel/tail loop.
The structure of this loop is probably <if then [else] end>.

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 = max(0,x) (Fortran instrinsic procedure)

gain
potential
hint
expert

Vectorization

Your loop is probably not vectorized. Only 14% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 30.50 to 3.81 cycles (8.00x speedup).

Details

Store and arithmetical 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 ftree-vectorize (included in O3) to enable loop vectorization (including 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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do i a%x(i) = b%x(i) (fast, stride 1)

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 30.50 to 12.75 cycles (2.39x speedup).

FMA

Detected 1 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

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).

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

VCOMISD: 0 occurrences

Slow data structures access

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

Details

Constant unknown stride: 12 occurrence(s)

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

Workaround

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

Type of elements and instruction set

34 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 24 FP arithmetical operations:

9: addition or subtraction (1 inside FMA instructions)
8.50: multiply (1 inside FMA instructions)
5.50: divide
1: square root

The binary loop is loading 112 bytes (14 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.21 FP operations per loaded or stored byte.

General properties

nb instructions	54.50
nb uops	55.50
loop length	269.50
used x86 registers	15
used mmx registers	0
used xmm registers	21
used ymm registers	0
used zmm registers	0
nb stack references	1
ADD-SUB / MUL ratio	1.07

Front-end

MACRO FUSION NOT POSSIBLE

micro-operation queue	6.94 cycles
front end	6.94 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.67	0.67	0.67	0.83	0.83	0.83	3.50	3.50	3.50	3.50	9.50	9.50	9.75	9.75	1.00	1.00
cycles	0.67	0.67	0.67	0.83	0.83	0.83	3.50	3.50	3.50	3.50	9.50	9.50	9.75	9.75	1.00	1.00

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	30.50
Longest recurrence chain latency (RecMII)	2.00

Cycles summary

Front-end	6.94
Dispatch	9.75
DIV/SQRT	30.50
Data deps.	2.00
Overall L1	30.50

Vectorization ratios

INT

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)
other	0%

all	15%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	40%

INT+FP

all	15%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	38%

Vector efficiency ratios

INT

all	12%
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)
other	12%

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

INT+FP

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 30.50 cycles. At this rate:

1% of peak load performance is reached (3.69 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))

Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source file and lines	calc_dt_kernel.f90:99-129
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = max(0,x) (Fortran instrinsic procedure)

gain
potential
hint
expert

Vectorization

Details

Workaround

Try another compiler or update/tune your current one:
- recompile with ftree-vectorize (included in O3) to enable loop vectorization (including 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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do i a%x(i) = b%x(i) (fast, stride 1)

Execution units bottlenecks

Performance is limited by execution of divide and square root operations (the divide/square root unit is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 32.50 to 10.00 cycles (3.25x speedup).

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact. This will be done by your compiler with ffast-math or Ofast
Check whether you really need double precision. If not, switch to single precision to speedup execution

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 32.50 to 13.00 cycles (2.50x speedup).

FMA

Detected 1 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

VCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/calc_dt_kernel.f90:123

Slow data structures access

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

Details

Constant unknown stride: 12 occurrence(s)

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

Workaround

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

Type of elements and instruction set

35 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 25 FP arithmetical operations:

9: addition or subtraction (1 inside FMA instructions)
9: multiply (1 inside FMA instructions)
6: divide
1: square root

The binary loop is loading 112 bytes (14 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.22 FP operations per loaded or stored byte.

General properties

nb instructions	56
nb uops	57
loop length	278
used x86 registers	15
used mmx registers	0
used xmm registers	22
used ymm registers	0
used zmm registers	0
nb stack references	1
ADD-SUB / MUL ratio	1.00

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	7.13 cycles
front end	7.13 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.67	0.67	0.67	1.00	1.00	1.00	3.50	3.50	3.50	3.50	10.00	10.00	10.00	10.00	1.00	1.00
cycles	0.67	0.67	0.67	1.00	1.00	1.00	3.50	3.50	3.50	3.50	10.00	10.00	10.00	10.00	1.00	1.00

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	32.50
Longest recurrence chain latency (RecMII)	2.00

Cycles summary

Front-end	7.13
Dispatch	10.00
DIV/SQRT	32.50
Data deps.	2.00
Overall L1	32.50

Vectorization ratios

INT

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)
other	0%

all	16%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	42%

INT+FP

all	16%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	40%

Vector efficiency ratios

INT

all	12%
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)
other	12%

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

INT+FP

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 32.50 cycles. At this rate:

1% of peak load performance is reached (3.45 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: 428ea7

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMINSD %XMM7,%XMM3,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
LEA 0x1(%RAX),%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.25	N/A
VMINSD %XMM1,%XMM7,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMINSD %XMM1,%XMM12,%XMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
CMP %RAX,-0x38(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25	N/A
JE 428fa0 <__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0+0x500>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOV %RDX,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R13,%RAX,8),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VMOVSD (%R14,%RAX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VADDSD %XMM3,%XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD (%R12,%RAX,8),%XMM5,%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VFMADD132SD %XMM2,%XMM0,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	4	0.50	scal (12.5%)
VMOVAPD %XMM4,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VMOVSD (%RBX,%RAX,8),%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%RAX,8),%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VANDPD %XMM6,%XMM0,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD 0x8(%R9,%RAX,8),%XMM7,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM5,%XMM11,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMULSD 0x8(%RDI,%RAX,8),%XMM1,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM10,%XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VANDPD %XMM6,%XMM4,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VMULSD %XMM5,%XMM15,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMOVAPD %XMM8,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VDIVSD %XMM1,%XMM7,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMOVSD 0x8(%RSI,%RAX,8),%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VADDSD %XMM8,%XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD (%R11,%RAX,8),%XMM1,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMOVAPD %XMM9,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VSUBSD %XMM3,%XMM4,%XMM30	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMOVSD 0x8(%RCX,%RAX,8),%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VANDPD %XMM6,%XMM3,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD %XMM9,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD (%R10,%RAX,8),%XMM1,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VSUBSD %XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VANDPD %XMM6,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD %XMM30,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VSQRTSD %XMM2,%XMM2,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	21	8.50	scal (12.5%)
VMAXSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM5,%XMM14,%XMM30	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM10,%XMM1,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMINSD (%R15,%RAX,8),%XMM18,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD %XMM10,%XMM30,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMAXSD %XMM2,%XMM11,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM17,%XMM3,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VADDSD %XMM5,%XMM5,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD %XMM2,%XMM10,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VDIVSD %XMM5,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMINSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VCOMISD %XMM0,%XMM16	2	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0.50	0.50	6	0.50	scal (12.5%)
JBE 428e80 <__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0+0x3e0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A
VMOVQ %XMM22,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
VDIVSD %XMM0,%XMM21,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VXORPD %XMM20,%XMM10,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VMULSD (%RDX),%XMM2,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
JMP 428e8a <__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0+0x3ea>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	0	1	N/A

Function	__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0
Source file and lines	calc_dt_kernel.f90:99-129
Module	exec

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = max(0,x) (Fortran instrinsic procedure)

gain
potential
hint
expert

Vectorization

Details

Workaround

Try another compiler or update/tune your current one:
- recompile with ftree-vectorize (included in O3) to enable loop vectorization (including 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: Fortran storage order is column-major: do i do j a(i,j) = b(i,j) (slow, non stride 1) => do i do j a(j,i) = b(i,j) (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): do i a(i)%x = b(i)%x (slow, non stride 1) => do i a%x(i) = b%x(i) (fast, stride 1)

Execution units bottlenecks

Workaround

Reduce the number of division or square root instructions:
- If denominator is constant over iterations, use reciprocal (replace x/y with x*(1/y)). Check precision impact. This will be done by your compiler with ffast-math or Ofast
Check whether you really need double precision. If not, switch to single precision to speedup execution

Expensive FP math instructions/calls

Detected performance impact from expensive FP math instructions/calls. By removing/reexpressing them, you can lower the cost of an iteration from 28.50 to 12.50 cycles (2.28x speedup).

FMA

Detected 1 FMA (fused multiply-add) operations. Presence of both ADD/SUB and MUL operations.

Workaround

Complex instructions

Detected COMPLEX INSTRUCTIONS.

Details

These instructions generate more than one micro-operation and only one of them can be decoded during a cycle and the extra micro-operations increase pressure on execution units.

VCOMISD: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/174-049-0948/intel/CloverLeaf1.3-FC/build/CloverLeaf1.3-FC/CloverLeaf_ref/kernels/calc_dt_kernel.f90:123

Slow data structures access

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

Details

Constant unknown stride: 12 occurrence(s)

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

Workaround

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

Type of elements and instruction set

33 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 23 FP arithmetical operations:

9: addition or subtraction (1 inside FMA instructions)
8: multiply (1 inside FMA instructions)
5: divide
1: square root

The binary loop is loading 112 bytes (14 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.21 FP operations per loaded or stored byte.

General properties

nb instructions	53
nb uops	54
loop length	261
used x86 registers	15
used mmx registers	0
used xmm registers	20
used ymm registers	0
used zmm registers	0
nb stack references	1
ADD-SUB / MUL ratio	1.14

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	6.75 cycles
front end	6.75 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.67	0.67	0.67	0.67	0.67	0.67	3.50	3.50	3.50	3.50	9.00	9.00	9.50	9.50	1.00	1.00
cycles	0.67	0.67	0.67	0.67	0.67	0.67	3.50	3.50	3.50	3.50	9.00	9.00	9.50	9.50	1.00	1.00

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	28.50
Longest recurrence chain latency (RecMII)	2.00

Cycles summary

Front-end	6.75
Dispatch	9.50
DIV/SQRT	28.50
Data deps.	2.00
Overall L1	28.50

Vectorization ratios

INT

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)
other	0%

all	14%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	38%

INT+FP

all	14%
load	0%
store	NA (no store vectorizable/vectorized instructions)
mul	0%
add-sub	0%
fma	0%
div/sqrt	0%
other	36%

Vector efficiency ratios

INT

all	12%
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)
other	12%

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

INT+FP

all	14%
load	12%
store	NA (no store vectorizable/vectorized instructions)
mul	12%
add-sub	12%
fma	12%
div/sqrt	12%
other	17%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 28.50 cycles. At this rate:

1% of peak load performance is reached (3.93 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: 428ea7

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMOVQ %XMM23,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	1	0.50	scal (12.5%)
VMOVSD (%RDX),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VMINSD %XMM7,%XMM3,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
LEA 0x1(%RAX),%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.25	N/A
VMINSD %XMM1,%XMM7,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMINSD %XMM1,%XMM12,%XMM12	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
CMP %RAX,-0x38(%RBP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25	N/A
JE 428fa0 <__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0+0x500>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A
MOV %RDX,%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R13,%RAX,8),%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VMOVSD (%R14,%RAX,8),%XMM2	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VADDSD %XMM3,%XMM3,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD (%R12,%RAX,8),%XMM5,%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VFMADD132SD %XMM2,%XMM0,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	4	0.50	scal (12.5%)
VMOVAPD %XMM4,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VMOVSD (%RBX,%RAX,8),%XMM5	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%RAX,8),%XMM7	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VANDPD %XMM6,%XMM0,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD 0x8(%R9,%RAX,8),%XMM7,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM5,%XMM11,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMULSD 0x8(%RDI,%RAX,8),%XMM1,%XMM4	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM10,%XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VANDPD %XMM6,%XMM4,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VMULSD %XMM5,%XMM15,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMOVAPD %XMM8,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VDIVSD %XMM1,%XMM7,%XMM7	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMOVSD 0x8(%RSI,%RAX,8),%XMM8	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VADDSD %XMM8,%XMM3,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD (%R11,%RAX,8),%XMM1,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMOVAPD %XMM9,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	vect (25.0%)
VSUBSD %XMM3,%XMM4,%XMM30	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMOVSD 0x8(%RCX,%RAX,8),%XMM9	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	scal (12.5%)
VANDPD %XMM6,%XMM3,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD %XMM9,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD (%R10,%RAX,8),%XMM1,%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VSUBSD %XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VANDPD %XMM6,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	vect (25.0%)
VADDSD %XMM30,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VSQRTSD %XMM2,%XMM2,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	21	8.50	scal (12.5%)
VMAXSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM5,%XMM14,%XMM30	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VMAXSD %XMM10,%XMM1,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMINSD (%R15,%RAX,8),%XMM18,%XMM3	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD %XMM10,%XMM30,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMAXSD %XMM2,%XMM11,%XMM2	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VMULSD %XMM17,%XMM3,%XMM10	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	3	0.50	scal (12.5%)
VADDSD %XMM5,%XMM5,%XMM5	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VDIVSD %XMM2,%XMM10,%XMM3	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VDIVSD %XMM5,%XMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	0	13	4	scal (12.5%)
VMINSD %XMM3,%XMM1,%XMM1	1	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	2	0.50	scal (12.5%)
VCOMISD %XMM0,%XMM16	2	0	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0.50	0.50	6	0.50	scal (12.5%)
JBE 428e80 <__calc_dt_kernel_module_MOD_calc_dt_kernel._omp_fn.0+0x3e0>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	1	0.50-1	N/A

Report Configuration

Vectorization

Details

Workaround

Expensive FP math instructions/calls

FMA

Workaround

Complex instructions

Details

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

FMA

Workaround

Complex instructions

Details

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

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

Workaround

Expensive FP math instructions/calls

FMA

Workaround

Complex instructions

Details

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code