OV - exec - Loop 174

0x42a880 MOV	%RCX,%RAX

0x42a883 CQTO

0x42a885 IDIV	%RBX

0x42a888 CLTQ

0x42a88a MOV	%RDI,%R12

0x42a88d IMUL	%RAX,%R12

0x42a891 ADD	%RDX,%R12

0x42a894 MOVQ	$0,(%R8,%R12,8)    [1]

0x42a89c MOV	%R10,%R12

0x42a89f IMUL	%RAX,%R12

0x42a8a3 ADD	%RDX,%R12

0x42a8a6 MOVQ	$0,(%R9,%R12,8)    [2]

0x42a8ae IMUL	%R15,%RAX

0x42a8b2 ADD	%RDX,%RAX

0x42a8b5 MOVQ	$0,(%R11,%RAX,8)    [3]

0x42a8bd INC	%RCX

0x42a8c0 CMP	%RSI,%RCX

0x42a8c3 JG	42a7c8

0x42a8c9 MOV	%RCX,%RAX

0x42a8cc OR	%RBX,%RAX

0x42a8cf SHR	$0x20,%RAX

0x42a8d3 JNE	42a880

0x42a8d5 MOV	%ECX,%EAX

0x42a8d7 XOR	%EDX,%EDX

0x42a8d9 DIV	%EBX

0x42a8db JMP	42a888

/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp: 144 - 149

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

144:     for (int j = (0); j < (yrange + 1); j++) {

145:       for (int i = (0); i < (xrange); i++) {

146:         field.vol_flux_y(i, j) = 0.0;

147:         field.mass_flux_y(i, j) = 0.0;

148:         field.yarea(i, j) = 0.0;

149:       }

/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/context.h: 69 - 69

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

69:   T &operator()(size_t i, size_t j) const { return data[i + j * sizeX]; }

Coverage (%)	Name	Source Location	Module
►100.00+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so

Path /

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	2.69
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	NA
Bottlenecks
Function	build_field(global_variables&) [clone .extracted.22]
Source	build_field.cpp:144-149,context.h:69-69
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	8.75
CQA cycles if no scalar integer	8.75
CQA cycles if FP arith vectorized	8.75
CQA cycles if fully vectorized	3.25
Front-end cycles	4.33
P0 cycles	2.35
P1 cycles	7.50
P2 cycles	0.00
P3 cycles	0.00
P4 cycles	1.50
P5 cycles	2.40
P6 cycles	2.35
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	2.40
P11 cycles	0.00
DIV/SQRT cycles	8.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	22.50
Nb uops	26.00
Nb loads	0.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	2.80
Bytes prefetched	0.00
Bytes loaded	0.00
Bytes stored	24.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	3.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	8.33
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	6.25
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	9.38
Vector-efficiency ratio other	10.94

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	5.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	NA
Bottlenecks	P1,
Function	build_field(global_variables&) [clone .extracted.22]
Source	build_field.cpp:144-149,context.h:69-69
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	7.50
CQA cycles if no scalar integer	7.50
CQA cycles if FP arith vectorized	7.50
CQA cycles if fully vectorized	1.50
Front-end cycles	4.33
P0 cycles	2.30
P1 cycles	7.50
P2 cycles	0.00
P3 cycles	0.00
P4 cycles	1.50
P5 cycles	2.20
P6 cycles	2.30
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	2.20
P11 cycles	0.00
DIV/SQRT cycles	6.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	23.00
Nb uops	26.00
Nb loads	0.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	3.20
Bytes prefetched	0.00
Bytes loaded	0.00
Bytes stored	24.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	3.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	7.29
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	6.25
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	6.25
Vector-efficiency ratio other	9.38

Metric	Value
CQA speedup if no scalar integer	1.00
CQA speedup if FP arith vectorized	1.00
CQA speedup if fully vectorized	2.00
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.33
Bottlenecks	P0,
Function	build_field(global_variables&) [clone .extracted.22]
Source	build_field.cpp:144-149,context.h:69-69
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	10.00
CQA cycles if no scalar integer	10.00
CQA cycles if FP arith vectorized	10.00
CQA cycles if fully vectorized	5.00
Front-end cycles	4.33
P0 cycles	2.40
P1 cycles	7.50
P2 cycles	0.00
P3 cycles	0.00
P4 cycles	1.50
P5 cycles	2.60
P6 cycles	2.40
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	2.60
P11 cycles	0.00
DIV/SQRT cycles	10.00
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	22.00
Nb uops	26.00
Nb loads	0.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	0.00
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	2.40
Bytes prefetched	0.00
Bytes loaded	0.00
Bytes stored	24.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	3.00
Stride indirect	0.00
Vectorization ratio all	0.00
Vectorization ratio load	NA
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	NA
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	9.38
Vector-efficiency ratio load	NA
Vector-efficiency ratio store	6.25
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	12.50

Path /

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

Function	build_field(global_variables&) [clone .extracted.22]
Source file and lines	build_field.cpp:144-149
Module	exec

The loop is defined in:

/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144-149
/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/context.h:69

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 = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 8% 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 8.75 to 3.25 cycles (2.69x 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)

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 8.75 to 4.50 cycles (1.94x speedup).

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.

DIV: 0 occurrences

Slow data structures access

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

Details

Constant unknown stride: 3 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)

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 0 occurrences

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

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 storing 24 bytes.

General properties

nb instructions	22.50
nb uops	26
loop length	85
used x86 registers	12
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE

micro-operation queue	4.33 cycles
front end	4.33 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	2.35	6.00	0.00	0.00	1.50	2.40	2.35	1.50	1.50	1.50	2.40	0.00
cycles	2.35	7.50	0.00	0.00	1.50	2.40	2.35	1.50	1.50	1.50	2.40	0.00

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions	8.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.33
Dispatch	7.50
DIV/SQRT	8.00
Data deps.	1.00
Overall L1	8.75

Vectorization ratios

all	0%
load	NA (no load vectorizable/vectorized instructions)
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	0%
other	0%

Vector efficiency ratios

all	8%
load	NA (no load vectorizable/vectorized instructions)
store	6%
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	9%
other	10%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 8.75 cycles. At this rate:

2% of peak store performance is reached (2.80 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Function	build_field(global_variables&) [clone .extracted.22]
Source file and lines	build_field.cpp:144-149
Module	exec

The loop is defined in:

/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144-149
/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/context.h:69

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 = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 7% 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 7.50 to 1.50 cycles (5.00x 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.

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 7.50 to 4.50 cycles (1.67x speedup).

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.

DIV: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144

Slow data structures access

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

Details

Constant unknown stride: 3 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)

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:146

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

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 storing 24 bytes.

General properties

nb instructions	23
nb uops	26
loop length	85
used x86 registers	12
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.33 cycles
front end	4.33 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	2.30	6.00	0.00	0.00	1.50	2.20	2.30	1.50	1.50	1.50	2.20	0.00
cycles	2.30	7.50	0.00	0.00	1.50	2.20	2.30	1.50	1.50	1.50	2.20	0.00

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions	6.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.33
Dispatch	7.50
DIV/SQRT	6.00
Data deps.	1.00
Overall L1	7.50

Vectorization ratios

all	0%
load	NA (no load vectorizable/vectorized instructions)
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	0%
other	0%

Vector efficiency ratios

all	7%
load	NA (no load vectorizable/vectorized instructions)
store	6%
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	6%
other	9%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 7.50 cycles. At this rate:

2% of peak store performance is reached (3.20 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 42a8c9

Instruction	Nb FU	P0	P1	P4	P5	P6	P7	P8	P9	P10	Latency	Recip. throughput	Vectorization
CLTQ	1	0	0.50	0	0.50	0	0	0	0	0	1	0.33	scal (12.5%)
MOV %RDI,%R12	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
IMUL %RAX,%R12	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%R12	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R8,%R12,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
MOV %R10,%R12	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
IMUL %RAX,%R12	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%R12	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R9,%R12,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
IMUL %R15,%RAX	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%RAX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R11,%RAX,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
INC %RCX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%RCX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
JG 42a7c8 <_Z11build_fieldR16global_variables.extracted.22+0x78>	1	0.50	0	0	0	0.50	0	0	0	0	0	0.50	N/A
MOV %RCX,%RAX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
OR %RBX,%RAX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1-2	0.20	N/A
SHR $0x20,%RAX	1	0.50	0	0	0	0.50	0	0	0	0	0-2	0.50	N/A
JNE 42a880 <_Z11build_fieldR16global_variables.extracted.22+0x130>	1	0.50	0	0	0	0.50	0	0	0	0	0	0.50	N/A
MOV %ECX,%EAX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
DIV %EBX	4	0	3	0	0	0	0	0	0	0	11-16	6	scal (6.3%)
JMP 42a888 <_Z11build_fieldR16global_variables.extracted.22+0x138>	1	0	0	0	0	0	0	0	0	0	0	5.84	N/A

Function	build_field(global_variables&) [clone .extracted.22]
Source file and lines	build_field.cpp:144-149
Module	exec

The loop is defined in:

/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144-149
/home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/context.h:69

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 = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Vectorization

Your loop is not vectorized. Only 9% 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 10.00 to 5.00 cycles (2.00x 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

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 10.00 to 7.50 cycles (1.33x 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

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 10.00 to 4.50 cycles (2.22x speedup).

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.

IDIV: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144

Slow data structures access

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

Details

Constant unknown stride: 3 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)

Conversion instructions

Detected expensive conversion instructions.

Details

CLTQ: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:146
CQTO: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-814-2343/intel/CloverLeaf2.0-CXX/build/CloverLeaf2.0-CXX/src/omp/build_field.cpp:144

Workaround

Avoid mixing data with different types. In particular, check if the type of constants is the same as array elements.

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 storing 24 bytes.

General properties

nb instructions	22
nb uops	26
loop length	85
used x86 registers	12
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	4.33 cycles
front end	4.33 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	2.40	6.00	0.00	0.00	1.50	2.60	2.40	1.50	1.50	1.50	2.60	0.00
cycles	2.40	7.50	0.00	0.00	1.50	2.60	2.40	1.50	1.50	1.50	2.60	0.00

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions	10.00
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	4.33
Dispatch	7.50
DIV/SQRT	10.00
Data deps.	1.00
Overall L1	10.00

Vectorization ratios

all	0%
load	NA (no load vectorizable/vectorized instructions)
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	0%
other	0%

Vector efficiency ratios

all	9%
load	NA (no load vectorizable/vectorized instructions)
store	6%
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	12%
other	12%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.00 cycles. At this rate:

1% of peak store performance is reached (2.40 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Found no such bottlenecks.

ASM code

In the binary file, the address of the loop is: 42a8c9

Instruction	Nb FU	P0	P1	P4	P5	P6	P7	P8	P9	P10	Latency	Recip. throughput	Vectorization
MOV %RCX,%RAX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CQTO	1	0.50	0	0	0	0.50	0	0	0	0	1	0.50	scal (12.5%)
IDIV %RBX	5	0	3	0	0	0	0	0	0	0	11-16	10	scal (12.5%)
CLTQ	1	0	0.50	0	0.50	0	0	0	0	0	1	0.33	scal (12.5%)
MOV %RDI,%R12	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
IMUL %RAX,%R12	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%R12	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R8,%R12,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
MOV %R10,%R12	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
IMUL %RAX,%R12	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%R12	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R9,%R12,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
IMUL %R15,%RAX	1	0	1	0	0	0	0	0	0	0	3	1	N/A
ADD %RDX,%RAX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
MOVQ $0,(%R11,%RAX,8)	1	0	0	0.50	0	0	0.50	0.50	0.50	0	1	0.50	scal (6.3%)
INC %RCX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %RSI,%RCX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1	0.20	N/A
JG 42a7c8 <_Z11build_fieldR16global_variables.extracted.22+0x78>	1	0.50	0	0	0	0.50	0	0	0	0	0	0.50	N/A
MOV %RCX,%RAX	1	0	0	0	0	0	0	0	0	0	1	0.17	N/A
OR %RBX,%RAX	1	0.20	0.20	0	0.20	0.20	0	0	0	0.20	1-2	0.20	N/A
SHR $0x20,%RAX	1	0.50	0	0	0	0.50	0	0	0	0	0-2	0.50	N/A
JNE 42a880 <_Z11build_fieldR16global_variables.extracted.22+0x130>	1	0.50	0	0	0	0.50	0	0	0	0	0	0.50	N/A

Report Configuration

Vectorization

Details

Workaround

Expensive FP math instructions/calls

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion instructions

Details

Workaround

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

Vectorization

Details

Workaround

Execution units bottlenecks

Expensive FP math instructions/calls

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion instructions

Details

Workaround

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

Vectorization

Details

Workaround

Execution units bottlenecks

Workaround

Expensive FP math instructions/calls

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion instructions

Details

Workaround

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