OV - exec - Function main.extracted.50

	Function: main.extracted.50	Module: exec	Source: stream.c:342-344	Coverage (incl. loops): 91.81% \| (excl. loops): 0.00%

/beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c: 342 - 344

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

342: #pragma omp parallel for

343: 	for (j=0; j<STREAM_ARRAY_SIZE; j++)

344: 	    c[j] = a[j]+b[j];

0x402ad0 PUSH	%RBP

0x402ad1 MOV	%RSP,%RBP

0x402ad4 PUSH	%RBX

0x402ad5 SUB	$0x28,%RSP

0x402ad9 MOVL	$0,-0xc(%RBP)

0x402ae0 MOV	(%RDI),%EBX

0x402ae2 MOVQ	$0,-0x20(%RBP)

0x402aea MOVQ	$0x3344ffff,-0x18(%RBP)

0x402af2 MOVQ	$0x1,-0x28(%RBP)

0x402afa SUB	$0x8,%RSP

0x402afe LEA	-0x28(%RBP),%RAX

0x402b02 MOV	$0x4ceb87e20,%RDI

0x402b0c MOV	$0x401050,%R10

0x402b16 LEA	-0xc(%RBP),%RCX

0x402b1a LEA	-0x20(%RBP),%R8

0x402b1e LEA	-0x18(%RBP),%R9

0x402b22 MOV	%EBX,%ESI

0x402b24 MOV	$0x22,%EDX

0x402b29 PUSH	$0x1

0x402b2b PUSH	$0x1

0x402b2d PUSH	%RAX

0x402b2e CALL	%R10

0x402b31 ADD	$0x20,%RSP

0x402b35 MOV	-0x20(%RBP),%RAX

0x402b39 MOV	-0x18(%RBP),%R8

0x402b3d SUB	%RAX,%R8

0x402b40 JB	402bf2

0x402b46 INC	%R8

0x402b49 MOV	%R8,%RDI

0x402b4c MOV	$0x19a685060,%RDX

0x402b56 MOV	$0x405060,%RSI

0x402b60 MOV	$0x334905060,%RCX

0x402b6a AND	$0x7ffffff8,%RDI

0x402b71 JE	402baf

0x402b73 LEA	(,%RAX,8),%R9

0x402b7b XOR	%R10D,%R10D

0x402b7e XCHG	%AX,%AX

(12) 0x402b80 LEA	(%R9,%R10,8),%R11

(12) 0x402b84 VMOVUPD	(%RDX,%R11,1),%ZMM0

(12) 0x402b8b VADDPD	(%RSI,%R11,1),%ZMM0,%ZMM0

(12) 0x402b92 VMOVUPD	%ZMM0,(%RCX,%R11,1)

(12) 0x402b99 ADD	$0x8,%R10

(12) 0x402b9d CMP	%RDI,%R10

(12) 0x402ba0 JB	402b80

0x402ba2 CMP	%RDI,%R8

0x402ba5 JE	402bf2

0x402ba7 VPBROADCASTQ	%R8,%ZMM0

0x402bad JMP	402bb7

0x402baf VPBROADCASTQ	%R8,%ZMM0

0x402bb5 XOR	%EDI,%EDI

0x402bb7 VPBROADCASTQ	%RDI,%ZMM1

0x402bbd VPSUBQ	%ZMM1,%ZMM0,%ZMM0

0x402bc3 MOV	$0x403180,%R8

0x402bcd VPCMPNLEUQ	(%R8),%ZMM0,%K1

0x402bd4 ADD	%RDI,%RAX

0x402bd7 VMOVUPD	(%RDX,%RAX,8),%ZMM0{%K1}{z}

0x402bde VMOVUPD	(%RSI,%RAX,8),%ZMM1{%K1}{z}

0x402be5 VADDPD	%ZMM1,%ZMM0,%ZMM0

0x402beb VMOVUPD	%ZMM0,(%RCX,%RAX,8){%K1}

0x402bf2 MOV	$0x4ceb87e40,%RDI

0x402bfc MOV	$0x401030,%RAX

0x402c06 MOV	%EBX,%ESI

0x402c08 ADD	$0x28,%RSP

0x402c0c POP	%RBX

0x402c0d POP	%RBP

0x402c0e VZEROUPPER

0x402c11 JMP	%RAX

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

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 /

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

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run icx_7

Source file and lines	stream.c:342-344
Module	exec

The function is defined in /beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c:342-344.

Warnings:
Ignoring paths for analysis

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 10.17 to 3.50 cycles (2.90x speedup).

Workaround

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

Vectorization

Your function is poorly vectorized. Only 40% of vector register length is used (average across all SSE/AVX instructions). By fully vectorizing your function, you can lower the cost of an iteration from 10.17 to 8.65 cycles (1.18x speedup).

Details

38% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

75% of SSE/AVX loads are used in vector version.
20% of SSE/AVX stores are used in vector version.
25% of SSE/AVX instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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.

Masked instructions

Detected masked instructions.

Details

Vector registers are partially exploited, which is expected if your function is irregular or mixes elements of different sizes.

Workaround

If your function is irregular, try to remove or hoist conditional structures out of your function. If it mixes elements of different sizes, try to uniformize them.

CALL instructions

Detected function call instructions.

Details

Calling (and then returning from) a function prevents many compiler optimizations (like vectorization), breaks control flow (which reduces pipeline performance) and executes extra instructions to save/restore the registers used inside it, which is very expensive (dozens of cycles). Consider to inline small functions.

unknown: 1 occurrences►
- /beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c:342

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.

VZEROUPPER: 1 occurrences►
- /beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c:342

Type of elements and instruction set

1 AVX-512 instructions are processing arithmetic or math operations on double precision FP elements in vector mode (eight at a time).

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

The binary function is composed of 8 FP arithmetical operations:

8: addition or subtraction

The binary function is loading 212 bytes (26 double precision FP elements). The binary function is storing 92 bytes (11 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	60
nb uops	61
loop length	289
used x86 registers	11
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	2
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	10.17 cycles
front end	10.17 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	4.70	4.60	2.33	2.33	5.50	4.50	4.60	5.50	5.50	5.50	4.60	2.33
cycles	4.70	4.60	2.33	2.33	5.50	4.50	4.60	5.50	5.50	5.50	4.60	2.33

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

Cycles summary

Front-end	10.17
Dispatch	5.50
Overall L1	10.17

Vectorization ratios

INT

all	21%
load	50%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
other	25%

all	100%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	38%
load	75%
store	20%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	25%

Vector efficiency ratios

INT

all	23%
load	53%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
other	24%

all	100%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	40%
load	76%
store	25%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	24%

Cycles and memory resources usage

Detected masked instructions: assuming all mask elements are active. Assuming all data fit into the L1 cache, each call to the function takes 10.17 cycles. At this rate:

10% of peak load performance is reached (20.85 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
7% of peak store performance is reached (9.05 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 10.17 to 5.50 cycles (1.85x speedup).

ASM code

In the binary file, the address of the function is: 402ad0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
PUSH %RBX	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
SUB $0x28,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVL $0,-0xc(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOV (%RDI),%EBX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (6.3%)
MOVQ $0,-0x20(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOVQ $0x3344ffff,-0x18(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOVQ $0x1,-0x28(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
SUB $0x8,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
LEA -0x28(%RBP),%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x4ceb87e20,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x401050,%R10	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	scal (12.5%)
LEA -0xc(%RBP),%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x20(%RBP),%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x18(%RBP),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV $0x22,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
PUSH $0x1	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH $0x1	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %RAX	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
CALL %R10	2	0.50	0	0	0	0.50	0	0.50	0.50	0.50	0.50	0	0	0	2.14	N/A
ADD $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV -0x20(%RBP),%RAX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV -0x18(%RBP),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SUB %RAX,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
JB 402bf2 <main.extracted.50+0x122>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
INC %R8	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %R8,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV $0x19a685060,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x405060,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x334905060,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
AND $0x7ffffff8,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
JE 402baf <main.extracted.50+0xdf>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (,%RAX,8),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
XOR %R10D,%R10D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
XCHG %AX,%AX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDI,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 402bf2 <main.extracted.50+0x122>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
VPBROADCASTQ %R8,%ZMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
JMP 402bb7 <main.extracted.50+0xe7>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
VPBROADCASTQ %R8,%ZMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VPBROADCASTQ %RDI,%ZMM1	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
VPSUBQ %ZMM1,%ZMM0,%ZMM0	1	0.50	0	0	0	0	0.50	0	0	0	0	0	0	0-1	0.50	vect (100.0%)
MOV $0x403180,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
VPCMPNLEUQ (%R8),%ZMM0,%K1																vect (100.0%)
ADD %RDI,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
VMOVUPD (%RDX,%RAX,8),%ZMM0{%K1}{z}	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	0-1	0.50	vect (100.0%)
VMOVUPD (%RSI,%RAX,8),%ZMM1{%K1}{z}	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	0-1	0.50	vect (100.0%)
VADDPD %ZMM1,%ZMM0,%ZMM0	1	0.50	0	0	0	0	0.50	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD %ZMM0,(%RCX,%RAX,8){%K1}	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	0-1	1	vect (100.0%)
MOV $0x4ceb87e40,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x401030,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x28,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
POP %RBX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %RBP	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
VZEROUPPER	2	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
JMP %RAX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	1.75	N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run icx_7

Source file and lines	stream.c:342-344
Module	exec

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 10.17 to 3.50 cycles (2.90x speedup).

Workaround

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

Vectorization

Details

38% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

75% of SSE/AVX loads are used in vector version.
20% of SSE/AVX stores are used in vector version.
25% of SSE/AVX instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Since your execution units are vector units, only a fully vectorized function can use their full power.

Workaround

Try another compiler or update/tune your current one
Make array accesses unit-stride:
- If your function 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.

Masked instructions

Detected masked instructions.

Details

Vector registers are partially exploited, which is expected if your function is irregular or mixes elements of different sizes.

Workaround

If your function is irregular, try to remove or hoist conditional structures out of your function. If it mixes elements of different sizes, try to uniformize them.

CALL instructions

Detected function call instructions.

Details

unknown: 1 occurrences►
- /beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c:342

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.

VZEROUPPER: 1 occurrences►
- /beegfs/hackathon/users/eoseret/qaas_runs_test/isix03.benchmarkcenter.megware.com/177-373-7190/stream-add/build/stream-add/src/stream.c:342

Type of elements and instruction set

1 AVX-512 instructions are processing arithmetic or math operations on double precision FP elements in vector mode (eight at a time).

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

The binary function is composed of 8 FP arithmetical operations:

8: addition or subtraction

The binary function is loading 212 bytes (26 double precision FP elements). The binary function is storing 92 bytes (11 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.03 FP operations per loaded or stored byte.

General properties

nb instructions	60
nb uops	61
loop length	289
used x86 registers	11
used mmx registers	0
used xmm registers	0
used ymm registers	0
used zmm registers	2
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	10.17 cycles
front end	10.17 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	4.70	4.60	2.33	2.33	5.50	4.50	4.60	5.50	5.50	5.50	4.60	2.33
cycles	4.70	4.60	2.33	2.33	5.50	4.50	4.60	5.50	5.50	5.50	4.60	2.33

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

Cycles summary

Front-end	10.17
Dispatch	5.50
Overall L1	10.17

Vectorization ratios

INT

all	21%
load	50%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
other	25%

all	100%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	38%
load	75%
store	20%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	25%

Vector efficiency ratios

INT

all	23%
load	53%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
other	24%

all	100%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	40%
load	76%
store	25%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	24%

Cycles and memory resources usage

Detected masked instructions: assuming all mask elements are active. Assuming all data fit into the L1 cache, each call to the function takes 10.17 cycles. At this rate:

10% of peak load performance is reached (20.85 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
7% of peak store performance is reached (9.05 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the function is: 402ad0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
PUSH %RBX	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
SUB $0x28,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVL $0,-0xc(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOV (%RDI),%EBX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (6.3%)
MOVQ $0,-0x20(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOVQ $0x3344ffff,-0x18(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
MOVQ $0x1,-0x28(%RBP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (6.3%)
SUB $0x8,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
LEA -0x28(%RBP),%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x4ceb87e20,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x401050,%R10	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	scal (12.5%)
LEA -0xc(%RBP),%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x20(%RBP),%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x18(%RBP),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV $0x22,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
PUSH $0x1	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH $0x1	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %RAX	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
CALL %R10	2	0.50	0	0	0	0.50	0	0.50	0.50	0.50	0.50	0	0	0	2.14	N/A
ADD $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV -0x20(%RBP),%RAX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV -0x18(%RBP),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SUB %RAX,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
JB 402bf2 <main.extracted.50+0x122>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
INC %R8	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %R8,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV $0x19a685060,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x405060,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x334905060,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
AND $0x7ffffff8,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
JE 402baf <main.extracted.50+0xdf>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (,%RAX,8),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
XOR %R10D,%R10D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
XCHG %AX,%AX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDI,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 402bf2 <main.extracted.50+0x122>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
VPBROADCASTQ %R8,%ZMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
JMP 402bb7 <main.extracted.50+0xe7>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
VPBROADCASTQ %R8,%ZMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VPBROADCASTQ %RDI,%ZMM1	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
VPSUBQ %ZMM1,%ZMM0,%ZMM0	1	0.50	0	0	0	0	0.50	0	0	0	0	0	0	0-1	0.50	vect (100.0%)
MOV $0x403180,%R8	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
VPCMPNLEUQ (%R8),%ZMM0,%K1																vect (100.0%)
ADD %RDI,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
VMOVUPD (%RDX,%RAX,8),%ZMM0{%K1}{z}	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	0-1	0.50	vect (100.0%)
VMOVUPD (%RSI,%RAX,8),%ZMM1{%K1}{z}	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	0-1	0.50	vect (100.0%)
VADDPD %ZMM1,%ZMM0,%ZMM0	1	0.50	0	0	0	0	0.50	0	0	0	0	0	0	3	0.50	vect (100.0%)
VMOVUPD %ZMM0,(%RCX,%RAX,8){%K1}	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	0-1	1	vect (100.0%)
MOV $0x4ceb87e40,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV $0x401030,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.28	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x28,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
POP %RBX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %RBP	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
VZEROUPPER	2	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
JMP %RAX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	1.75	N/A

Name	Coverage (%)	Time (s)
▼main.extracted.50–	91.81	4.27
○Loop 12 - stream.c:343-344 - exec	91.81	4.17

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Masked instructions

Details

Workaround

CALL instructions

Details

Complex instructions

Details

Type of elements and instruction set

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

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

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

Masked instructions

Details

Workaround

CALL instructions

Details

Complex instructions

Details

Type of elements and instruction set

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

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