OV - exec - Function hypre_CSRMatrixMatvecOutOfPlace.extracted.24

	Function: hypre_CSRMatrixMatvecOutOfPlace.extracted.24	Module: exec	Source: csr_matvec.c:158-161	Coverage (incl. loops): 0.31% \| (excl. loops): 0.00%

/home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c: 158 - 161

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

158: #pragma omp parallel for private(i) HYPRE_SMP_SCHEDULE

159: #endif

160:             for (i = 0; i < num_rows*num_vectors; i++)

161:                y_data[i] = b_data[i]*temp;

0x4df540 PUSH	%RBP

0x4df541 MOV	%RSP,%RBP

0x4df544 PUSH	%R15

0x4df546 PUSH	%R14

0x4df548 PUSH	%R12

0x4df54a PUSH	%RBX

0x4df54b SUB	$0x20,%RSP

0x4df54f MOVL	$0,-0x24(%RBP)

0x4df556 MOV	%R8,%R12

0x4df559 MOV	%RCX,%R14

0x4df55c MOV	%RDX,%R15

0x4df55f MOV	(%RDI),%EBX

0x4df561 MOVQ	$0,-0x30(%RBP)

0x4df569 MOVQ	$0x1,-0x38(%RBP)

0x4df571 SUB	$0x8,%RSP

0x4df575 LEA	-0x38(%RBP),%RAX

0x4df579 LEA	-0x24(%RBP),%RCX

0x4df57d LEA	-0x30(%RBP),%R8

0x4df581 LEA	0x10(%RBP),%R9

0x4df585 MOV	$0x536dd0,%EDI

0x4df58a MOV	%EBX,%ESI

0x4df58c MOV	$0x22,%EDX

0x4df591 PUSH	$0x1

0x4df593 PUSH	$0x1

0x4df595 PUSH	%RAX

0x4df596 CALL	410420 <__kmpc_for_static_init_8@plt>

0x4df59b ADD	$0x20,%RSP

0x4df59f MOV	-0x30(%RBP),%RCX

0x4df5a3 MOV	0x10(%RBP),%RAX

0x4df5a7 CMP	%RAX,%RCX

0x4df5aa JA	4df664

0x4df5b0 VMOVQ	%R12,%XMM0

0x4df5b5 LEA	(%R15,%RAX,8),%RDX

0x4df5b9 LEA	(%R14,%RCX,8),%RSI

0x4df5bd LEA	(%R15,%RCX,8),%RDI

0x4df5c1 CMP	%RSI,%RDX

0x4df5c4 JB	4df5f6

0x4df5c6 LEA	(%R14,%RAX,8),%RDX

0x4df5ca CMP	%RDI,%RDX

0x4df5cd JB	4df5f6

0x4df5cf INC	%RAX

0x4df5d2 NOPW	%CS:(%RAX,%RAX,1)

(4330) 0x4df5e0 VMULSD	(%R15,%RCX,8),%XMM0,%XMM1

(4330) 0x4df5e6 VMOVSD	%XMM1,(%R14,%RCX,8)

(4330) 0x4df5ec INC	%RCX

(4330) 0x4df5ef CMP	%RCX,%RAX

(4330) 0x4df5f2 JNE	4df5e0

0x4df5f4 JMP	4df664

0x4df5f6 INC	%RAX

0x4df5f9 MOV	%RAX,%R8

0x4df5fc SUB	%RCX,%R8

0x4df5ff MOV	%R8,%RDX

0x4df602 AND	$-0x4,%RDX

0x4df606 JE	4df63c

0x4df608 LEA	-0x1(%RDX),%R9

0x4df60c VBROADCASTSD	%XMM0,%YMM1

0x4df611 XOR	%R10D,%R10D

0x4df614 NOPW	%CS:(%RAX,%RAX,1)

(4329) 0x4df620 VMULPD	(%RDI,%R10,8),%YMM1,%YMM2

(4329) 0x4df626 VMOVUPD	%YMM2,(%RSI,%R10,8)

(4329) 0x4df62c ADD	$0x4,%R10

(4329) 0x4df630 CMP	%R9,%R10

(4329) 0x4df633 JLE	4df620

0x4df635 CMP	%RDX,%R8

0x4df638 JNE	4df63e

0x4df63a JMP	4df664

0x4df63c XOR	%EDX,%EDX

0x4df63e ADD	%RCX,%RDX

0x4df641 NOPW	%CS:(%RAX,%RAX,1)

(4328) 0x4df650 VMULSD	(%R15,%RDX,8),%XMM0,%XMM1

(4328) 0x4df656 VMOVSD	%XMM1,(%R14,%RDX,8)

(4328) 0x4df65c INC	%RDX

(4328) 0x4df65f CMP	%RDX,%RAX

(4328) 0x4df662 JNE	4df650

0x4df664 MOV	$0x536df0,%EDI

0x4df669 MOV	%EBX,%ESI

0x4df66b ADD	$0x20,%RSP

0x4df66f POP	%RBX

0x4df670 POP	%R12

0x4df672 POP	%R14

0x4df674 POP	%R15

0x4df676 POP	%RBP

0x4df677 VZEROUPPER

0x4df67a JMP	410170

0x4df67f NOP

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

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 orig_0

Source file and lines	csr_matvec.c:158-161
Module	exec

The function is defined in /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158-161.

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 12.00 to 2.50 cycles (4.80x speedup).

Workaround

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

Vectorization

Your function is probably not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions).

Details

Store and arithmetical SSE/AVX instructions are used in scalar version (process only one data element in vector registers).

Workaround

Read the "512-bits vectorization" report at "Potential" confidence level.

Execution units bottlenecks

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

512-bits vectorization

On some x86 processors supporting 512-bits vectorization, compilers are often too conservative and limit vectorization to 256 bits. Performance can then be improved by enforcing 512-bits vectorization, especially with many vectorized and high trip count loops. 512-bits vectorization performance overhead (compared to 256-bits) is generally lower on newer processors.

Workaround

Recompile with -mprefer-vector-width=512

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.

__kmpc_for_static_init_8@plt: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158

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►
- /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158

Type of elements and instruction set

No instructions are processing arithmetic or math operations on FP elements. This function is probably writing/copying data or processing integer elements.

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

The binary function does not contain any FP arithmetical operations. The binary function is loading 20 bytes. The binary function is storing 20 bytes.

General properties

nb instructions	70
nb uops	72
loop length	259
used x86 registers	14
used mmx registers	0
used xmm registers	1
used ymm registers	1
used zmm registers	0
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	12.00 cycles
front end	12.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.40	2.67	2.67	6.00	3.40	3.30	6.00	6.00	6.00	3.40	2.67
cycles	3.50	3.40	2.67	2.67	6.00	3.40	3.30	6.00	6.00	6.00	3.40	2.67

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	12.00
Dispatch	6.00
Overall L1	12.00

Vectorization ratios

INT

all	6%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	9%

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

INT+FP

all	5%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	8%

Vector efficiency ratios

INT

all	10%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	11%

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	10%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	11%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 12.00 cycles. At this rate:

0% of peak load performance is reached (1.67 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
1% of peak store performance is reached (1.67 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 12.00 to 6.00 cycles (2.00x speedup).

ASM code

In the binary file, the address of the function is: 4df540

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 %R15	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %R14	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %R12	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	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 $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVL $0,-0x24(%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 %R8,%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
MOV %RCX,%R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RDX,%R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
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,-0x30(%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,-0x38(%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 -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x24(%RBP),%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x30(%RBP),%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x10(%RBP),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x536dd0,%EDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
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 410420 <__kmpc_for_static_init_8@plt>	2	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	0	1	N/A
ADD $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV -0x30(%RBP),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%RBP),%RAX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP %RAX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JA 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
VMOVQ %R12,%XMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	1	1	scal (12.5%)
LEA (%R15,%RAX,8),%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA (%R14,%RCX,8),%RSI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA (%R15,%RCX,8),%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RSI,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JB 4df5f6 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xb6>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R14,%RAX,8),%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDI,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JB 4df5f6 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xb6>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JMP 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RAX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
SUB %RCX,%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%)
MOV %R8,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
AND $-0x4,%RDX	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 4df63c <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xfc>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA -0x1(%RDX),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VBROADCASTSD %XMM0,%YMM1	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
XOR %R10D,%R10D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDX,%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%)
JNE 4df63e <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xfe>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
JMP 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
ADD %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x536df0,%EDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
ADD $0x20,%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 %R12	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %R14	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %R15	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 410170 <__kmpc_for_static_fini@plt>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A
NOP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run orig_0

Source file and lines	csr_matvec.c:158-161
Module	exec

The function is defined in /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158-161.

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 12.00 to 2.50 cycles (4.80x speedup).

Workaround

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

Vectorization

Your function is probably not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions).

Details

Store and arithmetical SSE/AVX instructions are used in scalar version (process only one data element in vector registers).

Workaround

Read the "512-bits vectorization" report at "Potential" confidence level.

Execution units bottlenecks

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

512-bits vectorization

Workaround

Recompile with -mprefer-vector-width=512

CALL instructions

Detected function call instructions.

Details

__kmpc_for_static_init_8@plt: 1 occurrences►
- /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158

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►
- /home/eoseret/qaas_runs_GNR/173-927-0874/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:158

Type of elements and instruction set

No instructions are processing arithmetic or math operations on FP elements. This function is probably writing/copying data or processing integer elements.

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

The binary function does not contain any FP arithmetical operations. The binary function is loading 20 bytes. The binary function is storing 20 bytes.

General properties

nb instructions	70
nb uops	72
loop length	259
used x86 registers	14
used mmx registers	0
used xmm registers	1
used ymm registers	1
used zmm registers	0
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	12.00 cycles
front end	12.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.40	2.67	2.67	6.00	3.40	3.30	6.00	6.00	6.00	3.40	2.67
cycles	3.50	3.40	2.67	2.67	6.00	3.40	3.30	6.00	6.00	6.00	3.40	2.67

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	12.00
Dispatch	6.00
Overall L1	12.00

Vectorization ratios

INT

all	6%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	9%

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	0%

INT+FP

all	5%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	8%

Vector efficiency ratios

INT

all	10%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	11%

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)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	12%

INT+FP

all	10%
load	6%
store	6%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	11%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 12.00 cycles. At this rate:

0% of peak load performance is reached (1.67 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
1% of peak store performance is reached (1.67 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: 4df540

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 %R15	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %R14	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	N/A
PUSH %R12	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	5-12	0.50	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 $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOVL $0,-0x24(%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 %R8,%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
MOV %RCX,%R14	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RDX,%R15	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
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,-0x30(%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,-0x38(%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 -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x24(%RBP),%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA -0x30(%RBP),%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x10(%RBP),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x536dd0,%EDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
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 410420 <__kmpc_for_static_init_8@plt>	2	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	0	1	N/A
ADD $0x20,%RSP	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV -0x30(%RBP),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%RBP),%RAX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP %RAX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JA 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
VMOVQ %R12,%XMM0	1	0	0	0	0	0	1	0	0	0	0	0	0	1	1	scal (12.5%)
LEA (%R15,%RAX,8),%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA (%R14,%RCX,8),%RSI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA (%R15,%RCX,8),%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RSI,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JB 4df5f6 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xb6>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R14,%RAX,8),%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDI,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JB 4df5f6 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xb6>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JMP 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RAX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
SUB %RCX,%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%)
MOV %R8,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
AND $-0x4,%RDX	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 4df63c <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xfc>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA -0x1(%RDX),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VBROADCASTSD %XMM0,%YMM1	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
XOR %R10D,%R10D	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
CMP %RDX,%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%)
JNE 4df63e <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0xfe>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
JMP 4df664 <hypre_CSRMatrixMatvecOutOfPlace.extracted.24+0x124>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	5.84	N/A
XOR %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
ADD %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
NOPW %CS:(%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV $0x536df0,%EDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
MOV %EBX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
ADD $0x20,%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 %R12	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %R14	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1-6	0.33	N/A
POP %R15	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 410170 <__kmpc_for_static_fini@plt>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A
NOP	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A

Name	Coverage (%)	Time (s)
▼hypre_CSRMatrixMatvecOutOfPlace.extracted.24–	0.31	0.17
○Loop 4330 - csr_matvec.c:158-161 - exec	0.31	0.13
○Loop 4328 - csr_matvec.c:158-161 - exec	0.00	0.00
○Loop 4329 - csr_matvec.c:158-161 - exec	0.00	0.00

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

512-bits vectorization

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

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

512-bits vectorization

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

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code