OV - exec - Function hypre_CSRMatrixMatvecOutOfPlace.omp

	Function: hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13	Module: exec	Source: csr_matvec.c:234-237	Coverage (incl. loops): 0.39% \| (excl. loops): 0.00%

/home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c: 234 - 237

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

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

235: #endif

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

237:             y_data[i] *= alpha;

0x2f7e50 PUSH	%RBP

0x2f7e51 MOV	%RSP,%RBP

0x2f7e54 PUSH	%R15

0x2f7e56 PUSH	%R14

0x2f7e58 PUSH	%R12

0x2f7e5a PUSH	%RBX

0x2f7e5b SUB	$0x20,%RSP

0x2f7e5f MOV	(%RCX),%R12

0x2f7e62 IMUL	(%RDX),%R12

0x2f7e66 TEST	%R12,%R12

0x2f7e69 JLE	2f7fea

0x2f7e6f MOV	(%RDI),%R14D

0x2f7e72 DEC	%R12

0x2f7e75 MOV	%R9,%RBX

0x2f7e78 MOV	%R8,%R15

0x2f7e7b MOVQ	$0,-0x38(%RBP)

0x2f7e83 MOVQ	$0x1,-0x40(%RBP)

0x2f7e8b MOVL	$0,-0x2c(%RBP)

0x2f7e92 MOV	%R12,-0x28(%RBP)

0x2f7e96 SUB	$0x8,%RSP

0x2f7e9a LEA	-0x40(%RBP),%RAX

0x2f7e9e LEA	0x13893(%RIP),%RDI

0x2f7ea5 LEA	-0x2c(%RBP),%RCX

0x2f7ea9 LEA	-0x38(%RBP),%R8

0x2f7ead LEA	-0x28(%RBP),%R9

0x2f7eb1 MOV	%R14D,%ESI

0x2f7eb4 MOV	$0x22,%EDX

0x2f7eb9 PUSH	$0x1

0x2f7ebb PUSH	$0x1

0x2f7ebd PUSH	%RAX

0x2f7ebe CALL	305fa0 <@plt_start@+0x90>

0x2f7ec3 ADD	$0x20,%RSP

0x2f7ec7 MOV	-0x28(%RBP),%RAX

0x2f7ecb CMP	%R12,%RAX

0x2f7ece CMOVL	%RAX,%R12

0x2f7ed2 MOV	-0x38(%RBP),%RAX

0x2f7ed6 MOV	%R12,-0x28(%RBP)

0x2f7eda CMP	%R12,%RAX

0x2f7edd JG	2f7fd8

0x2f7ee3 MOV	(%R15),%RCX

0x2f7ee6 MOV	%R12,%RDX

0x2f7ee9 SUB	%RAX,%RDX

0x2f7eec INC	%RDX

0x2f7eef CMP	$0x4,%RDX

0x2f7ef3 JB	2f7fbd

0x2f7ef9 LEA	(%RCX,%RAX,8),%RSI

0x2f7efd LEA	0x8(%RBX),%RDI

0x2f7f01 CMP	%RDI,%RSI

0x2f7f04 JAE	2f7f14

0x2f7f06 LEA	0x8(%RCX,%R12,8),%RDI

0x2f7f0b CMP	%RBX,%RDI

0x2f7f0e JA	2f7fbd

0x2f7f14 CMP	$0x20,%RDX

0x2f7f18 JAE	2f7f1e

0x2f7f1a XOR	%EDI,%EDI

0x2f7f1c JMP	2f7f91

0x2f7f1e VBROADCASTSD	(%RBX),%ZMM0

0x2f7f24 LEA	0xc0(%RCX,%RAX,8),%R8

0x2f7f2c MOV	%RDX,%RDI

0x2f7f2f AND	$-0x20,%RDI

0x2f7f33 XOR	%R9D,%R9D

0x2f7f36 NOPW	%CS:(%RAX,%RAX,1)

(3598) 0x2f7f40 VMULPD	-0xc0(%R8,%R9,8),%ZMM0,%ZMM1

(3598) 0x2f7f48 VMULPD	-0x80(%R8,%R9,8),%ZMM0,%ZMM2

(3598) 0x2f7f50 VMULPD	-0x40(%R8,%R9,8),%ZMM0,%ZMM3

(3598) 0x2f7f58 VMULPD	(%R8,%R9,8),%ZMM0,%ZMM4

(3598) 0x2f7f5f VMOVUPD	%ZMM1,-0xc0(%R8,%R9,8)

(3598) 0x2f7f67 VMOVUPD	%ZMM2,-0x80(%R8,%R9,8)

(3598) 0x2f7f6f VMOVUPD	%ZMM3,-0x40(%R8,%R9,8)

(3598) 0x2f7f77 VMOVUPD	%ZMM4,(%R8,%R9,8)

(3598) 0x2f7f7e ADD	$0x20,%R9

(3598) 0x2f7f82 CMP	%R9,%RDI

(3598) 0x2f7f85 JNE	2f7f40

0x2f7f87 CMP	%RDI,%RDX

0x2f7f8a JE	2f7fd8

0x2f7f8c TEST	$0x1c,%DL

0x2f7f8f JE	2f7fba

0x2f7f91 VBROADCASTSD	(%RBX),%YMM0

0x2f7f96 MOV	%RDX,%R8

0x2f7f99 AND	$-0x4,%R8

0x2f7f9d ADD	%R8,%RAX

(3599) 0x2f7fa0 VMULPD	(%RSI,%RDI,8),%YMM0,%YMM1

(3599) 0x2f7fa5 VMOVUPD	%YMM1,(%RSI,%RDI,8)

(3599) 0x2f7faa ADD	$0x4,%RDI

(3599) 0x2f7fae CMP	%RDI,%R8

(3599) 0x2f7fb1 JNE	2f7fa0

0x2f7fb3 CMP	%R8,%RDX

0x2f7fb6 JNE	2f7fbd

0x2f7fb8 JMP	2f7fd8

0x2f7fba ADD	%RDI,%RAX

0x2f7fbd DEC	%RAX

(3597) 0x2f7fc0 VMOVSD	0x8(%RCX,%RAX,8),%XMM0

(3597) 0x2f7fc6 VMULSD	(%RBX),%XMM0,%XMM0

(3597) 0x2f7fca VMOVSD	%XMM0,0x8(%RCX,%RAX,8)

(3597) 0x2f7fd0 INC	%RAX

(3597) 0x2f7fd3 CMP	%R12,%RAX

(3597) 0x2f7fd6 JL	2f7fc0

0x2f7fd8 LEA	0x13771(%RIP),%RDI

0x2f7fdf MOV	%R14D,%ESI

0x2f7fe2 VZEROUPPER

0x2f7fe5 CALL	305fb0 <@plt_start@+0xa0>

0x2f7fea ADD	$0x20,%RSP

0x2f7fee POP	%RBX

0x2f7fef POP	%R12

0x2f7ff1 POP	%R14

0x2f7ff3 POP	%R15

0x2f7ff5 POP	%RBP

0x2f7ff6 RET

0x2f7ff7 NOPW	(%RAX,%RAX,1)

Coverage (%)	Name	Source Location	Module

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 aocc_2

Source file and lines	csr_matvec.c:234-237
Module	exec

The function is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234-237.

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 11.25 to 3.00 cycles (3.75x 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 11% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your function, you can lower the cost of an iteration from 11.25 to 0.84 cycles (13.38x 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 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.

No data for this section

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: 2 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234

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.

CMOVL: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236
LEA: 2 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236

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 60 bytes. The binary function is storing 36 bytes.

General properties

nb instructions	87
nb uops	90
loop length	318
used x86 registers	13
used mmx registers	0
used xmm registers	0
used ymm registers	1
used zmm registers	1
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	11.25 cycles
front end	11.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	8.33	8.33	8.33	8.33	8.33	8.33	4.25	4.25	4.25	4.25	0.00	0.00	0.00	0.00	0.00	0.00
cycles	8.33	8.33	8.33	8.33	8.33	8.33	4.25	4.25	4.25	4.25	0.00	0.00	0.00	0.00	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions

Cycles summary

Front-end	11.25
Dispatch	8.33
Overall L1	11.25

Vectorization ratios

INT

all	4%
load	0%
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)
other	6%

all	0%
load	0%
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	4%
load	0%
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	NA (no div/sqrt vectorizable/vectorized instructions)
other	5%

Vector efficiency ratios

INT

all	11%
load	6%
store	8%
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	12%
load	12%
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	11%
load	10%
store	8%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 11.25 cycles. At this rate:

2% of peak load performance is reached (5.33 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.20 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 11.25 to 8.33 cycles (1.35x speedup).

ASM code

In the binary file, the address of the function is: 2f7e50

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
PUSH %R15	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R14	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R12	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SUB $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV (%RCX),%R12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
IMUL (%RDX),%R12	1	0.33	0.33	0.33	0	0	0	0.25	0.25	0.25	0.25	3	0.33	N/A
TEST %R12,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JLE 2f7fea <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x19a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV (%RDI),%R14D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (6.3%)
DEC %R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV %R9,%RBX	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R8,%R15	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOVQ $0,-0x38(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOVQ $0x1,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOVL $0,-0x2c(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOV %R12,-0x28(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
SUB $0x8,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
LEA -0x40(%RBP),%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA 0x13893(%RIP),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x2c(%RBP),%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x38(%RBP),%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x28(%RBP),%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
MOV %R14D,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
MOV $0x22,%EDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
PUSH $0x1	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH $0x1	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RAX	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
CALL 305fa0 <@plt_start@+0x90>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0.50	N/A
ADD $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV -0x28(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
CMP %R12,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
CMOVL %RAX,%R12	2	0.33	0.33	0.33	0.33	0.33	0.33	0	0	0	0	1	0.33	N/A
MOV -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %R12,-0x28(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
CMP %R12,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JG 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV (%R15),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %R12,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
SUB %RAX,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
INC %RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP $0x4,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JB 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
LEA (%RCX,%RAX,8),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA 0x8(%RBX),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
CMP %RDI,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JAE 2f7f14 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0xc4>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
LEA 0x8(%RCX,%R12,8),%RDI	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	2	0.50	N/A
CMP %RBX,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JA 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
CMP $0x20,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JAE 2f7f1e <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0xce>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
JMP 2f7f91 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x141>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
VBROADCASTSD (%RBX),%ZMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
LEA 0xc0(%RCX,%RAX,8),%R8	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	2	0.50	N/A
MOV %RDX,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
AND $-0x20,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
XOR %R9D,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0.25	scal (6.3%)
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
CMP %RDI,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JE 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
TEST $0x1c,%DL	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
JE 2f7fba <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
VBROADCASTSD (%RBX),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV %RDX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (12.5%)
AND $-0x4,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
ADD %R8,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP %R8,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JNE 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
JMP 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
ADD %RDI,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
DEC %RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
LEA 0x13771(%RIP),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
MOV %R14D,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 305fb0 <@plt_start@+0xa0>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0.50	N/A
ADD $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R12	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R14	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R15	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
RET	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A

The code analyzed by CQA in that panel excludes loops and represents 0.00% of application time for run aocc_2

Source file and lines	csr_matvec.c:234-237
Module	exec

The function is defined in /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234-237.

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 11.25 to 3.00 cycles (3.75x speedup).

Workaround

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

Vectorization

Details

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.

No data for this section

CALL instructions

Detected function call instructions.

Details

unknown: 2 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:234

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.

CMOVL: 1 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236
LEA: 2 occurrences►
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236
- /home/eoseret/qaas_runs_ZEN5/173-989-6379/intel/AMG/build/AMG/AMG/seq_mv/csr_matvec.c:236

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 60 bytes. The binary function is storing 36 bytes.

General properties

nb instructions	87
nb uops	90
loop length	318
used x86 registers	13
used mmx registers	0
used xmm registers	0
used ymm registers	1
used zmm registers	1
nb stack references	4

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	11.25 cycles
front end	11.25 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	8.33	8.33	8.33	8.33	8.33	8.33	4.25	4.25	4.25	4.25	0.00	0.00	0.00	0.00	0.00	0.00
cycles	8.33	8.33	8.33	8.33	8.33	8.33	4.25	4.25	4.25	4.25	0.00	0.00	0.00	0.00	0.00	0.00

Execution ports to units layout:

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

Cycles executing div or sqrt instructions

Cycles summary

Front-end	11.25
Dispatch	8.33
Overall L1	11.25

Vectorization ratios

INT

all	4%
load	0%
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)
other	6%

all	0%
load	0%
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	4%
load	0%
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	NA (no div/sqrt vectorizable/vectorized instructions)
other	5%

Vector efficiency ratios

INT

all	11%
load	6%
store	8%
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	12%
load	12%
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	11%
load	10%
store	8%
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%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each call to the function takes 11.25 cycles. At this rate:

2% of peak load performance is reached (5.33 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.20 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: 2f7e50

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	Latency	Recip. throughput	Vectorization
PUSH %RBP	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
MOV %RSP,%RBP	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
PUSH %R15	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R14	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %R12	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RBX	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
SUB $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV (%RCX),%R12	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
IMUL (%RDX),%R12	1	0.33	0.33	0.33	0	0	0	0.25	0.25	0.25	0.25	3	0.33	N/A
TEST %R12,%R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JLE 2f7fea <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x19a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV (%RDI),%R14D	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	scal (6.3%)
DEC %R12	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV %R9,%RBX	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R8,%R15	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOVQ $0,-0x38(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOVQ $0x1,-0x40(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOVL $0,-0x2c(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (6.3%)
MOV %R12,-0x28(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
SUB $0x8,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
LEA -0x40(%RBP),%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA 0x13893(%RIP),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x2c(%RBP),%RCX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x38(%RBP),%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA -0x28(%RBP),%R9	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
MOV %R14D,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
MOV $0x22,%EDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
PUSH $0x1	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH $0x1	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
PUSH %RAX	1	0	0	0	0	0	0	0	0	0	0	1	0.50	N/A
CALL 305fa0 <@plt_start@+0x90>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0.50	N/A
ADD $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
MOV -0x28(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
CMP %R12,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
CMOVL %RAX,%R12	2	0.33	0.33	0.33	0.33	0.33	0.33	0	0	0	0	1	0.33	N/A
MOV -0x38(%RBP),%RAX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %R12,-0x28(%RBP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
CMP %R12,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JG 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
MOV (%R15),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	3	0.25	N/A
MOV %R12,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
SUB %RAX,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
INC %RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP $0x4,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JB 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
LEA (%RCX,%RAX,8),%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
LEA 0x8(%RBX),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
CMP %RDI,%RSI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JAE 2f7f14 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0xc4>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
LEA 0x8(%RCX,%R12,8),%RDI	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	2	0.50	N/A
CMP %RBX,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JA 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
CMP $0x20,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JAE 2f7f1e <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0xce>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
XOR %EDI,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0.25	N/A
JMP 2f7f91 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x141>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
VBROADCASTSD (%RBX),%ZMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
LEA 0xc0(%RCX,%RAX,8),%R8	2	0	0	0.50	0.50	0	0	0.25	0.25	0.25	0.25	2	0.50	N/A
MOV %RDX,%RDI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
AND $-0x20,%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
XOR %R9D,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0.25	scal (6.3%)
NOPW %CS:(%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A
CMP %RDI,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JE 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
TEST $0x1c,%DL	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
JE 2f7fba <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16a>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
VBROADCASTSD (%RBX),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	4	0.50	scal (12.5%)
MOV %RDX,%R8	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (12.5%)
AND $-0x4,%R8	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
ADD %R8,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
CMP %R8,%RDX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	scal (12.5%)
JNE 2f7fbd <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x16d>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50-1	N/A
JMP 2f7fd8 <hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13+0x188>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	1	N/A
ADD %RDI,%RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
DEC %RAX	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
LEA 0x13771(%RIP),%RDI	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.25	N/A
MOV %R14D,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0.17	scal (6.3%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 305fb0 <@plt_start@+0xa0>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0.50	N/A
ADD $0x20,%RSP	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	1	0.17	N/A
POP %RBX	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R12	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R14	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %R15	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
POP %RBP	1	0	0	0	0	0	0	0	0	0	0	1	0.33	N/A
RET	1	0	0	0	0.33	0.33	0.33	0	0	0	0	1	0.50	N/A
NOPW (%RAX,%RAX,1)	0	0	0	0	0	0	0	0	0	0	0	0	0.09	N/A

Name	Coverage (%)	Time (s)
▼hypre_CSRMatrixMatvecOutOfPlace.omp_outlined.13–	0.39	0.21
○Loop 3598 - csr_matvec.c:236-237 - exec	0.39	0.19
○Loop 3597 - csr_matvec.c:236-237 - exec	0.00	0.00
○Loop 3599 - csr_matvec.c:236-237 - exec	0.00	0.00

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

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

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