OV - attention-avx512 - Loop 42

	Loop Id: 42	Module: attention-avx512	Source: attention.cpp:55-56	Coverage: 0.10%

0x7cc0 VMOVUPS	(%RSI,%R13,4),%YMM0    [1]

0x7cc6 VSUBPS	0xc0(%RSP),%YMM0,%YMM0    [2]

0x7ccf VMOVAPS	%YMM0,0x1c0(%RSP)    [2]

0x7cd8 VEXTRACTF128	$0x1,%YMM0,%XMM0

0x7cde VMOVAPS	%XMM0,0x100(%RSP)    [2]

0x7ce7 VZEROUPPER

0x7cea CALL	89b0 <@plt_start@+0x20>

0x7cef VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7cf5 VMOVSHDUP	0x100(%RSP),%XMM0    [2]

0x7cfe CALL	89b0 <@plt_start@+0x20>

0x7d03 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7d09 VINSERTPS	$0x10,%XMM0,%XMM1,%XMM0

0x7d0f VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7d15 VPERMILPD	$0x1,0x100(%RSP),%XMM0    [2]

0x7d20 CALL	89b0 <@plt_start@+0x20>

0x7d25 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7d2b VINSERTPS	$0x20,%XMM0,%XMM1,%XMM0

0x7d31 VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7d37 VPERMILPS	$-0x1,0x100(%RSP),%XMM0    [2]

0x7d42 CALL	89b0 <@plt_start@+0x20>

0x7d47 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7d4d VINSERTPS	$0x30,%XMM0,%XMM1,%XMM0

0x7d53 VMOVAPS	%XMM0,0x100(%RSP)    [2]

0x7d5c VMOVAPS	0x1c0(%RSP),%YMM0    [2]

0x7d65 VZEROUPPER

0x7d68 CALL	89b0 <@plt_start@+0x20>

0x7d6d VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7d73 VMOVSHDUP	0x1c0(%RSP),%XMM0    [2]

0x7d7c CALL	89b0 <@plt_start@+0x20>

0x7d81 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7d87 VINSERTPS	$0x10,%XMM0,%XMM1,%XMM0

0x7d8d VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7d93 VPERMILPD	$0x1,0x1c0(%RSP),%XMM0    [2]

0x7d9e CALL	89b0 <@plt_start@+0x20>

0x7da3 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7da9 VINSERTPS	$0x20,%XMM0,%XMM1,%XMM0

0x7daf VMOVAPS	%XMM0,0x40(%RSP)    [2]

0x7db5 VPERMILPS	$-0x1,0x1c0(%RSP),%XMM0    [2]

0x7dc0 CALL	89b0 <@plt_start@+0x20>

0x7dc5 VMOVAPS	0x40(%RSP),%XMM1    [2]

0x7dcb MOV	0x220(%RSP),%RCX    [2]

0x7dd3 MOV	0x28(%RSP),%RSI    [2]

0x7dd8 VINSERTPS	$0x30,%XMM0,%XMM1,%XMM0

0x7dde VINSERTF128	$0x1,0x100(%RSP),%YMM0,%YMM0    [2]

0x7de9 VDIVPS	0x340(%RSP),%YMM0,%YMM0    [2]

0x7df2 VMOVUPS	%YMM0,(%RCX,%R13,4)    [3]

0x7df8 ADD	$0x8,%R13

0x7dfc CMP	%R13,0x338(%RSP)    [2]

0x7e04 JNE	7cc0

/home/eoseret/llm-attention/attention.cpp: 55 - 56

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

55:       for (int idx = 0; idx <= row; ++idx)  //vectorised

56:          P[row * N + idx] = expf(S_row[idx] - max_val) / sum;

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 /

Metric	Value
CQA speedup if no scalar integer	1.36
CQA speedup if FP arith vectorized	1.17
CQA speedup if fully vectorized	4.37
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.36
Bottlenecks	P6, P7, P8, P9,
Function	main
Source	attention.cpp:55-56
Source loop unroll info	unrolled by 16
Source loop unroll confidence level	max
Unroll/vectorization loop type	intermediate
Unroll factor	8
CQA cycles	9.50
CQA cycles if no scalar integer	7.00
CQA cycles if FP arith vectorized	8.13
CQA cycles if fully vectorized	2.17
Front-end cycles	7.00
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	3.00
P4 cycles	3.00
P5 cycles	3.00
P6 cycles	9.50
P7 cycles	9.50
P8 cycles	9.50
P9 cycles	9.50
P10 cycles	3.50
P11 cycles	3.50
P12 cycles	3.50
P13 cycles	3.50
P14 cycles	5.00
P15 cycles	5.00
DIV/SQRT cycles	2.50
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	49.00
Nb uops	56.00
Nb loads	20.00
Nb stores	10.00
Nb stack references	8.00
FLOP/cycle	1.68
Nb FLOP add-sub	8.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	8.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	58.11
Bytes prefetched	0.00
Bytes loaded	360.00
Bytes stored	192.00
Stride 0	1.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	83.33
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	57.14
Vector-efficiency ratio all	25.35
Vector-efficiency ratio load	29.41
Vector-efficiency ratio store	30.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	16.96

Metric	Value
CQA speedup if no scalar integer	1.36
CQA speedup if FP arith vectorized	1.17
CQA speedup if fully vectorized	4.37
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.36
Bottlenecks	P6, P7, P8, P9,
Function	main
Source	attention.cpp:55-56
Source loop unroll info	unrolled by 16
Source loop unroll confidence level	max
Unroll/vectorization loop type	intermediate
Unroll factor	8
CQA cycles	9.50
CQA cycles if no scalar integer	7.00
CQA cycles if FP arith vectorized	8.13
CQA cycles if fully vectorized	2.17
Front-end cycles	7.00
P0 cycles	0.33
P1 cycles	0.33
P2 cycles	0.33
P3 cycles	3.00
P4 cycles	3.00
P5 cycles	3.00
P6 cycles	9.50
P7 cycles	9.50
P8 cycles	9.50
P9 cycles	9.50
P10 cycles	3.50
P11 cycles	3.50
P12 cycles	3.50
P13 cycles	3.50
P14 cycles	5.00
P15 cycles	5.00
DIV/SQRT cycles	2.50
Inter-iter dependencies cycles	1
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	49.00
Nb uops	56.00
Nb loads	20.00
Nb stores	10.00
Nb stack references	8.00
FLOP/cycle	1.68
Nb FLOP add-sub	8.00
Nb FLOP mul	0.00
Nb FLOP fma	0.00
Nb FLOP div	8.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	58.11
Bytes prefetched	0.00
Bytes loaded	360.00
Bytes stored	192.00
Stride 0	1.00
Stride 1	2.00
Stride n	0.00
Stride unknown	0.00
Stride indirect	0.00
Vectorization ratio all	83.33
Vectorization ratio load	100.00
Vectorization ratio store	100.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	100.00
Vectorization ratio fma	NA
Vectorization ratio div_sqrt	100.00
Vectorization ratio other	57.14
Vector-efficiency ratio all	25.35
Vector-efficiency ratio load	29.41
Vector-efficiency ratio store	30.00
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	50.00
Vector-efficiency ratio fma	NA
Vector-efficiency ratio div_sqrt	50.00
Vector-efficiency ratio other	16.96

Path /

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

Function	main
Source file and lines	attention.cpp:55-56
Module	attention-avx512

The loop is defined in /home/eoseret/llm-attention/attention.cpp:55-56.

It is intermediate loop of related source loop which is unrolled by 16 (including vectorization).

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 9.50 to 7.00 cycles (1.36x speedup).

Workaround

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

Unrolling/vectorization cost

This loop is peel/tail of a unrolled/vectorized loop. If its cost is not negligible compared to the main (unrolled/vectorized) loop, unrolling/vectorization is counterproductive due to low trip count.

Details

The more iterations the main loop is processing, the higher the trip count must be to amortize peel/tail overhead.

Workaround

recompile with -fprofile-instr-generate, execute ,merge raw profiles with 'llvm-profdata merge -o default.profdata default.profraw' and recompile with -fprofile-instr-use (profile-guided optimization)
hardcode most frequent values of loop bounds by adding specialized paths.:
- For instance, replace for (i=0; i<n; i++) foo(i) with: switch (n) { case (4): for (i=0; i<4; i++) foo(i); break; case (6): for (i=0; i<6; i++) foo(i); break; default : for (i=0; i<n; i++) foo(i); break; }

Vectorization

Your loop is partially vectorized. Only 25% of vector register length is used (average across all SSE/AVX instructions).

Details

83% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

57% of SSE/AVX instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Execution units bottlenecks

Performance is limited by:

reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 9.50 to 7.00 cycles (1.36x speedup).

Workaround

Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop

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: 8 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56

Vector unaligned load/store instructions

Detected 2 suboptimal vector unaligned load/store instructions.

Details

VEXTRACTF128: 1 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56
VINSERTF128: 1 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56

Workaround

Pass to your compiler a micro-architecture specialization option:
- use march=native
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: replace { void *p = malloc (size); } with { void *p; posix_memalign (&p, 64, size); }.
2. inform your compiler that your arrays are vector aligned: if array 'foo' is 64 bytes-aligned, define a pointer 'p_foo' as __builtin_assume_aligned (foo, 64) and use it instead of 'foo' in the loop.

Type of elements and instruction set

2 AVX instructions are processing arithmetic or math operations on single precision FP elements in vector mode (eight at a time).

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

The binary loop is composed of 16 FP arithmetical operations:

8: addition or subtraction
8: divide

The binary loop is loading 360 bytes (90 single precision FP elements). The binary loop is storing 192 bytes (48 single precision FP elements).

Arithmetic intensity

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

General properties

nb instructions	49
nb uops	56
loop length	330
used x86 registers	4
used mmx registers	0
used xmm registers	2
used ymm registers	1
used zmm registers	0
nb stack references	8

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	7.00 cycles
front end	7.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	3.00	3.00	3.00	9.50	9.50	9.50	9.50	3.50	3.50	3.50	3.50	5.00	5.00
cycles	0.33	0.33	0.33	3.00	3.00	3.00	9.50	9.50	9.50	9.50	3.50	3.50	3.50	3.50	5.00	5.00

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	2.50
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	7.00
Dispatch	9.50
DIV/SQRT	2.50
Data deps.	1.00
Overall L1	9.50

Vectorization ratios

INT

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

all	82%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	100%
other	50%

INT+FP

all	83%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	100%
other	57%

Vector efficiency ratios

INT

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

all	25%
load	29%
store	30%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	50%
other	15%

INT+FP

all	25%
load	29%
store	30%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	50%
other	16%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 9.50 cycles. At this rate:

14% of peak load performance is reached (37.89 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
15% of peak store performance is reached (20.21 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: 7cc0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMOVUPS (%RSI,%R13,4),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (50.0%)
VSUBPS 0xc0(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (50.0%)
VMOVAPS %YMM0,0x1c0(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (50.0%)
VEXTRACTF128 $0x1,%YMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	3	0.50	vect (25.0%)
VMOVAPS %XMM0,0x100(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVSHDUP 0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	vect (12.5%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x10,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPD $0x1,0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x20,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPS $-0x1,0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x30,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x100(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVAPS 0x1c0(%RSP),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (50.0%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVSHDUP 0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	vect (12.5%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x10,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPD $0x1,0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x20,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPS $-0x1,0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
MOV 0x220(%RSP),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
MOV 0x28(%RSP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VINSERTPS $0x30,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VINSERTF128 $0x1,0x100(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0.50	0.50	0	0	0	1	0.50	vect (25.0%)
VDIVPS 0x340(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	10	2.50	vect (50.0%)
VMOVUPS %YMM0,(%RCX,%R13,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (50.0%)
ADD $0x8,%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %R13,0x338(%RSP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25	N/A
JNE 7cc0 <main+0x3270>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	0	0.33-1	N/A

Function	main
Source file and lines	attention.cpp:55-56
Module	attention-avx512

The loop is defined in /home/eoseret/llm-attention/attention.cpp:55-56.

It is intermediate loop of related source loop which is unrolled by 16 (including vectorization).

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 9.50 to 7.00 cycles (1.36x speedup).

Workaround

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

Unrolling/vectorization cost

Details

The more iterations the main loop is processing, the higher the trip count must be to amortize peel/tail overhead.

Workaround

recompile with -fprofile-instr-generate, execute ,merge raw profiles with 'llvm-profdata merge -o default.profdata default.profraw' and recompile with -fprofile-instr-use (profile-guided optimization)
hardcode most frequent values of loop bounds by adding specialized paths.:
- For instance, replace for (i=0; i<n; i++) foo(i) with: switch (n) { case (4): for (i=0; i<4; i++) foo(i); break; case (6): for (i=0; i<6; i++) foo(i); break; default : for (i=0; i<n; i++) foo(i); break; }

Vectorization

Your loop is partially vectorized. Only 25% of vector register length is used (average across all SSE/AVX instructions).

Details

83% of SSE/AVX instructions are used in vector version (process two or more data elements in vector registers):

57% of SSE/AVX instructions that are not load, store, addition, subtraction nor multiply instructions are used in vector version.

Execution units bottlenecks

Performance is limited by:

reading data from caches/RAM (load units are a bottleneck)
writing data to caches/RAM (the store unit is a bottleneck)

By removing all these bottlenecks, you can lower the cost of an iteration from 9.50 to 7.00 cycles (1.36x speedup).

Workaround

Read less array elements
Write less array elements
Provide more information to your compiler:
- hardcode the bounds of the corresponding 'for' loop

No data for this section

CALL instructions

Detected function call instructions.

Details

unknown: 8 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56
- /home/eoseret/llm-attention/attention.cpp:56

Vector unaligned load/store instructions

Detected 2 suboptimal vector unaligned load/store instructions.

Details

VEXTRACTF128: 1 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56
VINSERTF128: 1 occurrences►
- /home/eoseret/llm-attention/attention.cpp:56

Workaround

Pass to your compiler a micro-architecture specialization option:
- use march=native
Use vector aligned instructions:
1. align your arrays on 64 bytes boundaries: replace { void *p = malloc (size); } with { void *p; posix_memalign (&p, 64, size); }.
2. inform your compiler that your arrays are vector aligned: if array 'foo' is 64 bytes-aligned, define a pointer 'p_foo' as __builtin_assume_aligned (foo, 64) and use it instead of 'foo' in the loop.

Type of elements and instruction set

2 AVX instructions are processing arithmetic or math operations on single precision FP elements in vector mode (eight at a time).

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

The binary loop is composed of 16 FP arithmetical operations:

8: addition or subtraction
8: divide

The binary loop is loading 360 bytes (90 single precision FP elements). The binary loop is storing 192 bytes (48 single precision FP elements).

Arithmetic intensity

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

General properties

nb instructions	49
nb uops	56
loop length	330
used x86 registers	4
used mmx registers	0
used xmm registers	2
used ymm registers	1
used zmm registers	0
nb stack references	8

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	7.00 cycles
front end	7.00 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15
uops	0.33	0.33	0.33	3.00	3.00	3.00	9.50	9.50	9.50	9.50	3.50	3.50	3.50	3.50	5.00	5.00
cycles	0.33	0.33	0.33	3.00	3.00	3.00	9.50	9.50	9.50	9.50	3.50	3.50	3.50	3.50	5.00	5.00

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	2.50
Longest recurrence chain latency (RecMII)	1.00

Cycles summary

Front-end	7.00
Dispatch	9.50
DIV/SQRT	2.50
Data deps.	1.00
Overall L1	9.50

Vectorization ratios

INT

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

all	82%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	100%
other	50%

INT+FP

all	83%
load	100%
store	100%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	100%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	100%
other	57%

Vector efficiency ratios

INT

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

all	25%
load	29%
store	30%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	50%
other	15%

INT+FP

all	25%
load	29%
store	30%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	50%
fma	NA (no fma vectorizable/vectorized instructions)
div/sqrt	50%
other	16%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 9.50 cycles. At this rate:

14% of peak load performance is reached (37.89 out of 256.00 bytes loaded per cycle (GB/s @ 1GHz))
15% of peak store performance is reached (20.21 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: 7cc0

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	P12	P13	P14	P15	Latency	Recip. throughput	Vectorization
VMOVUPS (%RSI,%R13,4),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (50.0%)
VSUBPS 0xc0(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0.50	0.50	0	0	2	0.50	vect (50.0%)
VMOVAPS %YMM0,0x1c0(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (50.0%)
VEXTRACTF128 $0x1,%YMM0,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0	0.50	0.50	0	0	0	3	0.50	vect (25.0%)
VMOVAPS %XMM0,0x100(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVSHDUP 0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	vect (12.5%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x10,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPD $0x1,0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x20,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPS $-0x1,0x100(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x30,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x100(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVAPS 0x1c0(%RSP),%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (50.0%)
VZEROUPPER	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	1	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VMOVSHDUP 0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.50	vect (12.5%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x10,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPD $0x1,0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
VINSERTPS $0x20,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VMOVAPS %XMM0,0x40(%RSP)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (25.0%)
VPERMILPS $-0x1,0x1c0(%RSP),%XMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0.25	0	0	1	0.50	vect (25.0%)
CALL 89b0 <@plt_start@+0x20>	2	0	0	0	0.33	0.33	0.33	0.25	0.25	0.25	0.25	0	0	0	0	0	0	0	0.50	N/A
VMOVAPS 0x40(%RSP),%XMM1	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.50	vect (25.0%)
MOV 0x220(%RSP),%RCX	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
MOV 0x28(%RSP),%RSI	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0	0	3	0.25	N/A
VINSERTPS $0x30,%XMM0,%XMM1,%XMM0	1	0	0	0	0	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	1	0.25	scal (6.3%)
VINSERTF128 $0x1,0x100(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0.50	0.50	0	0	0	1	0.50	vect (25.0%)
VDIVPS 0x340(%RSP),%YMM0,%YMM0	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0.50	0.50	0	0	0	0	10	2.50	vect (50.0%)
VMOVUPS %YMM0,(%RCX,%R13,4)	1	0	0	0	0	0	0	0.25	0.25	0.25	0.25	0	0	0	0	0.50	0.50	4	0.50	vect (50.0%)
ADD $0x8,%R13	1	0.17	0.17	0.17	0.17	0.17	0.17	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
CMP %R13,0x338(%RSP)	1	0.17	0.17	0.17	0.17	0.17	0.17	0.25	0.25	0.25	0.25	0	0	0	0	0	0	1	0.25	N/A
JNE 7cc0 <main+0x3270>	1	0	0	0	0.33	0.33	0.33	0	0	0	0	0	0	0	0	0	0	0	0.33-1	N/A

Report Configuration

Code clean check

Workaround

Unrolling/vectorization cost

Details

Workaround

Vectorization

Details

Execution units bottlenecks

Workaround

CALL instructions

Details

Vector unaligned load/store instructions

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Code clean check

Workaround

Unrolling/vectorization cost

Details

Workaround

Vectorization

Details

Execution units bottlenecks

Workaround

CALL instructions

Details

Vector unaligned load/store instructions

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code