OV - attention-clang-gnr512 - Loop 91 - attention-clang-gnr512

	Loop Id: 91	Module: attention-clang-gnr512	Source: random.tcc:401-3367 [...]	Coverage: 0.04%

0x4cd0 MOV	%RBX,%RCX

0x4cd3 INC	%RBX

0x4cd6 MOV	%RBX,0x1938(%RSP)

0x4cde MOV	0x5b8(%RSP,%RCX,8),%RCX

0x4ce6 MOV	%RCX,%RDX

0x4ce9 SHR	$0xb,%RDX

0x4ced MOV	%EDX,%EDX

0x4cef XOR	%RCX,%RDX

0x4cf2 MOV	%EDX,%ECX

0x4cf4 SAL	$0x7,%ECX

0x4cf7 AND	$-0x62d3a980,%ECX

0x4cfd XOR	%RDX,%RCX

0x4d00 MOV	%ECX,%EDX

0x4d02 SAL	$0xf,%EDX

0x4d05 AND	$-0x103a0000,%EDX

0x4d0b XOR	%RCX,%RDX

0x4d0e MOV	%RDX,%RCX

0x4d11 SHR	$0x12,%RCX

0x4d15 XOR	%RDX,%RCX

0x4d18 VCVTUSI2SS	%RCX,%XMM15,%XMM2

0x4d1e VFMADD231SS	%XMM2,%XMM1,%XMM0

0x4d23 VMULSS	0x42dd(%RIP),%XMM1,%XMM1

0x4d2b DEC	%RAX

0x4d2e JE	50a0

0x4d34 CMP	$0x270,%RBX

0x4d3b JB	4cd0

0x4d3d VPBROADCASTQ	%R13,%ZMM2

0x4d43 XOR	%ECX,%ECX

0x4d45 VPBROADCASTQ	0x42f1(%RIP),%ZMM14

0x4d4f VPBROADCASTQ	0x42ef(%RIP),%ZMM15

0x4d59 VPBROADCASTQ	0x42ed(%RIP),%ZMM16

0x4d63 VPBROADCASTQ	0x42eb(%RIP),%ZMM17

0x4d6d NOPL	(%RAX)

  (92) 0x4d70 VMOVDQU64	0x5c0(%RSP,%RCX,8),%ZMM3

  (92) 0x4d78 VMOVDQU64	0x600(%RSP,%RCX,8),%ZMM4

  (92) 0x4d80 VMOVDQU64	0x640(%RSP,%RCX,8),%ZMM5

  (92) 0x4d88 VALIGNQ	$0x7,%ZMM2,%ZMM3,%ZMM6

  (92) 0x4d8f VMOVDQU64	0x680(%RSP,%RCX,8),%ZMM2

  (92) 0x4d97 VALIGNQ	$0x7,%ZMM3,%ZMM4,%ZMM7

  (92) 0x4d9e VALIGNQ	$0x7,%ZMM4,%ZMM5,%ZMM8

  (92) 0x4da5 VALIGNQ	$0x7,%ZMM5,%ZMM2,%ZMM9

  (92) 0x4dac VPANDQ	%ZMM15,%ZMM3,%ZMM10

  (92) 0x4db2 VPANDQ	%ZMM15,%ZMM4,%ZMM11

  (92) 0x4db8 VPANDQ	%ZMM15,%ZMM5,%ZMM12

  (92) 0x4dbe VPANDQ	%ZMM15,%ZMM2,%ZMM13

  (92) 0x4dc4 VPTERNLOGQ	$-0x8,%ZMM14,%ZMM6,%ZMM10

  (92) 0x4dcb VPTERNLOGQ	$-0x8,%ZMM14,%ZMM7,%ZMM11

  (92) 0x4dd2 VPTERNLOGQ	$-0x8,%ZMM14,%ZMM8,%ZMM12

  (92) 0x4dd9 VPTERNLOGQ	$-0x8,%ZMM14,%ZMM9,%ZMM13

  (92) 0x4de0 VPSRLQ	$0x1,%ZMM10,%ZMM6

  (92) 0x4de7 VPSRLQ	$0x1,%ZMM11,%ZMM7

  (92) 0x4dee VPSRLQ	$0x1,%ZMM12,%ZMM8

  (92) 0x4df5 VPSRLQ	$0x1,%ZMM13,%ZMM9

  (92) 0x4dfc VPXORQ	0x1220(%RSP,%RCX,8),%ZMM6,%ZMM6

  (92) 0x4e07 VPXORQ	0x1260(%RSP,%RCX,8),%ZMM7,%ZMM7

  (92) 0x4e12 VPXORQ	0x12a0(%RSP,%RCX,8),%ZMM8,%ZMM8

  (92) 0x4e1d VPXORQ	0x12e0(%RSP,%RCX,8),%ZMM9,%ZMM9

  (92) 0x4e28 VPTESTMQ	%ZMM16,%ZMM3,%K1

  (92) 0x4e2e VPTESTMQ	%ZMM16,%ZMM4,%K2

  (92) 0x4e34 VPTESTMQ	%ZMM16,%ZMM5,%K3

  (92) 0x4e3a VPTESTMQ	%ZMM16,%ZMM2,%K4

  (92) 0x4e40 VPXORQ	%ZMM17,%ZMM6,%ZMM6{%K1}

  (92) 0x4e46 VPXORQ	%ZMM17,%ZMM7,%ZMM7{%K2}

  (92) 0x4e4c VPXORQ	%ZMM17,%ZMM8,%ZMM8{%K3}

  (92) 0x4e52 VPXORQ	%ZMM17,%ZMM9,%ZMM9{%K4}

  (92) 0x4e58 VMOVDQU64	%ZMM6,0x5b8(%RSP,%RCX,8)

  (92) 0x4e63 VMOVDQU64	%ZMM7,0x5f8(%RSP,%RCX,8)

  (92) 0x4e6e VMOVDQU64	%ZMM8,0x638(%RSP,%RCX,8)

  (92) 0x4e79 VMOVDQU64	%ZMM9,0x678(%RSP,%RCX,8)

  (92) 0x4e84 ADD	$0x20,%RCX

  (92) 0x4e88 CMP	$0xe0,%RCX

  (92) 0x4e8f JNE	4d70

0x4e95 VEXTRACTI32X4	$0x3,%ZMM2,%XMM2

0x4e9c VPEXTRQ	$0x1,%XMM2,%RSI

0x4ea2 AND	$-0x80000000,%RSI

0x4ea9 MOV	0xcc0(%RSP),%RDX

0x4eb1 MOV	0xcc8(%RSP),%RCX

0x4eb9 MOV	%EDX,%EDI

0x4ebb AND	$0x7ffffffe,%EDI

0x4ec1 OR	%RSI,%RDI

0x4ec4 SHR	$0x1,%RDI

0x4ec7 XOR	0x1920(%RSP),%RDI

0x4ecf MOV	%EDX,%ESI

0x4ed1 AND	$0x1,%ESI

0x4ed4 NEG	%ESI

0x4ed6 AND	%R12D,%ESI

0x4ed9 XOR	%RDI,%RSI

0x4edc MOV	%RSI,0xcb8(%RSP)

0x4ee4 AND	$-0x80000000,%RDX

0x4eeb MOV	%ECX,%ESI

0x4eed AND	$0x7ffffffe,%ESI

0x4ef3 OR	%RDX,%RSI

0x4ef6 SHR	$0x1,%RSI

0x4ef9 XOR	0x1928(%RSP),%RSI

0x4f01 MOV	%ECX,%EDX

0x4f03 AND	$0x1,%EDX

0x4f06 NEG	%EDX

0x4f08 AND	%R12D,%EDX

0x4f0b XOR	%RSI,%RDX

0x4f0e MOV	%RDX,0xcc0(%RSP)

0x4f16 AND	$-0x80000000,%RCX

0x4f1d MOV	0xcd0(%RSP),%RDX

0x4f25 MOV	%EDX,%ESI

0x4f27 VPBROADCASTQ	%RDX,%XMM2

0x4f2d AND	$0x7ffffffe,%EDX

0x4f33 OR	%RCX,%RDX

0x4f36 SHR	$0x1,%RDX

0x4f39 XOR	0x1930(%RSP),%RDX

0x4f41 AND	$0x1,%ESI

0x4f44 NEG	%ESI

0x4f46 AND	%R12D,%ESI

0x4f49 XOR	%RDX,%RSI

0x4f4c MOV	%RSI,0xcc8(%RSP)

0x4f54 MOV	$0xe8,%ECX

0x4f59 VPBROADCASTQ	0x40de(%RIP),%XMM7

0x4f62 VPBROADCASTQ	0x40dd(%RIP),%XMM8

0x4f6b VPBROADCASTQ	0x40dc(%RIP),%XMM9

0x4f74 VPBROADCASTQ	0x40db(%RIP),%XMM10

0x4f7d NOPL	(%RAX)

  (93) 0x4f80 VMOVDQU	0x598(%RSP,%RCX,8),%XMM3

  (93) 0x4f89 VMOVDQU	0x5a8(%RSP,%RCX,8),%XMM4

  (93) 0x4f92 VPALIGNR	$0x8,%XMM2,%XMM3,%XMM2

  (93) 0x4f98 VMOVDQU	0x5b8(%RSP,%RCX,8),%XMM5

  (93) 0x4fa1 VPAND	%XMM3,%XMM8,%XMM6

  (93) 0x4fa5 VPTERNLOGQ	$-0x8,%XMM7,%XMM2,%XMM6

  (93) 0x4fac VPSRLQ	$0x1,%XMM6,%XMM2

  (93) 0x4fb1 VPXOR	-0x188(%RSP,%RCX,8),%XMM2,%XMM2

  (93) 0x4fba VPTESTMQ	%XMM9,%XMM3,%K1

  (93) 0x4fc0 VPXORQ	%XMM10,%XMM2,%XMM2{%K1}

  (93) 0x4fc6 VMOVDQU	%XMM2,0x590(%RSP,%RCX,8)

  (93) 0x4fcf VPALIGNR	$0x8,%XMM3,%XMM4,%XMM2

  (93) 0x4fd5 VPAND	%XMM4,%XMM8,%XMM3

  (93) 0x4fd9 VPTERNLOGQ	$-0x8,%XMM7,%XMM2,%XMM3

  (93) 0x4fe0 VPSRLQ	$0x1,%XMM3,%XMM2

  (93) 0x4fe5 VPXOR	-0x178(%RSP,%RCX,8),%XMM2,%XMM2

  (93) 0x4fee VPTESTMQ	%XMM9,%XMM4,%K1

  (93) 0x4ff4 VPXORQ	%XMM10,%XMM2,%XMM2{%K1}

  (93) 0x4ffa VMOVDQU	%XMM2,0x5a0(%RSP,%RCX,8)

  (93) 0x5003 VPALIGNR	$0x8,%XMM4,%XMM5,%XMM2

  (93) 0x5009 VPAND	%XMM5,%XMM8,%XMM3

  (93) 0x500d VPTERNLOGQ	$-0x8,%XMM7,%XMM2,%XMM3

  (93) 0x5014 VPSRLQ	$0x1,%XMM3,%XMM2

  (93) 0x5019 VPXOR	-0x168(%RSP,%RCX,8),%XMM2,%XMM2

  (93) 0x5022 VPTESTMQ	%XMM9,%XMM5,%K1

  (93) 0x5028 VPXORQ	%XMM10,%XMM2,%XMM2{%K1}

  (93) 0x502e VMOVDQU	%XMM2,0x5b0(%RSP,%RCX,8)

  (93) 0x5037 ADD	$0x6,%RCX

  (93) 0x503b VMOVDQA	%XMM5,%XMM2

  (93) 0x503f CMP	$0x274,%RCX

  (93) 0x5046 JNE	4f80

0x504c MOV	0x1930(%RSP),%RCX

0x5054 MOV	$-0x80000000,%RDX

0x505b AND	%RDX,%RCX

0x505e MOV	0x5b8(%RSP),%R13

0x5066 MOV	%R13D,%EDX

0x5069 AND	$0x7ffffffe,%EDX

0x506f OR	%RCX,%RDX

0x5072 SHR	$0x1,%RDX

0x5075 XOR	0x1218(%RSP),%RDX

0x507d MOV	%R13D,%ECX

0x5080 AND	$0x1,%ECX

0x5083 NEG	%ECX

0x5085 AND	%R12D,%ECX

0x5088 XOR	%RDX,%RCX

0x508b MOV	%RCX,0x1930(%RSP)

0x5093 XOR	%EBX,%EBX

0x5095 JMP	4cd0

/usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc: 401 - 3367

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

401:       for (size_t __k = 0; __k < (__n - __m); ++__k)

402:         {

403: 	  _UIntType __y = ((_M_x[__k] & __upper_mask)

404: 			   | (_M_x[__k + 1] & __lower_mask));

405: 	  _M_x[__k] = (_M_x[__k + __m] ^ (__y >> 1)

406: 		       ^ ((__y & 0x01) ? __a : 0));

407:         }

408:

409:       for (size_t __k = (__n - __m); __k < (__n - 1); ++__k)

410: 	{

411: 	  _UIntType __y = ((_M_x[__k] & __upper_mask)

412: 			   | (_M_x[__k + 1] & __lower_mask));

413: 	  _M_x[__k] = (_M_x[__k + (__m - __n)] ^ (__y >> 1)

414: 		       ^ ((__y & 0x01) ? __a : 0));

415: 	}

416:

417:       _UIntType __y = ((_M_x[__n - 1] & __upper_mask)

418: 		       | (_M_x[0] & __lower_mask));

419:       _M_x[__n - 1] = (_M_x[__m - 1] ^ (__y >> 1)

420: 		       ^ ((__y & 0x01) ? __a : 0));

[...]

455:       if (_M_p >= state_size)

456: 	_M_gen_rand();

457:

458:       // Calculate o(x(i)).

459:       result_type __z = _M_x[_M_p++];

460:       __z ^= (__z >> __u) & __d;

461:       __z ^= (__z << __s) & __b;

462:       __z ^= (__z << __t) & __c;

463:       __z ^= (__z >> __l);

[...]

3364:       for (size_t __k = __m; __k != 0; --__k)

3365: 	{

3366: 	  __sum += _RealType(__urng() - __urng.min()) * __tmp;

3367: 	  __tmp *= __r;

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	2.35
CQA speedup if FP arith vectorized	1.87
CQA speedup if fully vectorized	11.93
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.34
Bottlenecks
Function	main
Source	random.tcc:401-406,random.tcc:409-409,random.tcc:413-413,random.tcc:417-420,random.tcc:455-455,random.tcc:459-463,random.tcc:3364-3367
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	10.58
CQA cycles if no scalar integer	4.50
CQA cycles if FP arith vectorized	5.67
CQA cycles if fully vectorized	0.89
Front-end cycles	10.58
P0 cycles	7.85
P1 cycles	7.70
P2 cycles	3.50
P3 cycles	3.50
P4 cycles	1.50
P5 cycles	7.80
P6 cycles	7.85
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	7.80
P11 cycles	3.50
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	4
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	61.50
Nb uops	63.50
Nb loads	10.50
Nb stores	3.00
Nb stack references	5.50
FLOP/cycle	0.28
Nb FLOP add-sub	0.00
Nb FLOP mul	1.00
Nb FLOP fma	1.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	7.86
Bytes prefetched	0.00
Bytes loaded	80.00
Bytes stored	24.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	2.00
Stride indirect	0.00
Vectorization ratio all	0.74
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.86
Vector-efficiency ratio all	9.95
Vector-efficiency ratio load	9.18
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	6.25
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	6.25
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	9.93

Metric	Value
CQA speedup if no scalar integer	3.33
CQA speedup if FP arith vectorized	1.78
CQA speedup if fully vectorized	11.50
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.36
Bottlenecks	micro-operation queue,
Function	main
Source	random.tcc:401-406,random.tcc:409-409,random.tcc:413-413,random.tcc:417-420,random.tcc:455-455,random.tcc:459-463,random.tcc:3364-3367
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	16.67
CQA cycles if no scalar integer	5.00
CQA cycles if FP arith vectorized	9.34
CQA cycles if fully vectorized	1.45
Front-end cycles	16.67
P0 cycles	12.20
P1 cycles	12.10
P2 cycles	6.33
P3 cycles	6.33
P4 cycles	2.50
P5 cycles	12.20
P6 cycles	12.30
P7 cycles	2.50
P8 cycles	2.50
P9 cycles	2.50
P10 cycles	12.20
P11 cycles	6.33
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	4
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	97.00
Nb uops	100.00
Nb loads	19.00
Nb stores	5.00
Nb stack references	10.00
FLOP/cycle	0.18
Nb FLOP add-sub	0.00
Nb FLOP mul	1.00
Nb FLOP fma	1.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	11.28
Bytes prefetched	0.00
Bytes loaded	148.00
Bytes stored	40.00
Stride 0	0.00
Stride 1	0.00
Stride n	0.00
Stride unknown	2.00
Stride indirect	0.00
Vectorization ratio all	1.47
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	NA
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	NA
Vectorization ratio other	1.72
Vector-efficiency ratio all	10.29
Vector-efficiency ratio load	12.11
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	6.25
Vector-efficiency ratio add_sub	NA
Vector-efficiency ratio fma	6.25
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	10.13

Metric	Value
CQA speedup if no scalar integer	1.13
CQA speedup if FP arith vectorized	2.25
CQA speedup if fully vectorized	13.87
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.29
Bottlenecks	micro-operation queue,
Function	main
Source	random.tcc:401-406,random.tcc:409-409,random.tcc:413-413,random.tcc:417-420,random.tcc:455-455,random.tcc:459-463,random.tcc:3364-3367
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	4.50
CQA cycles if no scalar integer	4.00
CQA cycles if FP arith vectorized	2.00
CQA cycles if fully vectorized	0.32
Front-end cycles	4.50
P0 cycles	3.50
P1 cycles	3.30
P2 cycles	0.67
P3 cycles	0.67
P4 cycles	0.50
P5 cycles	3.40
P6 cycles	3.40
P7 cycles	0.50
P8 cycles	0.50
P9 cycles	0.50
P10 cycles	3.40
P11 cycles	0.67
DIV/SQRT cycles	0.00
Inter-iter dependencies cycles	4
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	26.00
Nb uops	27.00
Nb loads	2.00
Nb stores	1.00
Nb stack references	1.00
FLOP/cycle	0.67
Nb FLOP add-sub	0.00
Nb FLOP mul	1.00
Nb FLOP fma	1.00
Nb FLOP div	0.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	4.44
Bytes prefetched	0.00
Bytes loaded	12.00
Bytes stored	8.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	0.00
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	NA
Vectorization ratio other	0.00
Vector-efficiency ratio all	9.62
Vector-efficiency ratio load	6.25
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	6.25
Vector-efficiency ratio add_sub	12.50
Vector-efficiency ratio fma	6.25
Vector-efficiency ratio div_sqrt	NA
Vector-efficiency ratio other	9.72

Path /

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

Function	main
Source file and lines	random.tcc:401-3367
Module	attention-clang-gnr512

The loop is defined in /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:401-420,455-463,3364-3367.

Analyzed code is defined in /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:401-406,409,413,417-420,455,459-463,3364-3367.

Warnings:
Non-innermost loop: analyzing only self part (ignoring child loops).
The structure of this loop is probably <if then [else] end>.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

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

gain
potential
hint
expert

Vectorization

Your loop is probably not vectorized. Only 9% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 10.58 to 0.89 cycles (11.93x 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 loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

FMA

Detected 1 FMA (fused multiply-add) operations.

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.

VCVTUSI2SS: 0 occurrences
VPEXTRQ: 0 occurrences

Slow data structures access

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

Details

Constant unknown stride: 2 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

VCVTUSI2SS (INT32/64 to FP32, scalar): 0 occurrences

Workaround

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

Type of elements and instruction set

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

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

The binary loop is composed of 3 FP arithmetical operations:

1: addition or subtraction (all inside FMA instructions)
2: multiply (1 inside FMA instructions)

The binary loop is loading 80 bytes (20 single precision FP elements). The binary loop is storing 24 bytes (6 single precision FP elements).

Arithmetic intensity

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

General properties

nb instructions	61.50
nb uops	63.50
loop length	291
used x86 registers	7
used mmx registers	0
used xmm registers	6
used ymm registers	0
used zmm registers	2.50
nb stack references	5.50

Front-end

MACRO FUSION NOT POSSIBLE

micro-operation queue	10.58 cycles
front end	10.58 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	7.85	7.70	3.50	3.50	1.50	7.80	7.85	1.50	1.50	1.50	7.80	3.50
cycles	7.85	7.70	3.50	3.50	1.50	7.80	7.85	1.50	1.50	1.50	7.80	3.50

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	NA
Longest recurrence chain latency (RecMII)	4.00

Cycles summary

Front-end	10.58
Dispatch	7.90
Data deps.	4.00
Overall L1	10.58

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	10%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	9%

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

INT+FP

all	9%
load	9%
store	12%
mul	6%
add-sub	12%
fma	6%
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	9%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 10.58 cycles. At this rate:

3% of peak load performance is reached (5.77 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
1% of peak store performance is reached (2.09 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Function	main
Source file and lines	random.tcc:401-3367
Module	attention-clang-gnr512

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

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

gain
potential
hint
expert

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 16.67 to 5.00 cycles (3.33x speedup).

Workaround

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

Vectorization

Your loop is probably not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 16.67 to 1.45 cycles (11.50x speedup).

Details

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

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

FMA

Detected 1 FMA (fused multiply-add) operations.

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.

VCVTUSI2SS: 1 occurrences►
- /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:3366
VPEXTRQ: 1 occurrences►
- /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:404

Slow data structures access

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

Details

Constant unknown stride: 2 occurrence(s)

Non-unit stride (uncontiguous) accesses are not efficiently using data caches

Workaround

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

Conversion instructions

Detected expensive conversion instructions.

Details

VCVTUSI2SS (INT32/64 to FP32, scalar): 1 occurrences►
- /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:3366

Workaround

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

Type of elements and instruction set

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

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

The binary loop is composed of 3 FP arithmetical operations:

1: addition or subtraction (all inside FMA instructions)
2: multiply (1 inside FMA instructions)

The binary loop is loading 148 bytes (37 single precision FP elements). The binary loop is storing 40 bytes (10 single precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.02 FP operations per loaded or stored byte.

General properties

nb instructions	97
nb uops	100
loop length	473
used x86 registers	9
used mmx registers	0
used xmm registers	8
used ymm registers	0
used zmm registers	5
nb stack references	10

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	16.67 cycles
front end	16.67 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	12.20	12.10	6.33	6.33	2.50	12.20	12.30	2.50	2.50	2.50	12.20	6.33
cycles	12.20	12.10	6.33	6.33	2.50	12.20	12.30	2.50	2.50	2.50	12.20	6.33

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	NA
Longest recurrence chain latency (RecMII)	4.00

Cycles summary

Front-end	16.67
Dispatch	12.30
Data deps.	4.00
Overall L1	16.67

Vectorization ratios

INT

all	1%
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	1%

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

INT+FP

all	1%
load	0%
store	0%
mul	0%
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	1%

Vector efficiency ratios

INT

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

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

INT+FP

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

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 16.67 cycles. At this rate:

4% of peak load performance is reached (8.88 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
1% of peak store performance is reached (2.40 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 16.67 to 12.30 cycles (1.36x speedup).

ASM code

In the binary file, the address of the loop is: 4d34

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
MOV %RBX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
INC %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RBX,0x1938(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
MOV 0x5b8(%RSP,%RCX,8),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV %RCX,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
SHR $0xb,%RDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
MOV %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
XOR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %EDX,%ECX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
SAL $0x7,%ECX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
AND $-0x62d3a980,%ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
XOR %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
MOV %ECX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
SAL $0xf,%EDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (6.3%)
AND $-0x103a0000,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
XOR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %RDX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
SHR $0x12,%RCX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
XOR %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
VCVTUSI2SS %RCX,%XMM15,%XMM2	3	0.50	0.50	0	0	0	2	0	0	0	0	0	0	8	2	scal (12.5%)
VFMADD231SS %XMM2,%XMM1,%XMM0	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	4	0.50	scal (6.3%)
VMULSS 0x42dd(%RIP),%XMM1,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (6.3%)
DEC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
JE 50a0 <main+0x1820>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
CMP $0x270,%RBX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
JB 4cd0 <main+0x1450>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
VPBROADCASTQ %R13,%ZMM2	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
XOR %ECX,%ECX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VPBROADCASTQ 0x42f1(%RIP),%ZMM14	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x42ef(%RIP),%ZMM15	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x42ed(%RIP),%ZMM16	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x42eb(%RIP),%ZMM17	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
NOPL (%RAX)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VEXTRACTI32X4 $0x3,%ZMM2,%XMM2	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	vect (25.0%)
VPEXTRQ $0x1,%XMM2,%RSI	2	1	0.50	0	0	0	0.50	0	0	0	0	0	0	4	1	scal (12.5%)
AND $-0x80000000,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV 0xcc0(%RSP),%RDX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
MOV 0xcc8(%RSP),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV %EDX,%EDI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%EDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
OR %RSI,%RDI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
SHR $0x1,%RDI	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
XOR 0x1920(%RSP),%RDI	1	0.20	0.20	0.33	0.33	0	0.20	0.20	0	0	0	0.20	0.33	1-2	0.33	scal (12.5%)
MOV %EDX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x1,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NEG %ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
AND %R12D,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
XOR %RDI,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %RSI,0xcb8(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
AND $-0x80000000,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %ECX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
OR %RDX,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
SHR $0x1,%RSI	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
XOR 0x1928(%RSP),%RSI	1	0.20	0.20	0.33	0.33	0	0.20	0.20	0	0	0	0.20	0.33	1-2	0.33	scal (12.5%)
MOV %ECX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x1,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NEG %EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
AND %R12D,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
XOR %RSI,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %RDX,0xcc0(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
AND $-0x80000000,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
MOV 0xcd0(%RSP),%RDX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
MOV %EDX,%ESI	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
VPBROADCASTQ %RDX,%XMM2	1	0	0	0	0	0	1	0	0	0	0	0	0	3	1	scal (12.5%)
AND $0x7ffffffe,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
OR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
SHR $0x1,%RDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
XOR 0x1930(%RSP),%RDX	1	0.20	0.20	0.33	0.33	0	0.20	0.20	0	0	0	0.20	0.33	1-2	0.33	scal (12.5%)
AND $0x1,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NEG %ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
AND %R12D,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
XOR %RDX,%RSI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %RSI,0xcc8(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
MOV $0xe8,%ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
VPBROADCASTQ 0x40de(%RIP),%XMM7	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x40dd(%RIP),%XMM8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x40dc(%RIP),%XMM9	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
VPBROADCASTQ 0x40db(%RIP),%XMM10	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	3	0.33	scal (12.5%)
NOPL (%RAX)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV 0x1930(%RSP),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV $-0x80000000,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
AND %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
MOV 0x5b8(%RSP),%R13	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
MOV %R13D,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
OR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
SHR $0x1,%RDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
XOR 0x1218(%RSP),%RDX	1	0.20	0.20	0.33	0.33	0	0.20	0.20	0	0	0	0.20	0.33	1-2	0.33	scal (12.5%)
MOV %R13D,%ECX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
AND $0x1,%ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
NEG %ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
AND %R12D,%ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
XOR %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
MOV %RCX,0x1930(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
XOR %EBX,%EBX	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
JMP 4cd0 <main+0x1450>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A

Function	main
Source file and lines	random.tcc:401-3367
Module	attention-clang-gnr512

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

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

gain
potential
hint
expert

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 4.50 to 4.00 cycles (1.12x speedup).

Workaround

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

Vectorization

Your loop is not vectorized. Only 9% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 4.50 to 0.32 cycles (13.87x speedup).

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

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

FMA

Detected 1 FMA (fused multiply-add) operations.

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.

VCVTUSI2SS: 1 occurrences►
- /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:3366

Conversion instructions

Detected expensive conversion instructions.

Details

VCVTUSI2SS (INT32/64 to FP32, scalar): 1 occurrences►
- /usr/lib/gcc/x86_64-redhat-linux/11/../../../../include/c++/11/bits/random.tcc:3366

Workaround

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

Type of elements and instruction set

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

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

The binary loop is composed of 3 FP arithmetical operations:

1: addition or subtraction (all inside FMA instructions)
2: multiply (1 inside FMA instructions)

The binary loop is loading 12 bytes (3 single precision FP elements). The binary loop is storing 8 bytes (2 single precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.15 FP operations per loaded or stored byte.

General properties

nb instructions	26
nb uops	27
loop length	109
used x86 registers	5
used mmx registers	0
used xmm registers	4
used ymm registers	0
used zmm registers	0
nb stack references	1

Front-end

ASSUMED MACRO FUSION FIT IN UOP CACHE

micro-operation queue	4.50 cycles
front end	4.50 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.30	0.67	0.67	0.50	3.40	3.40	0.50	0.50	0.50	3.40	0.67
cycles	3.50	3.30	0.67	0.67	0.50	3.40	3.40	0.50	0.50	0.50	3.40	0.67

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	NA
Longest recurrence chain latency (RecMII)	4.00

Cycles summary

Front-end	4.50
Dispatch	3.50
Data deps.	4.00
Overall L1	4.50

Vectorization ratios

INT

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

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

INT+FP

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

Vector efficiency ratios

INT

all	10%
load	NA (no load vectorizable/vectorized instructions)
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	12%
fma	NA (no fma vectorizable/vectorized instructions)
other	9%

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

INT+FP

all	9%
load	6%
store	12%
mul	6%
add-sub	12%
fma	6%
div/sqrt	NA (no div/sqrt vectorizable/vectorized instructions)
other	9%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 4.50 cycles. At this rate:

1% of peak load performance is reached (2.67 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
1% of peak store performance is reached (1.78 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 4d34

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
MOV %RBX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
INC %RBX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV %RBX,0x1938(%RSP)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
MOV 0x5b8(%RSP,%RCX,8),%RCX	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV %RCX,%RDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
SHR $0xb,%RDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (12.5%)
MOV %EDX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
XOR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %EDX,%ECX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
SAL $0x7,%ECX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
AND $-0x62d3a980,%ECX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
XOR %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
MOV %ECX,%EDX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
SAL $0xf,%EDX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	scal (6.3%)
AND $-0x103a0000,%EDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
XOR %RCX,%RDX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (12.5%)
MOV %RDX,%RCX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
SHR $0x12,%RCX	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
XOR %RDX,%RCX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	N/A
VCVTUSI2SS %RCX,%XMM15,%XMM2	3	0.50	0.50	0	0	0	2	0	0	0	0	0	0	8	2	scal (12.5%)
VFMADD231SS %XMM2,%XMM1,%XMM0	1	0.50	0.50	0	0	0	0	0	0	0	0	0	0	4	0.50	scal (6.3%)
VMULSS 0x42dd(%RIP),%XMM1,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (6.3%)
DEC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (12.5%)
JE 50a0 <main+0x1820>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
CMP $0x270,%RBX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	N/A
JB 4cd0 <main+0x1450>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A

Report Configuration

Vectorization

Details

Workaround

FMA

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion instructions

Details

Workaround

Type of elements and instruction set

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

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Complex instructions

Details

Slow data structures access

Details

Workaround

Conversion instructions

Details

Workaround

Type of elements and instruction set

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

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

FMA

Complex instructions

Details

Conversion 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