Jump to content

AVX-512

From Wikipedia, the free encyclopedia
(Redirected from AVX512)

AVX-512 are 512-bit extensions to the 256-bit Advanced Vector Extensions SIMD instructions for x86 instruction set architecture (ISA) proposed by Intel in July 2013, and first implemented in the 2016 Intel Xeon Phi x200 (Knights Landing),[1] and then later in a number of AMD and other Intel CPUs (see list below). AVX-512 consists of multiple extensions that may be implemented independently.[2] This policy is a departure from the historical requirement of implementing the entire instruction block. Only the core extension AVX-512F (AVX-512 Foundation) is required by all AVX-512 implementations.

Besides widening most 256-bit instructions, the extensions introduce various new operations, such as new data conversions, scatter operations, and permutations.[2] The number of AVX registers is increased from 16 to 32, and eight new "mask registers" are added, which allow for variable selection and blending of the results of instructions. In CPUs with the vector length (VL) extension—included in most AVX-512-capable processors (see § CPUs with AVX-512)—these instructions may also be used on the 128-bit and 256-bit vector sizes.

AVX-512 is not the first 512-bit SIMD instruction set that Intel has introduced in processors: the earlier 512-bit SIMD instructions used in the first generation Xeon Phi coprocessors, derived from Intel's Larrabee project, are similar but not binary compatible and only partially source compatible.[1]

The successor to AVX-512 is AVX10, announced July 2023,[3] which will work on both performance and efficiency cores.

Instruction set

[edit]

The AVX-512 instruction set consists of several separate sets each having their own unique CPUID feature bit. However, they are typically grouped by the processor generation that implements them.

F, CD, ER, PF:  introduced with Xeon Phi x200 (Knights Landing) and Xeon Gold/Platinum (Skylake SP "Purley"), with the last two (ER and PF) being specific to Knights Landing.

  • AVX-512 Foundation (F) – expands most 32-bit and 64-bit based AVX instructions with the EVEX coding scheme to support 512-bit registers, operation masks, parameter broadcasting, and embedded rounding and exception control, implemented by Knights Landing and Skylake Xeon
  • AVX-512 Conflict Detection Instructions (CD) – efficient conflict detection to allow more loops to be vectorized, implemented by Knights Landing[1] and Skylake X
  • AVX-512 Exponential and Reciprocal Instructions (ER) – exponential and reciprocal operations designed to help implement transcendental operations, implemented by Knights Landing[1]
  • AVX-512 Prefetch Instructions (PF) – new prefetch capabilities, implemented by Knights Landing[1]

VL, DQ, BW:  introduced with Skylake X and Cannon Lake.

  • AVX-512 Vector Length Extensions (VL) – extends most AVX-512 operations to also operate on XMM (128-bit) and YMM (256-bit) registers[4]
  • AVX-512 Doubleword and Quadword Instructions (DQ) – adds new 32-bit and 64-bit AVX-512 instructions[4]
  • AVX-512 Byte and Word Instructions (BW) – extends AVX-512 to cover 8-bit and 16-bit integer operations[4]

IFMA, VBMI:  introduced with Cannon Lake.[5]

  • AVX-512 Integer Fused Multiply Add (IFMA) – fused multiply add of integers using 52-bit precision.
  • AVX-512 Vector Byte Manipulation Instructions (VBMI) adds vector byte permutation instructions which were not present in AVX-512BW.

4VNNIW, 4FMAPS:  introduced with Knights Mill.[6][7]

  • AVX-512 Vector Neural Network Instructions Word variable precision (4VNNIW) – vector instructions for deep learning, enhanced word, variable precision.
  • AVX-512 Fused Multiply Accumulation Packed Single precision (4FMAPS) – vector instructions for deep learning, floating point, single precision.

VPOPCNTDQ:  Vector population count instruction. Introduced with Knights Mill and Ice Lake.[8]

VNNI, VBMI2, BITALG:  introduced with Ice Lake.[8]

  • AVX-512 Vector Neural Network Instructions (VNNI) – vector instructions for deep learning.
  • AVX-512 Vector Byte Manipulation Instructions 2 (VBMI2) – byte/word load, store and concatenation with shift.
  • AVX-512 Bit Algorithms (BITALG) – byte/word bit manipulation instructions expanding VPOPCNTDQ.

VP2INTERSECT:  introduced with Tiger Lake.

  • AVX-512 Vector Pair Intersection to a Pair of Mask Registers (VP2INTERSECT).

GFNI, VPCLMULQDQ, VAES:  introduced with Ice Lake.[8]

  • These are not AVX-512 features per se. Together with AVX-512, they enable EVEX encoded versions of GFNI, PCLMULQDQ and AES instructions.

Encoding and features

[edit]

The VEX prefix used by AVX and AVX2, while flexible, did not leave enough room for the features Intel wanted to add to AVX-512. This has led them to define a new prefix called EVEX.

Compared to VEX, EVEX adds the following benefits:[7]

  • Expanded register encoding allowing 32 512-bit registers.
  • Adds 8 new opmask registers for masking most AVX-512 instructions.
  • Adds a new scalar memory mode that automatically performs a broadcast.
  • Adds room for explicit rounding control in each instruction.
  • Adds a new compressed displacement memory addressing mode.

The extended registers, SIMD width bit, and opmask registers of AVX-512 are mandatory and all require support from the OS.

SIMD modes

[edit]

The AVX-512 instructions are designed to mix with 128/256-bit AVX/AVX2 instructions without a performance penalty. However, AVX-512VL extensions allows the use of AVX-512 instructions on 128/256-bit registers XMM/YMM, so most SSE and AVX/AVX2 instructions have new AVX-512 versions encoded with the EVEX prefix which allow access to new features such as opmask and additional registers. Unlike AVX-256, the new instructions do not have new mnemonics but share namespace with AVX, making the distinction between VEX and EVEX encoded versions of an instruction ambiguous in the source code. Since AVX-512F only works on 32- and 64-bit values, SSE and AVX/AVX2 instructions that operate on bytes or words are available only with the AVX-512BW extension (byte & word support).[7]

Name Extension
sets 		Registers Types
Legacy SSE SSE–SSE4.2 xmm0–xmm15 single floats
from SSE2: bytes, words, doublewords, quadwords and double floats
AVX-128 (VEX) AVX, AVX2 xmm0–xmm15 bytes, words, doublewords, quadwords, single floats and double floats
AVX-256 (VEX) AVX, AVX2 ymm0–ymm15 single float and double float
from AVX2: bytes, words, doublewords, quadwords
AVX-128 (EVEX) AVX-512VL xmm0–xmm31
(k0–k7) 		doublewords, quadwords, single float and double float
with AVX512BW: bytes and words.
with AVX512-FP16: half float
AVX-256 (EVEX) AVX-512VL ymm0–ymm31
(k0–k7) 		doublewords, quadwords, single float and double float
with AVX512BW: bytes and words.
with AVX512-FP16: half float
AVX-512 (EVEX) AVX-512F zmm0–zmm31
(k0–k7) 		doublewords, quadwords, single float and double float
with AVX512BW: bytes and words
with AVX512-FP16: half float

Extended registers

[edit]
x64 AVX-512 register scheme as extension from the x64 AVX (YMM0–YMM15) and x64 SSE (XMM0–XMM15) registers
511 256 255 128 127 0
  ZMM0     YMM0     XMM0  
ZMM1 YMM1 XMM1
ZMM2 YMM2 XMM2
ZMM3 YMM3 XMM3
ZMM4 YMM4 XMM4
ZMM5 YMM5 XMM5
ZMM6 YMM6 XMM6
ZMM7 YMM7 XMM7
ZMM8 YMM8 XMM8
ZMM9 YMM9 XMM9
ZMM10 YMM10 XMM10
ZMM11 YMM11 XMM11
ZMM12 YMM12 XMM12
ZMM13 YMM13 XMM13
ZMM14 YMM14 XMM14
ZMM15 YMM15 XMM15
ZMM16 YMM16 XMM16
ZMM17 YMM17 XMM17
ZMM18 YMM18 XMM18
ZMM19 YMM19 XMM19
ZMM20 YMM20 XMM20
ZMM21 YMM21 XMM21
ZMM22 YMM22 XMM22
ZMM23 YMM23 XMM23
ZMM24 YMM24 XMM24
ZMM25 YMM25 XMM25
ZMM26 YMM26 XMM26
ZMM27 YMM27 XMM27
ZMM28 YMM28 XMM28
ZMM29 YMM29 XMM29
ZMM30 YMM30 XMM30
ZMM31 YMM31 XMM31

The width of the SIMD register file is increased from 256 bits to 512 bits, and expanded from 16 to a total of 32 registers ZMM0–ZMM31. These registers can be addressed as 256 bit YMM registers from AVX extensions and 128-bit XMM registers from Streaming SIMD Extensions, and legacy AVX and SSE instructions can be extended to operate on the 16 additional registers XMM16-XMM31 and YMM16-YMM31 when using EVEX encoded form.

Opmask registers

[edit]

AVX-512 vector instructions may indicate an opmask register to control which values are written to the destination, the instruction encoding supports 0–7 for this field, however, only opmask registers k1–k7 (of k0–k7) can be used as the mask corresponding to the value 1–7, whereas the value 0 is reserved for indicating no opmask register is used, i.e. a hardcoded constant (instead of 'k0') is used to indicate unmasked operations. The special opmask register 'k0' is still a functioning, valid register, it can be used in opmask register manipulation instructions or used as the destination opmask register.[9] A flag controls the opmask behavior, which can either be "zero", which zeros everything not selected by the mask, or "merge", which leaves everything not selected untouched. The merge behavior is identical to the blend instructions.

The opmask registers are normally 16 bits wide, but can be up to 64 bits with the AVX-512BW extension.[7] How many of the bits are actually used, though, depends on the vector type of the instructions masked. For the 32-bit single float or double words, 16 bits are used to mask the 16 elements in a 512-bit register. For double float and quad words, at most 8 mask bits are used.

The opmask register is the reason why several bitwise instructions which naturally have no element widths had them added in AVX-512. For instance, bitwise AND, OR or 128-bit shuffle now exist in both double-word and quad-word variants with the only difference being in the final masking.

New opmask instructions

[edit]

The opmask registers have a new mini extension of instructions operating directly on them. Unlike the rest of the AVX-512 instructions, these instructions are all VEX encoded. The initial opmask instructions are all 16-bit (Word) versions. With AVX-512DQ 8-bit (Byte) versions were added to better match the needs of masking 8 64-bit values, and with AVX-512BW 32-bit (Double) and 64-bit (Quad) versions were added so they can mask up to 64 8-bit values. The instructions KORTEST and KTEST can be used to set the x86 flags based on mask registers, so that they may be used together with non-SIMD x86 branch and conditional instructions.

Instruction Extension
set 		Description
KAND F Bitwise logical AND Masks
KANDN F Bitwise logical AND NOT Masks
KMOV F Move from and to Mask Registers or General Purpose Registers
KUNPCK F Unpack for Mask Registers
KNOT F NOT Mask Register
KOR F Bitwise logical OR Masks
KORTEST F OR Masks And Set Flags
KSHIFTL F Shift Left Mask Registers
KSHIFTR F Shift Right Mask Registers
KXNOR F Bitwise logical XNOR Masks
KXOR F Bitwise logical XOR Masks
KADD BW/DQ Add Two Masks
KTEST BW/DQ Bitwise comparison and set flags

New instructions in AVX-512 foundation

[edit]

Many AVX-512 instructions are simply EVEX versions of old SSE or AVX instructions. There are, however, several new instructions, and old instructions that have been replaced with new AVX-512 versions. The new or heavily reworked instructions are listed below. These foundation instructions also include the extensions from AVX-512VL and AVX-512BW since those extensions merely add new versions of these instructions instead of new instructions.

Blend using mask

[edit]

There are no EVEX-prefixed versions of the blend instructions from SSE4; instead, AVX-512 has a new set of blending instructions using mask registers as selectors. Together with the general compare into mask instructions below, these may be used to implement generic ternary operations or cmov, similar to XOP's VPCMOV.

Since blending is an integral part of the EVEX encoding, these instructions may also be considered basic move instructions. Using the zeroing blend mode, they can also be used as masking instructions.

Instruction Extension
set 		Description
VBLENDMPD F Blend float64 vectors using opmask control
VBLENDMPS F Blend float32 vectors using opmask control
VPBLENDMD F Blend int32 vectors using opmask control
VPBLENDMQ F Blend int64 vectors using opmask control
VPBLENDMB BW Blend byte integer vectors using opmask control
VPBLENDMW BW Blend word integer vectors using opmask control

Compare into mask

[edit]

AVX-512F has four new compare instructions. Like their XOP counterparts they use the immediate field to select between 8 different comparisons. Unlike their XOP inspiration, however, they save the result to a mask register and initially only support doubleword and quadword comparisons. The AVX-512BW extension provides the byte and word versions. Note that two mask registers may be specified for the instructions, one to write to and one to declare regular masking.[7]

Imme-
diate 		Compa-
rison 		Description
0 EQ Equal
1 LT Less than
2 LE Less than or equal
3 FALSE Set to zero
4 NEQ Not equal
5 NLT Greater than or equal
6 NLE Greater than
7 TRUE Set to one
Instruction Extension
set 		Description
VPCMPD, VPCMPUD F Compare signed/unsigned doublewords into mask
VPCMPQ, VPCMPUQ F Compare signed/unsigned quadwords into mask
VPCMPB, VPCMPUB BW Compare signed/unsigned bytes into mask
VPCMPW, VPCMPUW BW Compare signed/unsigned words into mask

Logical set mask

[edit]

The final way to set masks is using Logical Set Mask. These instructions perform either AND or NAND, and then set the destination opmask based on the result values being zero or non-zero. Note that like the comparison instructions, these take two opmask registers, one as destination and one a regular opmask.

Instruction Extension
set 		Description
VPTESTMD, VPTESTMQ F Logical AND and set mask for 32 or 64 bit integers.
VPTESTNMD, VPTESTNMQ F Logical NAND and set mask for 32 or 64 bit integers.
VPTESTMB, VPTESTMW BW Logical AND and set mask for 8 or 16 bit integers.
VPTESTNMB, VPTESTNMW BW Logical NAND and set mask for 8 or 16 bit integers.

Compress and expand

[edit]

The compress and expand instructions match the APL operations of the same name. They use the opmask in a slightly different way from other AVX-512 instructions. Compress only saves the values marked in the mask, but saves them compacted by skipping and not reserving space for unmarked values. Expand operates in the opposite way, by loading as many values as indicated in the mask and then spreading them to the selected positions.

Instruction Description
VCOMPRESSPD, VCOMPRESSPS Store sparse packed double/single-precision floating-point values into dense memory
VPCOMPRESSD, VPCOMPRESSQ Store sparse packed doubleword/quadword integer values into dense memory/register
VEXPANDPD, VEXPANDPS Load sparse packed double/single-precision floating-point values from dense memory
VPEXPANDD, VPEXPANDQ Load sparse packed doubleword/quadword integer values from dense memory/register

Permute

[edit]

A new set of permute instructions have been added for full two input permutations. They all take three arguments, two source registers and one index; the result is output by either overwriting the first source register or the index register. AVX-512BW extends the instructions to also include 16-bit (word) versions, and the AVX-512_VBMI extension defines the byte versions of the instructions.

Instruction Extension
set 		Description
VPERMB VBMI Permute packed bytes elements.
VPERMW BW Permute packed words elements.
VPERMT2B VBMI Full byte permute overwriting first source.
VPERMT2W BW Full word permute overwriting first source.
VPERMI2PD, VPERMI2PS F Full single/double floating-point permute overwriting the index.
VPERMI2D, VPERMI2Q F Full doubleword/quadword permute overwriting the index.
VPERMI2B VBMI Full byte permute overwriting the index.
VPERMI2W BW Full word permute overwriting the index.
VPERMT2PS, VPERMT2PD F Full single/double floating-point permute overwriting first source.
VPERMT2D, VPERMT2Q F Full doubleword/quadword permute overwriting first source.
VSHUFF32x4, VSHUFF64x2,
VSHUFI32x4, VSHUFI64x2 		F Shuffle four packed 128-bit lines.
VPMULTISHIFTQB VBMI Select packed unaligned bytes from quadword sources.

Bitwise ternary logic

[edit]

Two new instructions added can logically implement all possible bitwise operations between three inputs. They take three registers as input and an 8-bit immediate field. Each bit in the output is generated using a lookup of the three corresponding bits in the inputs to select one of the 8 positions in the 8-bit immediate. Since only 8 combinations are possible using three bits, this allow all possible 3 input bitwise operations to be performed.[7] These are the only bitwise vector instructions in AVX-512F; EVEX versions of the two source SSE and AVX bitwise vector instructions AND, ANDN, OR and XOR were added in AVX-512DQ.

The difference in the doubleword and quadword versions is only the application of the opmask.

Instruction Description
VPTERNLOGD, VPTERNLOGQ Bitwise Ternary Logic
Bitwise Ternary Logic Truth table
A0 A1 A2 Double AND
(0x80) 		Double OR
(0xFE) 		Bitwise blend
(0xCA)
0 0 0 0 0 0
0 0 1 0 1 1
0 1 0 0 1 0
0 1 1 0 1 1
1 0 0 0 1 0
1 0 1 0 1 0
1 1 0 0 1 1
1 1 1 1 1 1

Conversions

[edit]

A number of conversion or move instructions were added; these complete the set of conversion instructions available from SSE2.

Instruction Extension
set 		Description
VPMOVQD, VPMOVSQD, VPMOVUSQD,
VPMOVQW, VPMOVSQW, VPMOVUSQW,
VPMOVQB, VPMOVSQB, VPMOVUSQB,
VPMOVDW, VPMOVSDW, VPMOVUSDW,
VPMOVDB, VPMOVSDB, VPMOVUSDB 		F Down convert quadword or doubleword to doubleword, word or byte; unsaturated, saturated or saturated unsigned. The reverse of the sign/zero extend instructions from SSE4.1.
VPMOVWB, VPMOVSWB, VPMOVUSWB BW Down convert word to byte; unsaturated, saturated or saturated unsigned.
VCVTPS2UDQ, VCVTPD2UDQ,
VCVTTPS2UDQ, VCVTTPD2UDQ 		F Convert with or without truncation, packed single or double-precision floating point to packed unsigned doubleword integers.
VCVTSS2USI, VCVTSD2USI,
VCVTTSS2USI, VCVTTSD2USI 		F Convert with or without truncation, scalar single or double-precision floating point to unsigned doubleword integer.
VCVTPS2QQ, VCVTPD2QQ,
VCVTPS2UQQ, VCVTPD2UQQ,
VCVTTPS2QQ, VCVTTPD2QQ,
VCVTTPS2UQQ, VCVTTPD2UQQ 		DQ Convert with or without truncation, packed single or double-precision floating point to packed signed or unsigned quadword integers.
VCVTUDQ2PS, VCVTUDQ2PD F Convert packed unsigned doubleword integers to packed single or double-precision floating point.
VCVTUSI2PS, VCVTUSI2PD F Convert scalar unsigned doubleword integers to single or double-precision floating point.
VCVTUSI2SD, VCVTUSI2SS F Convert scalar unsigned integers to single or double-precision floating point.
VCVTUQQ2PS, VCVTUQQ2PD DQ Convert packed unsigned quadword integers to packed single or double-precision floating point.
VCVTQQ2PD, VCVTQQ2PS F Convert packed quadword integers to packed single or double-precision floating point.

Floating-point decomposition

[edit]

Among the unique new features in AVX-512F are instructions to decompose floating-point values and handle special floating-point values. Since these methods are completely new, they also exist in scalar versions.

Instruction Description
VGETEXPPD, VGETEXPPS Convert exponents of packed fp values into fp values
VGETEXPSD, VGETEXPSS Convert exponent of scalar fp value into fp value
VGETMANTPD, VGETMANTPS Extract vector of normalized mantissas from float32/float64 vector
VGETMANTSD, VGETMANTSS Extract float32/float64 of normalized mantissa from float32/float64 scalar
VFIXUPIMMPD, VFIXUPIMMPS Fix up special packed float32/float64 values
VFIXUPIMMSD, VFIXUPIMMSS Fix up special scalar float32/float64 value

Floating-point arithmetic

[edit]

This is the second set of new floating-point methods, which includes new scaling and approximate calculation of reciprocal, and reciprocal of square root. The approximate reciprocal instructions guarantee to have at most a relative error of 2−14.[7]

Instruction Description
VRCP14PD, VRCP14PS Compute approximate reciprocals of packed float32/float64 values
VRCP14SD, VRCP14SS Compute approximate reciprocals of scalar float32/float64 value
VRNDSCALEPS, VRNDSCALEPD Round packed float32/float64 values to include a given number of fraction bits
VRNDSCALESS, VRNDSCALESD Round scalar float32/float64 value to include a given number of fraction bits
VRSQRT14PD, VRSQRT14PS Compute approximate reciprocals of square roots of packed float32/float64 values
VRSQRT14SD, VRSQRT14SS Compute approximate reciprocal of square root of scalar float32/float64 value
VSCALEFPS, VSCALEFPD Scale packed float32/float64 values with float32/float64 values
VSCALEFSS, VSCALEFSD Scale scalar float32/float64 value with float32/float64 value

Broadcast

[edit]
Instruction Extension
set 		Description
VBROADCASTSS, VBROADCASTSD F, VL Broadcast single/double floating-point value
VPBROADCASTB, VPBROADCASTW,
VPBROADCASTD, VPBROADCASTQ 		F, VL, DQ, BW Broadcast a byte/word/doubleword/quadword integer value
VBROADCASTI32X2, VBROADCASTI64X2,
VBROADCASTI32X4, VBROADCASTI32X8,
VBROADCASTI64X4 		F, VL, DQ, BW Broadcast two or four doubleword/quadword integer values

Miscellaneous

[edit]
Instruction Extension
set 		Description
VALIGND, VALIGNQ F, VL Align doubleword or quadword vectors
VDBPSADBW BW Double block packed sum-absolute-differences (SAD) on unsigned bytes
VPABSQ F Packed absolute value quadword
VPMAXSQ, VPMAXUQ F Maximum of packed signed/unsigned quadword
VPMINSQ, VPMINUQ F Minimum of packed signed/unsigned quadword
VPROLD, VPROLVD, VPROLQ, VPROLVQ,
VPRORD, VPRORVD, VPRORQ, VPRORVQ 		F Bit rotate left or right
VPSCATTERDD, VPSCATTERDQ,
VPSCATTERQD, VPSCATTERQQ 		F Scatter packed doubleword/quadword with
signed doubleword and quadword indices
VSCATTERDPS, VSCATTERDPD,
VSCATTERQPS, VSCATTERQPD 		F Scatter packed float32/float64 with
signed doubleword and quadword indices

New instructions by sets

[edit]

Conflict detection

[edit]

The instructions in AVX-512 conflict detection (AVX-512CD) are designed to help efficiently calculate conflict-free subsets of elements in loops that normally could not be safely vectorized.[10]

Instruction Name Description
VPCONFLICTD,
VPCONFLICTQ 		Detect conflicts within vector of packed double- or quadwords values Compares each element in the first source, to all elements on same or earlier places in the second source and forms a bit vector of the results
VPLZCNTD,
VPLZCNTQ 		Count the number of leading zero bits for packed double- or quadword values Vectorized LZCNT instruction
VPBROADCASTMB2Q,
VPBROADCASTMW2D 		Broadcast mask to vector register Either 8-bit mask to quadword vector, or 16-bit mask to doubleword vector

Exponential and reciprocal

[edit]

AVX-512 exponential and reciprocal (AVX-512ER) instructions contain more accurate approximate reciprocal instructions than those in the AVX-512 foundation; relative error is at most 2−28. They also contain two new exponential functions that have a relative error of at most 2−23.[7]

Instruction Description
VEXP2PD, VEXP2PS Compute approximate exponential 2x of packed single or double-precision floating-point values
VRCP28PD, VRCP28PS Compute approximate reciprocals of packed single or double-precision floating-point values
VRCP28SD, VRCP28SS Compute approximate reciprocal of scalar single or double-precision floating-point value
VRSQRT28PD, VRSQRT28PS Compute approximate reciprocals of square roots of packed single or double-precision floating-point values
VRSQRT28SD, VRSQRT28SS Compute approximate reciprocal of square root of scalar single or double-precision floating-point value

Prefetch

[edit]

AVX-512 prefetch (AVX-512PF) instructions contain new prefetch operations for the new scatter and gather functionality introduced in AVX2 and AVX-512. T0 prefetch means prefetching into level 1 cache and T1 means prefetching into level 2 cache.

Instruction Description
VGATHERPF0DPS, VGATHERPF0QPS,
VGATHERPF0DPD, VGATHERPF0QPD 		Using signed dword/qword indices, prefetch sparse byte memory locations containing single/double-precision data using opmask k1 and T0 hint.
VGATHERPF1DPS, VGATHERPF1QPS,
VGATHERPF1DPD, VGATHERPF1QPD 		Using signed dword/qword indices, prefetch sparse byte memory locations containing single/double-precision data using opmask k1 and T1 hint.
VSCATTERPF0DPS, VSCATTERPF0QPS,
VSCATTERPF0DPD, VSCATTERPF0QPD 		Using signed dword/qword indices, prefetch sparse byte memory locations containing single/double-precision data using writemask k1 and T0 hint with intent to write.
VSCATTERPF1DPS, VSCATTERPF1QPS,
VSCATTERPF1DPD, VSCATTERPF1QPD 		Using signed dword/qword indices, prefetch sparse byte memory locations containing single/double precision data using writemask k1 and T1 hint with intent to write.

4FMAPS and 4VNNIW

[edit]

The two sets of instructions perform multiple iterations of processing. They are generally only found in Xeon Phi products.

Instruction Extension
set 		Description
V4FMADDPS,
V4FMADDSS 		4FMAPS Packed/scalar single-precision floating-point fused multiply-add (4-iterations)
V4FNMADDPS,
V4FNMADDSS 		4FMAPS Packed/scalar single-precision floating-point fused multiply-add and negate (4-iterations)
VP4DPWSSD 4VNNIW Dot product of signed words with double word accumulation (4-iterations)
VP4DPWSSDS 4VNNIW Dot product of signed words with double word accumulation and saturation (4-iterations)

BW, DQ and VBMI

[edit]

AVX-512DQ adds new doubleword and quadword instructions. AVX-512BW adds byte and words versions of the same instructions, and adds byte and word version of doubleword/quadword instructions in AVX-512F. A few instructions which get only word forms with AVX-512BW acquire byte forms with the AVX-512_VBMI extension (VPERMB, VPERMI2B, VPERMT2B, VPMULTISHIFTQB).

Two new instructions were added to the mask instructions set: KADD and KTEST (B and W forms with AVX-512DQ, D and Q with AVX-512BW). The rest of mask instructions, which had only word forms, got byte forms with AVX-512DQ and doubleword/quadword forms with AVX-512BW. KUNPCKBW was extended to KUNPCKWD and KUNPCKDQ by AVX-512BW.

Among the instructions added by AVX-512DQ are several SSE and AVX instructions that didn't get AVX-512 versions with AVX-512F, among those are all the two input bitwise instructions and extract/insert integer instructions.

Instructions that are completely new are covered below.

Floating-point instructions

[edit]

Three new floating-point operations are introduced. Since they are not only new to AVX-512 they have both packed/SIMD and scalar versions.

The VFPCLASS instructions tests if the floating-point value is one of eight special floating-point values, which of the eight values will trigger a bit in the output mask register is controlled by the immediate field. The VRANGE instructions perform minimum or maximum operations depending on the value of the immediate field, which can also control if the operation is done absolute or not and separately how the sign is handled. The VREDUCE instructions operate on a single source, and subtract from that the integer part of the source value plus a number of bits specified in the immediate field of its fraction.

Instruction Extension
set 		Description
VFPCLASSPS, VFPCLASSPD DQ Test types of packed single and double precision floating-point values.
VFPCLASSSS, VFPCLASSSD DQ Test types of scalar single and double precision floating-point values.
VRANGEPS, VRANGEPD DQ Range restriction calculation for packed floating-point values.
VRANGESS, VRANGESD DQ Range restriction calculation for scalar floating-point values.
VREDUCEPS, VREDUCEPD DQ Perform reduction transformation on packed floating-point values.
VREDUCESS, VREDUCESD DQ Perform reduction transformation on scalar floating-point values.

Other instructions

[edit]
Instruction Extension
set 		Description
VPMOVM2D, VPMOVM2Q DQ Convert mask register to double- or quad-word vector register.
VPMOVM2B, VPMOVM2W BW Convert mask register to byte or word vector register.
VPMOVD2M, VPMOVQ2M DQ Convert double- or quad-word vector register to mask register.
VPMOVB2M, VPMOVW2M BW Convert byte or word vector register to mask register.
VPMULLQ DQ Multiply packed quadword store low result. A quadword version of VPMULLD.

VBMI2

[edit]

Extend VPCOMPRESS and VPEXPAND with byte and word variants. Shift instructions are new.

Instruction Description
VPCOMPRESSB, VPCOMPRESSW Store sparse packed byte/word integer values into dense memory/register
VPEXPANDB, VPEXPANDW Load sparse packed byte/word integer values from dense memory/register
VPSHLD Concatenate and shift packed data left logical
VPSHLDV Concatenate and variable shift packed data left logical
VPSHRD Concatenate and shift packed data right logical
VPSHRDV Concatenate and variable shift packed data right logical

VNNI

[edit]

Vector Neural Network Instructions:[11] AVX512-VNNI adds EVEX-coded instructions described below. With AVX-512F, these instructions can operate on 512-bit vectors, and AVX-512VL further adds support for 128- and 256-bit vectors.

A later AVX-VNNI extension adds VEX encodings of these instructions which can only operate on 128- or 256-bit vectors. AVX-VNNI is not part of the AVX-512 suite, it does not require AVX-512F and can be implemented independently.

Instruction Description
VPDPBUSD Multiply and add unsigned and signed bytes
VPDPBUSDS Multiply and add unsigned and signed bytes with saturation
VPDPWSSD Multiply and add signed word integers
VPDPWSSDS Multiply and add word integers with saturation

IFMA

[edit]

Integer fused multiply-add instructions. AVX512-IFMA adds EVEX-coded instructions described below.

A separate AVX-IFMA instruction set extension defines VEX encoding of these instructions. This extension is not part of the AVX-512 suite and can be implemented independently.

Instruction Extension
set 		Description
VPMADD52LUQ IFMA Packed multiply of unsigned 52-bit integers and add the low 52-bit products to 64-bit accumulators
VPMADD52HUQ IFMA Packed multiply of unsigned 52-bit integers and add the high 52-bit products to 64-bit accumulators

VPOPCNTDQ and BITALG

[edit]
Instruction Extension set Description
VPOPCNTD, VPOPCNTQ VPOPCNTDQ Return the number of bits set to 1 in doubleword/quadword
VPOPCNTB, VPOPCNTW BITALG Return the number of bits set to 1 in byte/word
VPSHUFBITQMB BITALG Shuffle bits from quadword elements using byte indexes into mask

VP2INTERSECT

[edit]
Instruction Extension set Description
VP2INTERSECTD,
VP2INTERSECTQ 		VP2INTERSECT Compute intersection between doublewords/quadwords to a pair of mask registers

GFNI

[edit]

Galois field new instructions are useful for cryptography,[12] as they can be used to implement Rijndael-style S-boxes such as those used in AES, Camellia, and SM4.[13] These instructions may also be used for bit manipulation in networking and signal processing.[12]

GFNI is a standalone instruction set extension and can be enabled separately from AVX or AVX-512. Depending on whether AVX and AVX-512F support is indicated by the CPU, GFNI support enables legacy (SSE), VEX or EVEX-coded instructions operating on 128, 256 or 512-bit vectors.

Instruction Description
VGF2P8AFFINEINVQB Galois field affine transformation inverse
VGF2P8AFFINEQB Galois field affine transformation
VGF2P8MULB Galois field multiply bytes

VPCLMULQDQ

[edit]

VPCLMULQDQ with AVX-512F adds an EVEX-encoded 512-bit version of the PCLMULQDQ instruction. With AVX-512VL, it adds EVEX-encoded 256- and 128-bit versions. VPCLMULQDQ alone (that is, on non-AVX512 CPUs) adds only VEX-encoded 256-bit version. (Availability of the VEX-encoded 128-bit version is indicated by different CPUID bits: PCLMULQDQ and AVX.) The wider than 128-bit variations of the instruction perform the same operation on each 128-bit portion of input registers, but they do not extend it to select quadwords from different 128-bit fields (the meaning of imm8 operand is the same: either low or high quadword of the 128-bit field is selected).

Instruction Description
VPCLMULQDQ Carry-less multiplication quadword

VAES

[edit]
See also: AES instruction set

VEX- and EVEX-encoded AES instructions. The wider than 128-bit variations of the instruction perform the same operation on each 128-bit portion of input registers. The VEX versions can be used without AVX-512 support.

Instruction Description
VAESDEC Perform one round of an AES decryption flow
VAESDECLAST Perform last round of an AES decryption flow
VAESENC Perform one round of an AES encryption flow
VAESENCLAST Perform last round of an AES encryption flow

BF16

[edit]

AI acceleration instructions operating on the Bfloat16 numbers.

Instruction Description
VCVTNE2PS2BF16 Convert two vectors of packed single precision numbers into one vector of packed Bfloat16 numbers
VCVTNEPS2BF16 Convert one vector of packed single precision numbers to one vector of packed Bfloat16 numbers
VDPBF16PS Calculate dot product of two Bfloat16 pairs and accumulate the result into one packed single precision number

FP16

[edit]

An extension of the earlier F16C instruction set, adding comprehensive support for the binary16 floating-point numbers (also known as FP16, float16 or half-precision floating-point numbers). The new instructions implement most operations that were previously available for single and double-precision floating-point numbers and also introduce new complex number instructions and conversion instructions. Scalar and packed operations are supported.

Unlike the single and double-precision format instructions, the half-precision operands are neither conditionally flushed to zero (FTZ) nor conditionally treated as zero (DAZ) based on MXCSR settings. Subnormal values are processed at full speed by hardware to facilitate using the full dynamic range of the FP16 numbers. Instructions that create FP32 and FP64 numbers still respect the MXCSR.FTZ bit.[14]

Arithmetic instructions

[edit]
Instruction Description
VADDPH, VADDSH Add packed/scalar FP16 numbers.
VSUBPH, VSUBSH Subtract packed/scalar FP16 numbers.
VMULPH, VMULSH Multiply packed/scalar FP16 numbers.
VDIVPH, VDIVSH Divide packed/scalar FP16 numbers.
VSQRTPH, VSQRTSH Compute square root of packed/scalar FP16 numbers.
VFMADD{132, 213, 231}PH,
VFMADD{132, 213, 231}SH 		Multiply-add packed/scalar FP16 numbers.
VFNMADD{132, 213, 231}PH,
VFNMADD{132, 213, 231}SH 		Negated multiply-add packed/scalar FP16 numbers.
VFMSUB{132, 213, 231}PH,
VFMSUB{132, 213, 231}SH 		Multiply-subtract packed/scalar FP16 numbers.
VFNMSUB{132, 213, 231}PH,
VFNMSUB{132, 213, 231}SH 		Negated multiply-subtract packed/scalar FP16 numbers.
VFMADDSUB{132, 213, 231}PH Multiply-add (odd vector elements) or multiply-subtract (even vector elements) packed FP16 numbers.
VFMSUBADD{132, 213, 231}PH Multiply-subtract (odd vector elements) or multiply-add (even vector elements) packed FP16 numbers.
VREDUCEPH, VREDUCESH Perform reduction transformation of the packed/scalar FP16 numbers.
VRNDSCALEPH, VRNDSCALESH Round packed/scalar FP16 numbers to a given number of fraction bits.
VSCALEFPH, VSCALEFSH Scale packed/scalar FP16 numbers by multiplying it by a power of two.

Complex arithmetic instructions

[edit]
Instruction Description
VFMULCPH, VFMULCSH Multiply packed/scalar complex FP16 numbers.
VFCMULCPH, VFCMULCSH Multiply packed/scalar complex FP16 numbers. Complex conjugate form of the operation.
VFMADDCPH, VFMADDCSH Multiply-add packed/scalar complex FP16 numbers.
VFCMADDCPH, VFCMADDCSH Multiply-add packed/scalar complex FP16 numbers. Complex conjugate form of the operation.

Approximate reciprocal instructions

[edit]
Instruction Description
VRCPPH, VRCPSH Compute approximate reciprocal of the packed/scalar FP16 numbers. The maximum relative error of the approximation is less than 2−11 + 2−14.
VRSQRTPH, VRSQRTSH Compute approximate reciprocal square root of the packed/scalar FP16 numbers. The maximum relative error of the approximation is less than 2−14.

Comparison instructions

[edit]
Instruction Description
VCMPPH, VCMPSH Compare the packed/scalar FP16 numbers and store the result in a mask register.
VCOMISH Compare the scalar FP16 numbers and store the result in the flags register. Signals an exception if a source operand is QNaN or SNaN.
VUCOMISH Compare the scalar FP16 numbers and store the result in the flags register. Signals an exception only if a source operand is SNaN.
VMAXPH, VMAXSH Select the maximum of each vertical pair of the source packed/scalar FP16 numbers.
VMINPH, VMINSH Select the minimum of each vertical pair of the source packed/scalar FP16 numbers.
VFPCLASSPH, VFPCLASSSH Test packed/scalar FP16 numbers for special categories (NaN, infinity, negative zero, etc.) and store the result in a mask register.

Conversion instructions

[edit]
Instruction Description
VCVTW2PH Convert packed signed 16-bit integers to FP16 numbers.
VCVTUW2PH Convert packed unsigned 16-bit integers to FP16 numbers.
VCVTDQ2PH Convert packed signed 32-bit integers to FP16 numbers.
VCVTUDQ2PH Convert packed unsigned 32-bit integers to FP16 numbers.
VCVTQQ2PH Convert packed signed 64-bit integers to FP16 numbers.
VCVTUQQ2PH Convert packed unsigned 64-bit integers to FP16 numbers.
VCVTPS2PHX Convert packed FP32 numbers to FP16 numbers. Unlike VCVTPS2PH from F16C, VCVTPS2PHX has a different encoding that also supports broadcasting.
VCVTPD2PH Convert packed FP64 numbers to FP16 numbers.
VCVTSI2SH Convert a scalar signed 32-bit or 64-bit integer to FP16 number.
VCVTUSI2SH Convert a scalar unsigned 32-bit or 64-bit integer to FP16 number.
VCVTSS2SH Convert a scalar FP32 number to FP16 number.
VCVTSD2SH Convert a scalar FP64 number to FP16 number.
VCVTPH2W, VCVTTPH2W Convert packed FP16 numbers to signed 16-bit integers. VCVTPH2W rounds the value according to the MXCSR register. VCVTTPH2W rounds toward zero.
VCVTPH2UW, VCVTTPH2UW Convert packed FP16 numbers to unsigned 16-bit integers. VCVTPH2UW rounds the value according to the MXCSR register. VCVTTPH2UW rounds toward zero.
VCVTPH2DQ, VCVTTPH2DQ Convert packed FP16 numbers to signed 32-bit integers. VCVTPH2DQ rounds the value according to the MXCSR register. VCVTTPH2DQ rounds toward zero.
VCVTPH2UDQ, VCVTTPH2UDQ Convert packed FP16 numbers to unsigned 32-bit integers. VCVTPH2UDQ rounds the value according to the MXCSR register. VCVTTPH2UDQ rounds toward zero.
VCVTPH2QQ, VCVTTPH2QQ Convert packed FP16 numbers to signed 64-bit integers. VCVTPH2QQ rounds the value according to the MXCSR register. VCVTTPH2QQ rounds toward zero.
VCVTPH2UQQ, VCVTTPH2UQQ Convert packed FP16 numbers to unsigned 64-bit integers. VCVTPH2UQQ rounds the value according to the MXCSR register. VCVTTPH2UQQ rounds toward zero.
VCVTPH2PSX Convert packed FP16 numbers to FP32 numbers. Unlike VCVTPH2PS from F16C, VCVTPH2PSX has a different encoding that also supports broadcasting.
VCVTPH2PD Convert packed FP16 numbers to FP64 numbers.
VCVTSH2SI, VCVTTSH2SI Convert a scalar FP16 number to signed 32-bit or 64-bit integer. VCVTSH2SI rounds the value according to the MXCSR register. VCVTTSH2SI rounds toward zero.
VCVTSH2USI, VCVTTSH2USI Convert a scalar FP16 number to unsigned 32-bit or 64-bit integer. VCVTSH2USI rounds the value according to the MXCSR register. VCVTTSH2USI rounds toward zero.
VCVTSH2SS Convert a scalar FP16 number to FP32 number.
VCVTSH2SD Convert a scalar FP16 number to FP64 number.

Decomposition instructions

[edit]
Instruction Description
VGETEXPPH, VGETEXPSH Extract exponent components of packed/scalar FP16 numbers as FP16 numbers.
VGETMANTPH, VGETMANTSH Extract mantissa components of packed/scalar FP16 numbers as FP16 numbers.

Move instructions

[edit]
Instruction Description
VMOVSH Move scalar FP16 number to/from memory or between vector registers.
VMOVW Move scalar FP16 number to/from memory or general purpose register.

Legacy instructions with EVEX-encoded versions

[edit]
Group Legacy encoding Instructions AVX-512
extensions
SSE
SSE2
MMX 		AVX
SSE3
SSE4 		AVX2
FMA 		F VL BW DQ
VADD Yes Yes No VADDPD, VADDPS, VADDSD, VADDSS Y Y N N
VAND VANDPD, VANDPS, VANDNPD, VANDNPS N Y
VCMP VCMPPD, VCMPPS, VCMPSD, VCMPSS Y N N
VCOM VCOMISD, VCOMISS
VDIV VDIVPD, VDIVPS, VDIVSD, VDIVSS Y
VCVT VCVTDQ2PD, VCVTDQ2PS, VCVTPD2DQ, VCVTPD2PS, VCVTPH2PS, VCVTPS2PH, VCVTPS2DQ, VCVTPS2PD, VCVTSD2SI, VCVTSD2SS, VCVTSI2SD, VCVTSI2SS, VCVTSS2SD, VCVTSS2SI, VCVTTPD2DQ, VCVTTPS2DQ, VCVTTSD2SI, VCVTTSS2SI
VMAX VMAXPD, VMAXPS, VMAXSD, VMAXSS
VMIN VMINPD, VMINPS, VMINSD, VMINSS N
VMOV VMOVAPD, VMOVAPS, VMOVD, VMOVQ, VMOVDDUP, VMOVHLPS, VMOVHPD, VMOVHPS, VMOVLHPS, VMOVLPD, VMOVLPS, VMOVNTDQA, VMOVNTDQ, VMOVNTPD, VMOVNTPS, VMOVSD, VMOVSHDUP, VMOVSLDUP, VMOVSS, VMOVUPD, VMOVUPS, VMOVDQA32, VMOVDQA64, VMOVDQU8, VMOVDQU16, VMOVDQU32, VMOVDQU64 Y Y
VMUL VMULPD, VMULPS, VMULSD, VMULSS N
VOR VORPD, VORPS N Y
VSQRT VSQRTPD, VSQRTPS, VSQRTSD, VSQRTSS Y N
VSUB VSUBPD, VSUBPS, VSUBSD, VSUBSS
VUCOMI VUCOMISD, VUCOMISS N
VUNPCK VUNPCKHPD, VUNPCKHPS, VUNPCKLPD, VUNPCKLPS Y
VXOR VXORPD, VXORPS N Y
VEXTRACTPS No Yes No VEXTRACTPS Y N N
VINSERTPS VINSERTPS
VPEXTR VPEXTRB, VPEXTRW, VPEXTRD, VPEXTRQ N Y Y
VPINSR VPINSRB, VPINSRW, VPINSRD, VPINSRQ
VPACK Yes Yes Yes VPACKSSWB, VPACKSSDW, VPACKUSDW, VPACKUSWB Y N
VPADD VPADDB, VPADDW, VPADDD, VPADDQ, VPADDSB, VPADDSW, VPADDUSB, VPADDUSW Y
VPAND VPANDD, VPANDQ, VPANDND, VPANDNQ N
VPAVG VPAVGB, VPAVGW N Y
VPCMP VPCMPEQB, VPCMPEQW, VPCMPEQD, VPCMPEQQ, VPCMPGTB, VPCMPGTW, VPCMPGTD, VPCMPGTQ Y
VPMAX VPMAXSB, VPMAXSW, VPMAXSD, VPMAXSQ, VPMAXUB, VPMAXUW, VPMAXUD, VPMAXUQ
VPMIN VPMINSB, VPMINSW, VPMINSD, VPMINSQ, VPMINUB, VPMINUW, VPMINUD, VPMINUQ
VPMOV VPMOVSXBW, VPMOVSXBD, VPMOVSXBQ, VPMOVSXWD, VPMOVSXWQ, VPMOVSXDQ, VPMOVZXBW, VPMOVZXBD, VPMOVZXBQ, VPMOVZXWD, VPMOVZXWQ, VPMOVZXDQ
VPMUL VPMULDQ, VPMULUDQ, VPMULHRSW, VPMULHUW, VPMULHW, VPMULLD, VPMULLQ, VPMULLW
VPOR VPORD, VPORQ N
VPSUB VPSUBB, VPSUBW, VPSUBD, VPSUBQ, VPSUBSB, VPSUBSW, VPSUBUSB, VPSUBUSW Y
VPUNPCK VPUNPCKHBW, VPUNPCKHWD, VPUNPCKHDQ, VPUNPCKHQDQ, VPUNPCKLBW, VPUNPCKLWD, VPUNPCKLDQ, VPUNPCKLQDQ
VPXOR VPXORD, VPXORQ N
VPSADBW VPSADBW N Y
VPSHUF VPSHUFB, VPSHUFHW, VPSHUFLW, VPSHUFD, VPSLLDQ, VPSLLW, VPSLLD, VPSLLQ, VPSRAW, VPSRAD, VPSRAQ, VPSRLDQ, VPSRLW, VPSRLD, VPSRLQ, VPSLLVW, VPSLLVD, VPSLLVQ, VPSRLVW, VPSRLVD, VPSRLVQ, VPSHUFPD, VPSHUFPS Y
VEXTRACT No Yes Yes VEXTRACTF32X4, VEXTRACTF64X2, VEXTRACTF32X8, VEXTRACTF64X4, VEXTRACTI32X4, VEXTRACTI64X2, VEXTRACTI32X8, VEXTRACTI64X4 N Y
VINSERT VINSERTF32x4, VINSERTF64X2, VINSERTF32X8, VINSERTF64x4, VINSERTI32X4, VINSERTI64X2, VINSERTI32X8, VINSERTI64X4
VPABS VPABSB, VPABSW, VPABSD, VPABSQ Y N
VPALIGNR VPALIGNR N
VPERM VPERMD, VPERMILPD, VPERMILPS, VPERMPD, VPERMPS, VPERMQ Y N
VPMADD VPMADDUBSW VPMADDWD N Y
VFMADD No No Yes VFMADD132PD, VFMADD213PD, VFMADD231PD, VFMADD132PS, VFMADD213PS, VFMADD231PS, VFMADD132SD, VFMADD213SD, VFMADD231SD, VFMADD132SS, VFMADD213SS, VFMADD231SS Y N
VFMADDSUB VFMADDSUB132PD, VFMADDSUB213PD, VFMADDSUB231PD, VFMADDSUB132PS, VFMADDSUB213PS, VFMADDSUB231PS
VFMSUBADD VFMSUBADD132PD, VFMSUBADD213PD, VFMSUBADD231PD, VFMSUBADD132PS, VFMSUBADD213PS, VFMSUBADD231PS
VFMSUB VFMSUB132PD, VFMSUB213PD, VFMSUB231PD, VFMSUB132PS, VFMSUB213PS, VFMSUB231PS, VFMSUB132SD, VFMSUB213SD, VFMSUB231SD, VFMSUB132SS, VFMSUB213SS, VFMSUB231SS
VFNMADD VFNMADD132PD, VFNMADD213PD, VFNMADD231PD, VFNMADD132PS, VFNMADD213PS, VFNMADD231PS, VFNMADD132SD, VFNMADD213SD, VFNMADD231SD, VFNMADD132SS, VFNMADD213SS, VFNMADD231SS
VFNMSUB VFNMSUB132PD, VFNMSUB213PD, VFNMSUB231PD, VFNMSUB132PS, VFNMSUB213PS, VFNMSUB231PS, VFNMSUB132SD, VFNMSUB213SD, VFNMSUB231SD, VFNMSUB132SS, VFNMSUB213SS, VFNMSUB231SS
VGATHER VGATHERDPS, VGATHERDPD, VGATHERQPS, VGATHERQPD
VPGATHER VPGATHERDD, VPGATHERDQ, VPGATHERQD, VPGATHERQQ
VPSRAV VPSRAVW, VPSRAVD, VPSRAVQ Y

CPUs with AVX-512

[edit]
  • Intel
    • Knights Landing (Xeon Phi x200):[1][15] AVX-512 F, CD, ER, PF
    • Knights Mill (Xeon Phi x205):[8] AVX-512 F, CD, ER, PF, 4FMAPS, 4VNNIW, VPOPCNTDQ
    • Skylake-SP, Skylake-X:[16][17][18] AVX-512 F, CD, VL, DQ, BW
    • Cannon Lake:[8] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI
    • Cascade Lake: AVX-512 F, CD, VL, DQ, BW, VNNI
    • Cooper Lake: AVX-512 F, CD, VL, DQ, BW, VNNI, BF16
    • Ice Lake,[8] Rocket Lake:[19][20] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES
    • Tiger Lake (except Pentium and Celeron but some reviewer have the CPU-Z Screenshot of Celeron 6305 with AVX-512 support[21][22]):[23] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES, VP2INTERSECT
    • Alder Lake (never officially supported by Intel, completely removed in newer CPUsNote 1):[24][25] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES, BF16, VP2INTERSECT, FP16
    • Sapphire Rapids[26] and later P-core-only Xeon processors: AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES, BF16, FP16
  • Centaur Technology
    • "CNS" core (8c/8t):[27][28] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI
  • AMD
    • Zen 4:[29][30][31][32][33] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES, BF16
    • Zen 5:[34] AVX-512 F, CD, VL, DQ, BW, IFMA, VBMI, VBMI2, VPOPCNTDQ, BITALG, VNNI, VPCLMULQDQ, GFNI, VAES, BF16, VP2INTERSECT
Subset
F
CD
ER
PF
4FMAPS
4VNNIW
VPOPCNTDQ
VL
DQ
BW
IFMA
VBMI
VNNI
BF16
VBMI2
BITALG
VPCLMULQDQ
GFNI
VAES
VP2INTERSECT
FP16
Knights Landing (Xeon Phi x200, 2016) Yes Yes No
Knights Mill (Xeon Phi x205, 2017) Yes No
Skylake-SP, Skylake-X (2017) No No Yes No
Cannon Lake (2018) Yes No
Cascade Lake (2019) No Yes No
Cooper Lake (2020) Yes No
Ice Lake (2019) Yes No Yes No
Tiger Lake (2020) Yes No
Rocket Lake (2021) No
Alder Lake (2021) PartialNote 1 PartialNote 1
Zen 4 (2022) Yes Yes No
Sapphire Rapids (2023) No Yes
Zen 5 (2024) Yes No

^Note 1 : Intel does not officially support AVX-512 family of instructions on the Alder Lake microprocessors. In early 2022, Intel began disabling in silicon (fusing off) AVX-512 in Alder Lake microprocessors to prevent customers from enabling AVX-512.[35] In older Alder Lake family CPUs with some legacy combinations of BIOS and microcode revisions, it was possible to execute AVX-512 family instructions when disabling all the efficiency cores which do not contain the silicon for AVX-512.[36][37][24]

Performance

[edit]
See also: Advanced Vector Extensions § Downclocking

Intel Vectorization Advisor (starting from version 2017) supports native AVX-512 performance and vector code quality analysis (for "Core", Xeon and Intel Xeon Phi processors). Along with traditional hotspots profile, Advisor Recommendations and "seamless" integration of Intel Compiler vectorization diagnostics, Advisor Survey analysis also provides AVX-512 ISA metrics and new AVX-512-specific "traits", e.g. Scatter, Compress/Expand, mask utilization.[38][39]

On some processors (mostly pre-Ice Lake Intel), AVX-512 instructions can cause a frequency throttling even greater than its predecessors, causing a penalty for mixed workloads. The additional downclocking is triggered by the 512-bit width of vectors and depends on the nature of instructions being executed; using the 128 or 256-bit part of AVX-512 (AVX-512VL) does not trigger it. As a result, gcc and clang default to prefer using the 256-bit vectors for Intel targets.[40][41][42]

C/C++ compilers also automatically handle loop unrolling and preventing stalls in the pipeline in order to use AVX-512 most effectively, which means a programmer using language intrinsics to try to force use of AVX-512 can sometimes result in worse performance relative to the code generated by the compiler when it encounters loops plainly written in the source code.[43] In other cases, using AVX-512 intrinsics in C/C++ code can result in a performance improvement relative to plainly written C/C++.[44]

Reception

[edit]

There are many examples of AVX-512 applications, including media processing, cryptography, video games,[45] neural networks,[46] and even OpenJDK, which employs AVX-512 for sorting.[47]

In a much-cited quote from 2020, Linus Torvalds said "I hope AVX-512 dies a painful death, and that Intel starts fixing real problems instead of trying to create magic instructions to then create benchmarks that they can look good on,"[48] stating that he would prefer the transistor budget be spent on additional cores and integer performance instead, and that he "detests" floating point benchmarks.[49]

Numenta touts their "highly sparse"[50] neural network technology, which they say obviates the need for GPUs as their algorithms run on CPUs with AVX-512.[51] They claim a ten times speedup relative to A100 largely because their algorithms reduce the size of the neural network, while maintaining accuracy, by techniques such as the Sparse Evolutionary Training (SET) algorithm[52] and Foresight Pruning.[53]

See also

[edit]

References

[edit]
  • Kusswurm, Daniel (2022). Modern parallel programming with C++ and Assembly language : X86 SIMD development using AVX, AVX2, and AVX-512. New York, NY: Apress Media LLC. ISBN 978-1-4842-7918-2. OCLC 1304243196.
  1. ^ a b c d e f James Reinders (23 July 2013). "AVX-512 Instructions". Intel. Retrieved 20 August 2013.
  2. ^ a b Kusswurm 2022, p. 223.
  3. ^ Bonshor, Gavin (2023-07-25). "Intel Unveils AVX10 and APX Instruction Sets: Unifying AVX-512 For Hybrid Architectures". AnandTech. Retrieved 2024-08-21.
  4. ^ a b c James Reinders (17 July 2014). "Additional AVX-512 instructions". Intel. Retrieved 3 August 2014.
  5. ^ Anton Shilov. "Intel 'Skylake' processors for PCs will not support AVX-512 instructions". Kitguru.net. Retrieved 2015-03-17.
  6. ^ "Intel will add deep-learning instructions to its processors". 14 October 2016.
  7. ^ a b c d e f g h "Intel Architecture Instruction Set Extensions Programming Reference" (PDF). Intel. Retrieved 2014-01-29.
  8. ^ a b c d e f "Intel Architecture Instruction Set Extensions and Future Features Programming Reference". Intel. Retrieved 2017-10-16.
  9. ^ "Inline asm docs incorrectly state that k0 (X86 AVX-512) is hard-wired to zero · Issue #94977 · rust-lang/Rust". GitHub.
  10. ^ "AVX-512 Architecture/Demikhovsky Poster" (PDF). Intel. Retrieved 25 February 2014.
  11. ^ "Intel® Deep Learning Boost" (PDF). Intel. Retrieved 2021-10-11.
  12. ^ a b "Galois Field New Instructions (GFNI) Technology Guide". networkbuilders.intel.com.
  13. ^ Kivilinna, Jussi (19 April 2023). "camellia-simd-aesni". GitHub. Newer x86-64 processors also support Galois Field New Instructions (GFNI) which allow implementing Camellia s-box more straightforward manner and yield even better performance.
  14. ^ "Intel® AVX512-FP16 Architecture Specification, June 2021, Revision 1.0, Ref. 347407-001US" (PDF). Intel. 2021-06-30. Retrieved 2021-07-04.
  15. ^ "Intel Xeon Phi Processor product brief". Intel. Retrieved 12 October 2016.
  16. ^ "Intel unveils X-series platform: Up to 18 cores and 36 threads, from $242 to $2,000". Ars Technica. Retrieved 2017-05-30.
  17. ^ "Intel Advanced Vector Extensions 2015/2016: Support in GNU Compiler Collection" (PDF). Gcc.gnu.org. Retrieved 2016-10-20.
  18. ^ Patrizio, Andy (21 September 2015). "Intel's Xeon roadmap for 2016 leaks". Itworld.org. Archived from the original on 2016-10-21. Retrieved 2016-10-20.
  19. ^ "Intel Core i9-11900K Review - World's Fastest Gaming Processor?". www.techpowerup.com. 30 March 2021.
  20. ^ ""Add rocketlake to gcc" commit". gcc.gnu.org.
  21. ^ "Intel Celeron 6305 Processor (4M Cache, 1.80 GHz, with IPU) Product Specifications". ark.intel.com. Archived from the original on 2020-10-18. Retrieved 2020-11-10.
  22. ^ Laptop Murah Kinerja Boleh Diadu | HP 14S DQ2518TU, 18 June 2021, retrieved 2021-08-08
  23. ^ "Using the GNU Compiler Collection (GCC): x86 Options". GNU. Retrieved 2019-10-14.
  24. ^ a b Cutress, Ian; Frumusanu, Andrei. "The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity". www.anandtech.com. Retrieved 5 November 2021.
  25. ^ Larabel, Michael. "Intel Core i9 12900K "Alder Lake" AVX-512 On Linux". www.phoronix.com. Retrieved 2021-11-08.
  26. ^ Larabel, Michael. "AVX-512 Performance Comparison: AMD Genoa vs. Intel Sapphire Rapids & Ice Lake". www.phoronix.com. Retrieved 2023-01-19.
  27. ^ "The industry's first high-performance x86 SOC with server-class CPUs and integrated AI coprocessor technology". 2 August 2022. Archived from the original on December 12, 2019.{{cite web}}: CS1 maint: unfit URL (link)
  28. ^ "x86, x64 Instruction Latency, Memory Latency and CPUID dumps (instlatx64)". users.atw.hu.
  29. ^ "AMD Zen 4 Based Ryzen CPUs May Feature Up to 24 Cores, Support for AVX512 Vectors". Hardware Times. 2021-05-23. Retrieved 2021-09-02.
  30. ^ Hagedoorn, Hilbert (18 May 2021). "AMD working on a prodigious 96-core EPYC processor". Guru3D.com. Retrieved 2021-05-25.
  31. ^ clamchowder (2021-08-23). "Details on the Gigabyte Leak". Chips And Cheese. Retrieved 2022-06-10.
  32. ^ W1zzard (26 May 2022). "AMD Answers Our Zen 4 Tech Questions, with Robert Hallock". TechPowerUp. Retrieved 2022-05-29.
  33. ^ Larabel, Michael (2022-09-26). "AMD Zen 4 AVX-512 Performance Analysis On The Ryzen 9 7950X". www.phoronix.com.
  34. ^ Larabel, Michael (2024-02-10). "AMD Zen 5 Compiler Support Posted For GCC - Confirms New AVX Features & More". www.phoronix.com.
  35. ^ Alcorn, Paul (2022-03-02). "Intel Nukes Alder Lake's AVX-512 Support, Now Fuses It Off in Silicon". Tom's Hardware. Retrieved 2022-03-07.
  36. ^ Cutress, Ian; Frumusanu, Andrei (2021-08-19). "Intel Architecture Day 2021: Alder Lake, Golden Cove, and Gracemont Detailed". AnandTech. Retrieved 2021-08-25.
  37. ^ Alcorn, Paul (2021-08-19). "Intel Architecture Day 2021: Alder Lake Chips, Golden Cove and Gracemont Cores". Tom's Hardware. Retrieved 2021-08-21.
  38. ^ "Intel Advisor XE 2016 Update 3 What's new - Intel Software". Software.intel.com. Retrieved 2016-10-20.
  39. ^ "Intel Advisor - Intel Software". Software.intel.com. Retrieved 2016-10-20.
  40. ^ Cordes, Peter. "SIMD instructions lowering CPU frequency". Stack Overflow.
  41. ^ Cordes, Peter. "why does gcc auto-vectorization for tigerlake use ymm not zmm registers". Stack Overflow.
  42. ^ "LLVM 10.0.0 Release Notes".
  43. ^ Matthew Kolbe (2023-10-10). Lightning Talk: How to Leverage SIMD Intrinsics for Massive Slowdowns - Matthew Kolbe - CppNow 2023. C++Now. Retrieved 2023-10-15 – via YouTube.
  44. ^ Clausecker, Robert (2023-08-05). "Transcoding unicode characters with AVX-512 instructions". Software: Practice and Experience. 53 (12): 2430–2462. arXiv:2212.05098. doi:10.1002/spe.3261.
  45. ^ Szewczyk, Chris (2021-11-24). "The RPCS3 PS3 emulator gets a hefty boost on Intel Alder Lake CPUs with AVX-512 enabled". PC Gamer. Retrieved 2023-10-11.
  46. ^ Carneiro, André; Serpa, Matheus (2021-09-05). "Lightweight Deep Learning Applications on AVX-512". 2021 IEEE Symposium on Computers and Communications (ISCC). Athens: IEEE. pp. 1–6. doi:10.1109/ISCC53001.2021.9631464.
  47. ^ Parasa, Srinivas (2023-05-30). "JDK-8309130: x86_64 AVX512 intrinsics for Arrays.sort methods (int, long, float and double arrays)". OpenJDK. Retrieved 2023-10-11.
  48. ^ Tung, Liam (2020-07-13). "Linus Torvalds: I hope Intel's AVX-512 dies a painful death". ZDNet. Retrieved 2023-10-11.
  49. ^ Torvalds, Linus (2020-07-11). "Alder Lake and AVX-512". realworldtech.com. Retrieved 2023-10-11.
  50. ^ "Sparsity Enables 100x Performance Acceleration in Deep Learning Networks: A Technology Demonstration" (PDF). numenta.com. 2021-05-20. Retrieved 2023-10-11.
  51. ^ Afifi-Sabet, Keumars (2023-10-08). "A tiny startup has helped Intel trounce AMD and Nvidia in critical AI tests — is it game over already?". TechRadar. Retrieved 2023-10-11.
  52. ^ Mocanu, Decebal; Mocanu, Elena (2018). "Scalable training of artificial neural networks with adaptive sparse connectivity inspired by network science". Nature Communications. 9 (1): 2383. Bibcode:2018NatCo...9.2383M. doi:10.1038/s41467-018-04316-3. PMC 6008460. PMID 29921910.
  53. ^ Souza, Lucas (2020-10-30). "The Case for Sparsity in Neural Networks, Part 2: Dynamic Sparsity". numenta.com. Retrieved 2023-10-11.
Retrieved from "https://wiki.riteme.site/w/index.php?title=AVX-512&oldid=1258475329"