From WikiChip
Difference between revisions of "x86/avx512 vnni"
< x86

(Replaced support matrix, added missing intrinsics.)
 
(5 intermediate revisions by 2 users not shown)
Line 14: Line 14:
  
 
:[[File:vnni-vpdpwssd.svg|600px]]
 
:[[File:vnni-vpdpwssd.svg|600px]]
−
+
 
 
Likewise, for 8-bit values, three instructions are needed - <code>VPMADDUBSW</code> which is used to multiply two 8-bit pairs and add them together, followed by a <code>VPMADDWD</code> with the value <code>1</code> in order to simply up-convert the 16-bit values to  32-bit values, followed by the <code>VPADDD</code> instruction which adds the result to an accumulator.
 
Likewise, for 8-bit values, three instructions are needed - <code>VPMADDUBSW</code> which is used to multiply two 8-bit pairs and add them together, followed by a <code>VPMADDWD</code> with the value <code>1</code> in order to simply up-convert the 16-bit values to  32-bit values, followed by the <code>VPADDD</code> instruction which adds the result to an accumulator.
  
Line 24: Line 24:
  
 
== Detection ==
 
== Detection ==
 +
Support for these instructions is indicated by the AVX512_VNNI feature flag. 128- and 256-bit vectors are supported if the AVX512VL flag is set as well.
 +
 +
The {{x86|AVX-VNNI}} extension adds AVX (VEX encoded) versions of these instructions operating on 128- and 256-bit vectors.
 +
 
{| class="wikitable"
 
{| class="wikitable"
−
! colspan="2" | {{x86|CPUID}} !! rowspan="2" | Instruction Set  
+
! colspan="2" | {{x86|CPUID}} !! rowspan="2" | Instruction Set
 
|-
 
|-
 
! Input !! Output
 
! Input !! Output
 
|-
 
|-
−
| rowspan="14" | EAX=07H, ECX=0 || ECX[bit 11] || AVX512VNNI
+
| EAX=07H, ECX=0 || EBX[bit 31] || AVX512VL
 +
|-
 +
| EAX=07H, ECX=0 || ECX[bit 11] || AVX512_VNNI
 
|}
 
|}
  
 
== Microarchitecture support ==
 
== Microarchitecture support ==
−
{| class="wikitable"
+
<!-- Wrong/incomplete? Visit https://en.wikichip.org/wiki/Template:avx512_support_matrix -->
−
|-
+
{{avx512 support matrix|em=VL+VNNI}}
−
! Instructions !! Introduction
 
−
|-
 
−
| AVX512VNNI || {{intel|Cascade Lake|l=arch}}
 
−
|}
 
  
 
== Intrinsic functions ==
 
== Intrinsic functions ==
−
<source lang=asm>
+
<source lang=c>
−
# vpdpbusd
+
// VPDPBUSD
−
__m512i _mm512_dpbusd_epi32 (__m512i src, __m512i a, __m512i b)
+
__m128i _mm_dpbusd_epi32(__m128i, __m128i, __m128i);
−
__m512i _mm512_mask_dpbusd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b)
+
__m128i _mm_mask_dpbusd_epi32(__m128i, __mmask8, __m128i, __m128i);
−
__m512i _mm512_maskz_dpbusd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b)
+
__m128i _mm_maskz_dpbusd_epi32(__mmask8, __m128i, __m128i, __m128i);
−
# vpdpbusds
+
__m256i _mm256_dpbusd_epi32(__m256i, __m256i, __m256i);
−
__m512i _mm512_dpbusds_epi32 (__m512i src, __m512i a, __m512i b)
+
__m256i _mm256_mask_dpbusd_epi32(__m256i, __mmask8, __m256i, __m256i);
−
__m512i _mm512_mask_dpbusds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b)
+
__m256i _mm256_maskz_dpbusd_epi32(__mmask8, __m256i, __m256i, __m256i);
−
__m512i _mm512_maskz_dpbusds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b)
+
__m512i _mm512_dpbusd_epi32 (__m512i src, __m512i a, __m512i b);
−
# vpdpwssd
+
__m512i _mm512_mask_dpbusd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
−
__m512i _mm512_dpwssd_epi32 (__m512i src, __m512i a, __m512i b)
+
__m512i _mm512_maskz_dpbusd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
−
__m512i _mm512_mask_dpwssd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b)
+
// VPDPBUSDS
−
__m512i _mm512_maskz_dpwssd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b)
+
__m128i _mm_dpbusds_epi32(__m128i, __m128i, __m128i);
−
# vpdpwssds
+
__m128i _mm_mask_dpbusds_epi32(__m128i, __mmask8, __m128i, __m128i);
−
__m512i _mm512_dpwssds_epi32 (__m512i src, __m512i a, __m512i b)
+
__m128i _mm_maskz_dpbusds_epi32(__mmask8, __m128i, __m128i, __m128i);
−
__m512i _mm512_mask_dpwssds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b)
+
__m256i _mm256_dpbusds_epi32(__m256i, __m256i, __m256i);
−
__m512i _mm512_maskz_dpwssds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b)
+
__m256i _mm256_mask_dpbusds_epi32(__m256i, __mmask8, __m256i, __m256i);
 +
__m256i _mm256_maskz_dpbusds_epi32(__mmask8, __m256i, __m256i, __m256i);
 +
__m512i _mm512_dpbusds_epi32 (__m512i src, __m512i a, __m512i b);
 +
__m512i _mm512_mask_dpbusds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
 +
__m512i _mm512_maskz_dpbusds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
 +
// VPDPWSSD
 +
__m128i _mm_dpwssd_epi32(__m128i, __m128i, __m128i);
 +
__m128i _mm_mask_dpwssd_epi32(__m128i, __mmask8, __m128i, __m128i);
 +
__m128i _mm_maskz_dpwssd_epi32(__mmask8, __m128i, __m128i, __m128i);
 +
__m256i _mm256_dpwssd_epi32(__m256i, __m256i, __m256i);
 +
__m256i _mm256_mask_dpwssd_epi32(__m256i, __mmask8, __m256i, __m256i);
 +
__m256i _mm256_maskz_dpwssd_epi32(__mmask8, __m256i, __m256i, __m256i);
 +
__m512i _mm512_dpwssd_epi32 (__m512i src, __m512i a, __m512i b);
 +
__m512i _mm512_mask_dpwssd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
 +
__m512i _mm512_maskz_dpwssd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
 +
// VPDPWSSDS
 +
__m128i _mm_dpwssds_epi32(__m128i, __m128i, __m128i);
 +
__m128i _mm_mask_dpwssds_epi32(__m128i, __mmask8, __m128i, __m128i);
 +
__m128i _mm_maskz_dpwssds_epi32(__mmask8, __m128i, __m128i, __m128i);
 +
__m256i _mm256_dpwssds_epi32(__m256i, __m256i, __m256i);
 +
__m256i _mm256_mask_dpwssds_epi32(__m256i, __mmask8, __m256i, __m256i);
 +
__m256i _mm256_maskz_dpwssds_epi32(__mmask8, __m256i, __m256i, __m256i);
 +
__m512i _mm512_dpwssds_epi32 (__m512i src, __m512i a, __m512i b);
 +
__m512i _mm512_mask_dpwssds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
 +
__m512i _mm512_maskz_dpwssds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
 
</source>
 
</source>
  
 
== Bibliography ==
 
== Bibliography ==
−
* {{hcbib|30}}
+
* {{bib|hc|30|Intel}}
 
* Rodriguez, Andres, et al. "[https://ai.intel.com/nervana/wp-content/uploads/sites/53/2018/05/Lower-Numerical-Precision-Deep-Learning-Inference-Training.pdf Lower numerical precision deep learning inference and training]." Intel White Paper (2018).
 
* Rodriguez, Andres, et al. "[https://ai.intel.com/nervana/wp-content/uploads/sites/53/2018/05/Lower-Numerical-Precision-Deep-Learning-Inference-Training.pdf Lower numerical precision deep learning inference and training]." Intel White Paper (2018).
 +
* {{cite techdoc|title=Intel® Architecture Instruction Set Extensions and Future Features|url=https://www.intel.com/content/www/us/en/developer/articles/technical/intel-sdm.html|publ=Intel|pid=319433|rev=047|date=2022-12}}
  
−
[[category:x86]]
+
[[Category:x86_extensions]]

Latest revision as of 15:52, 15 March 2023

x86
Instruction Set Architecture
General
Variants
Topics
CPUIDs
Modes
Extensions(all)
v · d · e

AVX-512 Vector Neural Network Instructions (AVX512 VNNI) is an x86 extension, part of the AVX-512, designed to accelerate convolutional neural network-based algorithms.

Overview[edit]

The AVX512 VNNI x86 extension extends AVX-512 Foundation by introducing four new instructions for accelerating inner convolutional neural network loops.

  • VPDPBUSD - Multiplies the individual bytes (8-bit) of the first source operand by the corresponding bytes (8-bit) of the second source operand, producing intermediate word (16-bit) results which are summed and accumulated in the double word (32-bit) of the destination operand.
    • VPDPBUSDS - Same as above except on intermediate sum overflow which saturates to 0x7FFF_FFFF/0x8000_0000 for positive/negative numbers.
  • VPDPWSSD - Multiplies the individual words (16-bit) of the first source operand by the corresponding word (16-bit) of the second source operand, producing intermediate word results which are summed and accumulated in the double word (32-bit) of the destination operand.
    • VPDPWSSDS - Same as above except on intermediate sum overflow which saturates to 0x7FFF_FFFF/0x8000_0000 for positive/negative numbers.

Motivation[edit]

The major motivation behind the AVX512 VNNI extension is the observation that many tight convolutional neural network loops require the repeated multiplication of two 16-bit values or two 8-bit values and accumulate the result to a 32-bit accumulator. Using the foundation AVX-512, for 16-bit, this is possible using two instructions - VPMADDWD which is used to multiply two 16-bit pairs and add them together followed a VPADDD which adds the accumulate value.

vnni-vpdpwssd.svg

Likewise, for 8-bit values, three instructions are needed - VPMADDUBSW which is used to multiply two 8-bit pairs and add them together, followed by a VPMADDWD with the value 1 in order to simply up-convert the 16-bit values to 32-bit values, followed by the VPADDD instruction which adds the result to an accumulator.

vnni-vpdpbusd.svg

To address those two common operations, two new instructions were added (as well as two saturated versions): VPDPBUSD fuses VPMADDUBSW, VPMADDWD, and VPADDD and VPDPWSSD fuses VPMADDWD and VPADDD.

vnni-vpdpbusd-i.svg vnni-vpdpwssd-i.svg

Detection[edit]

Support for these instructions is indicated by the AVX512_VNNI feature flag. 128- and 256-bit vectors are supported if the AVX512VL flag is set as well.

The AVX-VNNI extension adds AVX (VEX encoded) versions of these instructions operating on 128- and 256-bit vectors.

CPUID Instruction Set
Input Output
EAX=07H, ECX=0 EBX[bit 31] AVX512VL
EAX=07H, ECX=0 ECX[bit 11] AVX512_VNNI

Microarchitecture support[edit]

Designer Microarchitecture Year Support Level
F CD ER PF BW DQ VL FP16 IFMA VBMI VBMI2 BITALG VPOPCNTDQ VP2INTERSECT 4VNNIW 4FMAPS VNNI BF16
Intel Knights Landing 2016 ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘
Knights Mill 2017 ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✔ ✘ ✔ ✔ ✘ ✘
Skylake (server) 2017 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘
Cannon Lake 2018 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘
Cascade Lake 2019 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✔ ✘
Cooper Lake 2020 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✔ ✔
Tiger Lake 2020 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✔ ✘
Rocket Lake 2021 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✔ ✘
Alder Lake 2021 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✔ ✔
Ice Lake (server) 2021 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✔ ✘
Sapphire Rapids 2023 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✔ ✔
AMD Zen 4 2022 ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✔ ✔ ✔ ✘ ✘ ✘ ✔ ✔
Centaur CHA ✔ ✔ ✘ ✘ ✔ ✔ ✔ ✘ ✔ ✔ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘

Intrinsic functions[edit]

// VPDPBUSD
__m128i _mm_dpbusd_epi32(__m128i, __m128i, __m128i);
__m128i _mm_mask_dpbusd_epi32(__m128i, __mmask8, __m128i, __m128i);
__m128i _mm_maskz_dpbusd_epi32(__mmask8, __m128i, __m128i, __m128i);
__m256i _mm256_dpbusd_epi32(__m256i, __m256i, __m256i);
__m256i _mm256_mask_dpbusd_epi32(__m256i, __mmask8, __m256i, __m256i);
__m256i _mm256_maskz_dpbusd_epi32(__mmask8, __m256i, __m256i, __m256i);
__m512i _mm512_dpbusd_epi32 (__m512i src, __m512i a, __m512i b);
__m512i _mm512_mask_dpbusd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
__m512i _mm512_maskz_dpbusd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
// VPDPBUSDS
__m128i _mm_dpbusds_epi32(__m128i, __m128i, __m128i);
__m128i _mm_mask_dpbusds_epi32(__m128i, __mmask8, __m128i, __m128i);
__m128i _mm_maskz_dpbusds_epi32(__mmask8, __m128i, __m128i, __m128i);
__m256i _mm256_dpbusds_epi32(__m256i, __m256i, __m256i);
__m256i _mm256_mask_dpbusds_epi32(__m256i, __mmask8, __m256i, __m256i);
__m256i _mm256_maskz_dpbusds_epi32(__mmask8, __m256i, __m256i, __m256i);
__m512i _mm512_dpbusds_epi32 (__m512i src, __m512i a, __m512i b);
__m512i _mm512_mask_dpbusds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
__m512i _mm512_maskz_dpbusds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
// VPDPWSSD
__m128i _mm_dpwssd_epi32(__m128i, __m128i, __m128i);
__m128i _mm_mask_dpwssd_epi32(__m128i, __mmask8, __m128i, __m128i);
__m128i _mm_maskz_dpwssd_epi32(__mmask8, __m128i, __m128i, __m128i);
__m256i _mm256_dpwssd_epi32(__m256i, __m256i, __m256i);
__m256i _mm256_mask_dpwssd_epi32(__m256i, __mmask8, __m256i, __m256i);
__m256i _mm256_maskz_dpwssd_epi32(__mmask8, __m256i, __m256i, __m256i);
__m512i _mm512_dpwssd_epi32 (__m512i src, __m512i a, __m512i b);
__m512i _mm512_mask_dpwssd_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
__m512i _mm512_maskz_dpwssd_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);
// VPDPWSSDS
__m128i _mm_dpwssds_epi32(__m128i, __m128i, __m128i);
__m128i _mm_mask_dpwssds_epi32(__m128i, __mmask8, __m128i, __m128i);
__m128i _mm_maskz_dpwssds_epi32(__mmask8, __m128i, __m128i, __m128i);
__m256i _mm256_dpwssds_epi32(__m256i, __m256i, __m256i);
__m256i _mm256_mask_dpwssds_epi32(__m256i, __mmask8, __m256i, __m256i);
__m256i _mm256_maskz_dpwssds_epi32(__mmask8, __m256i, __m256i, __m256i);
__m512i _mm512_dpwssds_epi32 (__m512i src, __m512i a, __m512i b);
__m512i _mm512_mask_dpwssds_epi32 (__m512i src, __mmask16 k, __m512i a, __m512i b);
__m512i _mm512_maskz_dpwssds_epi32 (__mmask16 k, __m512i src, __m512i a, __m512i b);

Bibliography[edit]

Retrieved from "https://en.wikichip.org/w/index.php?title=x86/avx512_vnni&oldid=101222"