Code

Feature : Add SYCL runtime support#747

Code

Open

abhilash1910 wants to merge 69 commits intobitsandbytes-foundation:sycl-refactorbitsandbytes-foundation/bitsandbytes:sycl-refactorfrom abhilash1910:syclabhilash1910/bitsandbytes:syclCopy head branch name to clipboard

+7,701-10

Commits

Commits on Sep 5, 2023

• init kernel headers

  abhilash1910

  committed
  75d9277

  Copy full SHA for 75d9277

Commits on Sep 20, 2023

• modify kernel header

  abhilash1910

  committed
  279fce4

  Copy full SHA for 279fce4

Commits on Sep 21, 2023

• modify headers

  abhilash1910

  committed
  d9dcee9

  Copy full SHA for d9dcee9

Commits on Sep 22, 2023

• update kernels

  abhilash1910

  committed
  ddae8ab

  Copy full SHA for ddae8ab

Commits on Nov 3, 2023

• Merge branch 'TimDettmers:main' into sycl

  abhilash1910

  authored
  adbffb7

  Copy full SHA for adbffb7

Commits on Mar 12, 2024

• Merge branch 'TimDettmers:main' into sycl

  abhilash1910

  authored
  26ad690

  Copy full SHA for 26ad690

Commits on Mar 20, 2024

• add partial kernels

  abhilash1910

  committed
  1f2c8bc

  Copy full SHA for 1f2c8bc

• add quant kernels

  abhilash1910

  committed
  dd935ad

  Copy full SHA for dd935ad

Commits on Mar 21, 2024

• add precondition optimizer kernel sycl

  abhilash1910

  committed
  d93fc2e

  Copy full SHA for d93fc2e

• add koptimizer 8bits sycl

  abhilash1910

  committed
  29e895f

  Copy full SHA for 29e895f

• add row col transform kernels

  abhilash1910

  committed
  98bd39b

  Copy full SHA for 98bd39b

Commits on Mar 22, 2024

• add gemm kernels

  abhilash1910

  committed
  8e15d5a

  Copy full SHA for 8e15d5a

• fix header path

  abhilash1910

  committed
  bd9e829

  Copy full SHA for bd9e829

• fix handler kernels

  abhilash1910

  committed
  45318d6

  Copy full SHA for 45318d6

Commits on Mar 30, 2024

• integrate kenrels with ops on kquant

  abhilash1910

  committed
  c9d606b

  Copy full SHA for c9d606b

Commits on Apr 1, 2024

• add 8 bit opt with ops

  abhilash1910

  committed
  adda60d

  Copy full SHA for adda60d

• add 8 bit 2 state adam kernel

  abhilash1910

  committed
  beffe78

  Copy full SHA for beffe78

• 8 bit optimizer ops integrate

  abhilash1910

  committed
  ed431fe

  Copy full SHA for ed431fe

• add all opt & quant kernels

  abhilash1910

  committed
  7ab3273

  Copy full SHA for 7ab3273

• add shared vars

  abhilash1910

  committed
  43a1899

  Copy full SHA for 43a1899

• add shared for spm

  abhilash1910

  committed
  38c2f8c

  Copy full SHA for 38c2f8c

• add gemm ops + kernel fix

  abhilash1910

  committed
  04a3ba1

  Copy full SHA for 04a3ba1

Commits on Apr 2, 2024

• update kernel headers

  abhilash1910

  committed
  06ba221

  Copy full SHA for 06ba221

Commits on Apr 3, 2024

• port transform kernels

  abhilash1910

  committed
  ee7997d

  Copy full SHA for ee7997d

Commits on Apr 4, 2024

• use ldg & sparse csr

  abhilash1910

  committed
  26863ea

  Copy full SHA for 26863ea

Commits on Apr 8, 2024

• add dnn prototype

  abhilash1910

  committed
  caa72f6

  Copy full SHA for caa72f6

• dnn kernel

  abhilash1910

  committed
  46f1f85

  Copy full SHA for 46f1f85

Commits on Apr 9, 2024

• add cmake initial

  abhilash1910

  committed
  a94c253

  Copy full SHA for a94c253

• cmake fix

  abhilash1910

  committed
  89fb73b

  Copy full SHA for 89fb73b

• fix build

  abhilash1910

  committed
  2644374

  Copy full SHA for 2644374

• fix build

  abhilash1910

  committed
  14f20e4

  Copy full SHA for 14f20e4

Commits on Apr 10, 2024

• fix build

  abhilash1910

  committed
  d151d0c

  Copy full SHA for d151d0c

• fix dnnl

  abhilash1910

  committed
  dfcd9d8

  Copy full SHA for dfcd9d8

Commits on May 10, 2024

• refactor with new api and use accessor

  abhilash1910

  committed
  886751c

  Copy full SHA for 886751c

Commits on May 13, 2024

• full 32 optimizer fixed update

  abhilash1910

  committed
  f90c06d

  Copy full SHA for f90c06d

• refine 8 bit optimizers

  abhilash1910

  committed
  873eadb

  Copy full SHA for 873eadb

• refine 8 bit blockwise opt

  abhilash1910

  committed
  f71243b

  Copy full SHA for f71243b

Commits on May 14, 2024

• refine k quantize blockwise

  abhilash1910

  committed
  493a5ec

  Copy full SHA for 493a5ec

• refine k quantize

  abhilash1910

  committed
  760a120

  Copy full SHA for 760a120

• refine percentile clipping

  abhilash1910

  committed
  c7d8326

  Copy full SHA for c7d8326

• refine estimate quantiles and k dequantize & headers

  abhilash1910

  committed
  693ca79

  Copy full SHA for 693ca79

Commits on May 15, 2024

• fix errors in k compress

  abhilash1910

  committed
  3a41f96

  Copy full SHA for 3a41f96

• refine template types for k quants

  abhilash1910

  committed
  37897ee

  Copy full SHA for 37897ee

• remove mma header

  abhilash1910

  committed
  7ddce74

  Copy full SHA for 7ddce74

• refine historgram

  abhilash1910

  committed
  c039aa1

  Copy full SHA for c039aa1

Commits on May 16, 2024

• refine row stats

  abhilash1910

  committed
  b9e9a9c

  Copy full SHA for b9e9a9c

• refine double row col quants

  abhilash1910

  committed
  c8473b5

  Copy full SHA for c8473b5

• refine helper functions

  abhilash1910

  committed
  c85baf9

  Copy full SHA for c85baf9

• complete refine non gemm kernels

  abhilash1910

  committed
  922786e

  Copy full SHA for 922786e

• fix extract function

  abhilash1910

  committed
  ee8225e

  Copy full SHA for ee8225e

Commits on May 20, 2024

• refine igemmlt kernels

  abhilash1910

  committed
  49ca1d7

  Copy full SHA for 49ca1d7

Commits on May 21, 2024

• refine dnn and gemm 4 bit kernels

  abhilash1910

  committed
  342f95c

  Copy full SHA for 342f95c

• refine spm experimental

  abhilash1910

  committed
  5e2611a

  Copy full SHA for 5e2611a

Commits on May 26, 2024

• fix host dereference issue on nv dequant & kquant

  abhilash1910

  committed
  6af46e2

  Copy full SHA for 6af46e2

Commits on Jun 6, 2024

• fix header name

  abhilash1910

  committed
  bcad0ea

  Copy full SHA for bcad0ea

• fix nv host dereference issue on 8 bit

  abhilash1910

  committed
  6d3ed26

  Copy full SHA for 6d3ed26

• fix nv issue k quants

  abhilash1910

  committed
  4f19d41

  Copy full SHA for 4f19d41

• fix nv issue on row col quant

  abhilash1910

  committed
  dd709ca

  Copy full SHA for dd709ca

Commits on Jun 7, 2024

• fix nv issue on 32 bit

  abhilash1910

  committed
  0fbedfe

  Copy full SHA for 0fbedfe

• fix nv issue on k dequant blockwise

  abhilash1910

  committed
  38312cb

  Copy full SHA for 38312cb

• fix nv issue on gemms and some quant kernels

  abhilash1910

  committed
  1d6f56b

  Copy full SHA for 1d6f56b

• refine

  abhilash1910

  committed
  b107b9c

  Copy full SHA for b107b9c

• refine

  abhilash1910

  committed
  7838151

  Copy full SHA for 7838151

• refine

  abhilash1910

  committed
  46d10ba

  Copy full SHA for 46d10ba

• add dnn build flag

  abhilash1910

  authored
  8795844

  Copy full SHA for 8795844

Commits on Jun 27, 2024

• update dnn linkage

  abhilash1910

  authored
  7007a02

  Copy full SHA for 7007a02

Commits on Jul 29, 2024

• Merge branch 'sycl-refactor' into sycl

  abhilash1910

  authored
  50a2838

  Copy full SHA for 50a2838

Commits on Aug 7, 2024

• fix cmake

  abhilash1910

  committed
  6f8ef48

  Copy full SHA for 6f8ef48

• upgrade to 2024.2 Intel LLVM compiler release, set Nvidia build flag

  abhilash1910

  committed
  a425d44

  Copy full SHA for a425d44