Fix deconv 3D post OPs segment fault issue. #130
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
luo-cheng2021
approved these changes
Jun 9, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Jul 22, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Sep 22, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Sep 23, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Sep 27, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Oct 4, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Oct 25, 2022
EgorDuplensky
pushed a commit
that referenced
this pull request
Feb 9, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Feb 13, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Feb 13, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Feb 23, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Mar 3, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Mar 10, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Mar 16, 2023
EgorDuplensky
pushed a commit
that referenced
this pull request
Apr 3, 2023
luweizhou2016
added a commit
that referenced
this pull request
May 25, 2023
luweizhou2016
added a commit
that referenced
this pull request
Jun 2, 2023
luweizhou2016
added a commit
that referenced
this pull request
Jun 8, 2023
luweizhou2016
added a commit
that referenced
this pull request
Jun 14, 2023
luweizhou2016
added a commit
that referenced
this pull request
Jun 15, 2023
[FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
pushed a commit
that referenced
this pull request
Jun 15, 2023
…ta pointers * Fix the deconv fused with depthwise issue in cpuFuncTests * Switch to use jit_uni_depthwise_injector API. * Fix potential conflicts in registers and YMM. * Update with optimization. * ONEDNN 3.2 migration squashed commits: - Preserve bf16emu scratch register when conflict with legacy post ops - fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16 - gemm convolution: fixed bias offset - Fix onednn gemm JIT/gemm kernel scale. - Luwei/fix deconv 3d postops bug (#136) - Fix deconv 3D post OPs segment fault issue. (#130) - Fix incorrect offset to rsp - fix depthwise nwc conv - fix legacyOps with stock src_zero_point in jit_avx512_core_amx - fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
luweizhou2016
pushed a commit
that referenced
this pull request
Jun 15, 2023
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
luweizhou2016
pushed a commit
that referenced
this pull request
Jun 29, 2023
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
luweizhou2016
pushed a commit
that referenced
this pull request
Dec 20, 2023
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
luweizhou2016
pushed a commit
that referenced
this pull request
Jan 10, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
luweizhou2016
pushed a commit
that referenced
this pull request
Jun 14, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 18, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 19, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
added a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 24, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 25, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 25, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
luweizhou2016
pushed a commit
that referenced
this pull request
Jul 25, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
luweizhou2016
added a commit
that referenced
this pull request
Jul 26, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
luweizhou2016
added a commit
that referenced
this pull request
Jul 29, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
luweizhou2016
added a commit
that referenced
this pull request
Jul 29, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
azhai219
pushed a commit
that referenced
this pull request
Dec 4, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (#128)
- Luwei/fix deconv 3d postops bug (#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
azhai219
pushed a commit
to azhai219/oneDNN
that referenced
this pull request
Dec 6, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (openvinotoolkit#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (openvinotoolkit#128)
- Luwei/fix deconv 3d postops bug (openvinotoolkit#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
azhai219
pushed a commit
to azhai219/oneDNN
that referenced
this pull request
Dec 9, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (openvinotoolkit#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (openvinotoolkit#128)
- Luwei/fix deconv 3d postops bug (openvinotoolkit#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
azhai219
pushed a commit
to azhai219/oneDNN
that referenced
this pull request
Dec 9, 2024
…ta pointers
ONEDNN 3.2 migration squashed commits:
- fix depthwise nwc conv
- Fix deconv 3D post OPs segment fault issue. (openvinotoolkit#130)
- fix in avx2 conv+fakequant post ops +nxc last channel wrong result (openvinotoolkit#128)
- Luwei/fix deconv 3d postops bug (openvinotoolkit#136)
-- Fix the deconv fused with depthwise issue in cpuFuncTests
-- Switch to use jit_uni_depthwise_injector API.
-- Fix potential conflicts in registers and YMM.
-- Update with optimization.
- fix legacyOps with stock src_zero_point in jit_avx512_core_amx
- Fix incorrect offset to rsp
- Preserve bf16emu scratch register when conflict with legacy post ops
- fix per-OC legacyPostOps for jit_avx512_dw_conv_fwd_kernel_bf16
- Fix segment fault caused by dest scale.
ONEDNN 3.5 migration squshed commmits:
[FIX][FORK][FEATURE] Introduced Depthwise and Quantization post ops
[FORK] [FIX] Fix legacy zero point issue on AVX2 + AVX512.
[FEATURE]Migrate legacy post ops and zero points on runtime data pointers
[FORK][FIX] jit_uni_dw_conv_row_f32: fixed post ops start idx
[FEATURE] Migrate legacy post ops and zero points on runtime data pointers
[Fix] Update the attr checking caused by forked onednn.
[FORK][FIX][x64] Add proper post op checks to gemm_conv is split
particially and squashed.
[FORK][FIX] jit_uni_dw_conv_kernel_f32: fixed register conflict
[FIX] SPlit the removed unused variables
[FIX] fix avx512 bf16 dw stack pointer
[ARM] Fixed legacy post-ops changes for ARM target
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix missing update post-ops argument offset to RSP when pushing.
OpenVINO pr:
openvinotoolkit/openvino#11836