Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Integrate LLVM at llvm/llvm-project@3e713575 #18155

Merged
merged 2 commits into from
Aug 8, 2024
Merged

Integrate LLVM at llvm/llvm-project@3e713575 #18155

merged 2 commits into from
Aug 8, 2024

Conversation

hanhanW
Copy link
Contributor

@hanhanW hanhanW commented Aug 8, 2024 •
edited
Loading

Cherry-pick commits:

@hanhanW
Copy link
Contributor Author

hanhanW commented Aug 8, 2024

llvm/llvm-project@da8778e does not handle the transposition with all unit dims case, I'm working on a fix in upstream.

@hanhanW
Copy link
Contributor Author

hanhanW commented Aug 8, 2024

llvm/llvm-project#102518 should fix the issue, cherry-picking the fix now.

@github-actions GitHub Actions
Copy link

github-actions bot commented Aug 8, 2024

Abbreviated Benchmark Summary

@ commit d8638e0dd1a0684e0510f4b7baf08d8418f475de (vs. base 8545650e0744a24919a3dc6d419c1ebfceef345e)

Data-Tiling Comparison Table

Click to show
Name No-DT (baseline) DT-Only DT-UK
BertLargeTF(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[30-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 765.832 (1.0X) 283.705 (2.7X) 228.205 (3.4X)
DeepLabV3_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 7.033 (1.0X) 9.314 (0.8X) 8.542 (0.8X)
EfficientNetV2STF(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 35.875 (1.0X) 36.266 (1.0X) 34.490 (1.0X)
EfficientNet_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 5.865 (1.0X) 10.993 (0.5X) 5.053 (1.2X)
GPT2_117M_TF_1X1XI32(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 9.192 (1.0X) 8.523 (1.1X) 8.518 (1.1X)
GPT2_117M_TF_1X4XI32(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 11.063 (1.0X) 9.093 (1.2X) 8.969 (1.2X)
MiniLML12H384Uncased(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 11.989 (1.0X) 15.508 (0.8X) 13.812 (0.9X)
MobileBertSquad_fp16(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 33.847 (1.0X) 65.523 (0.5X) 61.570 (0.5X)
MobileBertSquad_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 34.281 (1.0X) 66.382 (0.5X) 61.742 (0.6X)
MobileBertSquad_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[15-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 69.283 (1.0X) 134.169 (0.5X) 64.507 (1.1X)
MobileNetV1_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 4.890 (1.0X) 5.347 (0.9X) 4.584 (1.1X)
MobileNetV2_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 3.821 (1.0X) 5.359 (0.7X) 4.928 (0.8X)
MobileNetV2_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 5.934 (1.0X) 9.585 (0.6X) 5.462 (1.1X)
MobileNetV3Small_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 2.922 (1.0X) 3.442 (0.8X) 2.814 (1.0X)
MobileSSD_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 8.621 (1.0X) 11.025 (0.8X) 10.025 (0.9X)
PersonDetect_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 0.788 (1.0X) 1.395 (0.6X) 0.657 (1.2X)
PoseNet_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[8-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 4.208 (1.0X) 5.928 (0.7X) 5.321 (0.8X)
matmul_256x256x2048_i8_i4_i32_tile_config_default(linalg) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 7.577 (1.0X) 7.567 (1.0X) 7.564 (1.0X)
matmul_256x256x2048_i8_i8_i32_tile_config_default(linalg) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 6.661 (1.0X) 13.340 (0.5X) 1.811 (3.7X)
BertForMaskedLMTF(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[30-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 216.217 (1.0X) 137.353 (1.6X) 107.206 (2.0X)
DeepLabV3_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 32.549 (1.0X) 36.366 (0.9X) 30.102 (1.1X)
EfficientNetV2STF(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 278.824 (1.0X) 260.053 (1.1X) 229.744 (1.2X)
EfficientNet_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 26.981 (1.0X) 51.504 (0.5X) 13.111 (2.1X)
GPT2_117M_TF_1X1XI32(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 70.136 (1.0X) 38.710 (1.8X) 38.665 (1.8X)
GPT2_117M_TF_1X4XI32(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 88.122 (1.0X) 41.928 (2.1X) 40.977 (2.2X)
MiniLML12H384Uncased(stablehlo) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 79.662 (1.0X) 78.099 (1.0X) 59.044 (1.3X)
MobileBertSquad_fp16(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 180.305 (1.0X) 248.399 (0.7X) 186.979 (1.0X)
MobileBertSquad_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 181.735 (1.0X) 252.405 (0.7X) 191.767 (0.9X)
MobileBertSquad_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 520.530 (1.0X) 1088.054 (0.5X) 244.470 (2.1X)
MobileNetV1_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 25.762 (1.0X) 22.486 (1.1X) 17.808 (1.4X)
MobileNetV2_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 12.169 (1.0X) 14.829 (0.8X) 11.620 (1.0X)
MobileNetV2_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 21.778 (1.0X) 42.449 (0.5X) 11.884 (1.8X)
MobileNetV3Small_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 2.788 (1.0X) 3.264 (0.9X) 2.691 (1.0X)
MobileSSD_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 33.737 (1.0X) 39.183 (0.9X) 31.498 (1.1X)
PersonDetect_int8(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 0.715 (1.0X) 1.300 (0.6X) 0.578 (1.2X)
PoseNet_fp32(tflite) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_task(embedded_elf)[1-thread,full-inference,default-flags] with default @ c2-standard-60[cpu] 18.162 (1.0X) 23.844 (0.8X) 19.686 (0.9X)
matmul_1x256x2048_i8_i4_i32_tile_config_default(linalg) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 0.055 (1.0X) 0.055 (1.0X) 0.055 (1.0X)
matmul_1x256x2048_i8_i8_i32_tile_config_default(linalg) [x86_64-cascadelake-linux_gnu-llvm_cpu] local_sync(embedded_elf)[full-inference,default-flags] with default @ c2-standard-60[cpu] 0.043 (1.0X) 0.226 (0.2X) 0.022 (2.0X)

No improved or regressed benchmarks 🏖️

Improved Total Dispatch Sizes 🎉

Benchmark Name Total Dispatch Size (bytes)
MobileBertSquad\_fp16(tflite) [x86\_64-cascadelake-linux\_gnu-llvm\_cpu][experimental-flags,dt-only,compile-stats] 48496 (vs. 53552, 9.44%↓)
MobileBertSquad\_fp32(tflite) [x86\_64-cascadelake-linux\_gnu-llvm\_cpu][experimental-flags,dt-only,compile-stats] 48880 (vs. 53968, 9.43%↓)
GPT2\_117M\_TF\_1X1XI32(stablehlo) [x86\_64-cascadelake-linux\_gnu-llvm\_cpu][experimental-flags,dt-only,compile-stats] 10272 (vs. 11280, 8.94%↓)

For more information:

Source Workflow Run

@hanhanW
Copy link
Contributor Author

hanhanW commented Aug 8, 2024

The cherry-pick commit will be dropped in the next integrate. It is not merged into upstream repo yet, still waiting for LLVM CI jobs finished.

@hanhanW hanhanW merged commit 052c000 into main Aug 8, 2024
53 checks passed
@hanhanW hanhanW deleted the integrates/llvm-20240808 branch August 8, 2024 20:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ScottTodd ScottTodd ScottTodd approved these changes

@MaheshRavishankar MaheshRavishankar Awaiting requested review from MaheshRavishankar

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@hanhanW @ScottTodd