Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce #18019

bjacob · 2024-07-27T02:23:12Z

Advance to llvm/llvm-project@f4be6812

New local patch:

Reverting llvm/llvm-project@f6431f0c ([MLIR][GPUToNVVM] support fastMath and other non-supported mathOp llvm/llvm-project#99890). Discussion: https://discord.com/channels/689900678990135345/1266421307285966859
- This caused new compile failures in our CUDA flow. We suspect our usage of upstream passes/patterns need some updates in this repo. Reverting for now to give ourselves some time to investigate.

Existing local patches carried over:

Still carrying a revert of llvm/llvm-project@fa06668
Still carrying a revert of llvm/llvm-project@bbd4af5

Dropped local patches:

No longer carrying a local patch of llvm/llvm-project@4f80508

…l] fix for c194bc77a21d (Emilio Cota on 2024-07-25 22:52:31 -0400) (2 of 9) revert llvm#99890 Signed-off-by: Benoit Jacob <[email protected]>

github-actions · 2024-07-28T03:24:35Z

Abbreviated Benchmark Summary

@ commit f52f058439b6d9a5fd6a5cfc453505672c495ce6 (vs. base b8370b8c3a9edfdd95f2bbd2cc303751649e54c7)

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]	781.075 (1.0X)	270.127 (2.9X)	222.726 (3.5X)
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]	6.990 (1.0X)	9.326 (0.7X)	8.549 (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]	35.966 (1.0X)	36.528 (1.0X)	34.738 (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]	5.827 (1.0X)	10.996 (0.5X)	5.078 (1.1X)
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]	9.259 (1.0X)	8.568 (1.1X)	8.574 (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]	11.155 (1.0X)	9.079 (1.2X)	9.029 (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]	12.568 (1.0X)	15.596 (0.8X)	14.073 (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]	34.467 (1.0X)	65.886 (0.5X)	62.211 (0.6X)
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]	35.015 (1.0X)	66.371 (0.5X)	62.562 (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]	69.083 (1.0X)	134.613 (0.5X)	65.287 (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]	4.626 (1.0X)	5.351 (0.9X)	4.618 (1.0X)
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)	5.337 (0.7X)	4.964 (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]	6.035 (1.0X)	9.554 (0.6X)	5.465 (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]	2.941 (1.0X)	3.438 (0.9X)	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]	8.636 (1.0X)	11.009 (0.8X)	9.934 (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]	0.777 (1.0X)	1.387 (0.6X)	0.643 (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]	4.238 (1.0X)	5.845 (0.7X)	5.240 (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]	7.593 (1.0X)	7.581 (1.0X)	7.597 (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]	6.609 (1.0X)	13.307 (0.5X)	1.806 (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]	218.633 (1.0X)	136.563 (1.6X)	108.676 (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]	32.451 (1.0X)	36.550 (0.9X)	30.010 (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]	278.414 (1.0X)	257.941 (1.1X)	232.340 (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]	26.842 (1.0X)	51.405 (0.5X)	13.138 (2.0X)
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]	70.945 (1.0X)	38.577 (1.8X)	38.500 (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]	88.836 (1.0X)	41.827 (2.1X)	40.259 (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]	80.708 (1.0X)	77.052 (1.0X)	56.211 (1.4X)
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]	181.333 (1.0X)	250.037 (0.7X)	186.253 (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]	183.307 (1.0X)	252.253 (0.7X)	192.427 (1.0X)
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]	516.233 (1.0X)	1084.657 (0.5X)	240.499 (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]	26.758 (1.0X)	22.808 (1.2X)	17.823 (1.5X)
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]	12.156 (1.0X)	14.865 (0.8X)	11.742 (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]	21.675 (1.0X)	42.167 (0.5X)	11.920 (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]	2.806 (1.0X)	3.253 (0.9X)	2.693 (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]	35.461 (1.0X)	39.718 (0.9X)	31.481 (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]	0.711 (1.0X)	1.297 (0.5X)	0.575 (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]	18.017 (1.0X)	23.582 (0.8X)	19.178 (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]	0.054 (1.0X)	0.054 (1.0X)	0.054 (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]	0.042 (1.0X)	0.226 (0.2X)	0.021 (2.0X)

No improved or regressed benchmarks 🏖️

No improved or regressed compilation metrics 🏖️

For more information:

Source Workflow Run

bjacob force-pushed the integrates/llvm-20240726 branch 2 times, most recently from 5221fe5 to 9b30ede Compare July 27, 2024 02:24

bjacob changed the title ~~Integrates/llvm 20240726~~ Integrate LLVM at cb7d4a187a3ec1814f42612997e94acd460ddabf Jul 27, 2024

bjacob marked this pull request as ready for review July 27, 2024 02:36

bjacob marked this pull request as draft July 27, 2024 03:11

bjacob force-pushed the integrates/llvm-20240726 branch 2 times, most recently from b450dee to eece9bb Compare July 28, 2024 02:29

bjacob changed the title ~~Integrate LLVM at cb7d4a187a3ec1814f42612997e94acd460ddabf~~ Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce Jul 28, 2024

Advance LLVM to f4be6812e203690073280b9ac8d60092d75bbdce: [mlir][baze…

883b39d

…l] fix for c194bc77a21d (Emilio Cota on 2024-07-25 22:52:31 -0400) (2 of 9) revert llvm#99890 Signed-off-by: Benoit Jacob <[email protected]>

bjacob force-pushed the integrates/llvm-20240726 branch from eece9bb to 883b39d Compare July 28, 2024 02:38

bjacob marked this pull request as ready for review July 28, 2024 03:16

bjacob requested a review from ScottTodd July 28, 2024 03:26

Groverkss approved these changes Jul 29, 2024

View reviewed changes

kuhar approved these changes Jul 29, 2024

View reviewed changes

ScottTodd approved these changes Jul 29, 2024

View reviewed changes

bjacob merged commit 488cfc9 into main Jul 29, 2024
53 checks passed

bjacob deleted the integrates/llvm-20240726 branch July 29, 2024 15:19

bjacob mentioned this pull request Jul 29, 2024

Downstream builtin.unrealized_conversion_cast error following PR #99890, [MLIR][GPUToNVVM] support fastMath and other non-supported mathOp llvm/llvm-project#101030

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce #18019

Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce #18019

bjacob commented Jul 27, 2024 •

edited by ScottTodd

Loading

github-actions bot commented Jul 28, 2024

Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce #18019

Integrate LLVM at f4be6812e203690073280b9ac8d60092d75bbdce #18019

Conversation

bjacob commented Jul 27, 2024 • edited by ScottTodd Loading

github-actions bot commented Jul 28, 2024

Abbreviated Benchmark Summary

Data-Tiling Comparison Table

bjacob commented Jul 27, 2024 •

edited by ScottTodd

Loading