iree
iree copied to clipboard
Published
20 hours ago •
iree-org
iree-org
Update tiling sizes for ARM convolution configurations.
Abbreviated Benchmark Summary
@ commit d1902964989295540f4f88dde9dd2cf475a96800 (vs. base 979d6ea9457503a03003a5c6436a7dd08665423c)
Regressed Latencies 🚩
| Benchmark Name | Average Latency (ms) | Median Latency (ms) | Latency Standard Deviation (ms) |
|---|---|---|---|
| MobileBertSquad [fp32] (TFLite) 4-thread,big-core,full-inference,experimental-flags with IREE-LLVM-CPU @ Pixel-6-Pro (CPU-ARMv8.2-A) | 469.009 (vs. 434.511, 7.94%↑) | 469.624 | 4.756 |
| MobileNetV2 [fp32,imagenet] (TFLite) little-core,full-inference,default-flags with IREE-LLVM-CPU-Sync @ Pixel-6-Pro (CPU-ARMv8.2-A) | 230.635 (vs. 218.644, 5.48%↑) | 230.727 | 1.581 |
Improved Latencies 🎉
| Benchmark Name | Average Latency (ms) | Median Latency (ms) | Latency Standard Deviation (ms) |
|---|---|---|---|
| MobileSSD [fp32] (TFLite) 4-thread,big-core,full-inference,experimental-flags with IREE-LLVM-CPU @ Pixel-6-Pro (CPU-ARMv8.2-A) | 50.027 (vs. 53.675, 6.80%↓) | 50.066 | 1.425 |
| DeepLabV3 [fp32] (TFLite) little-core,full-inference,experimental-flags with IREE-LLVM-CPU-Sync @ Pixel-6-Pro (CPU-ARMv8.2-A) | 360.121 (vs. 385.234, 6.52%↓) | 360.063 | 0.770 |
| MobileBertSquad [fp32] (TFLite) big-core,full-inference,default-flags with IREE-LLVM-CPU-Sync @ Pixel-6-Pro (CPU-ARMv8.2-A) | 404.046 (vs. 428.573, 5.72%↓) | 405.934 | 5.182 |
[Top 3 out of 4 results showed]
No improved or regressed compilation metrics 🏖️
For more information:
