Benchmarks CUSTOM TVM LLVM O3 - tum-ei-eda/muriscv-nn GitHub Wiki
- Spike (
riscv-isa-sim
) (ISS, CPI=1)
- LLVM/Clang:
- TODO: Version
- Linker: lld (TODO)
- RISC-V GCC for Headers, libc,...
-
MLPerfTiny Benchmark
-
TODO: others!
-
MLonMCU : main
-
TVM : Nightly Pre-Build
-
Spike : 0bc176b3fca43560b9e8586cdbc41cfde073e17a
-
Spike PK : 7e9b671c0415dfd7b562ac934feb9380075d4aa2
- Used
-Os
flag for compilation. - Benchmarks generated using MLonMCU deployment tool with minimal efforts.
- Memory metrics are reported in Bytes
Cycles (Speedup) | Total ROM (rel.) | Total RAM (rel.) | VLEN | Layout | Kernels | Mode | Arch | Unroll | Auto-Vectorization |
---|---|---|---|---|---|---|---|---|---|
13659226 ( 1.0x ) |
128506 ( 1.232 ) |
59508 ( 3.097 ) |
0 | NCHW | TVM | Fallback | RV32GC | 1 | - |
15156140 ( 0.9x ) |
151698 ( 1.455 ) |
59508 ( 3.097 ) |
0 | NHWC | TVM | Fallback | RV32GC | 1 | - |
10931215 ( 1.3x ) |
113718 ( 1.091 ) |
51336 ( 2.672 ) |
0 | NCHW | TVM | Autotuned | RV32GC | 1 | - |
15156025 ( 0.9x ) |
151936 ( 1.457 ) |
59508 ( 3.097 ) |
0 | NHWC | TVM | Autotuned | RV32GC | 1 | - |
7134661 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
128 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
7134818 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
256 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
7139085 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
512 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
7142404 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
1024 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
7150732 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
2048 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
7167677 ( 1.9x ) |
128312 ( 1.231 ) |
59508 ( 3.097 ) |
4096 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
3675716 ( 3.8x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
128 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
2239792 ( 6.2x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
256 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
1522890 ( 9.1x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
512 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
1758669 ( 7.9x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
1024 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
22748654 ( 0.6x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
2048 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
22755432 ( 0.6x ) |
134828 ( 1.293 ) |
59508 ( 3.097 ) |
4096 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
4753190 ( 2.9x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
128 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
3801250 ( 3.6x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
256 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
3453530 ( 4.0x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
512 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
3467061 ( 4.0x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
1024 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
11384002 ( 1.2x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
2048 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
11387391 ( 1.2x ) |
114180 ( 1.095 ) |
51336 ( 2.672 ) |
4096 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
3675556 ( 3.8x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
128 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
2239696 ( 6.2x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
256 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1522826 ( 9.1x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
512 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1758621 ( 7.9x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
1024 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
22749585 ( 0.6x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
2048 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
22756363 ( 0.6x ) |
135082 ( 1.296 ) |
59508 ( 3.097 ) |
4096 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
13853219 ( Base ) |
104270 ( Base ) |
19212 ( Base ) |
0 | NHWC | muRISCV-NN | Scalar | RV32GC | 1 | - |
5959524 ( 2.3x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
5345011 ( 2.6x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
5040536 ( 2.7x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
4886604 ( 2.8x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
13861489 ( 1.0x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
13864878 ( 1.0x ) |
104326 ( 1.001 ) |
19212 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
5999392 ( 2.3x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
128 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4859646 ( 2.9x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
256 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4298376 ( 3.2x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
512 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4255892 ( 3.3x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
1024 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4259281 ( 3.3x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
2048 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4262670 ( 3.2x ) |
97100 ( 0.931 ) |
23676 ( 1.232 ) |
4096 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
Cycles (Speedup) | Total ROM (rel.) | Total RAM (rel.) | VLEN | Layout | Kernels | Mode | Arch | Unroll | Auto-Vectorization |
---|---|---|---|---|---|---|---|---|---|
61157543 ( 0.8x ) |
236458 ( 1.596 ) |
108420 ( 1.953 ) |
0 | NCHW | TVM | Fallback | RV32GC | 1 | - |
61510957 ( 0.8x ) |
222108 ( 1.499 ) |
108420 ( 1.953 ) |
0 | NHWC | TVM | Fallback | RV32GC | 1 | - |
46796243 ( 1.0x ) |
230842 ( 1.558 ) |
92236 ( 1.661 ) |
0 | NCHW | TVM | Autotuned | RV32GC | 1 | - |
61505795 ( 0.8x ) |
221884 ( 1.497 ) |
108420 ( 1.953 ) |
0 | NHWC | TVM | Autotuned | RV32GC | 1 | - |
67875486 ( 0.7x ) |
278038 ( 1.876 ) |
108420 ( 1.953 ) |
128 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
67837841 ( 0.7x ) |
278038 ( 1.876 ) |
108420 ( 1.953 ) |
256 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
68393194 ( 0.7x ) |
278038 ( 1.876 ) |
108420 ( 1.953 ) |
512 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
68399922 ( 0.7x ) |
278038 ( 1.876 ) |
108420 ( 1.953 ) |
1024 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
68418184 ( 0.7x ) |
278044 ( 1.876 ) |
108420 ( 1.953 ) |
2048 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
68448685 ( 0.7x ) |
278044 ( 1.876 ) |
108420 ( 1.953 ) |
4096 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
28921548 ( 1.6x ) |
225938 ( 1.525 ) |
108420 ( 1.953 ) |
128 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
25784893 ( 1.8x ) |
225938 ( 1.525 ) |
108420 ( 1.953 ) |
256 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
24261053 ( 1.9x ) |
225938 ( 1.525 ) |
108420 ( 1.953 ) |
512 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
53046146 ( 0.9x ) |
225938 ( 1.525 ) |
108420 ( 1.953 ) |
1024 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
72077810 ( 0.7x ) |
225934 ( 1.525 ) |
108420 ( 1.953 ) |
2048 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
72077810 ( 0.7x ) |
225934 ( 1.525 ) |
108420 ( 1.953 ) |
4096 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
16630876 ( 2.8x ) |
253738 ( 1.712 ) |
92236 ( 1.661 ) |
128 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
12966145 ( 3.6x ) |
253744 ( 1.712 ) |
92236 ( 1.661 ) |
256 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
29372386 ( 1.6x ) |
253742 ( 1.712 ) |
92236 ( 1.661 ) |
512 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
42794295 ( 1.1x ) |
253740 ( 1.712 ) |
92236 ( 1.661 ) |
1024 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
55150466 ( 0.9x ) |
253742 ( 1.712 ) |
92236 ( 1.661 ) |
2048 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
55191135 ( 0.9x ) |
253744 ( 1.712 ) |
92236 ( 1.661 ) |
4096 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
28921552 ( 1.6x ) |
225888 ( 1.524 ) |
108420 ( 1.953 ) |
128 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
25784913 ( 1.8x ) |
225884 ( 1.524 ) |
108420 ( 1.953 ) |
256 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
24261081 ( 1.9x ) |
225888 ( 1.524 ) |
108420 ( 1.953 ) |
512 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
53046178 ( 0.9x ) |
225884 ( 1.524 ) |
108420 ( 1.953 ) |
1024 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
72078686 ( 0.7x ) |
225888 ( 1.524 ) |
108420 ( 1.953 ) |
2048 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
72078686 ( 0.7x ) |
225884 ( 1.524 ) |
108420 ( 1.953 ) |
4096 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
47068782 ( Base ) |
148182 ( Base ) |
55516 ( Base ) |
0 | NHWC | muRISCV-NN | Scalar | RV32GC | 1 | - |
11895602 ( 4.0x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
11555669 ( 4.1x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
14581955 ( 3.2x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
16147489 ( 2.9x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
16851190 ( 2.8x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
36741489 ( 1.3x ) |
149850 ( 1.011 ) |
55516 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
15301556 ( 3.1x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
9716568 ( 4.8x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
7169834 ( 6.6x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
5927258 ( 7.9x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
5002159 ( 9.4x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
4756344 ( 9.9x ) |
142530 ( 0.962 ) |
55516 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
Cycles (Speedup) | Total ROM (rel.) | Total RAM (rel.) | VLEN | Layout | Kernels | Mode | Arch | Unroll | Auto-Vectorization |
---|---|---|---|---|---|---|---|---|---|
1598296 ( 0.9x ) |
588410 ( 1.845 ) |
5572 ( 1.168 ) |
0 | NCHW | TVM | Fallback | RV32GC | 1 | - |
1598296 ( 0.9x ) |
588410 ( 1.845 ) |
5572 ( 1.168 ) |
0 | NHWC | TVM | Fallback | RV32GC | 1 | - |
2242578 ( 0.6x ) |
614954 ( 1.929 ) |
6884 ( 1.443 ) |
0 | NCHW | TVM | Autotuned | RV32GC | 1 | - |
2242578 ( 0.6x ) |
614954 ( 1.929 ) |
6884 ( 1.443 ) |
0 | NHWC | TVM | Autotuned | RV32GC | 1 | - |
758695 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
128 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
757749 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
256 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
512 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
1024 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
2048 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
4096 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
758695 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
128 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
757749 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
256 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
512 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
1024 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
2048 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
767299 ( 1.8x ) |
588236 ( 1.845 ) |
5572 ( 1.168 ) |
4096 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
1181651 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
128 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1141790 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
256 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1123466 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
512 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1116913 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
1024 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1115331 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
2048 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1428115 ( 1.0x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
4096 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1181651 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
128 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1141790 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
256 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1123466 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
512 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1116913 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
1024 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1115331 ( 1.2x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
2048 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1428115 ( 1.0x ) |
603170 ( 1.892 ) |
6884 ( 1.443 ) |
4096 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
1379083 ( Base ) |
318848 ( Base ) |
4772 ( Base ) |
0 | NHWC | muRISCV-NN | Scalar | RV32GC | 1 | - |
574427 ( 2.4x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
488822 ( 2.8x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
443606 ( 3.1x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
420998 ( 3.3x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
409694 ( 3.4x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
1392002 ( 1.0x ) |
317962 ( 0.997 ) |
4772 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
620637 ( 2.2x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
503957 ( 2.7x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
445617 ( 3.1x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
416771 ( 3.3x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
413140 ( 3.3x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
411284 ( 3.4x ) |
318386 ( 0.999 ) |
4772 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
Cycles (Speedup) | Total ROM (rel.) | Total RAM (rel.) | VLEN | Layout | Kernels | Mode | Arch | Unroll | Auto-Vectorization |
---|---|---|---|---|---|---|---|---|---|
43216782 ( 0.9x ) |
603242 ( 1.787 ) |
181040 ( 2.113 ) |
0 | NCHW | TVM | Fallback | RV32GC | 1 | - |
44676670 ( 0.9x ) |
683242 ( 2.024 ) |
181040 ( 2.113 ) |
0 | NHWC | TVM | Fallback | RV32GC | 1 | - |
34081937 ( 1.1x ) |
560852 ( 1.662 ) |
181040 ( 2.113 ) |
0 | NCHW | TVM | Autotuned | RV32GC | 1 | - |
44676670 ( 0.9x ) |
683234 ( 2.024 ) |
181040 ( 2.113 ) |
0 | NHWC | TVM | Autotuned | RV32GC | 1 | - |
52860741 ( 0.7x ) |
708346 ( 2.098 ) |
181040 ( 2.113 ) |
128 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
52797989 ( 0.7x ) |
708350 ( 2.099 ) |
181040 ( 2.113 ) |
256 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
52801377 ( 0.7x ) |
708350 ( 2.099 ) |
181040 ( 2.113 ) |
512 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
52811544 ( 0.7x ) |
708350 ( 2.099 ) |
181040 ( 2.113 ) |
1024 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
52835265 ( 0.7x ) |
708346 ( 2.098 ) |
181040 ( 2.113 ) |
2048 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
52879324 ( 0.7x ) |
708348 ( 2.098 ) |
181040 ( 2.113 ) |
4096 | NCHW | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
12969532 ( 2.9x ) |
643768 ( 1.907 ) |
181040 ( 2.113 ) |
128 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
9304156 ( 4.1x ) |
643770 ( 1.907 ) |
181040 ( 2.113 ) |
256 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
10033438 ( 3.8x ) |
643762 ( 1.907 ) |
181040 ( 2.113 ) |
512 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
19222575 ( 2.0x ) |
643762 ( 1.907 ) |
181040 ( 2.113 ) |
1024 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
29248452 ( 1.3x ) |
643766 ( 1.907 ) |
181040 ( 2.113 ) |
2048 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
57479155 ( 0.7x ) |
643768 ( 1.907 ) |
181040 ( 2.113 ) |
4096 | NHWC | TVM | Fallback | RV32GCV | 1 | Loop+SLP |
16914104 ( 2.3x ) |
578528 ( 1.714 ) |
181040 ( 2.113 ) |
128 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
14873961 ( 2.6x ) |
578532 ( 1.714 ) |
181040 ( 2.113 ) |
256 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
15738494 ( 2.4x ) |
578534 ( 1.714 ) |
181040 ( 2.113 ) |
512 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
18618579 ( 2.1x ) |
578530 ( 1.714 ) |
181040 ( 2.113 ) |
1024 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
22818699 ( 1.7x ) |
578532 ( 1.714 ) |
181040 ( 2.113 ) |
2048 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
35125170 ( 1.1x ) |
578534 ( 1.714 ) |
181040 ( 2.113 ) |
4096 | NCHW | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
12969713 ( 2.9x ) |
643822 ( 1.907 ) |
181040 ( 2.113 ) |
128 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
9304129 ( 4.1x ) |
643820 ( 1.907 ) |
181040 ( 2.113 ) |
256 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
10033306 ( 3.8x ) |
643814 ( 1.907 ) |
181040 ( 2.113 ) |
512 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
19222391 ( 2.0x ) |
643820 ( 1.907 ) |
181040 ( 2.113 ) |
1024 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
29248250 ( 1.3x ) |
643820 ( 1.907 ) |
181040 ( 2.113 ) |
2048 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
57478941 ( 0.7x ) |
643820 ( 1.907 ) |
181040 ( 2.113 ) |
4096 | NHWC | TVM | Autotuned | RV32GCV | 1 | Loop+SLP |
38242307 ( Base ) |
337550 ( Base ) |
85664 ( Base ) |
0 | NHWC | muRISCV-NN | Scalar | RV32GC | 1 | - |
17816817 ( 2.1x ) |
338286 ( 1.002 ) |
85664 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
17351435 ( 2.2x ) |
338286 ( 1.002 ) |
85664 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
18602295 ( 2.1x ) |
338282 ( 1.002 ) |
85664 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
21936031 ( 1.7x ) |
338282 ( 1.002 ) |
85664 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
25672113 ( 1.5x ) |
338284 ( 1.002 ) |
85664 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
40494679 ( 0.9x ) |
338280 ( 1.002 ) |
85664 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Scalar | RV32GCV | 1 | Loop+SLP |
13624400 ( 2.8x ) |
331464 ( 0.982 ) |
85664 ( 1.0 ) |
128 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
10291784 ( 3.7x ) |
331464 ( 0.982 ) |
85664 ( 1.0 ) |
256 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
8995412 ( 4.3x ) |
331464 ( 0.982 ) |
85664 ( 1.0 ) |
512 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
8489902 ( 4.5x ) |
331466 ( 0.982 ) |
85664 ( 1.0 ) |
1024 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
8444904 ( 4.5x ) |
331464 ( 0.982 ) |
85664 ( 1.0 ) |
2048 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
8448293 ( 4.5x ) |
331466 ( 0.982 ) |
85664 ( 1.0 ) |
4096 | NHWC | muRISCV-NN | Vector | RV32GCV | 1 | - |
Click here to download the raw files for this benchmark.