Last active
June 24, 2020 20:16
-
-
Save robieta/50e37f3856e8c00f6cdc3dde610a6c42 to your computer and use it in GitHub Desktop.
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
gpu: 128 / 128 | |
27 samples were culled, 1893 remain | |
======================================== | |
== GPU ================================= | |
======================================== | |
Improved (>5%): 462 ( 24%) | |
Regressed (>5%): 266 ( 14%) | |
Within 5%: 1165 ( 62%) | |
Relative Δ Absolute Δ | numel dtype shape steps layout task specific | |
============================================================================================================================== | |
-98% 82295.6 us | 5984792 float32 [5984792] dim=0, k=2 | |
-98% 100001.1 us | 5984792 float64 [5984792] dim=0, k=2 | |
-98% 46101.1 us | 2665388 float64 [2665388] dim=0, k=32 | |
-98% 84188.3 us | 6042452 float32 [6042452] dim=0, k=47 | |
-97% 80714.1 us | 4681360 float64 [4681360] dim=0, k=1 | |
-97% 104044.8 us | 6042452 float64 [6042452] dim=0, k=47 | |
-97% 73750.6 us | 3770492 float64 [3770492] dim=0, k=37 | |
-97% 56073.5 us | 4681360 float32 [4681360] dim=0, k=1 | |
-97% 30142.0 us | 2665388 float32 [2665388] dim=0, k=32 | |
-97% 38740.5 us | 10404992 int8 [10404992] dim=0, k=115 | |
... | |
-0% 2.5 us | 2677090 float64 [26, 102965] [2, 1] [1 0] dim=0, k=21 | |
-0% 2.9 us | 5983808 float64 [93497, 64] dim=1, k=11 | |
-0% 5.5 us | 7906900 float64 [8548, 925] dim=1, k=445 | |
-0% 0.9 us | 2041569 float64 [11939, 171] dim=1, k=9 | |
-0% 22.0 us | 29843370 float32 [1605, 18594] dim=0, k=467 | |
-0% 1.3 us | 5570560 float32 [170, 128, 256] dim=2, k=1 | |
-0% 2.3 us | 2047760 float64 [1969, 16, 65] dim=1, k=6 | |
-0% 2.0 us | 3441152 float32 [44, 52, 1504] dim=1, k=2 | |
-0% 2.1 us | 2463691 float32 [17, 23, 6301] [2 0 1] dim=0, k=7 | |
-0% 0.4 us | 6607360 int8 [512, 12905] dim=1, k=5 | |
... | |
132% 64024.3 us | 20747019 float32 [309657, 67] [1, 8] dim=0, k=78611 | |
138% 7207.1 us | 10943150 int8 [188675, 58] [1 0] dim=0, k=12041 | |
179% 8825.3 us | 10850284 int8 [17, 638252] dim=1, k=57055 | |
185% 8181.9 us | 11534336 int8 [22, 524288] [1, 8] dim=1, k=20187 | |
204% 4163.6 us | 6291456 int8 [24, 262144] dim=1, k=17032 | |
236% 74717.7 us | 11788920 float64 [181368, 65] [1 0] dim=0, k=50492 | |
254% 38994.6 us | 20747019 int8 [309657, 67] [1, 8] dim=0, k=78611 | |
263% 70268.1 us | 11788920 float32 [181368, 65] [1 0] dim=0, k=50492 | |
274% 22039.7 us | 30076416 int8 [469944, 64] [4, 1] dim=0, k=11182 | |
471% 42955.7 us | 11788920 int8 [181368, 65] [1 0] dim=0, k=50492 | |
steps: | |
Indicates that `x` is sliced from a larger Tensor. For instance, if | |
shape is [12, 4] and steps are [2, 1], then a larger Tensor of size | |
[24, 4] was created, and then x = base_tensor[::2, ::1]. Omitted if | |
all elements are ones. | |
layout: | |
Indicates that `x` is not contiguous due to permutation. Invoking | |
`x.permute(steps)` (e.g. x.permute((2, 0, 1)) if steps = [2, 0, 1]) | |
would produce a Tensor whose shape matches memory order. (Though still | |
not contiguous if `steps` contains non-one elements.) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment