1211393 - Performance regression in Gallery

Reporter

Description

•

9 years ago

Gallery v2.2 cold launch: 1113ms
Gallery current cold launch: 1207ms (~90ms regression)
Gallery v2.2 USS: 17.71MB
Gallery current USS: 18.98MB (~1.25MB regression)

Bobby Chien

Reporter

Updated

•

9 years ago

blocking-b2g: --- → 2.5+

Priority: -- → P1

Punam Dahiya [:pdahiya]

Assignee

Updated

•

9 years ago

Assignee: nobody → pdahiya

Punam Dahiya [:pdahiya]

Assignee

Comment 2

•

9 years ago

Running raptor, after updating flame-kk with latest base image v18D_nightly_v4.zip and with latest buildid 20151009030234, here are raptor results after 10 runs

[Cold Launch: gallery.gaiamobile.org] Results from gallery.gaiamobile.org

| Metric                | Mean      | Median   | Min    | Max    | StdDev   | p95    |
| --------------------- | --------- | -------- | ------ | ------ | -------- | ------ |
| navigationLoaded      | 786.600   | 791.500  | 741    | 844    | 28.566   | 844    |
| navigationInteractive | 787.900   | 792.500  | 743    | 845    | 28.197   | 845    |
| visuallyLoaded        | 1215.400  | 1222     | 1176   | 1256   | 24.549   | 1256   |
| contentInteractive    | 1215.500  | 1222.500 | 1176   | 1256   | 24.586   | 1256   |
| mediaEnumerated       | 1675.300  | 1668     | 1615   | 1758   | 38.173   | 1758   |
| fullyLoaded           | 14650.300 | 14250    | 12598  | 16788  | 1388.377 | 16788  |
| pss                   | 22.083    | 22.215   | 20.988 | 22.405 | 0.395    | 22.405 |
| uss                   | 17.743    | 17.871   | 16.688 | 18.062 | 0.382    | 18.062 |
| rss                   | 34.943    | 35.072   | 33.801 | 35.371 | 0.416    | 35.371 |

Punam Dahiya [:pdahiya]

Assignee

Comment 3

•

9 years ago

There are very few changes that has landed in 2.5, main ones being l10n enhancement (Bug 1169030, 1170973), gaia -header tweaks (Bug 1198132) and jshint fixes (Bug 1158520), none of these changes hints launch time regression

Punam Dahiya [:pdahiya]

Assignee

Comment 4

•

9 years ago

Here are the launch time numbers on doing full flash on flame-kk 319 MB with version 2.2, platform 37.0 and buildId:20151009032504 and using 2.5 gaia on top of it.

[Cold Launch: gallery.gaiamobile.org] Results from gallery.gaiamobile.org

| Metric                | Mean   | Median | Min    | Max    | StdDev | p95    |
| --------------------- | ------ | ------ | ------ | ------ | ------ | ------ |
| navigationLoaded      | 483    | 483    | 483    | 483    | 0      | 483    |
| navigationInteractive | 484    | 484    | 484    | 484    | 0      | 484    |
| visuallyLoaded        | 911    | 911    | 911    | 911    | 0      | 911    |
| contentInteractive    | 911    | 911    | 911    | 911    | 0      | 911    |
| mediaEnumerated       | 1326   | 1326   | 1326   | 1326   | 0      | 1326   |
| fullyLoaded           | 9776   | 9776   | 9776   | 9776   | 0      | 9776   |
| uss                   | 13.500 | 13.500 | 13.500 | 13.500 | 0      | 13.500 |
| pss                   | 17.200 | 17.200 | 17.200 | 17.200 | 0      | 17.200 |
| rss                   | 29.200 | 29.200 | 29.200 | 29.200 | 0      | 29.200 |


Please note 2.5 gallery app doesn't run on as is on 2.2 and works after commenting Intl changes from bug 1170973. This shows there is underlying regression outside of gaia changes between 2.2 and 2.5

Punam Dahiya [:pdahiya]

Assignee

Comment 5

•

9 years ago

And for comparison here's the launch time numbers on doing full flash on flame-kk 319 MB with OS version 2.5, platform version 44.0a1 and BuildId: 201509030234 and with Intl changes from bug 1170973 commented.

[Cold Launch: gallery.gaiamobile.org] Results from gallery.gaiamobile.org

| Metric                | Mean      | Median | Min    | Max    | StdDev   | p95    |
| --------------------- | --------- | ------ | ------ | ------ | -------- | ------ |
| navigationLoaded      | 757.700   | 761    | 702    | 802    | 25.726   | 802    |
| navigationInteractive | 759       | 762    | 704    | 803    | 25.566   | 803    |
| visuallyLoaded        | 1102.900  | 1102   | 1066   | 1141   | 22.255   | 1141   |
| contentInteractive    | 1104      | 1104   | 1067   | 1142   | 22.009   | 1142   |
| mediaEnumerated       | 1578.900  | 1570   | 1494   | 1707   | 53.916   | 1707   |
| fullyLoaded           | 17683.500 | 17212  | 14130  | 26008  | 3240.776 | 26008  |
| uss                   | 16.531    | 17.224 | 11.344 | 17.637 | 1.817    | 17.637 |
| rss                   | 33.139    | 33.917 | 27.707 | 34.152 | 1.893    | 34.152 |
| pss                   | 20.671    | 21.390 | 15.412 | 21.759 | 1.844    | 21.759 |

Punam Dahiya [:pdahiya]

Assignee

Comment 6

•

9 years ago

Numbers in #comment 5 are after commenting Intl use but doesn't fully revert files thumbnail_date_group and thumbnail_item. For comparison I will update bug with numbers after revering patch bug1170973. Thanks

Punam Dahiya [:pdahiya]

Assignee

Comment 7

•

9 years ago

Here are cold launch time numbers after reverting patch for bug 1170973 and putting back shared/js/l10n_date.js in apps/gallery/index.html

[Cold Launch: gallery.gaiamobile.org] Results from gallery.gaiamobile.org

| Metric                | Mean      | Median  | Min    | Max    | StdDev   | p95    |
| --------------------- | --------- | ------- | ------ | ------ | -------- | ------ |
| navigationLoaded      | 758.600   | 765     | 695    | 787    | 27.886   | 787    |
| navigationInteractive | 759.900   | 766.500 | 696    | 788    | 27.984   | 788    |
| visuallyLoaded        | 1113.100  | 1111    | 1091   | 1140   | 15.751   | 1140   |
| contentInteractive    | 1114.200  | 1112    | 1092   | 1140   | 15.419   | 1140   |
| mediaEnumerated       | 1632.200  | 1621    | 1556   | 1772   | 60.496   | 1772   |
| fullyLoaded           | 14541.400 | 14598   | 12101  | 18847  | 1859.922 | 18847  |
| pss                   | 21.366    | 21.449  | 20.115 | 21.854 | 0.456    | 21.854 |
| rss                   | 33.968    | 34.035  | 32.684 | 34.434 | 0.473    | 34.434 |
| uss                   | 17.241    | 17.324  | 16.008 | 17.719 | 0.450    | 17.719 |

Punam Dahiya [:pdahiya]

Assignee

Comment 8

•

9 years ago

Including gandalf to share findings in comment#7. I will try using IntlHelper for setting group header and see if it helps with the launch time numbers. Thanks

Flags: needinfo?(gandalf)

Zibi Braniecki [:zbraniecki][:gandalf]

Comment 9

•

9 years ago

We started retesting with Eli and here's the new baseline for 2.2: https://gist.github.com/eliperelman/7868c7144a6805ec4f2d

Gallery:
| Metric                | Mean     | Median   | Min    | Max    | StdDev | 95% Bound |
| --------------------- | -------- | -------- | ------ | ------ | ------ | --------- |
| navigationLoaded      | 694.700  | 688.500  | 627    | 773    | 40.532 | 709.204   |
| navigationInteractive | 697.767  | 693.500  | 629    | 775    | 40.538 | 712.273   |
| visuallyLoaded        | 1118.833 | 1115     | 985    | 1188   | 40.577 | 1133.354  |
| contentInteractive    | 1119.867 | 1115     | 985    | 1188   | 40.403 | 1134.325  |
| mediaEnumerated       | 1487.333 | 1478     | 1386   | 1566   | 44.062 | 1503.101  |
| fullyLoaded           | 1852.433 | 1860.500 | 1685   | 1917   | 44.032 | 1868.190  |
| uss                   | 13.010   | 13       | 12.900 | 13.200 | 0.054  | 13.029    |
| rss                   | 30.043   | 30       | 29.900 | 30.200 | 0.067  | 30.067    |
| pss                   | 17.090   | 17.100   | 17     | 17.300 | 0.060  | 17.111    |

Using the same steps I tested master gallery;

| Metric                | Mean     | Median   | Min    | Max    | StdDev | 95% Bound |
| --------------------- | -------- | -------- | ------ | ------ | ------ | --------- |
| navigationLoaded      | 723.450  | 721      | 683    | 763    | 20.301 | 732.347   |
| navigationInteractive | 724.800  | 722.500  | 684    | 765    | 20.238 | 733.670   |
| visuallyLoaded        | 1131.850 | 1130.500 | 1093   | 1176   | 21.602 | 1141.317  |
| contentInteractive    | 1132.800 | 1131     | 1093   | 1181   | 22.569 | 1142.691  |
| mediaEnumerated       | 1583.600 | 1574.500 | 1535   | 1648   | 30.850 | 1597.121  |
| fullyLoaded           | 1881.350 | 1877     | 1808   | 1969   | 40.602 | 1899.145  |
| uss                   | 17.754   | 17.758   | 17.719 | 17.801 | 0.019  | 17.763    |
| rss                   | 37.688   | 37.690   | 37.652 | 37.734 | 0.019  | 37.696    |
| pss                   | 22.165   | 22.133   | 22.002 | 22.386 | 0.122  | 22.219    |

It seems to me like there is no performance regression (our number here is visuallyLoaded 95% Bound), but there is a significant memory regression (USS 95% Bound).

I'm not sure how Intl could regress it that much (4MB) since we create two objects, one on each class (static). I'd be really surprised if Intl was involved.

I would expect some underlying Gecko changes affecting Gallery database or sth like that more.

Flags: needinfo?(gandalf)

Zibi Braniecki [:zbraniecki][:gandalf]

Comment 10

•

9 years ago

New performance metrics for 2.2 vs master confirm my comment 9: https://gist.github.com/eliperelman/b81a27b7f7f3b3f75f49

There's a clear regression in memory, that may be attributed to Gecko or Gallery, and zero performance regression.

Punam Dahiya [:pdahiya]

Assignee

Comment 11

•

9 years ago

Thanks gandalf! I am glad to see new baseline numbers shared by Eli. Most of the apps are seeing ~ 2 MB regression in 2.5 hinting underlying gecko change. I will run about:memory report between 2.2 and 2.5 to identify memory regression area for gallery app.

Punam Dahiya [:pdahiya]

Assignee

Comment 12

•

9 years ago

about:memory gallery memory usage for 2.5

21.42 MB (100.0%) -- explicit
├───6.61 MB (30.87%) ++ js-non-window
├───5.71 MB (26.66%) -- images/content/raster/used
│   ├──0.29 MB (01.33%) -- image(213x213, blob:app://gallery.gaiamobile.org/15db0851-6e6d-4424-97f1-1e322d4178f8)
│   │  ├──0.27 MB (01.28%) -- locked
│   │  │  ├──0.18 MB (00.82%) -- surface(213x213@0)
│   │  │  │  ├──0.18 MB (00.82%) ── decoded-nonheap
│   │  │  │  └──0.00 MB (00.00%) ── decoded-heap
│   │  │  └──0.10 MB (00.46%) -- surface(159x159@0)
│   │  │     ├──0.10 MB (00.46%) ── decoded-nonheap
│   │  │     └──0.00 MB (00.00%) ── decoded-heap
│   │  └──0.01 MB (00.06%) ── source
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/2ff2336c-23ce-40fe-848b-8f72d0138786)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/4ab10d7d-cac5-4bff-bd73-380ee03e2228)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/4b988487-e1cc-4661-9b2d-b26e86582ecf)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/4f5d976a-cd16-46bf-a5f7-cde69a9ccb77)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/57b84091-c6e0-479a-90ea-d2634f26c8fb)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/5cbc9f92-c826-4a69-890c-e77252bc0932)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/71f6be31-d06a-498a-8b45-1128da1afa02)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/7c54a224-2848-4f9e-8414-e7b4ea7afdcc)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/8de8d174-5481-4a88-8a4e-5d21337b3568)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/9dacdb05-cb7f-4a4d-95a1-b645ffacf4f5)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/a184087d-1856-4e62-b9c9-6e6ef8b7f3d9)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/b5d89639-31eb-477d-98ab-9d2873119532)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/c74a6623-30a5-414a-92b6-6b0968bb184d)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/cf424474-9791-4fa8-b8a6-9f963f02584f)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/e06ee7a5-4eab-4ae1-9cda-d1f2d1d284ef)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/e9000cb3-ca68-493b-8d01-9b1d6b745842)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/ec3ca6e1-7aef-44b5-a7e5-76be071be299)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/edfd443e-1234-4963-a18b-a0160f0e07c3)
│   ├──0.29 MB (01.33%) ++ image(213x213, blob:app://gallery.gaiamobile.org/f361c2bd-f24d-49b2-8066-745d06fff151)
│   └──0.00 MB (00.01%) ++ <non-notable images>
├───3.78 MB (17.67%) ── heap-unclassified
├───1.46 MB (06.82%) ++ window-objects/top(app://gallery.gaiamobile.org/index.html, id=2147483649)
├───1.14 MB (05.33%) ++ heap-overhead
├───0.90 MB (04.19%) ++ layout
├───0.79 MB (03.69%) ++ (11 tiny)
├───0.39 MB (01.82%) ── freetype
├───0.33 MB (01.54%) ++ gfx
└───0.30 MB (01.42%) ++ xpcom

about:memory - gallery memory usage for 2.2

17.79 MB (100.0%) -- explicit
├───5.68 MB (31.90%) ++ js-non-window
├───3.84 MB (21.60%) -- images/content/raster/used
│   ├──0.19 MB (01.05%) -- image(blob:app://gallery.gaiamobile.org/0029c8c1-60ee-4eac-9866-32486aa6c264)
│   │  ├──0.18 MB (00.99%) ── decoded-nonheap
│   │  ├──0.01 MB (00.07%) ── source
│   │  └──0.00 MB (00.00%) ── decoded-heap
│   ├──0.19 MB (01.05%) -- image(blob:app://gallery.gaiamobile.org/24c4ef84-1d6e-4909-b769-661e9e8bbc7b)
│   │  ├──0.18 MB (00.99%) ── decoded-nonheap
│   │  ├──0.01 MB (00.07%) ── source
│   │  └──0.00 MB (00.00%) ── decoded-heap
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/33623a7d-379b-4977-80da-47eaeb0eb5fc)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/3f2cbe9a-5048-41e3-9a7c-8007d9274ac5)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/5549740e-89a4-4391-a5e7-ab528377cc83)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/59cca07f-2cc3-4a48-bc82-db75fa3e5dac)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/639c04ed-0bd7-4e52-a118-d77a5bdfd683)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/7d994bf2-1be5-4070-95ae-88232f6c71b2)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/9c3463ac-3a76-46c8-9f86-08875b22578a)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/9d9a49e7-e666-4b4c-9cec-908b33df7626)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/a2c3bbba-0feb-4cdc-a761-5ab988040259)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/aadf1d7c-4e25-471c-9962-1f14de0de80e)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/b2591008-6077-4a91-a540-3a5265748b88)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/bfab5c5e-369f-4374-a9c7-bef5ab0bc5f5)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/d14003df-e339-4b7e-b752-f76201c04599)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/ddc05480-d543-42b1-9626-553354b1b1cd)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/dfb192d3-209e-42a9-9a7a-2de378f81a08)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/e0ecb187-3d06-45a8-b0de-9f323df83ecf)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/f78b0be9-2d6e-4635-be9d-5b2b05bd8cfd)
│   ├──0.19 MB (01.05%) ++ image(blob:app://gallery.gaiamobile.org/fba4703d-2a26-40a9-9e13-311131151f5d)
│   └──0.09 MB (00.50%) -- (3 tiny)
│      ├──0.05 MB (00.26%) ++ image(resource://gre/res/text_caret@2.25x.png)
│      ├──0.02 MB (00.13%) ++ image(resource://gre/res/text_caret@1.5x.png)
│      └──0.02 MB (00.10%) ++ <non-notable images>
├───3.53 MB (19.83%) ── heap-unclassified
├───1.62 MB (09.10%) ++ heap-overhead
├───1.48 MB (08.33%) ++ window-objects/top(app://gallery.gaiamobile.org/index.html, id=1)
├───0.39 MB (02.17%) ++ (10 tiny)
├───0.30 MB (01.66%) ── freetype
├───0.29 MB (01.64%) ── layout/style-sheet-cache
├───0.25 MB (01.38%) ++ gfx
├───0.24 MB (01.37%) ++ xpcom
└───0.18 MB (01.02%) ── preferences

Punam Dahiya [:pdahiya]

Assignee

Comment 13

•

9 years ago

For 2.5 in addition to 213x213 sizes, we are generating 159x159 sizes which is taking 0.10 MB more per image blob. This accounts for 2 MB extra memory usage for images (20 images for light workload) by gallery in 2.5.

Setting NI flag for David and Seth to share the findings and get their inputs on why we are decoding images to 159x159 sizes in 2.5. Thanks!

Flags: needinfo?(seth)

Flags: needinfo?(dflanagan)

Punam Dahiya [:pdahiya]

Assignee

Comment 14

•

9 years ago

───6.61 MB (30.87%) -- js-non-window
│   ├──3.37 MB (15.75%) ++ runtime
│   ├──3.01 MB (14.06%) -- zones
│   │  ├──2.29 MB (10.70%) -- zone(0xb2853400)
│   │  │  ├──0.93 MB (04.34%) ++ compartment([System Principal])
│   │  │  ├──0.63 MB (02.95%) ++ (10 tiny)
│   │  │  ├──0.50 MB (02.31%) ++ compartment([System Principal], processChildGlobal) ********************
│   │  │  └──0.24 MB (01.10%) ++ strings
│   │  ├──0.55 MB (02.55%) ++ zone(0xb6820100)
│   │  └──0.17 MB (00.81%) ++ zone(0xb6820b00)


Additional overhead of 0.5 MB seen across apps in 2.5 with processChildGlobal

Punam Dahiya [:pdahiya]

Assignee

Comment 15

•

9 years ago

├───5.68 MB (31.90%) -- js-non-window
│   ├──2.99 MB (16.83%) ++ runtime
│   ├──2.21 MB (12.44%) -- zones
│   │  ├──1.64 MB (09.21%) -- zone(0xb3d13400)
│   │  │  ├──0.85 MB (04.79%) ++ compartment([System Principal])
│   │  │  ├──0.40 MB (02.22%) ++ (9 tiny)
│   │  │  └──0.39 MB (02.20%) ++ strings/string(<non-notable strings>)
│   │  ├──0.42 MB (02.35%) ++ zone(0xb30abc00)
│   │  └──0.16 MB (00.88%) ++ zone(0xb30af400)

which is not seen above in 2.2 memory report. Its underlying gecko change and could be one of the reason of memory increase seen across apps in 2.5

Seth Fowler [:seth] [:s2h]

Comment 16

•

9 years ago

(In reply to Punam Dahiya [:pdahiya] from comment #13)
> For 2.5 in addition to 213x213 sizes, we are generating 159x159 sizes which
> is taking 0.10 MB more per image blob. This accounts for 2 MB extra memory
> usage for images (20 images for light workload) by gallery in 2.5.
> 
> Setting NI flag for David and Seth to share the findings and get their
> inputs on why we are decoding images to 159x159 sizes in 2.5. Thanks!

We do not want the 213x213 versions. We only want the 159x159 versions. The 159x159 versions are the result of downscale-during-decode decoding the images to the size at which they will be displayed on the screen.

Unfortunately, we have not yet turned off the old, non-downscale-during-decode code path. The 213x213 versions are the full size versions of the images resulting from that code path. In order to turn that code path off, bug 1207355 needs to land. I'm currently waiting on review for those patches. Bug 1207355 will definitely land in Gecko 34 and will thus be a part of B2G 2.5.

I'm clearing the NI for David as I'm not sure there's really anything to add here; the problem is well understood.

Depends on: 1207355

Flags: needinfo?(seth)

Flags: needinfo?(dflanagan)

Punam Dahiya [:pdahiya]

Assignee

Comment 17

•

9 years ago

(In reply to Seth Fowler [:seth] [:s2h] from comment #16)
> (In reply to Punam Dahiya [:pdahiya] from comment #13)
> > For 2.5 in addition to 213x213 sizes, we are generating 159x159 sizes which
> > is taking 0.10 MB more per image blob. This accounts for 2 MB extra memory
> > usage for images (20 images for light workload) by gallery in 2.5.
> > 
> > Setting NI flag for David and Seth to share the findings and get their
> > inputs on why we are decoding images to 159x159 sizes in 2.5. Thanks!
> 
> We do not want the 213x213 versions. We only want the 159x159 versions. The
> 159x159 versions are the result of downscale-during-decode decoding the
> images to the size at which they will be displayed on the screen.
> 
Getting rid of 213 x 213 images should save ~ 3.6 MB and will be a huge improvement in memory cold launch numbers for gallery. Thanks Seth for clarifying and the update.

> Unfortunately, we have not yet turned off the old,
> non-downscale-during-decode code path. The 213x213 versions are the full
> size versions of the images resulting from that code path. In order to turn
> that code path off, bug 1207355 needs to land. I'm currently waiting on
> review for those patches. Bug 1207355 will definitely land in Gecko 34 and
> will thus be a part of B2G 2.5.
> 
> I'm clearing the NI for David as I'm not sure there's really anything to add
> here; the problem is well understood.

David Flanagan [:djf]

Comment 18

•

9 years ago

Moving the 2.5+ flag to bug 1207355 since there is nothing for us to do here other than wait for that bug to land.

Punam: would you email dev-b2g about the image size findings here? The double decode issues will affect the memory usage of any other app that displays a lot of images, so it could be causing memory regressions for the music and video apps among others.

blocking-b2g: 2.5+ → ---

Flags: needinfo?(pdahiya)

Punam Dahiya [:pdahiya]

Assignee

Comment 19

•

9 years ago

(In reply to David Flanagan [:djf] from comment #18)
> Moving the 2.5+ flag to bug 1207355 since there is nothing for us to do here
> other than wait for that bug to land.
> 
> Punam: would you email dev-b2g about the image size findings here? The
> double decode issues will affect the memory usage of any other app that
> displays a lot of images, so it could be causing memory regressions for the
> music and video apps among others.

Thanks David, just shared findings with dev-fxos mailing list.

Flags: needinfo?(pdahiya)

Bobby Chien

Reporter

Comment 20

•

9 years ago

Gallery is 976ms[1] on Flame and 492ms[2] on Aries.

[1] http://mzl.la/1M29Cyt
[2] http://mzl.la/1M29wa8

Status: NEW → RESOLVED

Closed: 9 years ago

Resolution: --- → FIXED

Target Milestone: --- → 2.6 S1 - 11/20

Bugzilla

Quick Search

Performance regression in Gallery

Categories

(Firefox OS Graveyard :: Gaia::Gallery, defect, P1)

Tracking

(Not tracked)

People

(Reporter: bobby.chien+bugzilla, Assigned: pdahiya)

References

(
URL
)

Details

(Keywords: perf, regression)

Crash Data

Security

(public)

User Story

Description

Updated

Updated

Comment 2

Comment 3

Comment 4

Comment 5

Comment 6

Comment 7

Comment 8

Comment 9

Comment 10

Comment 11

Comment 12

Comment 13

Comment 14

Comment 15

Comment 16

Comment 17

Comment 18

Comment 19

Comment 20