[WIP] Add support for double precision floats #288 #12299

fire · 2017-10-22T01:36:39Z

Work in progress. DO NOT MERGE.

fire · 2017-10-22T01:58:18Z

1>  cl /Fothirdparty\libpng\pngrtran.windows.opt.tools.64.obj /c thirdparty\libpng\pngrtran.c /nologo /O2 /DDEBUG_ENABLED /MT /Gd /GR /nologo "/IC:\Program Files (x86)\Windows Kits\10\/Include" /DWINDOWS_ENABLED /DOPENGL_ENABLED /DRTAUDIO_ENABLED /DWASAPI_ENABLED /DTYPED_METHOD_BIND /DWIN32 /DWINVER=0x0601 /D_WIN32_WINNT=0x0601 /D_WIN64 /DREAL_T_IS_DOUBLE "/IC:\Program Files (x86)\Microsoft Visual Studio\2017\Professional\VC\Tools\MSVC\14.11.25503\include" "/IC:\Program Files (x86)\Windows Kits\NETFXSDK\4.6.1\include\um" "/IC:\Program Files (x86)\Windows Kits\10\include\10.0.15063.0\ucrt" "/IC:\Program Files (x86)\Windows Kits\10\include\10.0.15063.0\shared" "/IC:\Program Files (x86)\Windows Kits\10\include\10.0.15063.0\um" "/IC:\Program Files (x86)\Windows Kits\10\include\10.0.15063.0\winrt" "/IC:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\include" /I "/IC:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\atlmfc\include" /I "/IC:\Program Files (x86)\Windows Kits\10\Include\10.0.10240.0\ucrt" /I /I "/IC:\Program Files (x86)\Windows Kits\8.1\Include\um" "/IC:\Program Files (x86)\Windows Kits\8.1\Include\shared" "/IC:\Program Files (x86)\Windows Kits\8.1\Include\winrt" /I /I /w /DMSVC /GR -DPTRCALL_ENABLED -DTOOLS_ENABLED -DGDSCRIPT_ENABLED -DMINIZIP_ENABLED -DXML_ENABLED -DGLAD_ENABLED -DGLES_OVER_GL -DPNG_ARM_NEON_OPT=0 /Icore /Icore\math /Ieditor /Idrivers /I. /Iplatform\windows /Ithirdparty\zstd /Ithirdparty\zstd\common /Ithirdparty\zlib /Ithirdparty\rtaudio /Ithirdparty\glad /Ithirdparty\libpng
1>C:\Program Files (x86)\Microsoft Visual Studio\2017\Professional\VC\Tools\MSVC\14.11.25503\include\type_traits(2342): warning C4577: 'noexcept' used with no exception handling mode specified; termination on exception is not guaranteed. Specify /EHsc
1>  pngmem.c
1>drivers\gles3\shader_gles3.cpp(98): error C2664: 'void (GLint,GLsizei,GLboolean,const GLfloat *)': cannot convert argument 4 from 'const real_t *' to 'const GLfloat *'
1>  drivers\gles3\shader_gles3.cpp(98): note: Types pointed to are unrelated; conversion requires reinterpret_cast, C-style cast or function-style cast
1>  pngpread.c

Blocked until I investigate this futher.

fire · 2017-10-22T02:02:01Z

https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_gpu_shader_fp64.txt Need to get the fp64 functions somehow.

fire · 2017-10-27T12:31:38Z

My plan is to split the patch into converting floats to real_t and @akien-mga What do you think?

That means trying to do the first #288 (comment) .

There is some support for using doubles in binary serialization,
but it's probably not complete. files such as resouce_format_binary.cpp
and marshalls.cpp would need to be checked and made sure that
doubles are written (and read) in a binary compatible way.

This is tested via float=64 option. Currently failing.

akien-mga · 2018-01-04T14:58:46Z

Thanks a lot for your contribution!
We are now entering a strict release freeze for Godot 3.0 and will only consider major bug fixes. We won't merge new features and enhancements anymore until 3.0 is released.

Moving this PR to the 3.1 milestone, to be reviewed once the release freeze is lifted. It could eventually be cherry-picked for a future 3.0.1 maintenance release if it doesn't change the user-facing APIs and doesn't compromise the engine's stability.

aaronfranke · 2018-01-15T00:52:20Z

@akien-mga This should definitely be considered a feature request and not a bug fix. It doesn't make too much sense to add to a 3.0.1 maintenance release, but the 3.1 milestone does make a lot of sense.

reduz · 2018-01-15T03:31:45Z

probably will make sense for 3.1, I definitely want this eventually added

…

On Sun, Jan 14, 2018 at 9:52 PM, Aaron Franke ***@***.***> wrote: @akien-mga <https://github.com/akien-mga> This should definitely be considered a feature request and not a bug fix. It doesn't make too much sense to add to a 3.0.1 maintenance release, but the 3.1 milestone does make a lot of sense. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#12299 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF-Z26idrw_QFMjV4Fj-tHhJesBbFMV8ks5tKqFOgaJpZM4QBxxo> .

fire · 2018-04-29T03:57:04Z

Was able to get a build running with many things converted to real_t (double).

Use scons p=windows vsproj=yes float=64 -j36 to build with floating point being 64 bit.

I had to disable ptrcall, bullet and gdnative. Would be interested in people looking why they break.

Note that the graphics driver still converts back to float.

fire · 2018-04-29T04:14:51Z

Hm, it was actually having linking errors :/ However if someone could help that would be so useful.

aaronfranke · 2018-04-29T04:51:16Z

Color math can be left as single-precision if it doesn't cause any issues. 23 significant digits is enough for HDR color math (should have 16 bits per channel) and most displays have 8-12 bits per channel.

aaronfranke · 2018-07-08T06:11:49Z

@fire I think it would be best if you created many smaller pull requests instead of this one big pull request. Not only does it get conflicts easily but also it's hard to review so many changes.

For example, you could make pull requests for "replace float with real_t in x and y files" or something. If you did this then we could merge these small pull requests one at a time to slowly bring the engine to be double-compatible and also have others test each small change as it gets into master.

fire · 2018-07-23T20:23:21Z

Got your message, I'll try to find effort to work on this.

aaronfranke · 2018-09-14T07:23:13Z

modules/bullet/bullet_physics_server.cpp

@@ -629,14 +629,14 @@ uint32_t BulletPhysicsServer::body_get_user_flags(RID p_body) const {
 	return 0;
 }

-void BulletPhysicsServer::body_set_param(RID p_body, BodyParameter p_param, float p_value) {
+void BulletPhysicsServer::body_set_param(RID p_body, BodyParameter p_param, real_t p_value) {


~~Shouldn't these changes~~

~~Be using Bullet's btScalar instead of Godot's real_t?~~

~~Be contributed upstream to Bullet directly?~~

modules/bullet is Godot's code, not upstream.

aaronfranke · 2018-11-14T04:59:27Z

This PR should probably be assigned the "salvageable" tag.

fire · 2019-04-06T15:20:03Z

@aaronfranke What are your plans for taking this forward? Doing some spring updates and cleaning on my pull requests.

aaronfranke · 2019-04-06T18:11:14Z

After #21922 is merged I plan to pick-and-test parts of this one, since this PR is functional-ish.

fire · 2019-04-30T14:50:19Z

@aaronfranke Updated to lastest. Still work in progress.

Anutrix · 2019-06-18T16:12:24Z

Any updates?

fire · 2019-07-12T15:45:59Z

Rebased.

akien-mga · 2019-07-18T13:36:13Z

Moving to 4.0 milestone as that's when we'll handle #288 (via this PR, #21922 and some more changes needed).

fire · 2019-10-22T07:04:02Z

This pull request is very out of date.

I'll take a look at rebasing on top of #21922 and Vulkan.

akien-mga changed the title ~~Add support for double precision floats #288~~ [WIP] Add support for double precision floats #288 Oct 22, 2017

akien-mga added enhancement topic:core labels Oct 22, 2017

fire force-pushed the double branch 2 times, most recently from 0bb8578 to 7026033 Compare October 22, 2017 17:41

fire force-pushed the double branch from 56a5812 to b33369e Compare October 27, 2017 13:46

akien-mga added this to the 3.1 milestone Jan 4, 2018

fire force-pushed the double branch from 4dacc20 to 9544808 Compare February 4, 2018 21:40

fire mentioned this pull request Feb 4, 2018

Add support for double precision floats #288

Closed

aaronfranke mentioned this pull request Mar 17, 2018

[Mono] Replace float with real_t, other misc C# improvements #17134

Merged

fire force-pushed the double branch from c4f8dac to 9181a3f Compare April 29, 2018 01:38

akien-mga modified the milestones: 3.1, 3.2 Jul 26, 2018

aaronfranke mentioned this pull request Sep 10, 2018

Some work on double-precision support #21922

Merged

aaronfranke reviewed Sep 14, 2018

View reviewed changes

fire force-pushed the double branch from ee6bdaf to 45896ca Compare April 30, 2019 14:49

fire requested review from AndreaCatania, BastiaanOlij, bojidar-bg, karroffel, reduz and vnen as code owners April 30, 2019 14:49

fire force-pushed the double branch 2 times, most recently from 2bd71a7 to 072a7d6 Compare April 30, 2019 15:14

akien-mga removed request for BastiaanOlij, karroffel, bojidar-bg, vnen and AndreaCatania April 30, 2019 15:27

fire force-pushed the double branch from 072a7d6 to 4acd873 Compare May 25, 2019 16:13

64 float support

078ae78

fire force-pushed the double branch from 4acd873 to 078ae78 Compare July 12, 2019 15:45

akien-mga modified the milestones: 3.2, 4.0 Jul 18, 2019

qarmin mentioned this pull request Oct 7, 2019

Failure of mathematical analysis in 3D scene #32604

Closed

fire closed this Oct 22, 2019

akien-mga added the archived label Oct 22, 2019

fire mentioned this pull request Sep 20, 2022

Emulate double precision for regular rendering operation when REAL_T_IS_DOUBLE #66178

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] Add support for double precision floats #288 #12299

[WIP] Add support for double precision floats #288 #12299

fire commented Oct 22, 2017 •

edited

Loading

fire commented Oct 22, 2017

fire commented Oct 22, 2017

fire commented Oct 27, 2017 •

edited

Loading

akien-mga commented Jan 4, 2018

aaronfranke commented Jan 15, 2018

reduz commented Jan 15, 2018 via email

fire commented Apr 29, 2018

fire commented Apr 29, 2018

aaronfranke commented Apr 29, 2018 •

edited

Loading

aaronfranke commented Jul 8, 2018 •

edited

Loading

fire commented Jul 23, 2018

aaronfranke Sep 14, 2018 •

edited

Loading

akien-mga Sep 14, 2018

aaronfranke commented Nov 14, 2018

fire commented Apr 6, 2019

aaronfranke commented Apr 6, 2019

fire commented Apr 30, 2019

Anutrix commented Jun 18, 2019

fire commented Jul 12, 2019

akien-mga commented Jul 18, 2019

fire commented Oct 22, 2019

[WIP] Add support for double precision floats #288 #12299

[WIP] Add support for double precision floats #288 #12299

Conversation

fire commented Oct 22, 2017 • edited Loading

fire commented Oct 22, 2017

fire commented Oct 22, 2017

fire commented Oct 27, 2017 • edited Loading

akien-mga commented Jan 4, 2018

aaronfranke commented Jan 15, 2018

reduz commented Jan 15, 2018 via email

fire commented Apr 29, 2018

fire commented Apr 29, 2018

aaronfranke commented Apr 29, 2018 • edited Loading

aaronfranke commented Jul 8, 2018 • edited Loading

fire commented Jul 23, 2018

aaronfranke Sep 14, 2018 • edited Loading

Choose a reason for hiding this comment

akien-mga Sep 14, 2018

Choose a reason for hiding this comment

aaronfranke commented Nov 14, 2018

fire commented Apr 6, 2019

aaronfranke commented Apr 6, 2019

fire commented Apr 30, 2019

Anutrix commented Jun 18, 2019

fire commented Jul 12, 2019

akien-mga commented Jul 18, 2019

fire commented Oct 22, 2019

fire commented Oct 22, 2017 •

edited

Loading

fire commented Oct 27, 2017 •

edited

Loading

aaronfranke commented Apr 29, 2018 •

edited

Loading

aaronfranke commented Jul 8, 2018 •

edited

Loading

aaronfranke Sep 14, 2018 •

edited

Loading