Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix up SSE1 support so that MathGeoLib unit tests and benchmarks work. #3010

Closed
juj opened this issue Nov 21, 2014 · 2 comments
Closed

Fix up SSE1 support so that MathGeoLib unit tests and benchmarks work. #3010

juj opened this issue Nov 21, 2014 · 2 comments
Labels
SIMD

Comments

@juj
Copy link
Collaborator

juj commented Nov 21, 2014

Building MathGeoLib for Emscripten with SSE1 support will hit a runtime error

exception thrown: TypeError: invalid arguments,__ZN4math6float4C2ERKNS_6float3Ef

where the line in question reads

$6 = SIMD_float32x4_shuffle($1,
        SIMD_float32x4_shuffle($3,
            SIMD_float32x4_withX(SIMD_float32x4_splat(f0), $w_),
            0, 4, -1, -1),
        0, 1, 4, 5);

To reproduce, follow the instructions to build the MathGeoLib SSE1 test suite at juj/MathGeoLib#36 and run it in Firefox Nightly.
@juj juj added the SIMD label Nov 21, 2014
@kripken
Copy link
Member

kripken commented Nov 21, 2014

cc @sunfishcode

@juj juj mentioned this issue Nov 21, 2014
Open
@juj juj mentioned this issue Jun 20, 2015
Merged
juj added a commit to juj/emscripten-fastcomp that referenced this issue Sep 3, 2015
@juj
When encountering float literals that have noncanonical NaN bit patte…
755141e 
…rns in a SIMD vector, don't attempt to create a float SIMD vector directly out of them, since JS source can't represent NaNs with noncanonical bits. Instead, create a int SIMD vector and cast it to float one to detour to the SIMD vector with correct bits intact. Fixes emscripten-core/emscripten#2840,  emscripten-core/emscripten#3560, emscripten-core/emscripten#3010 and emscripten-core/emscripten#3403.
juj added a commit to juj/emscripten-fastcomp that referenced this issue Sep 3, 2015
@juj
When encountering float literals that have noncanonical NaN bit patte…
f0da889 
…rns in a SIMD vector, don't attempt to create a float SIMD vector directly out of them, since JS source can't represent NaNs with noncanonical bits. Instead, create a int SIMD vector and cast it to float one to detour to the SIMD vector with correct bits intact. Fixes emscripten-core/emscripten#2840,  emscripten-core/emscripten#3560, emscripten-core/emscripten#3010 and emscripten-core/emscripten#3403.
@juj juj mentioned this issue Sep 3, 2015
Merged
juj added a commit to juj/emscripten-fastcomp that referenced this issue Sep 3, 2015
@juj
When encountering float literals that have noncanonical NaN bit patte…
0f9d3d1 
…rns in a SIMD vector, don't attempt to create a float SIMD vector directly out of them, since JS source can't represent NaNs with noncanonical bits. Instead, create a int SIMD vector and cast it to float one to detour to the SIMD vector with correct bits intact. Fixes emscripten-core/emscripten#2840,  emscripten-core/emscripten#3560, emscripten-core/emscripten#3010 and emscripten-core/emscripten#3403.
@juj
Copy link
Collaborator Author

juj commented Sep 18, 2015

Closing as fixed.

@juj juj closed this as completed Sep 18, 2015
juj added a commit to juj/emscripten-fastcomp that referenced this issue Sep 19, 2015
@juj
When encountering float literals that have noncanonical NaN bit patte…
7bc3933 
…rns in a SIMD vector, don't attempt to create a float SIMD vector directly out of them, since JS source can't represent NaNs with noncanonical bits. Instead, create a int SIMD vector and cast it to float one to detour to the SIMD vector with correct bits intact. Fixes emscripten-core/emscripten#2840,  emscripten-core/emscripten#3560, emscripten-core/emscripten#3010 and emscripten-core/emscripten#3403.
juj added a commit to juj/emscripten-fastcomp that referenced this issue Sep 22, 2015
@juj
When encountering float literals that have noncanonical NaN bit patte…
374e6dd 
…rns in a SIMD vector, don't attempt to create a float SIMD vector directly out of them, since JS source can't represent NaNs with noncanonical bits. Instead, create a int SIMD vector and cast it to float one to detour to the SIMD vector with correct bits intact. Fixes emscripten-core/emscripten#2840,  emscripten-core/emscripten#3560, emscripten-core/emscripten#3010 and emscripten-core/emscripten#3403.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
SIMD
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@kripken @juj