Remove copy from hsMatrix2SIMD in Metal renderer #1647
+23
−45
Conversation
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
Hoikas
reviewed
Dec 31, 2024
dpogue
reviewed
Dec 31, 2024
colincornaby
force-pushed
the
hsMatrix2SIMD-without-implicit-copy
branch
from
b11d7cf to
14d89ad
Compare
|
I've rebased this branch to include the Linux runner changes. @dpogue - I'm still considering your question. No real thoughts yet - I think the underlying fMap in the matrix is a pointer type which is why it was a pointer here. Shame the matrix type includes extra data beyond the matrix, but using the fMap as the raw underlying matrix data is a small inconvenience. |
colincornaby
force-pushed
the
hsMatrix2SIMD-without-implicit-copy
branch
from
14d89ad to
618ccc2
Compare
|
I've edited this commit to use reinterpret_cast and to return a reference instead of a pointer. |
hsMatrix2SIMD used to do a transpose of the input hsMatrix - but shaders and code were rewritten to make this unnecessary. hsMatrix2SIMD still does a vestigial memory copy. This commit removes that memory copy. This changes the meaning of hsMatrix2SIMD - so functions that need to make a copy now do so manually. Otherwise a direct cast is made. hsMatrix2SIMD has also been made constexpr. This should result in a small (~1%) performance improvement in the Metal renderer main loop.
colincornaby
force-pushed
the
hsMatrix2SIMD-without-implicit-copy
branch
from
618ccc2 to
0aa4708
Compare
Hoikas
reviewed
Jan 21, 2025
| matrix_float4x4* hsMatrix2SIMD(const hsMatrix44& src, matrix_float4x4* dst); | ||
| inline const matrix_float4x4& hsMatrix2SIMD(const hsMatrix44& src) | ||
| { | ||
| //reinterperate_cast not allowed in constexpr |
There was a problem hiding this comment.
Suggested change
| //reinterperate_cast not allowed in constexpr | |
| // reinterpret_cast not allowed in constexpr |
| inline const matrix_float4x4& hsMatrix2SIMD(const hsMatrix44& src) | ||
| { | ||
| //reinterperate_cast not allowed in constexpr | ||
| return *reinterpret_cast<const simd_float4x4 *>(src.fMap); |
There was a problem hiding this comment.
Suggested change
| return *reinterpret_cast<const simd_float4x4 *>(src.fMap); | |
| return *reinterpret_cast<const simd_float4x4*>(src.fMap); |
| matrix_float4x4 mat; | ||
| hsMatrix2SIMD(drawable->GetPaletteMatrix(span.fBaseMatrix + 1), &mat); | ||
| fDevice.CurrentRenderCommandEncoder()->setVertexBytes(&mat, sizeof(matrix_float4x4), VertexShaderArgumentBlendMatrix1); | ||
| const matrix_float4x4 *mat = &hsMatrix2SIMD(drawable->GetPaletteMatrix(span.fBaseMatrix + 1)); |
There was a problem hiding this comment.
Suggested change
| const matrix_float4x4 *mat = &hsMatrix2SIMD(drawable->GetPaletteMatrix(span.fBaseMatrix + 1)); | |
| const matrix_float4x4* mat = &hsMatrix2SIMD(drawable->GetPaletteMatrix(span.fBaseMatrix + 1)); |
| @@ -1315,8 +1314,7 @@ void plMetalPipeline::IRenderProjection(const plRenderPrimFunc& render, plLightI | |||
| fCurrentRenderPassUniforms->fogColor = {0.f, 0.f, 0.f}; | |||
| fCurrentRenderPassUniforms->diffuseCol = {1.f, 1.f, 1.f, 1.f}; | |||
|
|
|||
| matrix_float4x4 tXfm; | |||
| hsMatrix2SIMD(proj->GetTransform(), &tXfm); | |||
| const matrix_float4x4 &tXfm = hsMatrix2SIMD(proj->GetTransform()); | |||
There was a problem hiding this comment.
Suggested change
| const matrix_float4x4 &tXfm = hsMatrix2SIMD(proj->GetTransform()); | |
| const matrix_float4x4& tXfm = hsMatrix2SIMD(proj->GetTransform()); |
| if (weights[j]) { | ||
| float weight = weights[j]; | ||
| if (weight) { | ||
| const simd_float4x4 simdMatrix = hsMatrix2SIMD(matrixPalette[indices & 0xFF]); |
There was a problem hiding this comment.
Suggested change
| const simd_float4x4 simdMatrix = hsMatrix2SIMD(matrixPalette[indices & 0xFF]); | |
| const simd_float4x4& simdMatrix = hsMatrix2SIMD(matrixPalette[indices & 0xFF]); |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
hsMatrix2SIMD used to do a transpose of the input hsMatrix - but shaders and code were rewritten to make this unnecessary. hsMatrix2SIMD still does a vestigial memory copy. This commit removes that memory copy.
This changes the meaning of hsMatrix2SIMD - so functions that need to make a copy now do so manually. Otherwise a direct cast is made. hsMatrix2SIMD has also been made constexpr.
This should result in a small (~1%) performance improvement in the Metal renderer main loop.