aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Much more efficient half_to_float() code, via @rygorous.

Browse Source 
Also, switch deferred shading example to use it. (Rather than
the "fast" half to float that doesn't handle deforms, etc.)
This commit is contained in:
Matt Pharr
2012-03-21 10:42:27 -07:00
parent 316de0b880
commit 10c5ba140c
2 changed files with 35 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
examples/deferred/kernels.ispc
View File

@@ -327,8 +327,8 @@ ShadeTile(
// Reconstruct normal from G-buffer
float surface_normal_x, surface_normal_y, surface_normal_z;
float normal_x = half_to_float_fast(inputData.normalEncoded_x[gBufferOffset]);
float normal_y = half_to_float_fast(inputData.normalEncoded_y[gBufferOffset]);
float normal_x = half_to_float(inputData.normalEncoded_x[gBufferOffset]);
float normal_y = half_to_float(inputData.normalEncoded_y[gBufferOffset]);
float f = (normal_x - normal_x * normal_x) + (normal_y - normal_y * normal_y);
float m = sqrt(4.0f * f - 1.0f);
@@ -339,9 +339,9 @@ ShadeTile(
// Load other G-buffer parameters
float surface_specularAmount =
half_to_float_fast(inputData.specularAmount[gBufferOffset]);
half_to_float(inputData.specularAmount[gBufferOffset]);
float surface_specularPower =
half_to_float_fast(inputData.specularPower[gBufferOffset]);
half_to_float(inputData.specularPower[gBufferOffset]);
float surface_albedo_x = Unorm8ToFloat32(inputData.albedo_x[gBufferOffset]);
float surface_albedo_y = Unorm8ToFloat32(inputData.albedo_y[gBufferOffset]);
float surface_albedo_z = Unorm8ToFloat32(inputData.albedo_z[gBufferOffset]);