Vertex Shader constants

[fboivin 156]

In my HLSL vertex shader, I have an array of floats constants declared like this: float factors[6]; In my C++ code, I set the values of the array using the ID3DXConstantTable interface like this: ConstantTable->SetFloatArray( d3dDevice, "factors", factors, 6); However, what if I wanted to set the constants using the IDirect3DDevice9 SetVertexShaderConstantF method? The method only allows to upload vector4s so it's not possible to only upload 6 floats. Did I miss a method that allows to upload single floats in the IDirect3DDevice9 interface? If possible, I don't want to "patch" my array to 8 floats. Thanks! Francis

[circlesoft 1178]

Quote:

Original post by fboivin However, what if I wanted to set the constants using the IDirect3DDevice9 SetVertexShaderConstantF method? The method only allows to upload vector4s so it's not possible to only upload 6 floats.

It looks like SetVertexShaderConstantF() will only accept float4s. This is probably because a constant in HLSL is defined as a float4.

So you can do one of two things:

(1) Continue using the constant table
(2) Use ID3DXEffect

The SetVertexShaderConstant*() functions aren't used much anymore, especially for HLSL shaders.

[PAiNIC 128]

D3DDevice::SetVertexShaderConstantF is really there to deal with assembled shaders that don't have constant tables. GPU's are vector processors, so single floats don't really exist in hardware, they're actually represented as float4's. This would be why no single-float constant setting methods exist. The HLSL compiler can re-arrange single-float constants and pack four of them into a float4 constant implicitly.

Since you're using HLSL, it's probably best to continue using constant tables to set your VS constants.

[xsirxx 170]

Does SetFloatArray(...) not work? I use this on the PS' and VS'.... Does that not work for effects?

I grab the constant table, get the constant with GetConstantByName(...). Then do a SetFloatArray(..., 6);

[fboivin 156]

Playing a little more with this, I found out that if I declare a constant as "float my_array[6]" in HLSL, the compiler expands it to something like "float4 my_array[6]". Checking the output assemnly confirmed this as my_array's size was said to be 6.

Nice way to eat some bandwidth without noticing (ie: 24 floats when you think it's 6)!

[xsirxx 170]

Quote:

Original post by fboivin Playing a little more with this, I found out that if I declare a constant as "float my_array[6]" in HLSL, the compiler expands it to something like "float4 my_array[6]". Checking the output assemnly confirmed this as my_array's size was said to be 6. Nice way to eat some bandwidth without noticing (ie: 24 floats when you think it's 6)!

So if you declare:
float my_array[4]
float my_array2[4]

then its only 8?

[fboivin 156]

I'd have to test it out, but I'd think from what I've seen that this would be expanded to float4 my_array[4] and float4 my_array2[4].