vkd3d: Implement negative viewport height feature. #1611
Conversation
tests/d3d12_command.c
Outdated
| vbv.SizeInBytes = sizeof(quad); | ||
|
|
||
| set_viewport(&viewport, 0, context.render_target_desc.Height, | ||
| context.render_target_desc.Width, -(float)context.render_target_desc.Height, 0.0f, 1.0f); |
There was a problem hiding this comment.
I think a viewport test like this should test 4 draw calls targeting 4 quadrants to make sure the intended X / Y offsets end up where we expect.
Might as well throw InvertedViewportDepthFlipsZSupported in there and render different depths per quadrant.
There was a problem hiding this comment.
Sounds good, done. PTAL
null77
force-pushed
the
negative-viewport-height
branch
from
800f469
to
3200a4c
Compare
null77
force-pushed
the
negative-viewport-height
branch
from
3200a4c
to
1f3ef19
Compare
tests/d3d12_command.c
Outdated
| quads[4][4] = | ||
| { | ||
| { | ||
| {{0.0f, 1.0f, 1.0f, 1.0f}, {0.1f, 0.1f}}, |
There was a problem hiding this comment.
There is flat TEXCOORD input here to PS, so it does not verify that Y-flip actually happens.
There was a problem hiding this comment.
Changed to a variable TEXCOORD Y.
tests/d3d12_command.c
Outdated
| float4 main(float4 position : SV_Position, | ||
| float2 texcoord : TEXCOORD) : SV_Target | ||
| { | ||
| return float4(position.xyz, texcoord.x); |
There was a problem hiding this comment.
Should be texcoord.y here I think, since that will be "flipped" when viewport is flipped.
There was a problem hiding this comment.
Right, changed to output y.
tests/d3d12_command.c
Outdated
|
|
||
| for (i = 0; i < 4; i++) | ||
| { | ||
| ID3D12GraphicsCommandList_IASetVertexBuffers(command_list, 0, 1, &vbvs[i]); |
There was a problem hiding this comment.
I meant that we need 4 different RSSetViewports here rather than using the VBO to do the same thing.
These features correspond to Vulkan-native functionality.
null77
force-pushed
the
negative-viewport-height
branch
from
1f3ef19
to
f478f64
Compare
This mirrors a Vulkan feature and doesn't require any emulation.
@HansKristian-Work LMK if you'd prefer I de-duplicate the test code.