Navigate to the top level

Surface_as_vertex_array

Description
Tests rendering to GPU surfaces (render targets) and subsequently binding the rendered surface memory as a vertex attribute array (vertex stream).
This pattern takes advantage of Xbox UMA (Unified Memory Architecture) to perform GPU-driven vertex generation or displacement mapping without hardware vertex texture fetch.
Suite class def
src/tests/surface_as_vertex_array_tests.h

DynamicUpdateLoop

Sequentially renders a single color into a surface and immediately draws the corresponding quad in four successive passes within a single frame to verify surface update synchronization between draw passes. Expect to see four quads on a dark background: top-left Red, top-right Green, bottom-left Blue, bottom-right Yellow.

Image with alpha
Test result for Surface_as_vertex_array::DynamicUpdateLoop
Image as seen on display
No alpha test result for Surface_as_vertex_array::DynamicUpdateLoop

LinearDiffuseArray

Renders four solid color strips (Red, Green, Blue, Yellow) into the first row of a linear A8R8G8B8 surface. The surface memory is then bound as NV2A_VERTEX_ATTR_DIFFUSE with TYPE_UB_D3D format and drawn via DRAW_ARRAYS. Expect to see four quads on a dark background: top-left Red, top-right Green, bottom-left Blue, bottom-right Yellow.

Image with alpha
Test result for Surface_as_vertex_array::LinearDiffuseArray
Image as seen on display
No alpha test result for Surface_as_vertex_array::LinearDiffuseArray

MultiStream

Tests multiple simultaneous surface vertex streams. Surface A is rendered with Cyan and Black pixels and bound to NV2A_VERTEX_ATTR_DIFFUSE; Surface B is rendered with Black and Magenta pixels and bound to NV2A_VERTEX_ATTR_SPECULAR. A register combiner adds Diffuse and Specular colors. Expect to see two wide quads on a dark background: left Cyan (Cyan + Black) and right Magenta (Black + Magenta).

Image with alpha
Test result for Surface_as_vertex_array::MultiStream
Image as seen on display
No alpha test result for Surface_as_vertex_array::MultiStream

RenderScalePattern

Renders carrier triangle vertex indices and barycentric weights across swizzled surfaces A and B. The surfaces are bound as Diffuse and Specular vertex streams with a vertex shader that blends three carrier positions via address-register palette lookups (arl a0.x / c[A0+112]) to reconstruct a 3D mesh. Expect to see a clean, contiguous 3D mesh at 1x render scale. When emulator render scale > 1.0 downsamples/filters the surface with interpolation errors across diagonal seams, shifts in index values cause address-register lookups to select incorrect palette slots, visibly tearing the mesh into spikes (reproducing the behavior in the XDK DisplacementMap sample and xemu PR #2984).

Image with alpha
Test result for Surface_as_vertex_array::RenderScalePattern
Image as seen on display
No alpha test result for Surface_as_vertex_array::RenderScalePattern

SwizzledDiffuseArray

Renders four 2x2 color blocks into a swizzled A8R8G8B8 surface (paired with matching 32-bit depth buffer). Due to Morton order swizzling, the 2x2 blocks occupy contiguous 4-pixel chunks in memory. The surface is then bound as NV2A_VERTEX_ATTR_DIFFUSE. Expect to see four quads on a dark background: top-left Red, top-right Green, bottom-left Blue, bottom-right Yellow.

Image with alpha
Test result for Surface_as_vertex_array::SwizzledDiffuseArray
Image as seen on display
No alpha test result for Surface_as_vertex_array::SwizzledDiffuseArray