Shader Program Header Specification

The SPH field SphType sets the type of shader, where the type is either TYPE_01_VTG or TYPE_02_PS.

The SPH field Version sets is used during development to pick the version.

The SPH field ShaderType sets the type (e.g, VERTEX, TESSELLATION, GEOMETRY, or PIXEL) of shader for the shader program.

The SPH field MrtEnable, when TRUE indicates that the pixel shader outputs multiple colors (the number being controlled by the SPH Omap). It is always AND’d with SetCtMrtEnable.V(eff) to allow the driver to dynamically override the MRT (Multiple Render Target) behavior of the pixel shader. If the result is TRUE, then the pixel shader outputs will each be sent to its corresponding enabled target. If the result is FALSE, then pixel shader output 0 will be sent to each enabled target. This override of MRT is necessary to support OGL’s DrawBuffer call (which is inherently non-MRT) when an MRT enabled pixel shader is active. This field has no effect on the blending enables; that is, whether MrtEnable result is TRUE or FALSE, each color target still has an independent blend enable (unless SetSingleRopControl.Enable is TRUE). This SPH field is only used for pixel shaders.

The SPH field KillsPixels, if TRUE, enables pixel shader programs to kill pixels. When set to FALSE, pixel shaders KIL instructions become no-operations and trigger a hardware exception. Also, when this field is TRUE, EarlyZ is turned off, and Zcull’s visible pixel counting acceleration is turned off. This field has no effect on the texture color key operations. This SPH field is only used for pixels shaders.

The SPH field DoesGlobalStore indicates the shader might perform a global store.

The SPH field StreamOutMask selects which GS output streams are enbled as outputs from the GS. There are four GS output streams, numbered 0 to 3. If a stream is disabled in StreamOutMask, it is never written even if a buffer is bound to it.

The SPH field DoesLoadOrStore is used to enable power optimizations by disabling the load/store path if it is not being used. If a shader unit is only running pixel work that has DoesLoadOrStore set to FALSE, and it has declared no additional CallReturnStack by setting ShaderLocalMemoryCrsSize to zero, the load-store path can be safely shut down temporarily. When DoesLoadOrStore == FALSE, LD, ST, and all the variations thereof in the ISA, will be noop’ed by the HW.

The SPH field DoesFp64 is used power-off the double precision math if the compiler can guarantee it will never be used. If all of the work running on a given Shader unit has DoesFp64 set to FALSE, this math block will be powered down. Any double precision instruction encountered when DoesFp64 is FALSE will be noop’ed by the HW.

The SPH fields ShaderLocalMemoryLowSize and ShaderLocalMemoryHighSize set the required size of thread-private memory, for variable storage, needed by the shader program.

The SPH field PerPatchAttributeCount indicates the number of per-patch attributes that are written by the tesselation init shader (and read by the subsequent tesselation shader). Per-patch attributes are in addition to per-vertex attributes. This field is only used on tesselation init shaders.

The SPH field ThreadsPerInputPrimitive sets the maximum number of threads that are invoked for a primitive, thereby allowing the work of one shader to be divided amongst several shaders. This is the number of "instanced" shaders. This field has the following shader-specific meanings:

The SPH field ShaderLocalMemoryCrsSize sets the additional (off chip) call/return stack size (CRS_SZ). Units are in Bytes/Warp. Minimum value 0, maximum 1 megabyte. Must be multiples of 512 bytes.

The SPH field OutputTopology sets the primitive topology of the vertices that are output from the pipe stage. This field is only used with geometry shaders, where the value must be greater than zero and has a maximum of 1024. The allowed values are:

The SPH field MaxOutputVertexCount sets the maximum number of vertices that can be output by one shader thread. This field is only used with geometry shaders, where the value sets the maximum number of vertices output per thread, and OUT instructions beyond this are noop’ed.

The SPH fields StoreReqStart and StoreReqEnd set a range of attributes whose corresponding Odmap values of ST or ST_LAST are treated as ST_REQ. Normally, for an attribute whose Omap bit is TRUE and Odmap value is ST, when the shader writes data to this output, it can not count on being able to read it back, since the next downstream shader might have its Imap bit FALSE, thereby causing the Bmap bit to be FALSE. By including a ST type of attribute in the range of StoreReqStart and StoreReqEnd, the attribute’s Odmap value is treated as ST_REQ, so an Omap bit being TRUE causes the Bmap bit to be TRUE. This guarantees the shader program can output the value and then read it back later. This will save register space.

The SPH field StoreReqStart sets the first attribute whose ST or ST_LAST Odmap values are treated as ST_REQ. Note that Odmap values of discard are not affected.

The SPH field StoreReqEnd sets the last attribute whose ST of ST_LAST Odmap values are treated as ST_REQ. If no attributes are to have their Odmap value treated as ST_REQ, then the SPH needs to have StoreReqStart greater than StoreReqEnd.