Without this, we would reject programs which sampled multiple times from
registers defined in the same phase (block of instructions with the same
texture indirection count), as each sample would count as a new phase
beginning. Instead, keep track of which phases registers were written in,
and only bump phase when we're reading from one generated in this phase.
On the other hand, we failed to count oC or oD texture samples as being new
phases.
Bug #17865.
This set of code changes are for stencil code generation
support. Both one-sided and two-sided stenciling are supported.
In addition to the raw code generation changes, these changes had
to be made elsewhere in the system:
- Added new "register set" feature to the SPE assembly generation.
A "register set" is a way to allocate multiple registers and free
them all at the same time, delegating register allocation management
to the spe_function unit. It's quite useful in complex register
allocation schemes (like stenciling).
- Added and improved SPE macro calculations.
These are operations between registers and unsigned integer
immediates. In many cases, the calculation can be performed
with a single instruction; the macros will generate the
single instruction if possible, or generate a register load
and register-to-register operation if not. These macro
functions are: spe_load_uint() (which has new ways to
load a value in a single instruction), spe_and_uint(),
spe_xor_uint(), spe_compare_equal_uint(), and spe_compare_greater_uint().
- Added facing to fragment generation. While rendering, the rasterizer
needs to be able to determine front- and back-facing fragments, in order
to correctly apply two-sided stencil. That requires these changes:
- Added front_winding field to the cell_command_render block, so that
the state tracker could communicate to the rasterizer what it
considered to be the front-facing direction.
- Added fragment facing as an input to the fragment function.
- Calculated facing is passed during emit_quad().
Move GL_ARB_texture_non_power_of_two and GL_ATI_separate_stencil
from the generic extension list to the 965-specific list. Neither
extension is supported on i830-class hardware, and
GL_ATI_separate_stencil is not supported on i915-class hardare.
GL_ARB_texture_non_power_of_two is supported on i915-class hardare and
is already in the i915-specific list.
Keep track of enabled/active vertex attributes.
Keep track of potential vertex program outputs.
When generating fragment program, replace references to fragment attributes
which are effectively non-varying and non-computed passthrough attributes with
references to the new CURRENT_ATTRIB tracked state value.
Only downside is slight ugliness in VBO code where we need to validate state
twice in succession.
The ARB extension is a superset of the older SGIX extension. Any
hardware that can support the SGIX version can also support the ARB
version. In Mesa, any driver that supports one also supports the
other. This unification just simplifies some bits of code.
- Use a lookup table for log2.
- Compute (float) (1 << ipart) by tweaking with the exponent directly to
avoid integer overflow and float conversion.
- Also table negative exponents to avoid float division and branching.
- Implement util_fast_exp as function of util_fast_exp2.
