The current GLSL cmp instruction is incorrect, because:
- it ignores destination write mask
- it ignores source swizzle
- it ignores other source modifiers.
- it works incorrectly for src0 = 0
By default the FinalCombiner is setup to add the specular color to the
final result of the combiner operations. However, it should only do
this when the WINED3DRS_SPECULARENABLE renderstate is enabled.
- Implement if, else, endif, rep, endrep, break
- Implement ifc, breakc, using undocumented comparison bits in the instruction token
- Fix bug in main loop processing of codes with no dst token
- Fix bug in GLSL output modifier processing of codes with no dst token
- Fix bug in loop implementation (src1 contains the integer data, src0 is aL)
- Add versioning for all the instructions above, and remove
GLSL_REQUIRED thing, which is useless and should be removed from all
opcodes in general.
- move DEF, DEFI, DEFB handling into the register counting pass
- keep track of defined constants as a linked list (because there's a
few of them)
- apply immediate constants after global constants in the constant
loading function
- both types of constants now get loaded with array notation in the
shader (into the same array)
Do not attach non-GLSL shaders to the GLSL program, that will cause a
crash. Mix with ARB shaders is never going to happen, because the
selection code will always choose GLSL for both or ARB for both.
- currently half the shader selection code (GLSL vs ARB) is in
fillGLcaps. The parts that check for software shaders are in
GetDeviceCaps. That placement, will work, but is definitely not optimal.
FillGLcaps should detect support - it should not make decision as to
what's used, because that's not what the purpose of the function is.
GetDeviceCaps should report support as it has already been selected.
Instead, select shader mode in its own function, called in the
appropriate places.
- unifying pixel and vertex shaders into a single selection is a
mistake. A software vertex shader can be coupled with a hardware arb or
glsl pixel shader, or no shader at all. Split them back into two and add
a SHADER_NONE variant.
- drawprim is doing support checks for ARB_PROGRAM, and making shader
decisions based on that - that's wrong, support has already been
checked, and decided upon, and shaders can be implemented via software,
ARB_PROGRAm or GLSL, so that support check isn't valid.
- Store the shader selected mode into the shader itself. Different types
of shaders can be combined, so this is an improvement. In fact, storing
the mode into the settings globally is a mistake as well - it should be
done per device, since different cards have different capabilities.
This fixes the translations for a few instructions in GLSL and allows
Cubemap sampling in pixel shaders < 2.0. It makes some of the
lighting on textures in Half Life 2 look better, including some of the
water effects. It's not perfect yet, but much closer now.
Make wined3d use register combiners for texture stage operations. In
order to do that the texture unit index needs to be separated from the
texture stage index. For cards that don't support the
NV_register_combiners extension nothing should change.
On nVidia cards the value of GL_MAX_TEXTURE_UNITS is generally not
larger than 4. In Direct3D that would correspond to
MaxSimultaneousTextures in the caps, rather than MaxTextureBlendStages
(which can be much larger) to which it currently corresponds in
wined3d. Using register combiners we can get around that limitation
and get up to GL_MAX_GENERAL_COMBINERS_NV (typically 8) texture
stages. This patch adds code for doing the texture operations with
register combiners instead of ARB_texture_env_combine or
NV_texture_env_combine4, but doesn't make use of that code yet. That's
what the next patch will do.
The code for uploading / binding textures for use with pixel shaders
is slightly different from the one for uploading / binding textures
for use with the fixed function pipeline. It would be possible to keep
the code in a single function with a couple of conditionals, but in
combination with the changes needed for register combiners that would
become quite messy.
GL_LIMITS(textures) is currently used for both the number of texture
stages and the maximum number of simultaneous textures. In the current
code that's the same, but in a later patch that will be separated,
since a texture stage doesn't have to reference an actual
texture. Also, shaders can access a larger number of samplers than the
number of texture units the fixed function pipeline can access.
- Implement D3DSIO_DP2ADD, D3DSIO_TEXKILL, D3DSIO_TEXM3X3PAD
- Partially implement D3DSIO_TEXBEM, D3DSIO_TEXM3X3VSPEC (as much as
they are implemented in ARB_fragment_program at least).
- Stop copying the SHADER_PARSE_STATE struct in each ARB shader
routine - use a pointer instead.
We are only checking against GL_MAX_TEXTURES when binding samplers,
when we should be checking against the maximum number of samplers that
the card supports. Spotted by H. Verbeet.
- NVidia allows "const vec4 = {1.0, 2.0, 3.0, 4.0};", even though
that's not part of the spec.
- It should be "const vec4 = vecr4(1.0, 2.0, 3.0, 4.0);"
- This patch fixes this for D3DSIO_DEF and D3DSIO_DEFI.
- Separate the declaration phase of the shader string generator into
the arb and glsl specific files.
- Add declarations and recognition for application-sent constant
integers and booleans (locally defined ones will follow).
- Standardize capitilization of pixel/vertex specific variable names.
- Moves GLSL constant loading code into glsl_shader.c and out of the
over-populated drawprim.c.
- Creates a new file named arb_program_shader.c which will hold code
specific to ARB_vertex_program & ARB_fragment_program.
- Remove the constant loading calls from drawprim.c
- Implemented: D3DSIO_SGN, LOOP, ENDLOOP, LOGP, LIT, DST, SINCOS
- Process instruction-based modifiers (function existed, it just
wasn't being called)
- Add loop checking to register maps.
- Renamed "sng" to "sgn" for D3DSIO_SGN - it's not handled anywhere
except for GLSL, so won't matter.
There are a total of 17 instructions without a destination token. Of
those 9 have num_params != 0, which means that we will not process any
of them correctly, because we assume the first token (if present) is a
destination token.
Those are basically all the flow control instructions, which we plan to
support very soon. They have source tokens, and no destination. Add a
flag that marks them up to the ins table. Use this flag in the trace
pass, and generation pass.
- track sampler declarations and store the sampler usage in reg_maps structure
- store a fake sampler usage for 1.X shaders (defined as 2D sampler)
- re-sync glsl TEX implementation with the ARB one (no idea why they diverged..)
- use sampler type in new TEX implementation to support 2D, 3D, and Cube sampling
- change drawprim to bind pixel shader samplers
Additional improvements:
- rename texture limit to texcoord to prevent confusion
- add sampler limit, and use that for samplers - *not* the same as texcoord above
SM 3.0 can pack multiple "semantics" into 12 generic input/output registers.
To support that, define temporaries called IN and OUT, and use those as
the output registers. At the end of the vshader, unpack the OUT temps
into the proper GL variables. At the beginning of the pshader, pack the
GL variables back into 12 IN registers.
Various cleanups:
- do not use DWORD as a bitmask, that places artificial limit of 32 on
registers
- track attributes that are used and declare only those
- move declarations function call in pshader/vshader to allow us to
insert pixel or vertex specific code between the declarations and
the rest of the code
- remove redundant 0 intializers
- remove useless continue statement
Now that the declaration function is out of the way, the tracing pass,
which is very long and 100% the same can be shared between pixel and
vertex shaders.
The new function is called shader_trace_init(), and is responsible for:
- tracing the shader
- initializing the function length
- setting the shader version [needed very early]
The new function is called in pass 2 (getister counting/maps), and
it's now in baseshader. It operates on all INPUT and OUTPUT registers,
which, in addition to the old vertex shader input declarations covers
Shader Model 3.0 vshader output and pshader input declarations. The
result is stored into the reg_map structure.
Delete the entire namedArrays code path and all its dependencies (one
of which is quite long - storeOrder in drawprim is always FALSE, for
example). Delete declaredArrays, and make its code path the default.
I missed a register mask in the move to share the shader_hw_def()
function between pixel and vertex shaders for ARB shaders. Fixed
that, and made the GLSL version use the same mask for consistency.
- Based on comments from H. Verbeet
- Changed the distinction from .rgba & .xyzw masks to only use .xyzw
in GLSL shaders. They are interchangeable, and only served to make
the trace look more intuitive, but they don't always apply as-is, so
we'll just leave everything to .xyzw.
- Got rid of the "UseProgramObjectARB(0)" call in drawprim. If there
is no shader set on the next primitive, then that primitive will
call UseProgramObjectARB(0) when it begins to draw.
- Add functions to attach & detach shader objects, create and delete programs, and maintain the list of programs.
- Add a list of GLSL shader programs to the device which is initialized on Init3D(), and deleted on Release().
- Declare more variable names for GLSL programs.
- Some of these won't need to be declared eventually, but it doesn't hurt to do it for now.
- Correct output name for pixel shaders (gl_FragColor instead of glFragColor).
- Add a new file glsl_shader.c which contains almost every GLSL specific function we'll need
- Move print_glsl_info() into glsl_shader.c
- Move the shader_reg_maps struct info into the private header, and make it part of SHADER_OPCODE_ARG.
- Create a new shared ps/vs register map for float constants (future patch will make ARB programs use this, too)
loading float constants for GLSL.
- DrawPrim is just too big of a function. This separates the passing
of constants to the shader into new functions.
- Fixes an off-by-one error when loading vertex declaration constants
(should be <, not <=)
- Adds a function for GLSL loading of constants (aka Uniforms)
- Adds a GLSL program variable to the stateblock and sets it to 0 (a
future patch will actually create this program)
It is wrong to maintain a mapping from a constant index to a type
field, because different constant types do not share an index -
boolean constant 0 is supposed to co-exist with floating point
constant 0, not replace it. Drawprim and other code using the type
array to decide whether to look up a constant in bools, floats, or
ints is wrong - you can't make that decision based on the index.
Implement some basic opengl accelerated blts from and to render
targets. It's not perfect yet, but enought to make some D3D apps
happy. For now the only supported operations are:
- Full screen back -> Front buffer: Just call present
- Offscreen surface -> render target
- Render target -> offscreen surface(slow)
- render target colorfill
Each instruction can have a predication token. Account for it in the
trace pass, register count pass, and store it in the SHADER_OPCODE_ARG
structure for generation. MSDN claims the token is at the end of the
instruction, but that's not true - testing a demo, which lets me
manipulate the shader shows the predication token is the first source
token immediately following the destination token.
As previously mentioned, RASTOUT is invalid on pixel shaders.
On shaders 1.x, r0 is treated as the color output register:
http://www.gamedev.net/columns/hardcore/dxshader3/page2.asp
That's what we currently do in all cases, change it not to do so
for shaders >= 2.0. Support COLOROUT/DEPTHOUT instead.
Currently we hardcode a0.x, which I think is correct for shaders 1.0.
However, for shaders 2.0, we must look into the address token, and
print the register there. Handle both cases to correct the trace.
Change the trace pass, the register counting pass, and the hw
generator pass to take into account the new get_params() function. For
hw generation, store the address tokens into the SHADER_OPCODE_ARG
structure, so they're available to generator functions.
Add a new function to process parameters.
On shaders 1.0, processing parameters amounts to *pToken++.
On shaders 2.0+, we have a relative addressing token to account for.
This function should be used, instead of relying on num_params everywhere.
Share shader_dump_ins_modifiers(), and make vertex shaders use it.
The saturate modifer (_sat) is valid on vs_3_0+, and it isn't being
shown in the trace.
- unify per-version cases into a single sequence
- make use of get_register_name() for the destination register
- ignore destination write mask - it's only allowed on 2.0+ shaders,
and only the default mask is allowed there (xyzw)
- enable swizzle on texld 1.4 (it works together with dw/dz modifiers)
- implement for shaders 2.0-3.0
- add missing GL type and make GL cfg entries for A16B16G16R16
- correct L16 type to UNSIGNED_SHORT (..16 = short, not byte)
- change internal format for X8R8G8B8 from GL_RGB to GL_RGB8 which
seems more correct
- else, endif, break take 0 parameters [not 2]
- if, rep take 1 parameter [not 2]
- breakp takes 1 parameter [not 2]
- move all flow control instructions in one place
- sync software shaders signatures with this
Stefan Dösinger