I realized that it is better to lower casts to int to FLOOR+REINTERPET
instead of appending a FLOOR to all casts to int and assuming that this
is the case for all of them in d3dbc.c.
This in case we introduce new passes in the future that add casts that
we forget to lower, after the lower_casts_to_bool pass.
For temporary registers, SM1-SM3 integer types are internally
represented as floating point, so, in order to perform a cast
from ints to floats we need a mere MOV.
For constant integer registers "iN" there is no operation for casting
from a floating point register to them. For address registers "aN", and
the loop counting register "aL", vertex shaders have the "mova" operation
but we haven't used these registers in any way yet.
We probably would want to introduce these as synthetic variables
allocated in a special register set. In that case we have to remember to
use MOVA instead of MOV in the store operations, but they shouldn't be src
or dst of CAST operations.
Regarding constant integer registers, in some shaders, constants are
expected to be received formatted as an integer, such as:
int m;
float4 main() : sv_target
{
float4 res = {0, 0, 0, 0};
for (int k = 0; k < m; ++k)
res += k;
return res;
}
which compiles as:
// Registers:
//
// Name Reg Size
// ------------ ----- ----
// m i0 1
//
ps_3_0
def c0, 0, 1, 0, 0
mov r0, c0.x
mov r1.x, c0.x
rep i0
add r0, r0, r1.x
add r1.x, r1.x, c0.y
endrep
mov oC0, r0
but this only happens if the integer constant is used directly in an
instruction that needs it, and as I said there is no instruction that
allows converting them to a float representation.
Notice how a more complex shader, that performs operations with this
integer variable "m":
int m;
float4 main() : sv_target
{
float4 res = {0, 0, 0, 0};
for (int k = 0; k < m * m; ++k)
res += k;
return res;
}
gives the following output:
// Registers:
//
// Name Reg Size
// ------------ ----- ----
// m c0 1
//
ps_3_0
def c1, 0, 0, 1, 0
defi i0, 255, 0, 0, 0
mul r0.x, c0.x, c0.x
mov r1, c1.y
mov r0.y, c1.y
rep i0
mov r0.z, r0.x
break_ge r0.y, r0.z
add r1, r0.y, r1
add r0.y, r0.y, c1.z
endrep
mov oC0, r1
Meaning that the uniform "m" is just stored as a floating point in
"c0", the constant integer register "i0" is just set to 255 (hoping
it is a high enough value) using "defi", and the "break_ge"
involving c0 is used to break from the loop.
We could potentially use this approach to implement loops from SM3
without expecting the variables being received as constant integer
registers.
According to the D3D documentation, for SM1-SM3 constant integer
registers are only used by the 'loop' and 'rep' instructions.
These can be disassembled by D3DDisassemble() just fine, and perhaps
more importantly, shader model 1 vertex shaders do not require dcl_
instructions in Direct3D 8.
Currently, HLSL_RESOURCE_SAMPLE_LOD is not implemented for d3dbc,
but we are incorrectly writting a texld instruction to handle it.
This causes SM1 tests with the vulkan backend (in following patches)
to fail if VKD3D_SHADER_CONFIG="force_validation" is enabled.
For now a fixme is emited in these cases.
This was broken by commit e390bc35e2c9b0a2110370f916033eea2366317e; that
commit fixed the source count for these instructions, but didn't adjust
shader_sm1_skip_opcode(). Note that this only affects shader model 1;
later versions have a token count embedded in the initial opcode token.
This should have no effect, since in SM1 the allocation size is the
same as the bind count because there are no texture registers.
It is just done for consistency.
We have to distinguish between the "bind count" and the "allocation size"
of variables.
The "allocation size" affects the starting register id for the resource to
be allocated next, while the "bind count" is determined by the last field
actually used. The former may be larger than the latter.
What we are currently calling hlsl_reg.bind_count is actually the
"allocation size", so a rename is in order.
The real "bind count", which will be introduced in following patches,
is important because it is what should be shown in the RDEF table and
some resource allocation rules depend on it.
For instance, for this shader:
texture2D texs[3];
texture2D tex;
float4 main() : sv_target
{
return texs[0].Load(int3(0, 0, 0)) + tex.Load(int3(0, 0, 0));
}
the variable "texs" has a "bind count" of 1, but an "allocation size" of
3:
// Resource Bindings:
//
// Name Type Format Dim HLSL Bind Count
// ------------------------------ ---------- ------- ----------- -------------- ------
// texs texture float4 2d t0 1
// tex texture float4 2d t3 1
We want to be able to remap input signatures based on the signature index, but
signature normalization both reorders the signature, and requires the old
register index, so add a new field for this.
In SM1 we can expect all variables to always belong to a single regset.
structs in particular, should always be allocated to HLSL_REGSET_NUM,
since they are only allowed if all their components are numeric.
We are not covering the structs case because of the use of
hlsl_type_get_regset(), which is currently not defined for structs.
So the current shader
struct
{
float4 a;
float4 b;
} apple;
float4 main() : sv_target
{
return apple.a + apple.b;
}
fails with
vkd3d/libs/vkd3d-shader/hlsl.c:224: Aborting, reached unreachable code.
The solution is to iterate over all regsets to find the one where the
variable is allocated (if any), and ignore all others.
Variables that contain more than one object (arrays or structs) require
the allocation of contiguous registers in the respective object
register spaces.
This is largely derived from the parser in Wine/wined3d, as of wine-6.18.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
