Conor McCarthy
cfcd57209b
vkd3d-shader/ir: Convert SSAs to temps only if the block of origin does not dominate all uses.
2024-04-09 12:27:19 -05:00
Conor McCarthy
7bfc7410f2
vkd3d-shader/ir: Materialise phis to temps in the incoming blocks.
...
RADV converts temps to phi instructions, so converting phis to MOVC in
VSIR just translates back to phis feeding into a MOVC. This commit
eliminates the MOVC.
2024-04-09 12:27:15 -05:00
Francisco Casas
2621fc9572
vkd3d-shader/ir: Add missing src swizzle in vsir_program_lower_texkills().
2024-04-09 12:27:06 -05:00
Giovanni Mascellani
a278f32d3f
vkd3d-shader/ir: Synthesize selection constructs from conditional jumps.
2024-04-04 22:19:19 +02:00
Giovanni Mascellani
0d13b6afe9
vkd3d-shader/ir: Remove trailing `continue's.
2024-04-04 22:19:18 +02:00
Giovanni Mascellani
d8414302e8
vkd3d-shader/ir: Move `continue's to the false branch when possible.
2024-04-04 22:19:16 +02:00
Giovanni Mascellani
e6c20361ea
vkd3d-shader/ir: Drop a useless helper.
2024-04-03 22:23:08 +02:00
Giovanni Mascellani
9aff65c6d2
vkd3d-shader/ir: Open a loop when visiting the loop header.
...
Not when making the loop header available, which is too early and
might lead to open too many loops at once.
2024-04-03 22:23:06 +02:00
Giovanni Mascellani
d10031911d
vkd3d-shader/ir: Move vsir_program_normalise() to the end of the file.
...
vsir_program_normalise() has function calls to basically everything in
ir.c, so it's useful to have it in an easily reachable place to
quickly jump to wherever you need using your favorite code editor's
features.
2024-04-02 23:10:07 +02:00
Henri Verbeet
5de5f241a6
vkd3d-shader/ir: Pass a struct vsir_program to vkd3d_shader_normalise().
2024-03-19 22:57:56 +01:00
Giovanni Mascellani
e5f486edd3
vkd3d-shader/ir: Use type "unused" for labels.
...
The label itself is certainly an unsigned integer, but the register
has no meaningful data type. It cannot be evaluated to anything.
The goal of this is to reduce cluttering in the internal ASM dumps.
2024-03-18 23:07:45 +01:00
Giovanni Mascellani
8c15377ffc
vkd3d-shader/ir: Emit multilevel jumps in the structured program.
...
The new structurizer therefore reaches feature parity with the
older simple one, except for a couple of points:
* the old structurizer accepts any CFG, without requiring reducibility;
however, the DXIL specification requires the CFG to be reducible
anyway, so we're not really losing anything;
* the new structurizer additionally requires that no block has two
incoming back arrows; AFAIK this is condition that can happen,
but in practice it seems to be rare; also, it's not hard to add
support for it, as soon as it is decided it is useful.
On the other hand, the new structurizer makes use of the merging
information that are reconstructed from the CFG, which is important
for downstream optimization and fundamental for correctly emitting
tangled instructions.
Taking these considerations into account, the old structurizer is
considered superseded and is therefore removed.
2024-03-18 23:07:17 +01:00
Giovanni Mascellani
c2aa01de6e
vkd3d-shader/ir: Emit the reconstructed structured program.
...
Multilevel jumps are not supported yet, and trigger a fallback to the
simple structurizer.
2024-03-18 23:07:15 +01:00
Henri Verbeet
abf76372e0
vkd3d-shader/ir: Pass a struct vsir_program to vsir_validate().
2024-03-14 23:24:29 +01:00
Henri Verbeet
7af3e9300f
vkd3d-shader/ir: Pass a struct vsir_program to flatten_control_flow_constructs().
2024-03-14 23:24:28 +01:00
Henri Verbeet
268989f5b9
vkd3d-shader/ir: Pass a struct vsir_program to normalise_combined_samplers().
2024-03-14 23:24:27 +01:00
Henri Verbeet
674bebe9c9
vkd3d-shader/ir: Pass a struct vsir_program to shader_normalise_io_registers().
2024-03-14 23:24:24 +01:00
Giovanni Mascellani
d75dc76011
vkd3d-shader/ir: Dump the reconstructed structured program.
2024-03-14 22:48:53 +01:00
Giovanni Mascellani
2b3e21caeb
vkd3d-shader/ir: Add jumps to the structured programs.
2024-03-14 22:48:52 +01:00
Giovanni Mascellani
c25e4c47c3
vkd3d-shader/ir: Reconstruct a structured program.
...
For simplicity jumps are not currently processed.
2024-03-14 22:48:50 +01:00
Henri Verbeet
5f1f7ababb
vkd3d-shader/ir: Pass a struct vsir_program to remap_output_signature().
2024-03-13 22:00:42 +01:00
Henri Verbeet
0edf6d25f3
vkd3d-shader/ir: Pass a struct vsir_program to simple_structurizer_run().
2024-03-13 22:00:41 +01:00
Henri Verbeet
c3f3cb5ef9
vkd3d-shader/ir: Pass a struct vsir_program to materialize_ssas_to_temps().
2024-03-13 22:00:40 +01:00
Henri Verbeet
76791913d0
vkd3d-shader/ir: Pass a struct vsir_program to materialize_ssas_to_temps_process_reg().
2024-03-13 22:00:39 +01:00
Henri Verbeet
343a365c97
vkd3d-shader/ir: Pass a struct vsir_program to instruction_array_lower_texkills().
2024-03-13 22:00:37 +01:00
Giovanni Mascellani
f2b2040ff6
vkd3d-shader/ir: Sort loop intervals.
2024-03-12 22:14:47 +01:00
Giovanni Mascellani
070c4af8b3
vkd3d-shader/ir: Generate synthetic intervals for forward edges.
2024-03-12 22:14:46 +01:00
Giovanni Mascellani
1d60612340
vkd3d-shader/ir: Compute loop as intervals of the block order.
2024-03-12 22:14:45 +01:00
Giovanni Mascellani
204c8afb9c
vkd3d-shader/dxbc: Remove flag is_dxil.
...
The full shader profile is already available in vsir_program.
2024-03-11 22:09:31 +01:00
Giovanni Mascellani
470d83a9da
vkd3d-shader: Move shader signatures to vsir_program.
2024-03-11 22:09:31 +01:00
Giovanni Mascellani
b940d79aa8
vkd3d-shader/ir: Keep loops contiguous in the topological sort.
2024-03-08 23:35:48 +01:00
Giovanni Mascellani
cf9a023566
vkd3d-shader/ir: Dump the topological order of the control flow graph.
2024-03-08 23:35:47 +01:00
Giovanni Mascellani
ed40646c8a
vkd3d-shader/ir: Topologically sort the control flow graph.
2024-03-08 23:35:46 +01:00
Giovanni Mascellani
880d781b57
vkd3d-shader/ir: Allow adding to a block list without checking for duplicates.
2024-03-08 23:35:45 +01:00
Giovanni Mascellani
8723b5d266
vkd3d-shader/ir: Sort each loop by block label.
2024-03-07 23:08:36 +01:00
Giovanni Mascellani
3d4092d9dc
vkd3d-shader/ir: Dump the loops in the control flow graph.
2024-03-07 23:08:35 +01:00
Giovanni Mascellani
8241ca2b67
vkd3d-shader/ir: Keep track of loops by header block.
2024-03-07 23:08:34 +01:00
Giovanni Mascellani
088a6c49a2
vkd3d-shader/ir: Keep a reference to the message context inside struct vsir_cfg.
2024-03-07 23:08:33 +01:00
Giovanni Mascellani
9ac842b36b
vkd3d-shader/ir: Compute the loops in the control flow graph.
2024-03-07 23:08:32 +01:00
Giovanni Mascellani
de15f55477
vkd3d-shader/ir: Introduce a helper to express block domination.
2024-03-07 23:08:31 +01:00
Giovanni Mascellani
c3657c4799
vkd3d-shader/ir: Add a debug buffer to struct vsir_cfg.
2024-03-07 23:08:30 +01:00
Giovanni Mascellani
c5893288d9
vkd3d-shader/ir: Properly handle function-local indexable temps when flattening control flow.
...
They have to be considered code rather than declarations, as required
for instance by the SPIR-V backend.
2024-03-07 23:08:28 +01:00
Giovanni Mascellani
9c678532a7
vkd3d-shader/ir: Dump the domination relationship.
2024-02-22 22:45:16 +01:00
Giovanni Mascellani
04c2852c1a
vkd3d-shader/ir: Compute the domination relationship.
2024-02-22 22:45:15 +01:00
Giovanni Mascellani
2573d86ccf
vkd3d-shader/ir: Dump the control flow graph in the GraphViz format.
2024-02-22 22:45:14 +01:00
Giovanni Mascellani
da037b61ba
vkd3d-shader/ir: Build a representation of the control flow graph.
2024-02-22 22:45:13 +01:00
Giovanni Mascellani
51f13391e6
vkd3d-shader/ir: Introduce a simple control flow graph structurizer.
...
The structurizer is implemented along the lines of what is usually called
the "structured program theorem": the control flow is completely
virtualized by mean of an additional TEMP register which stores the
block index which is currently running. The whole program is then
converted to a huge switch construction enclosed in a loop, executing
at each iteration the appropriate block and updating the register
depending on block jump instruction.
The algorithm's generality is also its major weakness: it accepts any
input program, even if its CFG is not reducible, but the output
program lacks any useful convergence information. It satisfies the
letter of the SPIR-V requirements, but it is expected that it will
be very inefficient to run on a GPU (unless a downstream compiler is
able to devirtualize the control flow and do a proper convergence
analysis pass). The algorithm is however very simple, and good enough
to at least pass tests, enabling further development. A better
alternative is expected to be upstreamed incrementally.
Side note: the structured program theorem is often called the
Böhm-Jacopini theorem; Böhm and Jacopini did indeed prove a variation
of it, but their algorithm is different from what is commontly attributed
to them and implemented here, so I opted for not using their name.
2024-02-06 23:07:07 +01:00
Giovanni Mascellani
19aef21369
vkd3d-shader/ir: Handle PHI nodes when materializing SSA registers.
...
PHI nodes cannot be used with TEMP registers, so they have to be
converted to MOV/MOVC nodes and moved before the BRANCH node.
2024-02-06 23:07:03 +01:00
Giovanni Mascellani
e0d3e9c376
vkd3d-shader/ir: Materialize SSA registers to temporaries.
...
For simplicity PHI nodes are not currently handled.
The goal for this pass is to make the CFG structurizer simpler, because
it doesn't have to care about the more rigid rules SSA registers have
to satisfy than TEMP registers.
It is likely that the generated code will be harder for downstream
compilers to optimize and execute efficiently, so once a complete
structurizer is in place this pass should be removed, or at least
greatly reduced in scope.
2024-02-06 23:07:00 +01:00
Henri Verbeet
d38221bd68
vkd3d-shader/ir: Pass a struct vsir_program to shader_parser_get_src_params().
2024-02-01 00:08:23 +01:00
Henri Verbeet
47ec24ebad
vkd3d-shader/ir: Pass a struct vsir_program to shader_parser_get_dst_params().
2024-02-01 00:08:22 +01:00
Giovanni Mascellani
edc406d297
vkd3d-shader/ir: Check that SSA registers are used with compatible data types.
...
Specifically, accesses are always 32 bit or always 64 bit.
2024-01-29 22:33:34 +01:00
Giovanni Mascellani
026c502f31
vkd3d-shader/ir: Fixup PHI nodes when lowering switches to selection ladders.
...
A map between the blocks before and after the pass is built and then
used to fix the PHI nodes.
2024-01-29 22:33:17 +01:00
Giovanni Mascellani
378109051c
vkd3d-shader/ir: Lower monolithic switches to selection ladders.
...
PHI nodes must be fixed up after this pass, because the block references
might have become broken. For simplicitly this is not handled yet.
The goal for this pass is to make the CFG structurizer simpler, because
only conditional and unconditional branches must be supported.
Eventually this limitation might be lifted if there is advantage in
doing so.
2024-01-29 22:33:15 +01:00
Francisco Casas
cec45da45b
vkd3d-shader/ir: Update flatten_control_flow_constructs() comment.
2024-01-25 22:24:27 +01:00
Henri Verbeet
eabcaf15fb
vkd3d-shader/ir: Use vsir_instruction_is_dcl() in vsir_validate_instruction().
2024-01-24 22:38:36 +01:00
Henri Verbeet
365ad56888
vkd3d-shader/ir: Get rid of shader_instruction_is_dcl().
2024-01-24 22:38:34 +01:00
Francisco Casas
b92f6c448a
vkd3d-shader/ir: Lower texkill instructions to discard_nz.
2024-01-24 22:37:41 +01:00
Henri Verbeet
bf628f0c74
vkd3d-shader/ir: Store block names in struct vsir_program.
2024-01-23 20:27:35 +01:00
Henri Verbeet
f3c7d2d05c
vkd3d-shader/ir: Store the block count in struct vsir_program.
2024-01-23 20:27:34 +01:00
Giovanni Mascellani
de9725b6ba
vkd3d-shader/ir: Remove DCL_TEMPS instructions.
...
We have to do work to keep it updated across passes and we never read it.
2024-01-23 20:26:38 +01:00
Giovanni Mascellani
4b869f73bb
vkd3d-shader/ir: Run tracing and validation for DXIL code too.
2024-01-23 20:26:36 +01:00
Conor McCarthy
6df725718a
vkd3d-shader/ir: Refactor return code checking in vkd3d_shader_normalise().
...
We started with only one or two of these but it has become excessive.
Patch originally written by Conor McCarthy and updated by Giovanni Mascellani.
2024-01-23 20:26:34 +01:00
Giovanni Mascellani
232b2ad360
vkd3d-shader/ir: Validate PHI instructions.
2024-01-23 20:26:08 +01:00
Giovanni Mascellani
56f9057985
vkd3d-shader/ir: Do not allow IMMCONST and IMMCONST64 as destination registers.
2024-01-23 20:26:06 +01:00
Giovanni Mascellani
21633fcc52
vkd3d-shader/ir: Refactor register-type-specific code in parameter validation.
...
To better accommodate code for other register types.
2024-01-23 20:26:05 +01:00
Giovanni Mascellani
fb6409bda1
vkd3d-shader/ir: Check that all instructions appear in a block.
2024-01-23 20:26:03 +01:00
Henri Verbeet
adc02eada8
vkd3d-shader/ir: Store the temporary register count in struct vsir_program.
2024-01-22 22:18:53 +01:00
Henri Verbeet
94ca46916a
vkd3d-shader/ir: Store the SSA register count in struct vsir_program.
2024-01-22 22:18:52 +01:00
Henri Verbeet
7b85cd6a31
vkd3d-shader/ir: Store the "use_vocp" field in struct vsir_program.
2024-01-22 22:18:51 +01:00
Henri Verbeet
23dcd4f22b
vkd3d-shader/ir: Store the shader version in struct vsir_program.
2024-01-22 22:18:50 +01:00
Henri Verbeet
fc9043be3c
vkd3d-shader/ir: Introduce struct vsir_program.
2024-01-22 22:18:48 +01:00
Giovanni Mascellani
67902d4304
vkd3d-shader/ir: Validate RET instructions.
2024-01-18 23:15:58 +01:00
Giovanni Mascellani
af96730276
vkd3d-shader/ir: Validate SWITCH_MONOLITHIC instructions.
2024-01-18 23:15:57 +01:00
Giovanni Mascellani
46ebb404d7
vkd3d-shader/ir: Validate BRANCH instructions.
2024-01-18 23:15:56 +01:00
Giovanni Mascellani
bc9db34cdc
vkd3d-shader/ir: Validate LABEL instructions.
2024-01-18 23:15:55 +01:00
Giovanni Mascellani
dd1b2feec0
vkd3d-shader/ir: Validate LABEL registers.
2024-01-18 23:15:54 +01:00
Giovanni Mascellani
17f4afc2b5
vkd3d-shader/ir: Validate that structured CF does not appear in block-based shaders.
2024-01-18 23:15:53 +01:00
Conor McCarthy
83851133c5
vkd3d-shader/ir: Set merged signature element interpolation mode only from used elements.
2024-01-18 23:15:27 +01:00
Conor McCarthy
b8280e4210
vkd3d-shader/ir: Revert "Do not merge signature elements which have different interpolation modes.".
...
This reverts commit b5c067b41a
.
The commit causes regressions in other shaders because unused elements
do not have an interpolation mode.
2024-01-18 23:15:25 +01:00
Conor McCarthy
efe800f7f0
vkd3d-shader/dxil: Handle the DXIL SWITCH instruction.
2024-01-18 23:15:14 +01:00
Conor McCarthy
4c30b23821
vkd3d-shader: Make the control point count the outer dimension of I/O arrays.
...
The relative-addressed case in shader_register_normalise_arrayed_addressing()
leaves the control point id in idx[0], while for constant register
indices it is placed in idx[1]. The latter case could be fixed instead,
but placing the control point count in the outer dimension is more
logical.
2024-01-17 22:28:59 +01:00
Conor McCarthy
418e8d8100
vkd3d-shader/ir: Expand index range write masks to cover all element masks.
...
For example, this occurred in a shader:
reg_idx write_mask
0 xyz
1 xyzw
2 xyzw
3 xyz
The dcl_indexrange instruction covered only xyz, so once merged, searching for
xyzw failed.
It is impossible to declare an input array where elements have different
component counts, but the optimiser can create this case. One way for
this to occur is to dynamically index input values via a local array
containing copies of the input values. The optimiser converts this to
dynamically indexed inputs.
2024-01-17 22:28:56 +01:00
Conor McCarthy
559d9d4ee0
vkd3d-shader/ir: Include an initial label instruction in the first control flow block.
2024-01-17 22:28:41 +01:00
Conor McCarthy
37d9dba512
vkd3d-shader/ir: Store code block names in struct vkd3d_shader_desc.
2024-01-17 22:28:38 +01:00
Conor McCarthy
ffc65215ba
vkd3d-shader/ir: Flatten SWITCH/CASE/DEFAULT/ENDSWITCH control flow instructions.
2024-01-17 22:28:36 +01:00
Conor McCarthy
dcb8527327
vkd3d-shader/ir: Flatten LOOP/BREAK/CONTINUE/ENDLOOP control flow instructions.
2024-01-17 22:28:35 +01:00
Conor McCarthy
e1dddc01b7
vkd3d-shader/ir: Flatten IF/ELSE/ENDIF control flow instructions.
2024-01-17 22:28:34 +01:00
Giovanni Mascellani
b0c8a47f9d
vkd3d-shader/ir: Check that SSA registers are used validly.
...
Specifically, they are assigned only once and only assigned components
are used.
Right now we don't check that the assignment dominates all usages.
2024-01-15 19:57:01 +01:00
Giovanni Mascellani
26bebe8f91
vkd3d-shader/ir: Check that SSA registers have consistent dimensions.
2024-01-15 19:57:00 +01:00
Giovanni Mascellani
7503429555
vkd3d-shader/ir: Check that TEMP registers have consistent dimensions.
2024-01-15 19:56:59 +01:00
Giovanni Mascellani
fbd77486de
vkd3d-shader/ir: Use vkd3d_free() instead of free().
2024-01-15 19:56:58 +01:00
Giovanni Mascellani
26c6a87268
vkd3d-shader/ir: Simplify control flow in vsir_validate_register().
2024-01-15 19:56:56 +01:00
Giovanni Mascellani
78d95889bd
vkd3d-shader: Make *src and *dst mutable in vkd3d_shader_instruction.
...
They were originally made const because no optimization/normalization
pass existed. Now having to cast away const all the time is becoming
more and more burdening.
2024-01-11 23:05:11 +01:00
Giovanni Mascellani
bd9118cac4
vkd3d-shader/ir: Validate destination write masks depending on the dimension.
2024-01-04 22:23:25 +01:00
Giovanni Mascellani
15e7711786
vkd3d-shader/ir: Validate source swizzles depending on the dimension.
2024-01-04 22:23:24 +01:00
Giovanni Mascellani
e91da41dea
vkd3d-shader/ir: Validate SSA registers.
2024-01-04 22:23:23 +01:00
Zebediah Figura
1ce7e3d8b1
vkd3d-shader/d3dbc: Assign unique register indices for VKD3DSPR_RASTOUT.
2024-01-04 22:23:09 +01:00
Zebediah Figura
3f52fda8ad
vkd3d-shader: Lower shader model 1/2 inter-stage I/O to a flat array.
...
An alternative is that we stash the reg_type in the signature, but this seems
far simpler for the backend to deal with.
2024-01-04 22:23:07 +01:00
Henri Verbeet
e33d3b3954
vkd3d-shader/ir: Rename the "immconst_uint" field of struct vkd3d_shader_register to "immconst_u32".
...
For consistency with the rest of vkd3d-shader; e.g. put_u32() and
read_u32().
2024-01-03 22:37:39 +01:00