Add name annotations to the generated instrumentation code to
make it easier to understand. Example spirv-cross output:
vec4 _140;
if (0u < inst_bindless_direct_read_4(0u, 0u, 1u, uint(_19)))
{
_140 = texture(textures[nonuniformEXT(_19)], inUV);
}
else
{
inst_bindless_stream_write_4(50u, 1u, uint(_19), 0u);
_140 = vec4(0.0);
}
* Improve algorithm to reorder blocks in a function
In dead branch elimination, blocks can end up in a the wrong order, so
there is code to reorder the blocks in structured order. The problem is
that the algorithm to do that is very poor. It involves many searchs in
the function for the correct position to place the block, as well as
moving many block in the vector.
The solution is to write a specialized function in the function class
that will reorder the blocks in structured order. After computing the
structured order, reordering the block can be done in linear time, with
very little overhead.
Using SinglePassRunAndMatch<> instead of SinglePassRunAndCheck<>
makes tests more concise and makes it possible to use pattern
matching features.
Using Effcee stateful pattern matching to make it less repetitive
to check for generated functions and global variables.
This approach isn't worth
it for DebugPrintf functions because the generated code will change
depending on how many parameters are passed to every debugPrintfEXT()
call.
* spirv-opt: fix copy-propagate-arrays index opti on structs.
As per SPIR-V spec:
OpAccessChain indices must be OpConstant when indexing into a structure.
This optimization tried to remove load cascade. But in some scenario
failed:
```c
cbuffer MyStruct {
uint my_field;
};
uint main(uint index) {
const uint my_array[1] = { my_field };
return my_array[index]
}
```
This is valid as the struct is indexed with a constant index, and then
the array is indexed using a dynamic index.
The optimization would consider the local array to be useless and
generated a load directly into the struct.
* spirv-opt: prevent creation of unused instructions
Copy-propagate-arrays optimization pass would create unused constants,
even if the optimization not completed.
This was caused by the way we handled OpAccessChain squashing: we
only referenced constants, and had to create them upfront.
Fixes#4887
Signed-off-by: Nathan Gauër <brioche@google.com>
Specifically, DebugSourceContinued, DebugCompilationUnit, and
DebugEntryPoint. These instructions are top-level instructions
which do not or may not have a user except for the tool and so
should not be eliminated.
When folding a vector shuffle with an undef literal, it is possible that the
literal is adjusted so that it will then be interpreted as an index into
the input operands. This is fixed by special casing that case, and not
adjusting those operands.
Fixes#4859
An access chain instruction interpretes its index operands as signed.
The composite insert and extract instruction interpret their index
operands as unsigned, so it is not possible to represent a negative
number.
This commit adds a check to the local-access-chain-convert pass to check
for a negative number in the access chain and to not do the conversion.
Fixes#4856
This reverts commit d18d0d92e5.
This is reverted because it causes a 7X slowdown when legalizing
SPIR-V with NonSemantic.Shader.DebugInfo.100 instructions.
This is due to the creation of very large UseLists for several
heavily used operands for this extension combined with the fact
that the original commit changed the performance of Uselists to O(n).
Arrays do not have to have a size that is known at compile time. It
could be a spec constant. In these cases, treat the array
as if it is arbitrarily long. This commit will treat it like it is an
array of size UINT32_MAX.
Fixes https://crbug.com/oss-fuzz/47397.
If the `instruction` operand in an extended instruction instruction is
too large, it causes undefined behaviour when that value is cast to the
enum for the corresponding set. This is done with the
NonSemanticDebug100 instruction set. We need to avoid the undefined
behaviour.
Fixes#4727
Which functions are processed is determined by which ones are on the
call tree from the entry points before dead code is removed. So it is
possible that a function is process because it is called from an entry
point, but the CFG is not cleaned up because the call to the function
was removed.
The fix is to process and cleanup every function in the module. Since
all of the dead functions would have already been removed in an earlier
step of DCE, it should not make a different in compile time.
Fixes#4731
* Structural dominance introduced in SPIR-V 1.6 rev2
* Changes the structured cfg validation to use structural dominance
* structural dominance is based on a cfg where merge and continue
declarations are counted as graph edges
* Basic blocks now track structural predecessors and structural
successors
* Add validation for entry into a loop
* Fixed an issue with inlining a single block loop
* The continue target needs to be moved to the latch block
* Simplify the calculation of structured exits
* no longer requires block depth
* Update many invalid tests
We currently build the structured order for all nodes reachable from the
loop header when unrolling a loop. However, unrolling only needs the
nodes in the loop and possibly the merge node.
To avoid needless computation, I have implemented a search that will
stop at the merge node.
Fixes#4827
The code in `CFG::SplitLoopHeader` assumes the loop header is not the
latch. This leads to it not being able to find the latch block. This
has been fixed, and a test added.
Fixes#4527
If the predecessor blocks are the same, then there is only 1 value for the
OpPhi. The simplition pass will simplify it, and it causes problems for
if-conversion. In these cases, if-conversion can just punt.
Fixes#3554.
An access chain could have a constant index that is an out of bounds
access. This is valid spir-v, even if it can cause problems at runtime.
However, it is not valid to have an OpCompositeExtract with an out of
bounds access. This means we have to stop local-access-chain-convert
from making that change.
Fixes#4605
This change adds a folding rule which transforms x * y - a and a - x * y
into FMA(x, y, -a) and FMA(-x, y, a), respectively.
While the SPIR-V instruction count remains the same, target instruction
sets typically feature FMA instruction variants that can negate an
operand. Also this transformation may unlock further optimizations which
eliminate the negation.
(Google bug: b/226145988)
* Add move folding for composite instructions
Fold chains of insert into construct
If a chain of OpCompositeInsert instruction write to every element of a
composite object, then we can replace it with an OpCompositeConstruct.
Fold a construct fed by extracts to a single extract
We already fold an OpCompositeConstruct when it is simlpy reconstructing
an object that was decomposed by a series of OpCompositeExtract
instructions. However, we do not do that if that object is an element
of a larger object.
I have updated the rule, so that if the original object is a an element
of a larger object, then the OpCompositeConstruct is replaced with a
single OpCompositeExtract from the larger object.
Fixes#4371.
* Handle 64-bit integers in local access chain convert
The local access chain convert pass does on run on module that have
64-bit integers, even if they have nothing to to with access chains.
This is very limiting because other passes rely on the access chains
being removed. So this commit will add this functionality to the pass.
spirv validation require OpFunctionCall with memory object, usually this
is non issue as all the functions are inlined.
This pass deal with some case for
DontInline function. accesschain input operand would be replaced new
created variable
This reverts commit 671f6e633f.
PR #4783 was reverted because it caused OpenCL CTS failures for clvk.
The was in clspv, which was not adding the no contract decoration when
it was required. This has been fixed in
https://github.com/google/clspv/pull/845. We can now reapply #4783.
Adding Fma instruction can speed up the code. This was requested by
swiftshader, so they do not have to do this analysis themselves. It can
also help reduce the code size, and the work the ICD compilers have to
do.
spread-volatile-semantics pass spreads Volatile semantics for builtin
variables used by certain execution models based on
VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V"). Therefore, shaders without execution model (e.g., used only
for linkage) are not the target of the pass. This commit lets the pass
just return SuccessWithoutChange in that case.
If the body of the module does not have any ids change, compact ids will
not change the id bound. This can cause problems because the id bound
could be much higher than the largest id in that is used. It should be
reset any time it is not the larger id used + 1.
Fixes#4604
When folding a vector shuffle feeding a vector shuffle, we do not
propagate an 0xFFFFFFFF, which has a special meaning, correctly. We
adjust the value making it lose it meaning as an undefined value.
Fixes#4581
CCP does not want to fold an instruction unless it folds to a constant.
There is an asser to check for this. The question if a spec constant
counts as a constant. The constant folder considers a spec constant a
constand, but CCP does not. I've fixed the assert in CCP to match what
the folder does. It should not require any new changes to CCP.
Swift shader needs a way to inline all functions, even those marked as
DontInline. See https://github.com/KhronosGroup/SPIRV-Tools/pull/4471.
This implements the suggestion I made in the PR. We add a pass that
will remove the DontInline function control, so that the inlining passes
will inline them.
SwiftShader will still have to modify their code to add this pass before
the other passes are run.
The function `BuildInvalideAnalyses` will be rebuilt for every analysis that
has been requested, but it is not necessary. It also can cause problems
because if the CFG needs to be rebuilt, so do the dominator trees.
This change will make the functionality match the description of the
function.
* Optimize DefUseManager allocations
Saves around 30-35% of compilation time.
For inst->use_ids, use a pool linked list instead of allocating vectors for every instruction. For inst->uses, use a "PooledLinkedList"' -- a linked list that has shared storage for all nodes. Neither re-use nodes, instead we do a bulk compaction operation when too much memory is being wasted (tuneable).
Includes separate PooledLinkedList templated datastructure, a very special case construct, but split out to make the code a little easier to understand.
Scalar replacement generates a null when there value for a member will
not be used. The null is used to make sure things are
deterministic in case there is an error.
However, some type cannot be null, so we will change that to use undef.
To keep the code simpler we will always use the undef.
Fixes#3996
The handling of the RayQueryKHR type is not complete in the type
manager. The tests were not picking this up. I've added a test to make
sure that the `GenerateAllTypes` function actually does generate all of
the types. Once it is added there other tests should pick up on the
other parts that were missing.
Add a pass to spread Volatile semantics to variables with SMIDNV,
WarpIDNV, SubgroupSize, SubgroupLocalInvocationId, SubgroupEqMask,
SubgroupGeMask, SubgroupGtMask, SubgroupLeMask, or SubgroupLtMask BuiltIn
decorations or OpLoad for them when the shader model is the ray
generation, closest hit, miss, intersection, or callable shaders. This
pass can be used for VUID-StandaloneSpirv-VulkanMemoryModel-04678 and
VUID-StandaloneSpirv-VulkanMemoryModel-04679 (See "Standalone SPIR-V
Validation" section of Vulkan spec "Appendix A: Vulkan Environment for
SPIR-V").
Handle variables used by multiple entry points:
1. Update error check to make it working regardless of the order of
entry points.
2. For a variable, if it is used by two entry points E1 and E2 and
it needs the Volatile semantics for E1 while it does not for E2
- If VulkanMemoryModel capability is enabled, which means we have to
set memory operation of load instructions for the variable, we
update load instructions in E1, but do not update the ones in E2.
- If VulkanMemoryModel capability is disabled, which means we have
to add Volatile decoration for the variable, we report an error
because E1 needs to add Volatile decoration for the variable while
E2 does not.
For the simplicity of the implementation, we assume that all functions
other than entry point functions are inlined.
