This commit adds forward iterator, and renames functions to
it matches the std::unordered_set/std::set better.
This goes against the SPIR-V coding style, but might be better in
the long run, especially when this set is used along real STL
sets.
(Right now, they are not compatible, and requires 2 syntaxes).
This container could in theory handle bidirectional
iterator, but for now, only forward seemed required for
our use-cases.
Signed-off-by: Nathan Gauër <brioche@google.com>
The current EnumSet implementation is only efficient for enums with
values < than 64. The reason is the first 63 values are stored as a
bitmask in a 64 bit unsigned integer, and the other values are stored
in a std::set.
For small enums, this is fine (most SPIR-V enums have IDs < than 64),
but performance starts to drop with larger enums (Capabilities,
opcodes).
Design considerations:
----------------------
This PR changes the internal behavior of the EnumSet to handle enums
with arbitrary values while staying performant.
The idea is to extend the 64-bits buckets sparsely:
- each bucket can store 64 value, starting from a multiplier of 64.
This could be considered as a hashset with linear probing.
- For small enums, there is a slight memory overhead due to the bucket
storage, but lookup is still constant.
- For linearly distributed values, lookup is constant.
- Worse case for storage are for enums with values which are multiples of 64.
But lookup is constant.
- Worse case for lookup are enums with a lot of small ranges scattered in
the space (requires linear probing).
For enums like capabilities/opcodes, this bucketing is useful as values
are usually scatters in distinct, but almost contiguous blocks.
(vendors usually have allocated ranges, like [5000;5500], while [1000;5000]
is mostly unused).
Benchmarking:
-------------
Benchmarking was done in 2 ways:
- a benchmark built for the occasion, which only measure the EnumSet
performance.
- SPIRV-Tools tests, to measure a more realist scenario.
Running SPIR-V tests with both implementations shows the same
performance (delta < noise). So seems like we have no regressions.
This method is noisy by nature (I/O, etc), but the most representative
of a real-life scenario.
Protocol:
- run spirv-tests with no stdout using perf, multiple times.
Result:
- measure noise is larger than the observed difference.
The custom benchmark was testing EnumSet interfaces using SPIRV enums.
Doing thousand of insertion/deletion/lookup, with 2 kind of scenarios:
- add once, lookup many times.
- add/delete/loopkup many time.
For small enums, results are similar (delta < noise). Seems relevant
with the previously observed results as most SPIRV enums are small, and
SPIRV-Tools is not doing that many intensive operations on EnumSets.
Performance on large enums (opcode/capabilities) shows an improvement:
+-----------------------------+---------+---------+---------+
| Metric | Old | New | Delta % |
+-----------------------------+---------+---------+---------+
| Execution time | 27s | 7s | -72% |
| Instruction count | 174b | 129b | -25% |
| Branch count | 28b | 33b | +17% |
| Branch miss | 490m | 26m | -94% |
| Cache-misses | 149k | 26k | -82% |
+-----------------------------+---------+---------+---------+
Future work
-----------
This was by-design an NFC change to compare apples-to-apples.
The next PR aims to add STL-like iterators to the EnumSet to allow
using it with STL algorithms, and range-based for loops.
Signed-off-by: Nathan Gauër <brioche@google.com>
* Validate layouts for PhysicalStorageBuffer pointers
Fixes#5282
* These pointers may not orginate from a variable so standard layout
validation misses them
* Now checks every instructions that results in a physical storage
buffer pointer
* May not start from a Block-decorated struct so that part is fudged
with a valid layout
* formatting
* SPV_KHR_cooperative_matrix
* Update DEPS with headers
* Update according to review recommendations
* Bugfix and formatting
* Formatting missed or damaged by VS2022
Follow up to #5249
* glslang tests that physical storage buffer pointers can be used as
varyings in shaders
* Allow physical storage buffer pointers in IO interfaces as a 64-bit
type
* Update fuzz tests to not use invalid combinations of LoopMerge + BranchConditional
New IDs were selected to make clear the transformation being done here, instead
of reordering all IDs in between or refactoring the SPIR-V in the test.
* spirv-val: Conditional Branch without an exit is invalid in loop header
From 2.16.2, for CFG:
Selections must be structured. That is, an OpSelectionMerge
instruction is required to precede:
- an OpSwitch instruction
- an OpBranchConditional instruction that has different True Label
and False Label operands where neither are declared merge blocks
or Continue Targets.
Contributes to https://github.com/KhronosGroup/glslang/issues/2439
* When OpTypeStruct is used in Vulkan env and its last member
is a RuntimeArray, check if the struct is decorated with
Block or BufferBlock, as required by VUID-...-04680.
* Validate version 5 of clspv reflection
* Add validation for instructions in versions 2 through 5
* Change the instruction reported for remaining indices on an access
chain to the access chain itself for clarity
* update spirv-headers
* Fix layout validation
Fixes#5010
* Unless scalar block layout is enabled, no member can reside at an
offset between the end of the previous member that is a struct or
array and the next multiple of that alignment
* Remove a dead if that was introduced
Co-authored-by: David Neto <dneto@google.com>
Fixes#4671
Fixes https://crbug.com/oss-fuzz/43265
* Only validate full layout in a vulkan environment
* Universal validation still checks that the right decorations are
present, but the values are only considered for vulkan
* One exception is that invalid overlaps are only checked for vulkan
* This is a pragmatic choice as SPIR-V doesn't define the size of
types so the amount of universal checking would be quite limited
* Removed redundant check for GLSLShared and GLSLPacked decorations
* Should never have been validated as part of universal validation
* make conditionals independent
The always-friendly messages make it harder to debug when the
disassembly is later generated without friendly names.
Additionally, the friendly-name-mapper is slow. Disabling it improves
performance of an ANGLE test that creates numerous shaders by ~5%.
