including SPV generation using SPV_EXT_fragment_invocation_density.
This is an alias of the functionality in SPV_NV_shading_rate, and thus in some
cases we can only have one set of the tokens present (switch statements), so
we have picked the EXT version. This required updating the expected test
results for SPV_NV_shading_rate.
Also updated the known-good for spirv-headers so that the validator in
spirv-tools knows about the new extension.
Includes the following:
Add Vulkan 1.1 capability sets
Don't merge types of resources
Remove stores of undef.
Make sure the constant folder get the correct type.
Includes the following spirv-opt improvements:
Preserve inst-to-block and def-use in passes.
Add store for var initializer in inlining.
Handle types with self references.
Includes:
Fold divide and multiply by same value.
Allow ADCE to remove more instructions.
Remove dead Workgroup variables in ADCE.
Fold fclamp feeding compare.
Add pass to fold a load feeding an extract.
Remove redundant stores.
SROA: Only create symbols that are loaded.
Get ADCE to handle OpCopyMemory
Remove the limit on struct size in SROA.
Notably, this includes the following:
- Fix exponential performance of DeadInsertElim
- Constant folding of OpVectorTimesScalar
- Add Vector DCE
- Add store-store elimination to LocalSingleBlockElim
- Improve folding of OpVectorTimesScalar and OpPhi
- Fixes handling of kill and unreachable ops in inlining
Changes include:
Fix SSA rewrite for nested loops.
Add support for two new extensions:
- SPV_NV_shader_subgroup_partitioned
- SPV_EXT_descriptor_indexing
Legalize OpImageTexelPointer
Copy propagate arrays
Contains the following improvements:
Fixes 1341: Atomics (and other instructions with side-effects) removed.
Add folding for redundant add/sub/mul/div/mix operations
Add constant folding rules for floating-point comparison
Add folding of redundant OpSelect insns
Add folding of OpCompositeExtract and OpConstantComposite
Update SPIRV-Tools. Relevant functional changes:
- Optimizer enhancements:
- ADCE now removes OpSwitch
- Block merging occurs in more cases
- Optimizer fixes:
- Constant propagation (CCP): support matrix constants
- #1199: Optimizer: Fix CCP: don't propagate spec constants.
- #1203: Optimizer: Fix common uniform elim bug introduced by refactoring.
- #1210: Optimizer: Aggressive dead code elimination: Fix 'break' identification.
- #1212: Optimizer: Aggressive dead code elimination: Was skipping too many instructions.
- #1214: Optimizer: Aggressive dead code elimination: Fix infinite loop.
- #1228: Optimizer: Fix CCP: Handling of varying Phi nodes; was resulting in infinite loop.
- #1245: Optimizer: Dead branch elimination: Avoid a null pointer dereference.
- #1250: Optimizer: Dead branch elimination: Avoid spuriously reporting a change.
- #1262: Support building on VisualStudio 2013 again
Update SPIRV-Headers, with "unified1" directory.
Updated one Glslang legalization test base result due to better block merging.
Also added known-good mechanism to fetch latest validated spirv-tools.
Also added -Od and -Os to disable optimizer and optimize for size.
Fetching spirv-tools is optional for both glsl and hlsl. Legalization
of hlsl is done by default if spirv-opt is present at cmake time.
Optimization for glsl is currently done through the option -Os.
Legalization testing is currently only done on four existing shaders.
A separate baseLegalResults directory holds those results. All previous
testing is done with the optimizer disabled.
