The `sema.inst_map` datastructure is very often accessed. All
instructions that reference the result of other instructions does a
lookup into this field. Because of this, a significant amount of time,
is spent in `std.HashMap.get`.
This commit replaces the `HashMap` with a simpler data structure that
uses the zir indexes to index into a slice for the result. See the data
structure doc comment for more info.
The TODO comment in safetyPanic mentions introducing the concept of
reference-counted decls. That sounds like Zig current semantics for
normal declarations. By placing the panic messages in builtin there is
no need for another concept in the compiler.
spirv: introduce SpvModule.Fn to generate function code into
spirv: assembler error message setup
spirv: runtime spec info
spirv: inline assembly tokenizer
spirv: inline assembly lhs result/opcode parsing
spirv: forgot to fmt
spirv: tokenize opcodes and assigned result-ids
spirv: operand parsing setup
spirv: assembler string literals
spirv: assembler integer literals
spirv: assembler value enums
spirv: assembler bit masks
spirv: update assembler to new asm air format
spirv: target 1.5 for now
Current vulkan sdk version (1.3.204) ships spirv tools targetting 1.5,
and so these do not work with binaries targetting 1.6 yet. In the
future, this version number should be decided by the target.
spirv: store operands in flat arraylist.
Instead of having dedicated Operand variants for variadic operands,
just flatten them and store them in the normal inst.operands list.
This is a little simpler, but is not easily decodable in the operand
data representation.
spirv: parse variadic assembly operands
spirv: improve assembler result-id tokenization
spirv: begin instruction processing
spirv: only remove decl if it was actually allocated
spirv: work around weird miscompilation
Seems like there are problems with switch in anonymous struct literals.
spirv: begin resolving some types in assembler
spirv: improve instruction processing
spirv: rename some types + process OpTypeInt
spirv: process OpTypeVector
spirv: process OpTypeMatrix and OpTypeSampler
spirv: add opcode class to spec, remove @exclude'd instructions
spirv: process more type instructions
spirv: OpTypeFunction
spirv: OpTypeOpaque
spirv: parse LiteralContextDependentNumber operands
spirv: emit assembly instruction into right section
spirv: parse OpPhi parameters
spirv: inline assembly inputs
spirv: also copy air types
spirv: inline assembly outputs
spirv: spir-v address spaces
spirv: basic vector constants/types and shuffle
spirv: assembler OpTypeImage
spirv: some stuff
spirv: remove spirv address spaces for now
I wasn't able to create a reduced test case for this but the reasoning
can be seen in `abiAlignmentAdvancedUnion` where if `strat` was lazy
`hasRuntimeBitsAdvanced` would be given `null` instead of `sema`
which would cause eager evaluation when it is not valid or desired.
* point to init part of field delc when that's where the error occurs
* update test to reflect fixed error message
* only lookup source location in case of error
These parameters are only ever needed when `std.builtin` is out of sync
with the compiler in which case panicking is the only valid operation
anyways. Removing them causes a domino effect of functions no longer
needing a `src` and/or a `block` parameter resulting in handling
compilation errors where they are actually meaningful becoming simpler.
Previously sema only checked that the private declaration was in the same
file as the lookup but now it also checks that the namespace where
the decl was included from was also in the same file.
Closes#13077
* Add tagName to Value which behaves like @tagName.
* Add hashUncoerced to Value as an alternative to hash when we want to
produce the same hash for value that can coerce to each other.
* Hash owner_decl instead of module_fn in Sema.instantiateGenericCall
since Module.Decl.Index is not affected by ASLR like *Module.Fn was,
and also because GenericCallAdapter.eql was already doing this.
* Use Value.hashUncoerced in Sema.instantiateGenericCall because
GenericCallAdapter.eql uses Value.eqlAdvanced to compare args, which
ignores coersions.
* Add revealed missing cases to Value.eqlAdvanced.
Without these changes, we were breaking the hash contract for
monomorphed_funcs, and were generating different hashes for values that
compared equal. This resulted in a 0.2% chance when compiling
self-hosted of producing a different output, which depended on
fingerprint collisions of hashes that were affected by ASLR. Normally,
the different hashes would have resulted in equal checks being skipped,
but in the case of a fingerprint collision, the truth would be revealed
and the compiler's behavior would diverge.
These functions have a very error-prone API. They are essentially
`all(cmp(op, ...))` but that's not reflected in the name.
This renames these functions to `compareAllAgainstZero...` etc.
for clarity and fixes >20 locations where the predicate was
incorrect.
In the future, the scalar `compare` should probably be split off
from the vector comparison. Rank-polymorphic programming is great,
but a proper implementation in Zig would decouple comparison and
reduction, which then needs a way to fuse ops at comptime.
This is a follow-up to 9dc98fba, which made comptime initialization
patterns for union/struct more robust, especially when storing to
comptime-known pointers (and globals).
Resolves#13063.
