* compiler, runtime, reflect: generate type-specific hash/equal for composite map keys
For map keys that are not trivially binary-comparable, the compiler now
generates type-specific hash and equal functions as LLVM IR instead of
going through the interface+reflection path. This covers comparable
types: strings, floats, complex numbers, interfaces, channels, and
composites containing any mix of these.
Previously, maps with composite keys containing strings or floats
converted the key to interface{}, hashed via reflection, and compared
through interface equality. Now the compiler walks struct fields and
array elements directly, dispatching to the right runtime helper for
each field type and storing keys at their actual type.
Struct keys are always handled field-by-field so padding bytes do not
affect equality or hashing. Blank fields are ignored, matching Go
equality. Generated hash/equal function names use canonical underlying
type structure so structurally identical key types can share generated
functions. Padding zeroing before map operations is no longer needed
because structs no longer use the binary key path.
Also fix reflect map iteration for interface-keyed maps: MapIter.Key
returns an interface Value for map[interface{}] keys instead of
unpacking to the concrete key kind.
* compiler: generate loops for array map key hash/equal
Previously, array key hash and equal functions were unrolled at compile
time, generating one block of IR per element. For large arrays like
[1000]int inside a struct with non-binary fields, this caused code
explosion.
Now, binary-element arrays dispatch directly to hash32/memequal for the
whole array. Non-binary-element arrays generate an LLVM IR loop. The
equal loop short-circuits on the first mismatch.
Small arrays are still unrolled instead of looping, keeping the simple
cases compact.
* reflect: fix at-runtime map issues from review, and more found locally
Maps created through reflect.MakeMap need hash/equal behavior that
matches compiler-created maps. Add hashmapMakeReflect for composite key
types, using runtime closures that reconstruct interface{} values from
raw key bytes and delegate to the interface hash and equality paths.
Interface-keyed maps are already stored as interface values, so use the
existing interface hash/equal helpers directly for those. This keeps
reflect insert, lookup, delete, and compiled lookup paths consistent.
Also fix addressable small values used as interface map keys or
interface map values. loadSmallValue puts small indirect values back in
the pointer-sized interface data field the same way valueInterfaceUnsafe
does.
* compiler, interp, reflect: fix pointer map literals; remove interface fallback
Package-level map literals with pointer keys (both *T and
unsafe.Pointer) crash the compiler: the interp pass panics when trying
to hash pointer data as raw bytes, because pointer values in the interp
memory model are symbolic identities that do not fit in a byte.
Fix this by setting a recoverable error flag instead of panicking. The
interp detects the error after each instruction and defers the map
insert to runtime init code, where real addresses are available for
hashing. This matches how the interp already handles other operations
it cannot evaluate at compile time.
With this fix, unsafe.Pointer can also be classified as a binary map
key, which was the last type requiring the interface-based fallback.
Since all comparable types now use either the binary or the
compiler-generated hash/equal path, remove the interface fallback from
the compiler and reflect packages.
* compiler, transform: always pass hash/equal function pointers to hashmapMakeGeneric
The compiler now always resolves the hash and equal functions at compile
time and passes them directly to hashmapMakeGeneric, instead of passing
an algorithm enum to hashmapMake and resolving at runtime. For string
keys, the runtime hashmapStringPtrHash/hashmapStringEqual functions are
referenced directly. For binary keys, hash32/memequal are referenced.
The old hashmapMake with alg enum is retained for reflect, which still
needs runtime resolution when creating maps dynamically.
The OptimizeMaps transform pass is updated to handle both hashmapMake
and hashmapMakeGeneric, and to recognize hashmapGenericSet in addition
to hashmapBinarySet and hashmapStringSet. The now-unused
hashmapCanGenerateHashEqual helper is removed.
* runtime: store large map keys and values indirectly
When a map key or value exceeds 128 bytes, the bucket now stores a
pointer to separately allocated memory instead of the data inline. This
matches Go's MapMaxKeyBytes/MapMaxElemBytes threshold and prevents
bucket sizes from exploding for large key/value types.
For example, map[[256]byte]int previously used 2128 bytes per bucket
(16 header + 256*8 keys + 8*8 values); now it uses 144 bytes per bucket
(16 header + 8*8 pointers + 8*8 values).
The indirection is fully encapsulated in the runtime via helper
functions. Store the computed key and value slot sizes on the hashmap so
all runtime and reflect paths use the same bucket layout, including
non-indirect keys and values.
Add big-key golden coverage and benchmarks. Make the benchmark vary
enough key bytes to exercise hashing.
When a string variable already had a source-level default value, the -ldflags
"-X" flag was silently ignored and the variable remained the default value
at runtime.
Fix by stripping the InitOrder entry for each -X variable before
LoadSSA() is called. With no entry in InitOrder, go/ssa emits no init
store, so the global stays zero-valued in the IR. makeGlobalsModule
still injects the actual -X values at final link time.
The -X values remain out of the per-package build cache with only the
variable names appearing in the cache key.
Signed-off-by: deadprogram <ron@hybridgroup.com>
This adds support for `-gc=boehm` on `-target=wasip1` and `-target=wasm`
(in a browser or NodeJS). Notably it does *not* add Boehm GC support for
`-target=wasip2`, since that target doesn't have a real libc.
older behavior for wasi modules to not return an exit code as if they were reactors.
See #4726 for some details on what this is intended to address.
Signed-off-by: deadprogram <ron@hybridgroup.com>
These should make sure basic functionality is still working.
Using the `-short` flag to avoid taking too long to run all tests (and
to install all the necessary emulators), and because some targets might
not work in older Go/LLVM versions (such as WASI).
This does _not_ run tests and checks against expected IR, because LLVM
IR changes a lot across versions.
While there are some browser tests, Node.js is just a lot better for
testing this kind of stuff because it's much faster and we don't need a
browser for this.
This adds support for //go:wasmexport with `-target=wasm` (in the
browser). This follows the //go:wasmexport proposal, meaning that
blocking functions are not allowed.
Both `-buildmode=default` and `-buildmode=c-shared` are supported. The
latter allows calling exported functions after `go.run()` has returned.
This adds support for the `//go:wasmexport` pragma as proposed here:
https://github.com/golang/go/issues/65199
It is currently implemented only for wasip1 and wasm-unknown, but it is
certainly possible to extend it to other targets like GOOS=js and
wasip2.
The reflect package needs to know the endianness of the system in a few
places. Before this patch, it assumed little-endian systems. But with
GOARCH=mips we now have a big-endian system which also needs to be
supported. So this patch fixes the reflect package to work on big-endian
systems.
Also, I've updated the tests for MIPS: instead of running the
little-endian tests, I've changed it to run the big-endian tests
instead. The two are very similar except for endianness so this should
be fine. To be sure we won't accidentally break little-endian support,
I've kept a single MIPS little-endian test (the CGo test, which doesn't
yet work on big-endian systems anyway).
The interp package was assuming that all targets were little-endian. But
that's not true: we now have a big-endian target (GOARCH=mips).
This fixes the interp package to use the appropriate byte order for a
given target.
Previously, the compiler would default to hardfloat. This is not
supported by some MIPS CPUs.
This took me much longer than it should have because of a quirk in the
LLVM Mips backend: if the target-features string is not set (like during
LTO), the Mips backend picks the first function in the module and uses
that. Unfortunately, in the case of TinyGo this first function is
`llvm.dbg.value`, which is an LLVM intrinsic and doesn't have the
target-features string. I fixed it by adding a `-mllvm -mattr=` flag to
the linker.
This shows nicely formatted error messages for missing symbol names and
for out-of-flash, out-of-RAM conditions (on microcontrollers with
limited flash/RAM).
Unfortunately the missing symbol name errors aren't available on Windows
and WebAssembly because the linker doesn't report source locations yet.
This is something that I could perhaps improve in LLD.
This adds linux/mipsle (little endian Mips) support to TinyGo.
It also adds experimental linux/mips (big-endian) support. It doesn't
quite work yet, some parts of the standard library (like the reflect
package) currently seem to assume a little-endian system.
This is a refactor, which should (in theory) not change the behavior of
the compiler. But since this is a pretty large change, there is a chance
there will be some regressions. For that reason, the previous commits
added a bunch of tests to make sure most error messages will not be
changed due to this refactor.
This is needed for some improvements I'm going to make next.
This commit also refactors error handling slightly to make it more
easily testable, this should hopefully not result in any actual changes
in behavior.
Uses a vendored "internal/diff" from Big Go to show the differences
between the expected and actual outputs when tests fail.
Signed-off-by: L. Pereira <l.pereira@fastly.com>
This means it's possible to test just a particular OS/arch with a
command like this:
go test -run=Build -target=linux/arm
I found it useful while working on MIPS support.
Support for `-panic=trap` was previously a pass in the optimization
pipeline. This change moves it to the compiler and runtime, which in my
opinion is a much better place.
As a side effect, it also fixes
https://github.com/tinygo-org/tinygo/issues/4161 by trapping inside
runtime.runtimePanicAt and not just runtime.runtimePanic.
This change also adds a test for the list of imported functions. This is
a more generic test where it's easy to add more tests for WebAssembly
file properties, such as exported functions.
This fixes the new loop variable behavior in Go 1.22.
Specifically:
* The compiler (actually, the x/tools/go/ssa package) now correctly
picks up the Go version.
* If a module doesn't specify the Go version, the current Go version
(from the `go` tool and standard library) is used. This fixes
`go run`.
* The tests in testdata/ that use a separate directory are now
actually run in that directory. This makes it possible to use a
go.mod file there.
* There is a test to make sure old Go modules still work with the old
Go behavior, even on a newer Go version.
This adds true GOOS=wasip1 support in addition to our existing
-target=wasi support. The old support for WASI isn't removed, but should
be treated as deprecated and will likely be removed eventually to reduce
the test burden.
Add the timers test because they now work correctly on AVR, probably as
a result of the reflect refactor: https://github.com/tinygo-org/tinygo/pull/2640
I've also updated a few of the other tests to indicate the new status
and why they don't work. It's no longer because of compiler errors, but
because of linker or runtime errors (which is at least some progress).
For example, I found that testdata/reflect.go works if you disable
`testAppendSlice` and increase the stack size.
Before this patch, a compile error would prevent the 'ok' or 'FAIL' line
to be printed. That's unexpected. This patch changes the code in such a
way that it's obvious a test result line is printed in all cases.
To be able to also print the package name, I had to make sure the build
result is passed through everywhere even on all the failure paths. This
results in a bit of churn, but it's all relatively straightforward.
Found while working on Go 1.20.
This commit adds support for time.NewTimer and time.NewTicker. It also
adds support for the Stop() method on time.Timer, but doesn't (yet) add
support for the Reset() method.
The implementation has been carefully written so that programs that
don't use these timers will normally not see an increase in RAM or
binary size. None of the examples in the drivers repo change as a result
of this commit. This comes at the cost of slightly more complex code and
possibly slower execution of the timers when they are used.
Jake Bailey
deadprogram
Nia Waldvogel
Elias Naur
Ayke van Laethem
Damian Gryski
L. Pereira
Randy Reddig
Kenneth Bell
Elliott Sales de Andrade