Add full flash XIP support for ESP32, enabling code and read-only data to
execute/load directly from flash via the MMU cache rather than consuming
precious SRAM. This increases available RAM from ~328KB to effectively
unlimited for code/rodata, while keeping ~121KB for the Go heap.
Changes:
- src/device/esp/esp32.S: Add MMU initialization in call_start_cpu0
- Call ROM bootloader mmu_init() and cache_flash_mmu_set() to map DROM/IROM
- Enable flash cache via ROM Cache_Read_Enable()
- Fix tinygo_scanCurrentStack to spill all register windows for GC
- targets/esp32-interrupts.S: Add exception diagnostics
- targets/esp32.ld: Major linker script restructure for XIP
- Add DROM (4MB @ 0x3F400000) and IROM (4MB @ 0x400D0000) regions
- Move .rodata to DROM, main .text to IROM (both flash-mapped)
- Keep boot code, vectors, and WiFi blob IRAM sections in SRAM0
- Create WiFi arena in SRAM1 pool 7/6 (64KB @ 0x3FFF0000)
- Move .bss and heap to SRAM2 (200KB @ 0x3FFAE000), avoiding ROM/MAC regions
- Add _drom_flash_addr variable (patched by builder with flash offset)
- targets/esp32.json: Add linker wrap flags for malloc/free and WiFi functions
Signed-off-by: deadprogram <ron@hybridgroup.com>
Goexit cannot run deferred functions on architectures without panic
unwinding. Query runtime.supportsRecover before calling it so unsupported
targets report an error instead of leaving the test runner blocked forever.
LLVM ComputeValueVTs recursively expands arrays and structs into one
value type per scalar leaf. SelectionDAG call lowering allocates data
structures proportional to this count, which makes very large values
exhaust memory or crash LLVM.
Count scalar leaves and use pointers for internal parameters and results
when the count exceeds 1024. A result pointer is the first parameter,
and aggregate parameters point to read-only memory. Exported function
types are unchanged.
Keep these SSA values in memory and copy them with memcpy when needed.
Handle calls, interfaces, maps, channels, selects, defers, goroutines,
phis, and multiple results. Update the expected compiler IR and re-enable
the native compress/flate tests.
Add compiler coverage for large aggregate parameters and results,
including function values, interfaces, maps, channels, selects, deferred
calls, goroutines, phis, and multiple results.
createInstruction records each LLVM value in locals, emits stack object
tracking, and constructs function returns inline.
Move these operations to setValue and createReturn. This keeps the
instruction switch from having to know how an SSA value or function
result is emitted.
createFunctionCall and createGo currently lower arguments before they
determine whether the call is direct, an interface invoke, or through a
function value.
Resolve the callee, function type, and context first, then append the
arguments in the same order as before. This does not change the
generated LLVM IR.
getFunction and getLLVMFunctionType duplicate the construction of LLVM
result types for functions with zero, one, or multiple results.
Move this code to getLLVMResultType. Also split the existing parameter
expansion code into expandDirectFormalParamType so callers can request
the current flattened parameter types directly.
Map, channel, index, and goroutine lowering each create allocas, store
SSA values into them, load results, and emit lifetime intrinsics.
Add helpers for these operations and use runtimeValueResult for runtime
calls that write a value and an optional comma-ok result. The generated
LLVM IR is unchanged.
createDefer builds an LLVM struct containing the deferred function and
its arguments. createRunDefers separately reconstructs the same struct
type before loading the fields.
Keep each LLVM value together with its type while building a deferred
call record, and load all argument fields through one helper. This
removes the duplicate lists of field types and field loads.
The shared timekeeping code assumes the TIMG0 timer counts at 40MHz,
but on the ESP32-C6 the timer clock defaults to the 40MHz XTAL, giving
a 20MHz tick with the /2 prescaler and making all delays run twice as
long. Select PLL_F80M (80MHz) and enable the timer group clock so the
25ns/tick assumption holds.
Signed-off-by: deadprogram <ron@hybridgroup.com>
Currently, calling any sync.Map method from inside the sync.Map.Range
callback f deadlocks. Moreover, Go's sync.Map explicitly permits the
Range callback to call other methods on the map ("Range does not
block other methods on the receiver; even f itself may call any
method on m").
This commit prevents the deadlock by changing sync.Map.Range to:
- copy the map's keys under the lock
- release the lock
- iterate over the key snapshot
A snapshot satisfies Go's sync.Map.Range contract, which only
requires that no key is visited more than once and may reflect any
mapping from any point during the call.
Using a snapshot keeps the implementation simple, in line with this
file's stated scope ("no more efficient than a map with a lock").
Also added TestMapRangeAndDelete regression test, which deletes map
entries from inside the map's Range callback.
Replace movi instructions with constants outside the 12-bit signed
range (-2048..2047) with movi+slli sequences. Large constants cause
the assembler to emit auto-generated .literal section entries, which
triggers an lld bug where l32r PC-relative offsets are miscalculated
when .literal.* sections are merged with .text.* sections.
Signed-off-by: deadprogram <ron@hybridgroup.com>
This fixes the builder for ESP32 (original) by separating the RAM segments loadable by the
ROM bootloader from flash-mapped segments (DROM/IROM) which require MMU setup by startup code.
These changes are needed for to allow for ESP32 to correctly flashing large programs which
as a result require XIP support.
Signed-off-by: deadprogram <ron@hybridgroup.com>
Define the board's external crystal (HSE) frequency `xtalHz` in the
respective `board_*.go` files. Per-topology PLL tables (F1, F4, F7)
will compute the register dividers from it.
Moving this parameter to the board definition level cleanly isolates
board hardware characteristics from general MCU chip configurations,
eliminating the need for custom target build tags. Targets using
HSE-clocked STM32 chips must define `xtalHz` or fail to build.
Clang on i386 mingw emits calls to _alloca (decorated as __alloca)
for stack probing. This was provided by chkstk2.S, not chkstk.S
which only provides __chkstk_ms.
Signed-off-by: deadprogram <ron@hybridgroup.com>
cc1as: change AssemblerInvocation.Triple from std::string to
llvm::Triple, matching upstream LLVM 22 and fixing deprecation warnings
for lookupTarget, createMCRegInfo, createMCAsmInfo, and
createMCSubtargetInfo.
interp: always use ConstNamedStruct in rawValue.toLLVMValue instead of
falling back to ConstStruct for unnamed structs. LLVM 22 uses anonymous
identified struct types where previous versions used literal structs;
ConstStruct creates a literal type that doesn't match, causing an
"initializer type mismatch" panic for globals like fmt.ppFree.
compileopts: add build-tag-guarded feature patching for pre-LLVM 20
(strip +bulk-memory-opt and +call-indirect-overlong from wasm features)
and restructure into three files covering all LLVM version ranges.
targets: strip redundant negative features from RISC-V target JSONs
(esp32c3, esp32c6, fe310, k210, riscv-qemu, tkey). These ~190 negative
features per target listed every extension LLVM knows about that the
target doesn't use, but LLVM disables them by default. They became stale
across LLVM versions, causing "not a recognized feature" warnings. Keep
only positive features and -relax.
Signed-off-by: deadprogram <ron@hybridgroup.com>
Builder:
- Update clang.cpp for LLVM 22 API changes: DiagnosticOptions is no
longer ref-counted, TextDiagnosticPrinter and DiagnosticsEngine take
references instead of pointers, createDiagnostics() signature changed.
- Update cc1as.cpp: clang/Driver/Options.h moved to
clang/Options/Options.h, namespace changed from clang::driver::options
to clang::options, MCInstPrinter now passed as unique_ptr.
- Add new LLVM 22 libraries to GNUmakefile: clangAnalysisLifetimeSafety,
clangOptions, LLVMDTLTO, and dtlto component.
- Add llvm22 build tag to all build/test commands.
CGo:
- Handle CXType_Unexposed in libclang.go by resolving via canonical
type. LLVM 22 reports builtin type aliases (e.g. __size_t) as
Unexposed instead of Typedef.
- Always make C typedefs into Go type aliases. LLVM 22 changed
getTypedefDeclUnderlyingType to return CXType_Enum directly instead
of wrapping in an elaborated type.
Compileopts:
- Add build-tag-guarded ClangTriple() to substitute wasm32-unknown-wasi
with wasm32-unknown-wasip1 for LLVM 22 (deprecated triple).
- Add build-tag-guarded patchFeatures() to map renamed Xtensa features
(atomctl, memctl, timerint, esp32s3) for LLVM 22.
Targets:
- Remove -zca from RISC-V target feature strings (esp32c3, esp32c6,
fe310, k210, riscv-qemu, tkey). LLVM 22 now implies +zca from +c.
- Remove -zcd from k210. LLVM 22 now implies +zcd from +c,+d.
Tests:
- Change TestClangAttributes to check individual feature flags instead
of exact string match, allowing new LLVM features without failures.
- Update TestBinarySize expected values for LLVM 22 codegen.
Signed-off-by: deadprogram <ron@hybridgroup.com>
Written entirely by Claude (Anthropic's Claude Code), at the request of
and under the direction of dgryski, on the dgryski/llvm23 branch (LLVM
22 support, in preparation for the eventual LLVM 23 release; the
corresponding go-llvm branch of the same name adds the matching
binding support this depends on).
Found while running real-world Go test suites through a tinygo built
against LLVM 22 (as a smoke test of the earlier captures(none)/
nocapture and lifetime-intrinsic fixes on this branch): `tinygo test`
on github.com/dgryski/go-rc5 panicked at compile time with "unknown
value" inside interp.(*runner).getValue, while compiling
(*sync.Once).Do. Bisecting against LLVM 21 (unaffected) narrowed it to
a change introduced in LLVM 22 specifically, unrelated to the earlier
captures/lifetime work.
Root cause: LLVM 22 stopped exposing switch-instruction case values as
regular instruction operands -- only the condition and
destination-block operands remain. interp/compiler.go's `case
llvm.Switch:` handling walked raw operands assuming the old
alternating (value, label) layout, so under LLVM 22 it read a
destination *block* where it expected an integer case value,
eventually panicking deep inside a getValue call it couldn't resolve.
This only manifested on functions actually compiled by the interp
package (TinyGo's compile-time evaluator for global initializers) that
happen to reach a switch-based dispatch, such as sync.Once's
defer-based Do method -- hence it needed a real, moderately complex
package (crypto/cipher via go-rc5) to surface, and never showed up in
smaller smoke tests.
Fix: use the new version-independent Value.SuccessorsCount()/
Value.Successor(i) and Value.GetSwitchCaseValue(i) helpers just added
to go-llvm, instead of raw Operand() indexing.
Also applies the same captures(none)/nocapture golden-IR normalization
already used in transform/transform_test.go and compiler/compiler_test.go
to interp/interp_test.go's separate copy of the same comparator helper,
which was missed in the earlier pass and started failing two of its
subtests once actually run against LLVM 22.
Verified: `tinygo test` on go-rc5 now passes reliably (repeated runs)
against both LLVM 20 (default) and LLVM 22; full transform/compiler/
interp/cgo/goenv test suites pass on both versions. The `builder`
package has pre-existing, environment-related test failures (stale
GOROOT cache, local clang/target-feature drift) confirmed identical
against an unmodified LLVM 20 build, so unrelated to this change.
Written entirely by Claude (Anthropic's Claude Code), at the request of
and under the direction of dgryski, as part of an effort to get TinyGo
building against upcoming LLVM releases (this branch currently targets
LLVM 22, verified against real LLVM 22.1.8; a corresponding go-llvm
branch of the same name adds the matching binding support).
LLVM 21 replaced the boolean 'nocapture' enum attribute with the more
expressive 'captures' int attribute, where captures(none) (value 0) is
the equivalent of the old nocapture. This matters for
transform.OptimizeAllocs, which relies on reading this attribute for
its interprocedural escape analysis, and for compiler/symbol.go, which
emits it on a number of runtime/generated functions. Confirmed
empirically (via `opt -passes=function-attrs`) that the cutoff is
LLVM 20 emits/expects nocapture, LLVM 21+ emits/expects
captures(none). Since TinyGo must keep working with LLVM 20, both
sites now go through new version-gated helpers in
compiler/llvmutil (NoCaptureAttrName/IsNoCapture) rather than switching
unconditionally.
Also fixes a second, unrelated but load-bearing break found while
testing against LLVM 22: llvm.lifetime.start/end dropped their i64
size argument (confirmed via `opt -passes=verify`, the cutoff here is
one version later, at LLVM 22). compiler/llvmutil now builds the
right call signature based on version.
Adds llvm21 and llvm22 build-tag config files to the cgo package,
which parses cgo fragments via libclang and had never been updated
past LLVM 20 even though the compiler package itself already gained
LLVM 21 support previously -- a latent gap that would have caused a
version mismatch between the cgo preprocessor and the rest of the
compiler when building with -tags llvm21 or llvm22.
Finally, updates the golden-IR test comparators in
transform/transform_test.go and compiler/compiler_test.go to
normalize a few cosmetic LLVM 21/22 output differences (the
captures(none) rename/reordering, a new 'nocreateundeforpoison'
intrinsic attribute, and the lifetime intrinsic arity change) so a
single golden file continues to match output from either LLVM
version.
Verified by building a full (non-byollvm) tinygo binary against real
LLVM 22.1.8 and running a compiled Go program end-to-end (exercising
OptimizeAllocs' stack-allocation path), and by running the
transform/compiler/cgo test suites against both LLVM 20 (default) and
LLVM 22.
Not yet addressed: the byollvm embedded-clang/lld build path hits
separate, unrelated Clang C++ API breakage against LLVM 22
(DiagnosticOptions reference-to-pointer change, missing headers) --
that is a larger follow-up effort.
* feat(machine): add UART line inversion support.
Add InvertTX and InvertRX to UARTConfig to allow enabling hardware
line inversion on supported targets. Added hardware implementation for
RP2 (RP2040, RP2350), STM32 (newer families), SAM (SAMD51, SAME5x),
and ESP (ESP32, ESP32-C3, ESP32-C6).
* refactor(machine): Refactor UART inversion with pin setter helpers.
RP2: Extract setOutOver/setInOver methods on Pin, removing inline
IO control register manipulation from UART configure.
SAM: Switch to SetCTRLA_TXINV/SetCTRLA_RXINV methods, dropping
the unused device/sam import and raw SetBits/ClearBits calls.
---------
Co-authored-by: Konstantin Sharlaimov <ksharlaimov@inavflight.com>
Go 1.27 jsonv2 pushes testdata/json.go beyond the LM3S6965 flash
budget. Keep coverage on larger emulated targets while skipping only the
too-small Cortex-M QEMU configuration.
Go 1.27 packages can emit stack probes on Windows amd64 and arm64.
Link the compiler-rt stack probe builtins and use compiler-rt for those
targets so these packages build.
Go 1.27 adds generic methods (golang/go#77273). golang.org/x/tools
go/ssa v0.42.0 does not instantiate them: objectMethod only applies
receiverTypeArgs and ignores method-level type parameters, so a call
such as (*math/rand/v2.Rand).N resolves to the abstract generic method
and its body reaches createConst with a type-parameter-typed zero
constant, triggering "panic: expected nil interface constant".
Upgrade golang.org/x/tools to v0.47.0, whose go/ssa returns nil from
MethodValue for generic methods and instantiates method-level type
parameters. No compiler changes are required.
Since x/tools v0.47.0 requires Go 1.25, raise the minimum supported Go
version from 1.24 to 1.25.
Signed-off-by: deadprogram <ron@hybridgroup.com>
The interface lowering pass used a separate formatter for method
signatures. Same-named local aliases could therefore make distinct generic
methods compare equal. Reuse getTypeCodeName so interface checks preserve
type identity.
x/tools SSA names and go/types strings can preserve the source spelling of
type arguments, so aliases can make distinct instances collide. Use the
canonical type encoding for function names, synthetic local type owners,
instantiated named types, and method sets.
Add cases where same-named local aliases instantiate generic functions,
local types, and methods with float32 and float64. Record the current
collisions and incorrect results.
Go 1.27 introduced the //go:linknamestd directive, a standard-library
variant of //go:linkname that does not require importing "unsafe". The
iter package switched to it for referencing runtime.newcoro and
runtime.coroswitch, which caused "linker could not find symbol
iter.newcoro / iter.coroswitch" errors when building with Go 1.27.
Handle //go:linknamestd the same as //go:linkname, bypassing the unsafe
import requirement, and add test coverage in the pragma compiler test.
Signed-off-by: deadprogram <ron@hybridgroup.com>
Modern golang.org/x/sys/unix (v0.36+) declares its linknames
detached from function declarations, e.g.:
func syscall_syscall(...)
//go:linkname syscall_syscall syscall.syscall
TinyGo's pragma parser only inspected function doc comments and
therefore missed these, producing link errors like:
undefined symbol: _golang.org/x/sys/unix.syscall_syscall
Extend parsePragmas to also walk the enclosing *ast.File's
free-standing comments for //go:linkname directives matching the
function's name. Function-attached directives still take
precedence. The existing 'unsafe' import gate is preserved.
Fixes#4395, #5365
STM32F4 and F7 share the same OTG FS IP but have no USB driver in
TinyGo. This adds a full device-mode driver covering CDC, HID, and
MSC with working examples.
- OTG FS has 4 physical EPs (0–3); virtual indices 4–7 fold onto
them via physEP() so the existing machine/usb API is unchanged
- F4 bypasses VBUS sensing via GCCFG.NOVBUSSENS; F7 uses the USB
voltage regulator + GOTGCTL B-valid override instead
- STM32F7 PLL_Q changed 2→9 to produce the 48 MHz clock required
by USB/RNG/SDMMC; CK48MSEL cleared to select main PLL as source
- HID and MSC descriptors remapped at init() to physical endpoint
addresses (EP2/EP1 for HID, EP2/EP3 for MSC)
- usb-storage example replaced machine.Flash with a FAT12 RAM disk
so the host mounts without reformatting
- MSC sendCSW sets queuedBytes before state transition to fix a
missed byte-count on the status phase
Treat panicState as a bitmask so a recovered panic during Goexit
can clear the panic bit while preserving the pending Goexit bit.
This removes the separate deferFrame Goexit field and restores the
previous defer frame size.
deadprogram
Jake Bailey
Pat Whittingslow
Nia Waldvogel
Faye Amacker
Konstantin Sharlaimov
rdon(あーるどん)
Damian Gryski
felipegenef
rdon-key
Evan Wies