This is a small change to make it easier to support architectures that
need to restore more than just the sp and pc registers. In particular,
it is needed for the AVR architecture that needs to restore the frame
pointer (Y register).
If an interrupt happens between the writes to SPL and SPH, the stack
pointer is inconsistent and terrible things will happen. Therefore,
disable interrupts while updating the stack pointer.
Interrupts are restored _before_ the write to SPH. This is safe, because
interrupts are re-enabled with a one cycle delay. The avr-gcc and Clang
compilers do the same thing when they need to update the stack pointer.
It's almost impossible to test for this bug, but it should make firmware
just a little bit more reliable.
This adds early Go 1.19 support. There are a number of things that don't
work yet, but the smoke tests pass so it's at least working for a
significant subset of programs.
This change also switches from CircleCI convenience images to upstream
Go images. This makes it a bit easier to use the latest Go versions.
Also, the convenience images are not updated anymore.
For example, this commit moves the 'throw' branch of an assertion (nil
check, slice index check, etc) to the end of the function while
inserting the "continue" branch right after the insert location. This
makes the resulting IR easier to follow.
For some reason, this also reduces code size a bit on average. The
TinyGo smoke tests saw a reduction of 0.22%, mainly from WebAssembly.
The drivers repo saw little average change in code size (-0.01%).
This commit also adds a few compiler tests for the defer keyword.
llvm.AddBasicBlock should never be used. Instead, we should use the
AddBasicBlock method of the current LLVM context.
This didn't lead to any bugs... yet. But probably would, eventually.
Previously we used to scan between _edata and _end. This is not correct:
the .data section starts *before* _edata.
Fixing this would mean changing _edata to _etext, but that didn't quite
work either. It appears that there are inaccessible pages between _etext
and _end on ARM. Therefore, a different solution was needed.
What I've implemented is similar to Windows and MacOS: namely, finding
writable segments by parsing the program header of the currently running
program. It's a lot more verbose, but it should be correct on all
architectures. It probably also reduces the globals to scan to those
that _really_ need to be scanned.
This bug didn't result in issues in CI, but did result in a bug in the
recover branch: https://github.com/tinygo-org/tinygo/pull/2331. This
patch fixes this bug.
Show the correct error message when trying to strip debug information.
Also, remove the special case for GOOS=linux that was probably dead
code: it was only reachable on baremetal systems which were already
checked before.
tests/testing/recurse has two directories with tests;
"make smoketest" now does "tinygo test ./..." in that directory
and fails if it does not run both directories' tests.
Previously, the MakeGCStackSlots pass would attempt to pop the stack chain before a tail call.
This resulted in use-after-free bugs when the tail call allocated memory and used a value allocated by its caller.
Instead of trying to move the stack chain pop, remove the tail flag from the call.
This change triggers a revert whenever a basic block runs instructions at runtime twice.
As a result, a loop body with runtime-only instructions will no longer be unrolled.
This should help some extreme cases where loops can be expanded into hundreds or thousands of instructions.
Precise globals require a whole program optimization pass that is hard
to support when building packages separately. This patch removes support
for these globals by converting the last use (Linux) to use
linker-defined symbols instead.
For details, see: https://github.com/tinygo-org/tinygo/issues/2870
This shrinks transform.Optimize() a little bit, working towards the goal
of https://github.com/tinygo-org/tinygo/issues/2870. I ran the smoke
tests and there is no practical downside: one test got smaller (??) and
one had a different .hex hash, but other than that there was no
difference.
This should also make TinyGo a liiitle bit faster but it's probably not
even measurable.
The transform package is the more appropriate location for package-level
optimizations, to match `transform.Optimize` for whole-program
optimizations.
This is just a refactor, to make later changes easier to read.
Move it from *builder to *compilerContext, so that it can be called in
more places. This is necessary to create a string value (for the file
name) in createEmbedGlobal.
This commit moves the calculation of the package action ID (cache key)
into a separate job. At the moment, this won't have a big effect but
this change is necessary for some future changes I want to make.
This replaces "precise" global scanning in LLVM with conservative
scanning of writable MachO segments. Eventually I'd like to get rid of
the AddGlobalsBitmap pass, and this is one step towards that goal.
This is necessary for the next commit. The next commit would otherwise
cause an issue with the following constant operation:
i64 add (i64 ptrtoint (%runtime.machHeader* @_mh_execute_header to i64), i64 32)
Simplify the interrupt-based timer code in a few ways:
- Do not recalibrate the timer every 100ms. Instead, rely on the fact
that the machine package will calbrate the timer if necessary if it
makes changes to Timer0.
- Do not configure Timer0 and then set nanosecondsInTick based on that
value. Instead, use a fixed value.
These two changes together mean that in code that doesn't use PWM,
nanosecondsInTick will be constant which makes the TIMER0_OVF interrupt
handler a lot smaller.
Together this reduces the code size of AVR binaries by about 1200 bytes,
making it pretty close to the pre-timer code size (only about 250 bytes
larger).
It also somehow fixes a problem with
tinygo.org/x/drivers/examples/ws2812 on the Arduino Uno. I'm not quite
sure what was going wrong, but bisecting pointed towards the timer code
(https://github.com/tinygo-org/tinygo/pull/2428) and with this
simplification the bug appears to be gone.
Scan globals conservatively by reading writable sections from the PE
header.
I'd like to get rid of needing to precisely scan globals eventually, and
this brings us one step closer. It also avoids a bug with ThinLTO on
Windows.
This is just basic support. It doesn't add support for reading DWARF,
because that's a bit complicated on MacOS (it isn't stored in the file
itself but separately in the object files). But at least this change
makes it possible to easily print executable sizes by section type like
for other operating systems.
Bug:
1. fn.locals[v.value] returns 0 (the default value) if v.value is not
part of the fn.locals map.
2. locals[fn.locals[v.value]] then returns the first local value, which
is usually non-nil
3. This incorrect value is then used as the operand value.
The manifestation of this convoluted bug was
https://github.com/tinygo-org/tinygo/issues/2842. It didn't occur more
often probably because it only seems to happen in practice with inline
assembly.
Fixes https://github.com/tinygo-org/tinygo/issues/2842
Ayke van Laethem
Yurii Soldak
Damian Gryski
Dan Kegel
sago35
José Carlos Chávez
Nia Waldvogel
Krzysztof Jagiello
Olaf Flebbe
Steven Kabbes
deadprogram
Dylan Arbour