What each target compiles#

Each construct below was compiled for every target, and this page records what EmbCC did: compiled it, accepted it with a warning, or refused it with an error that names it. EmbCC never compiles a construct it refuses into code that does something else (see Status and known limitations). A construct that compiles may still have limits on a target; each target's section of Targets gives them. tests/golden/limits.sh checks every cell of this page against the compiler.

Where a message names its own target, it is written TARGET here. Every compile for x86_64-windows-gnu also warns that the target is not yet the Microsoft x64 ABI (Windows).

`__int128`#

Refused on 17 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, avr, mipsel-none-elf, mips-none-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf, m68k-none-elf:
```
__int128 does not exist on this target (it needs 64-bit registers; use long long)
```

Refused on x86_64-windows-gnu:

__int128 in the signature of 'f' is not supported for a Windows target yet: EmbCC lowers its arithmetic to libgcc helpers whose arguments it places in the System V registers

`long double` arithmetic#

Refused on 1 of the 25 targets.

Refused on x86_64-windows-gnu:

long double in the signature of 'f' is not supported for a Windows target yet: there it travels by reference and returns through a hidden pointer, and EmbCC passes it on the stack by value

`_Float16`#

Compiles on every target.

`__fp16`#

Compiles on every target.

`__bf16`#

Compiles on every target.

a struct holding a `_Float16`, by value#

Refused on 1 of the 25 targets.

Refused on thumbv7em-none-eabihf:

'f': a struct holding a _Float16, passed or returned by value, is not supported yet with the hard-float calling convention, which puts each binary16 member in a floating-point register of its own; pass a pointer to it

a `_Float16` through `...`#

Refused on 25 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, riscv64-unknown-elf, avr, mipsel-none-elf, mips-none-elf, mips64el-none-elf, mips64-none-elf, loongarch64-unknown-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf, m68k-none-elf, x86_64-elf, aarch64-elf, x86_64-linux-gnu, aarch64-apple-darwin, x86_64-windows-gnu:
```
'g': passing a _Float16 through '...' is not supported yet: it is not promoted there, and EmbCC does not yet place an unpromoted 16-bit float among the variable arguments; convert it to double, as va_arg will read it
```

`_BitInt(N)`#

Compiles on every target.

`_Complex int`#

Compiles on every target.

GNU vector types (`vector_size`)#

Compiles on every target.

one-byte atomic read-modify-write#

Compiles on every target.

eight-byte atomic read-modify-write#

Compiles on every target.

sixteen-byte compare-and-swap#

Refused on 21 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, avr, mipsel-none-elf, mips-none-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf, m68k-none-elf:
```
__int128 does not exist on this target (it needs 64-bit registers; use long long)
```

Refused on riscv64-unknown-elf:

the RV64 backend cannot lower a 128-bit value yet (function f) [cas16 w=16 size=16]

Refused on mips64el-none-elf, mips64-none-elf:

the MIPS64 backend cannot lower a 16-byte atomic (MIPS64's lld/scd are a doubleword; there is no 128-bit ll/sc) yet (function f) [cas16 w=16 size=16]

Refused on loongarch64-unknown-elf:

the LoongArch64 backend cannot lower a sixteen-byte atomic (the LA64 base ISA has no 128-bit ll/sc or am* instruction) yet (function f) [cas16 w=16 size=16]

Refused on x86_64-windows-gnu:

__int128 in the signature of 'f' is not supported for a Windows target yet: EmbCC lowers its arithmetic to libgcc helpers whose arguments it places in the System V registers

`_Thread_local`#

Refused on 2 of the 25 targets.

Refused on aarch64-apple-darwin:

__thread is not supported for a Darwin target yet: Mach-O addresses a thread-local through a __thread_vars descriptor, which this writer does not emit

Refused on x86_64-windows-gnu:

__thread is not supported for a Windows target yet: Windows reaches a thread-local through a _tls_index and a TLS directory this writer does not emit

variable-length arrays#

Compiles on every target.

a local aligned to 32 bytes#

Compiles on every target.

`__builtin_frame_address(0)`#

Compiles on every target.

`__builtin_return_address(1)`#

Refused on 19 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, riscv64-unknown-elf, avr, mipsel-none-elf, mips-none-elf, mips64el-none-elf, mips64-none-elf, loongarch64-unknown-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf:
```
__builtin_return_address(1) is not supported on TARGET: code for this target keeps no frame-pointer chain, so only level 0 (this function's own frame) can be found
```

`__builtin_setjmp`#

Refused on 1 of the 25 targets.

Refused on x86_64-windows-gnu:

__builtin_setjmp is not supported for TARGET: MinGW's setjmp takes a frame for its unwinder, and EmbCC does not lower onto it

computed `goto`#

Compiles on every target.

`asm goto`#

Refused on 25 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, riscv64-unknown-elf, avr, mipsel-none-elf, mips-none-elf, mips64el-none-elf, mips64-none-elf, loongarch64-unknown-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf, m68k-none-elf, x86_64-elf, aarch64-elf, x86_64-linux-gnu, aarch64-apple-darwin, x86_64-windows-gnu:
```
'asm goto' is not supported: its template branches to a label, which needs a patchable placeholder in each backend's inline assembler and CFG edges the optimizer honours. Use a normal asm that sets a value and branch on that
```

nested functions#

Refused on 25 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi, riscv32-unknown-elf, riscv64-unknown-elf, avr, mipsel-none-elf, mips-none-elf, mips64el-none-elf, mips64-none-elf, loongarch64-unknown-elf, xtensa-none-elf, tricore-none-elf, rx-none-elf, powerpc-none-eabi, sparc-none-elf, m68k-none-elf, x86_64-elf, aarch64-elf, x86_64-linux-gnu, aarch64-apple-darwin, x86_64-windows-gnu:
```
expected ';' before '{' [E0002]
```

`attribute((interrupt))`#

Refused on 9 of the 25 targets.

Refused on loongarch64-unknown-elf, xtensa-none-elf, powerpc-none-eabi, sparc-none-elf, x86_64-elf, aarch64-elf, x86_64-linux-gnu, aarch64-apple-darwin, x86_64-windows-gnu:

__attribute__((interrupt)) is not supported: the handler would return with an ordinary return instead of the interrupt return the CPU needs, and without saving the registers (on ARMv7-M it needs neither, and is accepted; on AVR, RISC-V, MIPS, RX, ColdFire, TriCore and ARMv7-A it is implemented)

`attribute((naked))`#

Refused on 4 of the 25 targets.

Refused on x86_64-elf, x86_64-linux-gnu, aarch64-apple-darwin, x86_64-windows-gnu:

__attribute__((naked)) is not supported: on this target the body could only be assembled by the file-scope assembler's few instructions; it is supported on every target but x86-64 and AArch64 for Darwin or Windows

an assembly source (`.s`) with an instruction#

Refused on 3 of the 25 targets.

Refused on x86_64-elf, x86_64-linux-gnu, x86_64-windows-gnu:
```
no assembly-file support for TARGET yet
```

`-S` (assembly output)#

Compiles on every target.

`-g` (DWARF)#

Refused on 1 of the 25 targets.

Refused on x86_64-windows-gnu:

-g is not supported for a Windows target yet: its debug information goes in CodeView records this does not write, and emitting DWARF under a COFF name would be worse than refusing

A warning on aarch64-apple-darwin:

-g: no debug information for TARGET yet; tests/golden/out/limits/src/plain.c is compiled without it

`--unwind-tables`#

Refused on 21 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi:
```
unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no ARM EHABI unwind tables (.ARM.exidx)
```

Refused on riscv32-unknown-elf, riscv64-unknown-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no RISC-V .eh_frame

Refused on avr:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no AVR .eh_frame

Refused on mipsel-none-elf, mips-none-elf, mips64el-none-elf, mips64-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no MIPS .eh_frame

Refused on loongarch64-unknown-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no LoongArch .eh_frame

Refused on xtensa-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no Xtensa .eh_frame

Refused on tricore-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no TriCore .eh_frame

Refused on rx-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no RX .eh_frame

Refused on powerpc-none-eabi:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no PowerPC .eh_frame

Refused on sparc-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no SPARC .eh_frame

Refused on m68k-none-elf:

unwind tables are not supported for TARGET yet (--unwind-tables, --exceptions): EmbCC writes no ColdFire .eh_frame

Refused on x86_64-windows-gnu:

C++ exceptions are not supported for a Windows target yet: the unwind tables go in .pdata and .xdata and neither is written

C++ (without exceptions)#

Refused on 2 of the 25 targets.

Refused on avr:

C++ is not yet supported for TARGET: the C++ lowering assumes a 32-bit int and 4-byte pointers (integral promotions, member pointers, RTTI), and this target's are 16 bits

Refused on x86_64-windows-gnu:

C++ exceptions are not supported for a Windows target yet: the unwind tables go in .pdata and .xdata and neither is written

C++ exceptions#

Refused on 21 of the 25 targets.

Refused on thumbv6m-none-eabi, thumbv7m-none-eabi, thumbv7em-none-eabihf, thumbv8m.base-none-eabi, thumbv8m.main-none-eabi, armv7a-none-eabi:
```
C++ exceptions are not supported for TARGET yet: EmbCC writes no ARM EHABI unwind tables (.ARM.exidx); compile with --no-exceptions
```

Refused on riscv32-unknown-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no RISC-V .eh_frame; compile with --no-exceptions

Refused on riscv64-unknown-elf:

the RV64 backend cannot lower this operation yet (function _Z1fv) [landing w=8 size=4]

Refused on avr:

C++ is not yet supported for TARGET: the C++ lowering assumes a 32-bit int and 4-byte pointers (integral promotions, member pointers, RTTI), and this target's are 16 bits

Refused on mipsel-none-elf, mips-none-elf, mips64-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no MIPS .eh_frame; compile with --no-exceptions

Refused on mips64el-none-elf:

the MIPS64 backend cannot lower this operation yet (function _Z1fv) [landing w=8 size=4]

Refused on loongarch64-unknown-elf:

the LoongArch64 backend cannot lower this operation yet (function _Z1fv) [landing w=8 size=4]

Refused on xtensa-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no Xtensa .eh_frame; compile with --no-exceptions

Refused on tricore-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no TriCore .eh_frame; compile with --no-exceptions

Refused on rx-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no RX .eh_frame; compile with --no-exceptions

Refused on powerpc-none-eabi:

C++ exceptions are not supported for TARGET yet: EmbCC writes no PowerPC .eh_frame; compile with --no-exceptions

Refused on sparc-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no SPARC .eh_frame; compile with --no-exceptions

Refused on m68k-none-elf:

C++ exceptions are not supported for TARGET yet: EmbCC writes no ColdFire .eh_frame; compile with --no-exceptions

Refused on x86_64-windows-gnu:

C++ exceptions are not supported for a Windows target yet: the unwind tables go in .pdata and .xdata and neither is written

The whole matrix#

Target	int128	ldouble	float16	fp16	bf16	float16struct	float16va	bitint	complexint	vector	atomic1	atomic8	atomic16	tls	vla	align32	frameaddr	retaddr1	setjmp	cgoto	asmgoto	nested	interrupt	naked	asmfile	dashS	debug	unwind	cxx	cxxeh
`thumbv6m-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`thumbv7m-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`thumbv7em-none-eabihf`	no	yes	yes	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`thumbv8m.base-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`thumbv8m.main-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`armv7a-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`riscv32-unknown-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`riscv64-unknown-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`avr`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	no	no
`mipsel-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`mips-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`mips64el-none-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`mips64-none-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`loongarch64-unknown-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	no	yes	yes	yes	yes	no	yes	no
`xtensa-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	no	yes	yes	yes	yes	no	yes	no
`tricore-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`rx-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`powerpc-none-eabi`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	no	yes	yes	yes	yes	no	yes	no
`sparc-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	no	yes	yes	no	no	no	yes	yes	yes	yes	no	yes	no
`m68k-none-elf`	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	yes	no	no	yes	yes	yes	yes	yes	no	yes	no
`x86_64-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	no	no	no	no	no	yes	yes	yes	yes	yes
`aarch64-elf`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	no	no	no	yes	yes	yes	yes	yes	yes	yes
`x86_64-linux-gnu`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes	no	no	no	no	no	yes	yes	yes	yes	yes
`aarch64-apple-darwin`	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	yes	no	no	no	no	yes	yes	warns	yes	yes	yes
`x86_64-windows-gnu`	no	no	yes	yes	yes	yes	no	yes	yes	yes	yes	yes	no	no	yes	yes	yes	yes	no	yes	no	no	no	no	no	yes	no	no	no	no

The columns are the sections above, in order:

int128: __int128
ldouble: long double arithmetic
float16: _Float16
fp16: __fp16
bf16: __bf16
float16struct: a struct holding a _Float16, by value
float16va: a _Float16 through ...
bitint: _BitInt(N)
complexint: _Complex int
vector: GNU vector types (vector_size)
atomic1: one-byte atomic read-modify-write
atomic8: eight-byte atomic read-modify-write
atomic16: sixteen-byte compare-and-swap
tls: _Thread_local
vla: variable-length arrays
align32: a local aligned to 32 bytes
frameaddr: __builtin_frame_address(0)
retaddr1: __builtin_return_address(1)
setjmp: __builtin_setjmp
cgoto: computed goto
asmgoto: asm goto
nested: nested functions
interrupt: __attribute__((interrupt))
naked: __attribute__((naked))
asmfile: an assembly source (.s) with an instruction
dashS: -S (assembly output)
debug: -g (DWARF)
unwind: --unwind-tables
cxx: C++ (without exceptions)
cxxeh: C++ exceptions

What each target compiles#

__int128#

long double arithmetic#

_Float16#

__fp16#

__bf16#

a struct holding a _Float16, by value#

a _Float16 through ...#

_BitInt(N)#

_Complex int#

GNU vector types (vector_size)#

one-byte atomic read-modify-write#

eight-byte atomic read-modify-write#

sixteen-byte compare-and-swap#

_Thread_local#

variable-length arrays#

a local aligned to 32 bytes#

__builtin_frame_address(0)#

__builtin_return_address(1)#

__builtin_setjmp#

computed goto#

asm goto#

nested functions#

__attribute__((interrupt))#

__attribute__((naked))#

an assembly source (.s) with an instruction#

-S (assembly output)#

-g (DWARF)#

--unwind-tables#

C++ (without exceptions)#

C++ exceptions#

The whole matrix#

`__int128`#

`long double` arithmetic#

`_Float16`#

`__fp16`#

`__bf16`#

a struct holding a `_Float16`, by value#

a `_Float16` through `...`#

`_BitInt(N)`#

`_Complex int`#

GNU vector types (`vector_size`)#

`_Thread_local`#

`__builtin_frame_address(0)`#

`__builtin_return_address(1)`#

`__builtin_setjmp`#

computed `goto`#

`asm goto`#

`attribute((interrupt))`#

`attribute((naked))`#

an assembly source (`.s`) with an instruction#

`-S` (assembly output)#

`-g` (DWARF)#

`--unwind-tables`#