diff options
| author | Ayke van Laethem <[email protected]> | 2023-09-23 15:06:27 +0200 |
|---|---|---|
| committer | Ron Evans <[email protected]> | 2024-01-05 21:13:30 +0100 |
| commit | 6984af43a08b7fa1c1020700c42ce8c417d1f542 (patch) | |
| tree | 79e9b9bfe07c1d9a67913585a39ec17321715ce4 /compileopts | |
| parent | 81c56c3ab89fe0598361c00f03eb0498d2ea0924 (diff) | |
| download | tinygo-6984af43a08b7fa1c1020700c42ce8c417d1f542.tar.gz tinygo-6984af43a08b7fa1c1020700c42ce8c417d1f542.zip | |
all: statically link to LLVM 17 instead of LLVM 16
We can now finally do it, now that Espressif has updated their fork.
Diffstat (limited to 'compileopts')
| -rw-r--r-- | compileopts/target.go | 22 |
1 files changed, 19 insertions, 3 deletions
diff --git a/compileopts/target.go b/compileopts/target.go index f7b8a065f..66f44b7cc 100644 --- a/compileopts/target.go +++ b/compileopts/target.go @@ -301,10 +301,10 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) { switch goarch { case "386": spec.CPU = "pentium4" - spec.Features = "+cx8,+fxsr,+mmx,+sse,+sse2,+x87" + spec.Features = "+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" case "amd64": spec.CPU = "x86-64" - spec.Features = "+cx8,+fxsr,+mmx,+sse,+sse2,+x87" + spec.Features = "+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" case "arm": spec.CPU = "generic" spec.CFlags = append(spec.CFlags, "-fno-unwind-tables", "-fno-asynchronous-unwind-tables") @@ -320,8 +320,10 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) { spec.CPU = "generic" if goos == "darwin" { spec.Features = "+neon" - } else { // windows, linux + } else if goos == "windows" { spec.Features = "+neon,-fmv" + } else { // linux + spec.Features = "+neon,-fmv,-outline-atomics" } case "wasm": spec.CPU = "generic" @@ -349,6 +351,20 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) { spec.RTLib = "compiler-rt" spec.Libc = "musl" spec.LDFlags = append(spec.LDFlags, "--gc-sections") + if goarch == "arm64" { + // Disable outline atomics. For details, see: + // https://cpufun.substack.com/p/atomics-in-aarch64 + // A better way would be to fully support outline atomics, which + // makes atomics slightly more efficient on systems with many cores. + // But the instructions are only supported on newer aarch64 CPUs, so + // this feature is normally put in a system library which does + // feature detection for you. + // We take the lazy way out and simply disable this feature, instead + // of enabling it in compiler-rt (which is a bit more complicated). + // We don't really need this feature anyway as we don't even support + // proper threading. + spec.CFlags = append(spec.CFlags, "-mno-outline-atomics") + } } else if goos == "windows" { spec.Linker = "ld.lld" spec.Libc = "mingw-w64" |