c-b: Replace the no-asm feature with an opposite-polarity arch

Make `libm` and `compiler-builtins` use the same configuration:

* By default there is an `arch` feature which allows `core::arch` and
  inline assembly.
* Disabling the `arch` feature turns these off.

Positive feature flags are easier to reason about than negative, so this
will allow for some build script cleanup as well.

library/compiler-builtins/builtins-shim/Cargo.toml

@@ -39,17 +39,16 @@ test = false
 cc = { version = "1.2", optional = true }
 
 [features]
-default = []
+default = ["arch"]
+
+# Enable architecture-specific features such as SIMD or assembly routines. If
+# disabled, the generic version can be tested on any platform.
+arch = []
 
 # Enable compilation of C code in compiler-rt, filling in some more optimized
 # implementations and also filling in unimplemented intrinsics
 c = ["dep:cc"]
 
-# For implementations where there is both a generic version and a platform-
-# specific version, use the generic version. This is meant to enable testing
-# the generic versions on all platforms.
-no-asm = []
-
 # Flag this library as the unstable compiler-builtins lib. This must be enabled
 # when using as `std`'s dependency.'
 compiler-builtins = ["unmangled-names"]

library/compiler-builtins/builtins-test/Cargo.toml

@@ -6,7 +6,7 @@ publish = false
 license = "MIT AND Apache-2.0 WITH LLVM-exception AND (MIT OR Apache-2.0)"
 
 [dependencies]
-compiler_builtins = { workspace = true, features = ["unstable-public-internals"] }
+compiler_builtins = { workspace = true, default-features = false, features = ["unstable-public-internals"] }
 
 # For fuzzing tests we want a deterministic seedable RNG. We also eliminate potential
 # problems with system RNGs on the variety of platforms this crate is tested on.
@@ -24,9 +24,10 @@ gungraun = { workspace = true, optional = true }
 paste.workspace = true
 
 [features]
-default = []
+# Defaults should match the defaults in compiler-builtins since we have that
+# dependency with `default-features=false`.
+default = ["compiler_builtins/arch"]
 c = ["compiler_builtins/c"]
-no-asm = ["compiler_builtins/no-asm"]
 
 # Enable icount benchmarks (requires gungraun-runner and valgrind locally)
 icount = ["dep:gungraun"]
@@ -36,11 +37,6 @@ icount = ["dep:gungraun"]
 benchmarking-reports = ["walltime", "criterion/plotters", "criterion/html_reports"]
 walltime = ["dep:criterion"]
 
-# NOTE: benchmarks must be run with `--no-default-features` or with
-# `-p builtins-test`, otherwise the default `compiler-builtins` feature
-# of the `compiler_builtins` crate gets activated, resulting in linker
-# errors.
-
 [[bench]]
 name = "float_add"
 harness = false

library/compiler-builtins/ci/bench-icount.sh

@@ -79,7 +79,7 @@ function run_icount_benchmarks() {
 
 # Run once with softfloats, once with arch instructions enabled
 run_icount_benchmarks --features force-soft-floats -- --save-baseline=softfloat
-run_icount_benchmarks -- --save-baseline=hardfloat
+run_icount_benchmarks --features arch -- --save-baseline=hardfloat
 
 if [ "$failed" != "0" ]; then
     echo "One or more benchmarks failed"

library/compiler-builtins/ci/miri.sh

@@ -13,10 +13,11 @@ targets=(
     s390x-unknown-linux-gnu
 )
 for target in "${targets[@]}"; do
-    # Only run the `mem` tests to avoid this taking too long.
+    # Only run the `mem` tests to avoid this taking too long. Disable default
+    # features to turn off `arch` and avoid inline assembly.
     cargo miri test \
         --manifest-path builtins-test/Cargo.toml \
-        --features no-asm \
+        --no-default-features \
         --target "$target" \
         -- mem
 done

library/compiler-builtins/ci/run.sh

@@ -55,10 +55,10 @@ else
     "${test_builtins[@]}" --release
     "${test_builtins[@]}" --features c
     "${test_builtins[@]}" --features c --release
-    "${test_builtins[@]}" --features no-asm
-    "${test_builtins[@]}" --features no-asm --release
     "${test_builtins[@]}" --benches
     "${test_builtins[@]}" --benches --release
+    "${test_builtins[@]}" --no-default-features
+    "${test_builtins[@]}" --no-default-features --release
 
     # Validate that having a verbatim path for the target directory works
     # (trivial to regress using `/` in paths to build artifacts rather than
@@ -85,8 +85,8 @@ symcheck_cb_args=(-- --package compiler_builtins --features compiler-builtins)
 "${symcheck[@]}" "${symcheck_cb_args[@]}" --release
 "${symcheck[@]}" "${symcheck_cb_args[@]}" --features c
 "${symcheck[@]}" "${symcheck_cb_args[@]}" --features c --release
-"${symcheck[@]}" "${symcheck_cb_args[@]}" --features no-asm
-"${symcheck[@]}" "${symcheck_cb_args[@]}" --features no-asm --release
+"${symcheck[@]}" "${symcheck_cb_args[@]}" --no-default-features
+"${symcheck[@]}" "${symcheck_cb_args[@]}" --no-default-features --release
 
 run_intrinsics_test() {
     build_args=(--verbose --manifest-path builtins-test-intrinsics/Cargo.toml)

library/compiler-builtins/compiler-builtins/Cargo.toml

@@ -34,17 +34,16 @@ core = { path = "../../core", optional = true }
 cc = { version = "1.2", optional = true }
 
 [features]
-default = []
+default = ["arch"]
+
+# Enable architecture-specific features such as SIMD or assembly routines. If
+# disabled, the generic version can be tested on any platform.
+arch = []
 
 # Enable compilation of C code in compiler-rt, filling in some more optimized
 # implementations and also filling in unimplemented intrinsics
 c = ["dep:cc"]
 
-# For implementations where there is both a generic version and a platform-
-# specific version, use the generic version. This is meant to enable testing
-# the generic versions on all platforms.
-no-asm = []
-
 # Flag this library as the unstable compiler-builtins lib. This must be enabled
 # when using as `std`'s dependency.'
 compiler-builtins = ["dep:core", "unmangled-names"]

library/compiler-builtins/compiler-builtins/build.rs

@@ -90,7 +90,7 @@ fn configure_libm(target: &Target) {
     set_cfg("intrinsics_enabled", true);
 
     // The arch module may contain assembly.
-    set_cfg("arch_enabled", !cfg!(feature = "no-asm"));
+    set_cfg("arch_enabled", cfg!(feature = "arch"));
 
     let opt = !matches!(target.opt_level.as_str(), "0" | "1");
     set_cfg("optimizations_enabled", opt);

library/compiler-builtins/compiler-builtins/src/int/specialized_div_rem/mod.rs

@@ -144,7 +144,7 @@ fn u64_by_u64_div_rem(duo: u64, div: u64) -> (u64, u64) {
         target_family = "wasm",
         not(any(target_pointer_width = "16", target_pointer_width = "32")),
     ),
-    not(all(not(feature = "no-asm"), target_arch = "x86_64")),
+    not(all(feature = "arch", target_arch = "x86_64")),
     not(any(target_arch = "sparc", target_arch = "sparc64"))
 ))]
 impl_trifecta!(
@@ -165,7 +165,7 @@ fn u64_by_u64_div_rem(duo: u64, div: u64) -> (u64, u64) {
         target_family = "wasm",
         not(any(target_pointer_width = "16", target_pointer_width = "32")),
     )),
-    not(all(not(feature = "no-asm"), target_arch = "x86_64")),
+    not(all(feature = "arch", target_arch = "x86_64")),
     not(any(target_arch = "sparc", target_arch = "sparc64"))
 ))]
 impl_delegate!(
@@ -186,7 +186,7 @@ fn u64_by_u64_div_rem(duo: u64, div: u64) -> (u64, u64) {
 ///
 /// If the quotient does not fit in a `u64`, a floating point exception occurs.
 /// If `div == 0`, then a division by zero exception occurs.
-#[cfg(all(not(feature = "no-asm"), target_arch = "x86_64"))]
+#[cfg(all(feature = "arch", target_arch = "x86_64"))]
 #[inline]
 unsafe fn u128_by_u64_div_rem(duo: u128, div: u64) -> (u64, u64) {
     let duo_lo = duo as u64;
@@ -208,7 +208,7 @@ unsafe fn u128_by_u64_div_rem(duo: u128, div: u64) -> (u64, u64) {
 }
 
 // use `asymmetric` instead of `trifecta` on x86_64
-#[cfg(all(not(feature = "no-asm"), target_arch = "x86_64"))]
+#[cfg(all(feature = "arch", target_arch = "x86_64"))]
 impl_asymmetric!(
     u128_div_rem,
     zero_div_fn,
@@ -237,7 +237,7 @@ fn u32_by_u32_div_rem(duo: u32, div: u32) -> (u32, u32) {
 // When not on x86 and the pointer width is not 64, use `delegate` since the division size is larger
 // than register size.
 #[cfg(all(
-    not(all(not(feature = "no-asm"), target_arch = "x86")),
+    not(all(feature = "arch", target_arch = "x86")),
     not(target_pointer_width = "64")
 ))]
 impl_delegate!(
@@ -254,7 +254,7 @@ fn u32_by_u32_div_rem(duo: u32, div: u32) -> (u32, u32) {
 
 // When not on x86 and the pointer width is 64, use `binary_long`.
 #[cfg(all(
-    not(all(not(feature = "no-asm"), target_arch = "x86")),
+    not(all(feature = "arch", target_arch = "x86")),
     target_pointer_width = "64"
 ))]
 impl_binary_long!(
@@ -272,7 +272,7 @@ fn u32_by_u32_div_rem(duo: u32, div: u32) -> (u32, u32) {
 ///
 /// If the quotient does not fit in a `u32`, a floating point exception occurs.
 /// If `div == 0`, then a division by zero exception occurs.
-#[cfg(all(not(feature = "no-asm"), target_arch = "x86"))]
+#[cfg(all(feature = "arch", target_arch = "x86"))]
 #[inline]
 unsafe fn u64_by_u32_div_rem(duo: u64, div: u32) -> (u32, u32) {
     let duo_lo = duo as u32;
@@ -294,7 +294,7 @@ unsafe fn u64_by_u32_div_rem(duo: u64, div: u32) -> (u32, u32) {
 }
 
 // use `asymmetric` instead of `delegate` on x86
-#[cfg(all(not(feature = "no-asm"), target_arch = "x86"))]
+#[cfg(all(feature = "arch", target_arch = "x86"))]
 impl_asymmetric!(
     u64_div_rem,
     zero_div_fn,

library/compiler-builtins/compiler-builtins/src/mem/mod.rs

@@ -4,10 +4,7 @@
 #![allow(unsafe_op_in_unsafe_fn)]
 
 // memcpy/memmove/memset have optimized implementations on some architectures
-#[cfg_attr(
-    all(not(feature = "no-asm"), target_arch = "x86_64"),
-    path = "x86_64.rs"
-)]
+#[cfg_attr(all(feature = "arch", target_arch = "x86_64"), path = "x86_64.rs")]
 mod impls;
 
 intrinsics! {

library/compiler-builtins/libm-test/Cargo.toml

@@ -36,7 +36,9 @@ rand = { workspace = true, optional = true }
 libtest-mimic.workspace = true
 
 [features]
-default = ["build-mpfr", "unstable-float"]
+# Defaults should match the defaults in compiler-builtins since we have that
+# dependency with `default-features=false`.
+default = ["build-mpfr", "unstable-float", "compiler_builtins/arch"]
 
 # Propagated from libm because this affects which functions we test.
 unstable-float = ["libm/unstable-float", "rug?/nightly-float"]