Implement RISC-V Zb intrinsics

ci/docker/riscv64gc-unknown-linux-gnu/Dockerfile

@@ -5,5 +5,5 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
         gcc-riscv64-linux-gnu libc6-dev-riscv64-cross
 
 ENV CARGO_TARGET_RISCV64GC_UNKNOWN_LINUX_GNU_LINKER=riscv64-linux-gnu-gcc \
-    CARGO_TARGET_RISCV64GC_UNKNOWN_LINUX_GNU_RUNNER="qemu-riscv64 -L /usr/riscv64-linux-gnu -cpu rv64,zk=true" \
+    CARGO_TARGET_RISCV64GC_UNKNOWN_LINUX_GNU_RUNNER="qemu-riscv64 -L /usr/riscv64-linux-gnu -cpu rv64,zk=true,zbb=true,zbc=true" \
     OBJDUMP=riscv64-linux-gnu-objdump

ci/run.sh

@@ -47,7 +47,7 @@ case ${TARGET} in
     # Some of our test dependencies use the deprecated `gcc` crates which
     # doesn't detect RISC-V compilers automatically, so do it manually here.
     riscv64*)
-        export RUSTFLAGS="${RUSTFLAGS} -Ctarget-feature=+zk"
+        export RUSTFLAGS="${RUSTFLAGS} -Ctarget-feature=+zk,+zbb,+zbc"
         export TARGET_CC="riscv64-linux-gnu-gcc"
         ;;
 esac

crates/core_arch/src/riscv_shared/mod.rs

@@ -1,10 +1,12 @@
 //! Shared RISC-V intrinsics
 
 mod p;
+mod zb;
 mod zk;
 
 #[unstable(feature = "stdsimd", issue = "27731")]
 pub use p::*;
+pub use zb::*;
 pub use zk::*;
 
 use crate::arch::asm;

crates/core_arch/src/riscv_shared/zb.rs

@@ -0,0 +1,150 @@
+#[cfg(test)]
+use stdarch_test::assert_instr;
+
+#[cfg(target_arch = "riscv32")]
+extern "unadjusted" {
+    #[link_name = "llvm.riscv.orc.b.i32"]
+    fn _orc_b_32(rs: i32) -> i32;
+
+    #[link_name = "llvm.riscv.clmul.i32"]
+    fn _clmul_32(rs1: i32, rs2: i32) -> i32;
+
+    #[link_name = "llvm.riscv.clmulh.i32"]
+    fn _clmulh_32(rs1: i32, rs2: i32) -> i32;
+
+    #[link_name = "llvm.riscv.clmulr.i32"]
+    fn _clmulr_32(rs1: i32, rs2: i32) -> i32;
+}
+
+#[cfg(target_arch = "riscv64")]
+extern "unadjusted" {
+    #[link_name = "llvm.riscv.orc.b.i64"]
+    fn _orc_b_64(rs1: i64) -> i64;
+
+    #[link_name = "llvm.riscv.clmul.i64"]
+    fn _clmul_64(rs1: i64, rs2: i64) -> i64;
+
+    #[link_name = "llvm.riscv.clmulh.i64"]
+    fn _clmulh_64(rs1: i64, rs2: i64) -> i64;
+
+    #[link_name = "llvm.riscv.clmulr.i64"]
+    fn _clmulr_64(rs1: i64, rs2: i64) -> i64;
+}
+
+/// Bitwise OR-Combine, byte granule
+///
+/// Combines the bits within every byte through a reciprocal bitwise logical OR. This sets the bits of each byte in
+/// the result rd to all zeros if no bit within the respective byte of rs is set, or to all ones if any bit within the
+/// respective byte of rs is set.
+///
+/// Source: RISC-V Bit-Manipulation ISA-extensions
+///
+/// Version: v1.0.0
+///
+/// Section: 2.24
+///
+/// # Safety
+///
+/// This function is safe to use if the `zbb` target feature is present.
+#[target_feature(enable = "zbb")]
+// See #1464
+// #[cfg_attr(test, assert_instr(orc.b))]
+#[inline]
+pub unsafe fn orc_b(rs: usize) -> usize {
+    #[cfg(target_arch = "riscv32")]
+    {
+        _orc_b_32(rs as i32) as usize
+    }
+
+    #[cfg(target_arch = "riscv64")]
+    {
+        _orc_b_64(rs as i64) as usize
+    }
+}
+
+/// Carry-less multiply (low-part)
+///
+/// clmul produces the lower half of the 2·XLEN carry-less product.
+///
+/// Source: RISC-V Bit-Manipulation ISA-extensions
+///
+/// Version: v1.0.0
+///
+/// Section: 2.11
+///
+/// # Safety
+///
+/// This function is safe to use if the `zbc` target feature is present.
+#[target_feature(enable = "zbc")]
+// See #1464
+// #[cfg_attr(test, assert_instr(clmul))]
+#[inline]
+pub unsafe fn clmul(rs1: usize, rs2: usize) -> usize {
+    #[cfg(target_arch = "riscv32")]
+    {
+        _clmul_32(rs1 as i32, rs2 as i32) as usize
+    }
+
+    #[cfg(target_arch = "riscv64")]
+    {
+        _clmul_64(rs1 as i64, rs2 as i64) as usize
+    }
+}
+
+/// Carry-less multiply (high-part)
+///
+/// clmulh produces the upper half of the 2·XLEN carry-less product.
+///
+/// Source: RISC-V Bit-Manipulation ISA-extensions
+///
+/// Version: v1.0.0
+///
+/// Section: 2.12
+///
+/// # Safety
+///
+/// This function is safe to use if the `zbc` target feature is present.
+#[target_feature(enable = "zbc")]
+// See #1464
+// #[cfg_attr(test, assert_instr(clmulh))]
+#[inline]
+pub unsafe fn clmulh(rs1: usize, rs2: usize) -> usize {
+    #[cfg(target_arch = "riscv32")]
+    {
+        _clmulh_32(rs1 as i32, rs2 as i32) as usize
+    }
+
+    #[cfg(target_arch = "riscv64")]
+    {
+        _clmulh_64(rs1 as i64, rs2 as i64) as usize
+    }
+}
+
+/// Carry-less multiply (reversed)
+///
+/// clmulr produces bits 2·XLEN−2:XLEN-1 of the 2·XLEN carry-less product.
+///
+/// Source: RISC-V Bit-Manipulation ISA-extensions
+///
+/// Version: v1.0.0
+///
+/// Section: 2.13
+///
+/// # Safety
+///
+/// This function is safe to use if the `zbc` target feature is present.
+#[target_feature(enable = "zbc")]
+// See #1464
+// #[cfg_attr(test, assert_instr(clmulr))]
+#[inline]
+pub unsafe fn clmulr(rs1: usize, rs2: usize) -> usize {
+    #[cfg(target_arch = "riscv32")]
+    {
+        _clmulr_32(rs1 as i32, rs2 as i32) as usize
+    }
+
+    #[cfg(target_arch = "riscv64")]
+    {
+        _clmulr_64(rs1 as i64, rs2 as i64) as usize
+    }
+}