Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implementation for Aarch64 TME intrinsics #855

Merged
merged 3 commits into from
May 29, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ci/docker/aarch64-unknown-linux-gnu/Dockerfile
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
FROM ubuntu:18.04
FROM ubuntu:20.04
RUN apt-get update && apt-get install -y --no-install-recommends \
gcc \
ca-certificates \
Expand Down
3 changes: 3 additions & 0 deletions crates/core_arch/src/aarch64/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,9 @@ pub use self::neon::*;
mod crypto;
pub use self::crypto::*;

mod tme;
pub use self::tme::*;

mod crc;
pub use self::crc::*;

Expand Down
183 changes: 183 additions & 0 deletions crates/core_arch/src/aarch64/tme.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,183 @@
//! ARM's Transactional Memory Extensions (TME).
//!
//! This CPU feature is available on Aarch64 - A architecture profile.
//! This feature is in the non-neon feature set. TME specific vendor documentation can
//! be found [TME Intrinsics Introduction][tme_intrinsics_intro].
//!
//! The reference is [ACLE Q4 2019][acle_q4_2019_ref].
//!
//! ACLE has a section for TME extensions and state masks for aborts and failure codes.
//! [ARM A64 Architecture Register Datasheet][a_profile_future] also describes possible failure code scenarios.
//!
//! [acle_q4_2019_ref]: https://static.docs.arm.com/101028/0010/ACLE_2019Q4_release-0010.pdf
//! [tme_intrinsics_intro]: https://developer.arm.com/docs/101028/0010/transactional-memory-extension-tme-intrinsics
//! [llvm_aarch64_int]: https://github.com/llvm/llvm-project/commit/a36d31478c182903523e04eb271bbf102bfab2cc#diff-ff24e1c35f4d54f1110ce5d90c709319R626-R646
//! [a_profile_future]: https://static.docs.arm.com/ddi0601/a/SysReg_xml_futureA-2019-04.pdf?_ga=2.116560387.441514988.1590524918-1110153136.1588469296

#[cfg(test)]
use stdarch_test::assert_instr;

extern "C" {
#[link_name = "llvm.aarch64.tstart"]
fn aarch64_tstart() -> u64;
#[link_name = "llvm.aarch64.tcommit"]
fn aarch64_tcommit() -> ();
#[link_name = "llvm.aarch64.tcancel"]
fn aarch64_tcancel(imm0: u64) -> ();
#[link_name = "llvm.aarch64.ttest"]
fn aarch64_ttest() -> u64;
}

/// Transaction successfully started.
pub const _TMSTART_SUCCESS: u64 = 0x00_u64;

/// Extraction mask for failure reason
pub const _TMFAILURE_REASON: u64 = 0x00007FFF_u64;

/// Transaction retry is possible.
pub const _TMFAILURE_RTRY: u64 = 1 << 15;

/// Transaction executed a TCANCEL instruction
pub const _TMFAILURE_CNCL: u64 = 1 << 16;

/// Transaction aborted because a conflict occurred
pub const _TMFAILURE_MEM: u64 = 1 << 17;

/// Fallback error type for any other reason
pub const _TMFAILURE_IMP: u64 = 1 << 18;

/// Transaction aborted because a non-permissible operation was attempted
pub const _TMFAILURE_ERR: u64 = 1 << 19;

/// Transaction aborted due to read or write set limit was exceeded
pub const _TMFAILURE_SIZE: u64 = 1 << 20;

/// Transaction aborted due to transactional nesting level was exceeded
pub const _TMFAILURE_NEST: u64 = 1 << 21;

/// Transaction aborted due to a debug trap.
pub const _TMFAILURE_DBG: u64 = 1 << 22;

/// Transaction failed from interrupt
pub const _TMFAILURE_INT: u64 = 1 << 23;

/// Indicates a TRIVIAL version of TM is available
pub const _TMFAILURE_TRIVIAL: u64 = 1 << 24;

/// Starts a new transaction. When the transaction starts successfully the return value is 0.
/// If the transaction fails, all state modifications are discarded and a cause of the failure
/// is encoded in the return value.
///
/// [ARM TME Intrinsics](https://developer.arm.com/docs/101028/0010/transactional-memory-extension-tme-intrinsics).
#[inline]
#[target_feature(enable = "tme")]
#[cfg_attr(test, assert_instr(tstart))]
pub unsafe fn __tstart() -> u64 {
aarch64_tstart()
}

/// Commits the current transaction. For a nested transaction, the only effect is that the
/// transactional nesting depth is decreased. For an outer transaction, the state modifications
/// performed transactionally are committed to the architectural state.
///
/// [ARM TME Intrinsics](https://developer.arm.com/docs/101028/0010/transactional-memory-extension-tme-intrinsics).
#[inline]
#[target_feature(enable = "tme")]
#[cfg_attr(test, assert_instr(tcommit))]
pub unsafe fn __tcommit() {
aarch64_tcommit()
}

/// Cancels the current transaction and discards all state modifications that were performed transactionally.
///
/// [ARM TME Intrinsics](https://developer.arm.com/docs/101028/0010/transactional-memory-extension-tme-intrinsics).
#[inline]
#[target_feature(enable = "tme")]
#[cfg_attr(test, assert_instr(tcancel, imm0 = 0x0))]
#[rustc_args_required_const(0)]
pub unsafe fn __tcancel(imm0: u64) {
macro_rules! call {
($imm0:expr) => {
aarch64_tcancel($imm0)
};
}
constify_imm8!(imm0, call)
}

/// Tests if executing inside a transaction. If no transaction is currently executing,
/// the return value is 0. Otherwise, this intrinsic returns the depth of the transaction.
///
/// [ARM TME Intrinsics](https://developer.arm.com/docs/101028/0010/transactional-memory-extension-tme-intrinsics).
#[inline]
#[target_feature(enable = "tme")]
#[cfg_attr(test, assert_instr(ttest))]
pub unsafe fn __ttest() -> u64 {
aarch64_ttest()
}

#[cfg(test)]
mod tests {
use stdarch_test::simd_test;

use crate::core_arch::aarch64::*;

const CANCEL_CODE: u64 = (0 | (0x123 & _TMFAILURE_REASON) as u64) as u64;

#[simd_test(enable = "tme")]
unsafe fn test_tstart() {
let mut x = 0;
for i in 0..10 {
let code = tme::__tstart();
if code == _TMSTART_SUCCESS {
x += 1;
assert_eq!(x, i + 1);
break;
}
assert_eq!(x, 0);
}
}

#[simd_test(enable = "tme")]
unsafe fn test_tcommit() {
let mut x = 0;
for i in 0..10 {
let code = tme::__tstart();
if code == _TMSTART_SUCCESS {
x += 1;
assert_eq!(x, i + 1);
tme::__tcommit();
}
assert_eq!(x, i + 1);
}
}

#[simd_test(enable = "tme")]
unsafe fn test_tcancel() {
let mut x = 0;

for i in 0..10 {
let code = tme::__tstart();
if code == _TMSTART_SUCCESS {
x += 1;
assert_eq!(x, i + 1);
tme::__tcancel(CANCEL_CODE);
break;
}
}

assert_eq!(x, 0);
}

#[simd_test(enable = "tme")]
unsafe fn test_ttest() {
for _ in 0..10 {
let code = tme::__tstart();
if code == _TMSTART_SUCCESS {
if tme::__ttest() == 2 {
tme::__tcancel(CANCEL_CODE);
break;
}
}
}
}
}
5 changes: 4 additions & 1 deletion crates/simd-test-macro/src/lib.rs
Original file line number Diff line number Diff line change
Expand Up @@ -156,5 +156,8 @@ fn find_name(item: TokenStream) -> Ident {
}
}

tokens.next().and_then(get_ident).expect("failed to find function name")
tokens
.next()
.and_then(get_ident)
.expect("failed to find function name")
}
2 changes: 2 additions & 0 deletions crates/std_detect/src/detect/arch/aarch64.rs
Original file line number Diff line number Diff line change
Expand Up @@ -33,4 +33,6 @@ features! {
/// Release consistent Processor consistent (RcPc)
@FEATURE: #[unstable(feature = "stdsimd", issue = "27731")] dotprod: "dotprod";
/// Vector Dot-Product (ASIMDDP)
@FEATURE: #[unstable(feature = "stdsimd", issue = "27731")] tme: "tme";
/// Transactional Memory Extensions (TME)
}
1 change: 1 addition & 0 deletions crates/std_detect/src/detect/os/aarch64.rs
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ pub(crate) fn detect_features() -> cache::Initializer {
enable_feature(Feature::pmull, pmull);
// Crypto is specified as AES + PMULL + SHA1 + SHA2 per LLVM/hosts.cpp
enable_feature(Feature::crypto, aes && pmull && sha1 && sha2);
enable_feature(Feature::tme, bits_shift(aa64isar0, 27, 24) == 1);
enable_feature(Feature::lse, bits_shift(aa64isar0, 23, 20) >= 1);
enable_feature(Feature::crc, bits_shift(aa64isar0, 19, 16) >= 1);

Expand Down
1 change: 1 addition & 0 deletions crates/std_detect/src/detect/os/freebsd/aarch64.rs
Original file line number Diff line number Diff line change
Expand Up @@ -17,5 +17,6 @@ mod tests {
println!("rdm: {:?}", is_aarch64_feature_detected!("rdm"));
println!("rcpc: {:?}", is_aarch64_feature_detected!("rcpc"));
println!("dotprod: {:?}", is_aarch64_feature_detected!("dotprod"));
println!("tme: {:?}", is_aarch64_feature_detected!("tme"));
}
}
1 change: 1 addition & 0 deletions crates/std_detect/tests/cpu-detection.rs
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,7 @@ fn aarch64_linux() {
println!("rdm: {}", is_aarch64_feature_detected!("rdm"));
println!("rcpc: {}", is_aarch64_feature_detected!("rcpc"));
println!("dotprod: {}", is_aarch64_feature_detected!("dotprod"));
println!("tme: {}", is_aarch64_feature_detected!("tme"));
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You need to implement detection for this feature. Do a search for dotprod to see how it is done.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updated with feature detection, didn't find the feature tbh. Even instructions and signatures are the same. llvm/llvm-project@a36d314#diff-3674dac5c4241890431fece8594c3143R25-R34

What am I doing wrong? (I still suspect from qemu, checked source code, it doesn't exist there yet. Might be wrong.)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The current feature detection is fine. Linux doesn't have the TME flag in /proc/cpuinfo or AT_HWCAPS yet, so there's nothing we can do about that for now.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, when it enters to kernel I will add here.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Addressed your comments. Thanks for the help! @Amanieu

}

#[test]
Expand Down
1 change: 1 addition & 0 deletions crates/stdarch-verify/tests/arm.rs
Original file line number Diff line number Diff line change
Expand Up @@ -366,6 +366,7 @@ fn verify_all_signatures() {
&& !rust.file.ends_with("v6.rs\"")
&& !rust.file.ends_with("v7.rs\"")
&& !rust.file.ends_with("v8.rs\"")
&& !rust.file.ends_with("tme.rs\"")
{
println!(
"missing arm definition for {:?} in {}",
Expand Down