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.

Sign up for GitHub

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

Add arm_intrinsics.json for intrinsic-test and stdarch-verify #1427

Merged
merged 2 commits into from
May 15, 2023
Merged

Add arm_intrinsics.json for intrinsic-test and stdarch-verify #1427

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6bb6f23
Remove ACLE submodule
adamgemmell Apr 11, 2023
dcc70ef
Use arm_intrinsics.json in stdarch-verify
adamgemmell Apr 20, 2023
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
3 changes: 0 additions & 3 deletions .gitmodules
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +0,0 @@
[submodule "crates/intrinsic-test/acle"]
path = crates/intrinsic-test/acle
url = https://github.com/ARM-software/acle.git
2 changes: 1 addition & 1 deletion ci/docker/aarch64-unknown-linux-gnu/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
qemu-user \
make \
file \
clang-13 \
clang-15 \
lld

ENV CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER=aarch64-linux-gnu-gcc \
Expand Down
2 changes: 1 addition & 1 deletion ci/docker/armv7-unknown-linux-gnueabihf/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
qemu-user \
make \
file \
clang-13 \
clang-15 \
lld
ENV CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_LINKER=arm-linux-gnueabihf-gcc \
CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_RUNNER="qemu-arm -L /usr/arm-linux-gnueabihf" \
Expand Down
4 changes: 2 additions & 2 deletions ci/run.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,10 +137,10 @@ esac

if [ "${TARGET}" = "aarch64-unknown-linux-gnu" ]; then
export CPPFLAGS="-fuse-ld=lld -I/usr/aarch64-linux-gnu/include/ -I/usr/aarch64-linux-gnu/include/c++/9/aarch64-linux-gnu/"
RUST_LOG=warn cargo run ${INTRINSIC_TEST} --release --bin intrinsic-test -- crates/intrinsic-test/acle/tools/intrinsic_db/advsimd.csv --runner "${CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_RUNNER}" --cppcompiler "clang++-13" --skip crates/intrinsic-test/missing_aarch64.txt
RUST_LOG=warn cargo run ${INTRINSIC_TEST} --release --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json --runner "${CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_RUNNER}" --cppcompiler "clang++-15" --skip crates/intrinsic-test/missing_aarch64.txt
elif [ "${TARGET}" = "armv7-unknown-linux-gnueabihf" ]; then
export CPPFLAGS="-fuse-ld=lld -I/usr/arm-linux-gnueabihf/include/ -I/usr/arm-linux-gnueabihf/include/c++/9/arm-linux-gnueabihf/"
RUST_LOG=warn cargo run ${INTRINSIC_TEST} --release --bin intrinsic-test -- crates/intrinsic-test/acle/tools/intrinsic_db/advsimd.csv --runner "${CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_RUNNER}" --cppcompiler "clang++-13" --skip crates/intrinsic-test/missing_arm.txt --a32
RUST_LOG=warn cargo run ${INTRINSIC_TEST} --release --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json --runner "${CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_RUNNER}" --cppcompiler "clang++-15" --skip crates/intrinsic-test/missing_arm.txt --a32
fi

if [ "$NORUN" != "1" ] && [ "$NOSTD" != 1 ]; then
Expand Down
7 changes: 5 additions & 2 deletions crates/intrinsic-test/Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,13 +1,16 @@
[package]
name = "intrinsic-test"
version = "0.1.0"
authors = ["Jamie Cunliffe <Jamie.Cunliffe@arm.com>"]
edition = "2021"
authors = ["Jamie Cunliffe <Jamie.Cunliffe@arm.com>",
"James McGregor <James.McGregor2@arm.com",
"Adam Gemmell <Adam.Gemmell@arm.com"]
license = "MIT OR Apache-2.0"
edition = "2021"

[dependencies]
lazy_static = "1.4.0"
serde = { version = "1", features = ["derive"] }
serde_json = "1.0"
csv = "1.1"
clap = "2.33.3"
regex = "1.4.2"
Expand Down
1 change: 0 additions & 1 deletion crates/intrinsic-test/acle
Open in desktop
Submodule acle deleted from 5626f8
72 changes: 13 additions & 59 deletions crates/intrinsic-test/missing_aarch64.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,39 +23,6 @@ vusdotq_lane_s32
vusdotq_s32
vusdot_s32

# Implemented in Clang but missing from CSV
vcmla_f64
vcmla_lane_f64
vcmla_laneq_f64
vcmlaq_lane_f64
vcmlaq_laneq_f64
vcmlaq_rot180_lane_f64
vcmlaq_rot180_laneq_f64
vcmlaq_rot270_lane_f64
vcmlaq_rot270_laneq_f64
vcmlaq_rot90_lane_f64
vcmlaq_rot90_laneq_f64
vcmla_rot180_f64
vcmla_rot180_lane_f64
vcmla_rot180_laneq_f64
vcmla_rot270_f64
vcmla_rot270_lane_f64
vcmla_rot270_laneq_f64
vcmla_rot90_f64
vcmla_rot90_lane_f64
vcmla_rot90_laneq_f64

# Implemented in Clang and stdarch but missing from CSV
vmov_n_p64
vmovq_n_p64
vreinterpret_f32_p64
vreinterpret_p64_s64
vreinterpretq_f32_p128
vreinterpretq_f32_p64
vreinterpretq_p128_p64
vreinterpretq_p64_p128
vtst_p16
vtstq_p16

# Missing from both Clang and stdarch
vrnd32x_f64
Expand All @@ -67,30 +34,17 @@ vrnd64xq_f64
vrnd64z_f64
vrnd64zq_f64

# QEMU 6.0 doesn't support these instructions
vmmlaq_s32
vmmlaq_u32
vsm3partw1q_u32
vsm3partw2q_u32
vsm3ss1q_u32
vsm3tt1aq_u32
vsm3tt1bq_u32
vsm3tt2aq_u32
vsm3tt2bq_u32
vsm4ekeyq_u32
vsm4eq_u32
vusmmlaq_s32

# LLVM select error in debug builds
vqshlu_n_s16
vqshlu_n_s32
vqshlu_n_s64
vqshlu_n_s8
vqshlub_n_s8
vqshlud_n_s64
vqshluh_n_s16
vqshluq_n_s16
vqshluq_n_s32
vqshluq_n_s64
vqshluq_n_s8
vqshlus_n_s32
#vqshlu_n_s16
#vqshlu_n_s32
#vqshlu_n_s64
#vqshlu_n_s8
#vqshlub_n_s8
#vqshlud_n_s64
#vqshluh_n_s16
#vqshluq_n_s16
#vqshluq_n_s32
#vqshluq_n_s64
#vqshluq_n_s8
#vqshlus_n_s32
Copy link
Member

Choose a reason for hiding this comment

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

My understanding is that this issue is still present if intrinsic-test is changed to run the Rust code without --release. The problem is due to the way we generate LLVM IR: we emit the vector argument as a constant and then load it, but LLVM's intrinsic only works if the vector is passed as an immediate to the intrinsics.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, it's still there - hence why I've only commented these intrinsics out instead of removing them. We have a ticket to fix this in the future. The test tool currently runs in --release mode and I think it makes sense to extend coverage by including this set.


142 changes: 26 additions & 116 deletions crates/intrinsic-test/missing_arm.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,15 +23,6 @@ vusdotq_lane_s32
vusdotq_s32
vusdot_s32

# Implemented in Clang and stdarch but missing from CSV
vtst_p16
vtstq_p16

# QEMU 6.0 doesn't support these instructions
vmmlaq_s32
vmmlaq_u32
vusmmlaq_s32

# Implemented in Clang and stdarch for A64 only even though CSV claims A32 support
__crc32d
__crc32cd
Expand Down Expand Up @@ -214,110 +205,29 @@ vrndx_f32
vrndxq_f32

# LLVM select error in debug builds
vqrshrn_n_s16
vqrshrn_n_s32
vqrshrn_n_s64
vqrshrn_n_u16
vqrshrn_n_u32
vqrshrn_n_u64
vqrshrun_n_s16
vqrshrun_n_s32
vqrshrun_n_s64
vqshrn_n_s16
vqshrn_n_s32
vqshrn_n_s64
vqshrn_n_u16
vqshrn_n_u32
vqshrn_n_u64
vqshrun_n_s16
vqshrun_n_s32
vqshrun_n_s64
vrshrn_n_s16
vrshrn_n_s32
vrshrn_n_s64
vrshrn_n_u16
vrshrn_n_u32
vrshrn_n_u64
vshrq_n_u64
vshr_n_u64

# Failing tests: stdarch has incorrect results compared to Clang
vqshlu_n_s16
vqshlu_n_s32
vqshlu_n_s64
vqshlu_n_s8
vqshluq_n_s16
vqshluq_n_s32
vqshluq_n_s64
vqshluq_n_s8
vsli_n_p16
vsli_n_p8
vsli_n_s16
vsli_n_s32
vsli_n_s64
vsli_n_s8
vsli_n_u16
vsli_n_u32
vsli_n_u64
vsli_n_u8
vsliq_n_p16
vsliq_n_p8
vsliq_n_s16
vsliq_n_s32
vsliq_n_s64
vsliq_n_s8
vsliq_n_u16
vsliq_n_u32
vsliq_n_u64
vsliq_n_u8
vsri_n_p16
vsri_n_p8
vsri_n_s16
vsri_n_s32
vsri_n_s64
vsri_n_s8
vsri_n_u16
vsri_n_u32
vsri_n_u64
vsri_n_u8
vsriq_n_p16
vsriq_n_p8
vsriq_n_s16
vsriq_n_s32
vsriq_n_s64
vsriq_n_s8
vsriq_n_u16
vsriq_n_u32
vsriq_n_u64
vsriq_n_u8

# These produce a different result on Clang depending on the optimization level.
# This is definitely a bug in LLVM.
vadd_f32
vaddq_f32
vcvt_s32_f32
vcvt_u32_f32
vcvtq_s32_f32
vcvtq_u32_f32
vfma_f32
vfma_n_f32
vfmaq_f32
vfmaq_n_f32
vfms_f32
vfmsq_f32
vmla_f32
vmla_lane_f32
vmla_n_f32
vmlaq_f32
vmlaq_lane_f32
vmlaq_n_f32
vmls_f32
vmls_lane_f32
vmls_n_f32
vmlsq_f32
vmlsq_lane_f32
vmlsq_n_f32
vmul_lane_f32
vmul_n_f32
vmulq_lane_f32
vmulq_n_f32
Copy link
Member

Choose a reason for hiding this comment

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

Have you checked this by trying the C version without the -O2 flag?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I have, everything passes on aarch64 and armv7 with Rust --release and clang -O0

#vqrshrn_n_s16
#vqrshrn_n_s32
#vqrshrn_n_s64
#vqrshrn_n_u16
#vqrshrn_n_u32
#vqrshrn_n_u64
#vqrshrun_n_s16
#vqrshrun_n_s32
#vqrshrun_n_s64
#vqshrn_n_s16
#vqshrn_n_s32
#vqshrn_n_s64
#vqshrn_n_u16
#vqshrn_n_u32
#vqshrn_n_u64
#vqshrun_n_s16
#vqshrun_n_s32
#vqshrun_n_s64
#vrshrn_n_s16
#vrshrn_n_s32
#vrshrn_n_s64
#vrshrn_n_u16
#vrshrn_n_u32
#vrshrn_n_u64
#vshrq_n_u64
#vshr_n_u64
Loading