Intrinsics, guarded by __ARM_FEATURE_SVE_MATMUL_INT8:
- svusdot[_s32]
- svusdot[_n_s32]
- svusdot_lane[_s32]
- svsudot[_s32]
- svsudot[_n_s32]
- svsudot_lane[_s32]
Paths
| Differential D79877
[clang][SveEmitter] SVE builtins for `svusdot` and `svsudot` ACLE. ClosedPublic Authored by fpetrogalli on May 13 2020, 9:20 AM.
Details Summary Intrinsics, guarded by __ARM_FEATURE_SVE_MATMUL_INT8:
Diff Detail
Event Timelinefpetrogalli added a parent revision: D79639: [SveEmitter] Builtins for SVE matrix multiply `mmla`..May 13 2020, 9:38 AM Comment Actions LGTM with one small comment.
This revision is now accepted and ready to land.May 13 2020, 11:21 AM fpetrogalli added inline comments.
fpetrogalli added inline comments. Closed by commit rGb593bfd4d8e3: [clang][SveEmitter] SVE builtins for `svusdot` and `svsudot` ACLE. (authored by fpetrogalli). · Explain WhyMay 18 2020, 4:17 PM This revision was automatically updated to reflect the committed changes.
Revision Contents
Diff 264750 clang/include/clang/Basic/TargetBuiltins.h
clang/include/clang/Basic/arm_sve.td
clang/lib/CodeGen/CGBuiltin.cpp
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_sudot.c
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_usdot.c
clang/utils/TableGen/SveEmitter.cpp
|
These intrinsics aren't overloaded; maybe consider writing the actual type int8_t/uint8_t, instead of introducing "unsigned scalar of 1/4 width element type"? If there's some reason the current form is better, that's fine, though.