Description
Feature gate: #![feature(stdarch_s390x)]
This is a tracking issue for the s390x (aka SystemZ) intrinsics in core::arch::s390x.
Public API
Everything in core::arch::s390x.
Missing instructions
based on the clang vecintrin.h (roungly similar to the linkable GCC vecintrin.h).
vec_any_nges390x: final batch of intrinsics stdarch#1743vec_any_ngts390x: final batch of intrinsics stdarch#1743vec_any_nles390x: final batch of intrinsics stdarch#1743vec_any_nlts390x: final batch of intrinsics stdarch#1743vec_all_nges390x: final batch of intrinsics stdarch#1743vec_all_ngts390x: final batch of intrinsics stdarch#1743vec_all_nles390x: final batch of intrinsics stdarch#1743vec_all_nlts390x: final batch of intrinsics stdarch#1743vec_all_nans390x: final batch of intrinsics stdarch#1743vec_all_numerics390x: final batch of intrinsics stdarch#1743vec_any_nans390x: final batch of intrinsics stdarch#1743vec_any_numerics390x: final batch of intrinsics stdarch#1743vec_genmasks390x: add more intrinsics stdarch#1728vec_genmasks_8s390x: add more intrinsics stdarch#1728vec_genmasks_16s390x: add more intrinsics stdarch#1728vec_genmasks_32s390x: add more intrinsics stdarch#1728vec_genmasks_64s390x: add more intrinsics stdarch#1728vec_splat_u8s390x: add more intrinsics stdarch#1728vec_splat_s8s390x: add more intrinsics stdarch#1728vec_splat_u16s390x: add more intrinsics stdarch#1728vec_splat_s16s390x: add more intrinsics stdarch#1728vec_splat_u32s390x: add more intrinsics stdarch#1728vec_splat_s32s390x: add more intrinsics stdarch#1728vec_splat_u64s390x: add more intrinsics stdarch#1728vec_splat_s64s390x: add more intrinsics stdarch#1728vec_checksums390x: another batch of intrinsics stdarch#1738vec_gfmsum_128s390x: another batch of intrinsics stdarch#1738vec_gfmsum_accum_128s390x: another batch of intrinsics stdarch#1738vec_ceilS390x float rounding stdarch#1712vec_roundpS390x float rounding stdarch#1712vec_floorS390x float rounding stdarch#1712vec_roundmS390x float rounding stdarch#1712vec_truncS390x float rounding stdarch#1712vec_roundzS390x float rounding stdarch#1712vec_rintS390x float rounding stdarch#1712vec_roundcS390x float rounding stdarch#1712vec_roundS390x float rounding stdarch#1712vec_doublees390x: final batch of intrinsics stdarch#1743vec_add_u128s390x: another batch of intrinsics stdarch#1738vec_addc_u128s390x: another batch of intrinsics stdarch#1738vec_adde_u128s390x: another batch of intrinsics stdarch#1738vec_addec_u128s390x: another batch of intrinsics stdarch#1738vec_bperm_u128s390x: final batch of intrinsics stdarch#1743vec_cmpeq_idxs390x: final batch of intrinsics stdarch#1743vec_cmpeq_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmpeq_or_0_idxs390x: final batch of intrinsics stdarch#1743vec_cmpeq_or_0_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmpne_idxs390x: final batch of intrinsics stdarch#1743vec_cmpne_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmpne_or_0_idxs390x: final batch of intrinsics stdarch#1743vec_cmpne_or_0_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmpnrg_ccs390x: final batch of intrinsics stdarch#1743vec_cmpnrg_idxs390x: final batch of intrinsics stdarch#1743vec_cmpnrg_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmpnrg_or_0_idxs390x: final batch of intrinsics stdarch#1743vec_cmpnrg_or_0_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmprg_ccs390x: final batch of intrinsics stdarch#1743vec_cmprg_idxs390x: final batch of intrinsics stdarch#1743vec_cmprg_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cmprg_or_0_idxs390x: final batch of intrinsics stdarch#1743vec_cmprg_or_0_idx_ccs390x: final batch of intrinsics stdarch#1743vec_cp_until_zeros390x: final batch of intrinsics stdarch#1743vec_cp_until_zero_ccs390x: final batch of intrinsics stdarch#1743vec_extend_s64s390x: final batch of intrinsics stdarch#1743vec_find_any_eqs390x: add more intrinsics stdarch#1728vec_find_any_eq_ccs390x: add more intrinsics stdarch#1728vec_find_any_eq_idxs390x: add more intrinsics stdarch#1728vec_find_any_eq_idx_ccs390x: add more intrinsics stdarch#1728vec_find_any_eq_or_0_idxs390x: add more intrinsics stdarch#1728vec_find_any_eq_or_0_idx_ccs390x: add more intrinsics stdarch#1728vec_find_any_nes390x: add more intrinsics stdarch#1728vec_find_any_ne_ccs390x: add more intrinsics stdarch#1728vec_find_any_ne_idxs390x: add more intrinsics stdarch#1728vec_find_any_ne_idx_ccs390x: add more intrinsics stdarch#1728vec_find_any_ne_or_0_idxs390x: add more intrinsics stdarch#1728vec_find_any_ne_or_0_idx_ccs390x: add more intrinsics stdarch#1728vec_fp_test_data_classs390x: final batch of intrinsics stdarch#1743vec_gather_elements390x: final batch of intrinsics stdarch#1743vec_gfmsum_accums390x: another batch of intrinsics stdarch#1738vec_load_bndrys390x: another batch of intrinsics stdarch#1738vec_load_lens390x: another batch of intrinsics stdarch#1738vec_load_len_rs390x: another batch of intrinsics stdarch#1738vec_load_pairs390x: another batch of intrinsics stdarch#1738vec_mergehs390x: add more intrinsics stdarch#1728vec_mergels390x: add more intrinsics stdarch#1728vec_msum_u128s390x: final batch of intrinsics stdarch#1743vec_packs_ccs390x: another batch of intrinsics stdarch#1738vec_packsu_ccs390x: another batch of intrinsics stdarch#1738vec_popcntS390x vector bitwise operations stdarch#1709vec_rl_masks390x: add more intrinsics stdarch#1728vec_scatter_elements390x: final batch of intrinsics stdarch#1743vec_search_string_ccs390x: final batch of intrinsics stdarch#1743vec_search_string_until_zero_ccs390x: final batch of intrinsics stdarch#1743vec_splats390x: add more intrinsics stdarch#1728vec_store_lens390x: another batch of intrinsics stdarch#1738vec_store_len_rs390x: another batch of intrinsics stdarch#1738vec_sub_u128s390x: add more intrinsics stdarch#1728vec_subc_u128s390x: add more intrinsics stdarch#1728vec_sube_u128s390x: add more intrinsics stdarch#1728vec_subec_u128s390x: add more intrinsics stdarch#1728vec_sum_u128s390x: add more intrinsics stdarch#1728vec_test_masks390x: final batch of intrinsics stdarch#1743vec_unpackhs390x: another batch of intrinsics stdarch#1738vec_unpackls390x: another batch of intrinsics stdarch#1738vec_unsigneds390x: final batch of intrinsics stdarch#1743
blocked on #137447
vec_insert_and_zeroaddvec_extract,vec_insert,vec_promoteandvec_insert_and_zerostdarch#1772
from nnp-assist, current qemu traps on these
vec_extend_to_fp32_hivec_extend_to_fp32_lovec_round_from_fp32vec_convert_to_fp16vec_convert_from_fp16
deprecated functions
vec_ctdvec_ctd_s64vec_ctd_u64vec_ctslvec_ctulvec_ld2fvec_st2fvec_xstd2vec_xstw4vec_xld2vec_xlw4vec_permi
Steps / History
Unresolved Questions
- None yet.
@rustbot label O-SystemZ
general s390x vector/intrinsics progress is tracked at #130869
cc @taiki-e
Footnotes
Metadata
Metadata
Assignees
Labels
Type
Projects
Milestone
Relationships
Development
Participants
Activity
vec_addfors390xrust-lang/stdarch#1703s390x: addvec_sub,vec_mul,vec_min,vec_max,vec_absandvec_splatsrust-lang/stdarch#1704vec_extract,vec_insert,vec_promoteandvec_insert_and_zerorust-lang/stdarch#1772uweigand commentedon May 16, 2025
The new machines IBM z17 and IBM LinuxONE Emperor 5 were recently announced. These machines implement the
arch15level of the z/Architecture. Support for this has been added to LLVM here: llvm/llvm-project@8424bf2Support for the new architecture level also comes with a new revision of the vector intrinsics (implemented across GCC, LLVM, and the IBM compilers). It would be good to update the Rust implementation to match.
The new
vecintrin.hfile can be seen e.g. here: https://github.com/llvm/llvm-project/blob/8424bf207efd89eacf2fe893b67be98d535e1db6/clang/lib/Headers/vecintrin.h This implements the following set of changes compared to the previous version:Generic cleanup
While reviewing the new changes, we noticed a number of inconsistencies and deficiencies in the existing intrinsics, which were cleaned up as part of the new revision. Specifically:
vec_and,vec_or, andvec_xorintrinsics. These are mostly redundant with the&,|, and^operators, but can also be used with floating-point vector arguments.vector unsigned char. Also, the operand to be shifted should not be of anyvector booltype since the result may not necessarily be a valid bool vector value. Added the following intrinsics:and deprecated those intrinsics:
vec_load_len/vec_store_lenandvec_load_len_r/vec_store_len_r. Both now support onlyvector signed charandvector unsigned char. Added intrinsics:and deprecated the existing intrinsics:
Support for 128-bit integer vector types
One main feature of the
arch15ISA is support for a full set of arithmetical operations on 128-bit integer values held in vector registers. This is used to a large extent implicitly by the code-generator back end. However, there are also a number of operations that required intrinsics to fully exploit. We decided to add the following new vector types to be used with those intrinsics:Note that since the vector length is only 128 bits, these vector types only contain a single element. They are still useful as they use a different ABI (passed in vector registers and not in memory), and it seems cleaner to consistently use "vector" types with the vector intrinsics.
Note that many operations on these types can actually be performed with prior versions of the ISA, so the types have been made available unconditionally. Many existing intrinsics have been extended to support the new types:
Some other intrinsics also now support the new types, but only when the
vector-enhancements-3feature is present:Finally, a number of existing intrinsics already operated on 128-bit integer types, but used
vector unsigned charto represent those values in the absence of a better type. These have now all been deprecated:and replaced by the following new intrinsics:
and new overloads of existing intrinsics:
Other new ISA capabilities
In addition to the above, the new ISA provides an extended set of new multiplication operations on 64-bit and 128-bit
integers, including 64->128 and 128->256 widening multiply. These have been added as new overloads to the following intrinsics (only available with
vector-enhancements-3):Finally, there are a few completely new intrinsics to support new operations (with
vector-enhancements-3):FYI @folkertdev @taiki-e @cuviper @fneddy