BSD Commit Log Search

emscripten: Some py scripts may require a shell wrapper at runtime, starting with emar

Revert " [SPIRV] Addition of extension SPV_KHR_non_semantic_info and SPV_KHR_relaxed_extended_instruction" (#177093)

Reverts llvm/llvm-project#169643 due to build issues

 [SPIRV] Addition of extension SPV_KHR_non_semantic_info and SPV_KHR_relaxed_extended_instruction (#169643)

--Added support for the extension SPV_KHR_non_semantic_info
--Added support for the extension SPV_KHR_relaxed_extended_instruction 
--Added instructions from the documentation of the extension. 
--Added supporting tests for the same.

Same as #165302

---------

Co-authored-by: Michal Paszkowski <michal at michalpaszkowski.com>

[RISC-V][MC] Handle YGPR registers in RISCVMCInstrAnalysis

This previously triggered an assertion in a test for RVY's AUIPC
instruction that will be submitted as a follow up
(https://github.com/llvm/llvm-project/pull/177074).

Pull Request: https://github.com/llvm/llvm-project/pull/177071

emscripten: default cache is in install area, set it to $HOME/.cache

[MLIR][XeGPU] setUnitDim bug fix and add documentation (#173521)

This PR fix a bug in setUnitDimData and setUnitDimLayout, and adds
documentation and test.
It also cleans up the shapecast op pattern in the wg distribution to use
local temporary layout instead of getting from definition op's result
(one TODO item from PR
[#172125](https://github.com/llvm/llvm-project/pull/172125)).

[clang-tidy] Fix assert failure in modernize-use-std-format when args are in macros (#176684)

Part of [#175183](https://github.com/llvm/llvm-project/issues/175183)

   doc: Updated lang/pear to 1.10.17nb1

   lang/pear: fix conflicting with base php

   Rename README.pkgsrc to README-pear.pkgsrc at install.

   Bump PKGREVISION.

[AMDGPU] Use subclass query for VGPR_16 RA hinting. NFCI (#177017)

[MLIR][NVVM] Add tensormap.replace NVVM Op (#174926)

This change adds the `nvvm.tensormap.replace` Op to the NVVM
dialect for the `tensormap.replace` PTX instruction.

PTX ISA Reference:
https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-tensormap-replace

emscripten: requires docs directory at runtime, added to the package

emscripten: Wrapper scripts added to $prefix/bin

powerpc/mpc85xx: Fix PCI attach error cleanup

If an error occurs during attach after ofw_pcib_init() runs, the device
is torn down, leaving the rmans embedded in the softc attached to the
rman list, thus corrupting the rman list.  Fix this by undoing
everything that was done by this point.

MFC after:      1 week

powerpc/mpc85xx: Set pc_hwref to the primary thread ID

On multithreaded cores (e6500) the CPU ID in the device tree (reg[0]) is
the primary core, which may not match the cpuid, until Book-E threading
is added.

dpaa: Simplify CPU binding for bman and qman

If cpu-handle property doesn't exist simply iterate and assign the CPUs
in sequence rather than following the convoluted search which may not
bear fruit in some cases.  If cpu-handle doesn't exist for one portal it
probably doesn't exist for any of them.

[AMDGPU] Update documentation for wave reduction intrinsics

[AMDGPU] Add builtins for wave reduction intrinsics

[MachineCombiner] Defer pattern order verfication. NFC (#177081)

The `verifyPatternOrder` function was created to check if machine
combiner patterns are ordered by their latency reduction degree. This
function can be merged into the following loop, which iterates through
all the patterns, such that (1) `TII::genAlternativeCodeSequence`, which
can be quite expensive, won't be called twice (2) since it's now placed
after we print out the candidate sequence through debug print, if
anything goes wrong during the verification, we at least know what the
sequence looks like (3) we don't need to verify all patterns if we will
eventually exit early.

This should be a NFC

   drm/amdkfd: Fix improper NULL termination of queue restore SMI event string

   From Brian Kocoloski
   47206d70d1fba05470a2bd00ae3d66d27487c195 in linux-6.12.y/6.12.66
   969faea4e9d01787c58bab4d945f7ad82dad222d in mainline linux

   drm/amd/display: Fix DP no audio issue

   From Charlene Liu
   f609041424d56f673e278f38ebbc71e05564b2ea in linux-6.12.y/6.12.66
   3886b198bd6e49c801fe9552fcfbfc387a49fbbc in mainline linux

   drm/amd/display: shrink struct members

   From Rosen Penev
   e83af97d5c3913c5d0bb81dcf0188f7c48731215 in linux-6.12.y/6.12.66
   7329417fc9ac128729c3a092b006c8f1fd0d04a6 in mainline linux

Address review comments

Created using spr 1.3.6-beta.1

[𝘀𝗽𝗿] changes introduced through rebase

Created using spr 1.3.6-beta.1

[skip ci]

Rebase

Created using spr 1.3.6-beta.1

   drm/radeon: Remove __counted_by from ClockInfoArray.clockInfo[]

   From Alex Deucher
   7500ab83bad207341916950bb21248af0a1ee21e in linux-6.12.y/6.12.66
   19158c7332468bc28572bdca428e89c7954ee1b1 in mainline linux

   drm/amdgpu: Fix query for VPE block_type and ip_count

   From Alan Liu
   90b4b130a20d80decfc9b115c187f818ab83a30a in linux-6.12.y/6.12.66
   72d7f4573660287f1b66c30319efecd6fcde92ee in mainline linux

cron: Implement full PAM session lifecycle for user jobs

Extend PAM integration beyond account checks to include credential
establishment and session management, allowing PAM modules to configure
the execution environment for user cron jobs.

Previously, cron only called pam_acct_mgmt() to verify account validity
but immediately terminated the PAM handle before job execution. This
prevented PAM modules from establishing sessions, setting credentials
(e.g., Kerberos tickets), or exporting environment variables needed by
jobs.

The PAM handle now persists in the intermediate process throughout the
job execution, enabling proper session open/close pairing. Credentials
are established and sessions opened while still running as root, before
dropping privileges in the grandchild. PAM environment variables are
exported in the job process with user crontab variables taking precedence.

A session rule (pam_permit.so) is added to /etc/pam.d/cron to enable

    [9 lines not shown]

Allow toggling stig settings

[LoongArch] Remove DAG combination for extractelement

Combination for `trunc+extend+extractelement` to a single
`extractelement` may occur error, because the high bits of the
extract index truncated by `trunc` operation are reserved
after the combination.

This commit remove this combination and the issue
https://github.com/llvm/llvm-project/issues/176839 will never
appear.