Thurston Dang 4220538e25

[msan] Handle multiply-add-accumulate; apply to AVX Vector Neural Network Instructions (VNNI) (#153927) ...

This extends the pmadd handler (recently improved in https://github.com/llvm/llvm-project/pull/153353) to three-operand intrinsics (multiply-add-accumulate), and applies it to the AVX Vector Neural Network Instructions.

Updates the tests from https://github.com/llvm/llvm-project/pull/153135

2025-08-18 13:18:27 -07:00

..

AArch64

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

ARM32

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

i386

[msan] Handle SSE/AVX pshuf intrinsic by applying to shadow (#153895)

2025-08-15 20:28:30 -07:00

LoongArch

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

Mips

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

Mips32

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

PowerPC

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

PowerPC32

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

RISCV32

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

SystemZ

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

X86

[msan] Handle multiply-add-accumulate; apply to AVX Vector Neural Network Instructions (VNNI) (#153927)

2025-08-18 13:18:27 -07:00

abs-vector.ll

[msan] Model is_int_min_poison to avoid false negative in abs (#148069)

2025-07-10 16:47:53 -07:00

alloca.ll

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

array_types.ll

…

atomics.ll

…

attributes.ll

…

bitreverse.ll

…

bmi.ll

…

byval-alignment.ll

…

byval.ll

…

check_access_address.ll

…

check-array.ll

…

check-constant-shadow.ll

…

check-struct.ll

…

clmul.ll

…

count-zeroes.ll

[msan] Sharpen instrumentation of Intrinsic::{ctlz,cttz} (#145609)

2025-06-25 09:29:59 -07:00

csr.ll

…

disambiguate-origin.ll

…

expand-experimental-reductions.ll

[msan] Apply handleVectorReduceIntrinsic to max/min vector instructions (#129819)

2025-03-08 19:31:48 -08:00

freeze.ll

…

funnel_shift.ll

…

instrumentation-with-call-threshold.ll

…

invalidate_global_aa.ll

…

is-fpclass.ll

…

libatomic.ll

…

manual-shadow.ll

…

masked-store-load.ll

…

missing_origin.ll

…

msan_asm_conservative.ll

…

msan_basic.ll

…

msan_debug_info.ll

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

msan_eager.ll

…

msan_invalidate.ll

…

msan_kernel_basic.ll

…

msan_llvm_is_constant.ll

…

msan_llvm_launder_invariant.ll

[IR] Remove size argument from lifetime intrinsics (#150248)

2025-08-08 11:09:34 +02:00

msan_llvm_strip_invariant.ll

…

msan-disable-checks.ll

…

msan-pass-second-run.ll

…

mul_by_constant.ll

…

no-check-rt-unaligned.ll

…

nosanitize.ll

…

opaque-ptr.ll

…

or.ll

[msan] Re-fix disjoint OR instrumentation from #145990 (#148760)

2025-07-15 15:32:15 -07:00

origin-alignment.ll

…

origin-array.ll

…

overflow.ll

…

partial-poison.ll

[msan] Add off-by-default flag to fix false negatives from partially undefined constant fixed-length vectors (#143837)

2025-06-20 10:11:12 -07:00

pr32842.ll

…

reduce.ll

…

return_from_main.ll

…

saturating.ll

…

scmp.ll

…

stable_set_alloca_origin.ll

…

store-long-origin.ll

…

store-origin.ll

…

str-nobuiltin.ll

…

ucmp.ll

…

unreachable.ll

…

unsized_type.ll

…

vector_arith.ll

[msan] Reland with even more improvement: Improve packed multiply-add instrumentation (#153353)

2025-08-15 16:35:42 -07:00

vector_cmp.ll

…

vector_cvt.ll

…

vector_pack.ll

…

vector_shift.ll

…

vector-load-store.ll

[LLVM][IRBuilder] Use NUW arithmetic for Create{ElementCount,TypeSize}. (#143532)

2025-06-19 13:24:39 +01:00

vector-reduce-fadd.ll

…

vector-reduce-fmul.ll

…

vector-track-origins-neon.ll

…

vector-track-origins-struct.ll

…

vscale.ll

[LLVM][IRBuilder] Use NUW arithmetic for Create{ElementCount,TypeSize}. (#143532)

2025-06-19 13:24:39 +01:00

with-call-type-size.ll

Reapply "[msan] Automatically print shadow for failing outlined checks" (#145611) (#145615)

2025-06-24 20:33:11 -07:00