VPABSQ

VPABSQ replaces each quadword with its absolute value. It is AVX-512F’s quadword form of VPABSD, whose page covers the most negative value and the sequences that AVX2 code uses instead, since AVX2 has no absolute value of quadwords. VPABSQ exists in EVEX encodings only, with masking, a broadcast source and the 512-bit width. The cases recorded on Zen 5 show the same edge as for doublewords: 8000000000000000h stayed 8000000000000000h, from a register and from a broadcast, while FFFFFFFFFFFFFFFFh became 1 and 7FFFFFFFFFFFFFFFh stayed itself. Taken as uint64_t, every result is the right magnitude.

In compiled code#

llabs(a[i] - b[i]) over int64_t arrays becomes a VPSUBQ and VPABSQ with GCC 14 and Clang 23 at -O2 and -march=x86-64-v4, and with a sum around it, the distance between two vectors, a VPADDQ into an accumulator as well. The difference may overflow there, as it may in the C source.

When the input is narrower, the compilers prefer the narrow instruction: for llabs((int64_t)a[i]) with int32_t elements, both take VPABSD of the doublewords and widen the results with VPMOVZXDQ rather than sign-extending first and using VPABSQ. Zero extension is right there, because the absolute value of a 32-bit integer always fits 32 unsigned bits: 80000000h, which VPABSD returns for -2147483648, becomes 2147483648.

For unsigned numbers there is no absolute value to take; the distance a > b ? a - b : b - a is the larger minus the smaller, as the VPMINUQ page shows.

AMD’s latency spreadsheet for Zen 5 lists VPABSQ as it lists VPABSD: one macro-op with a latency of 2 cycles on the pipes FP0 and FP3.

Forms

Latency and throughput

Data flow

Where the elements of the sources go, element 0 on the right; results that depend on a control value are drawn for an example. How to read.

Operation

What the forms do, in pseudocode that our tests run against results recorded on real processors. How to read.

for i in range(VL // 64):
    if k1[i]:
        zmm1.q[i] = abs(src2.sq[i])
    elif zeroing:
        zmm1.q[i] = 0
    # else: zmm1.q[i] keeps its value
# m64bcst: src2.q[i] is the m64 for every i
# zmm1 is zeroed from bit VL up

Operands

Flags

None: XED records no flag that the forms test or change.

Exceptions

The exception class is XED's name for the set of exception conditions that a form follows. How to read.

All forms

AVX512-E4

Attributes

What else XED records about the forms. How to read.

All forms

write masking

Memory forms

embedded broadcast (tuple Full); memory fault suppression

Intrinsics

The C intrinsics of Clang 23.1.2 that compile to VPABSQ, each with the form it compiles to and the Clang options it needs. The forms come from compiling every intrinsic and decoding the result with Intel XED; all intrinsics A–Z explains the method.

CPUID and processors

Each form belongs to an XED ISA set. A form can be used when the processor reports every CPUID bit of one of its ISA set's alternatives.

More AVX10.1 instructions

• VMOVNTPS
• VMOVQ
• VMOVSD
• VMOVSH
• VMOVSHDUP
• VMOVSLDUP
• VMOVSS
• VMOVUPD
• VMOVUPS
• VMOVW
• VMULPD
• VMULPH
• VMULPS
• VMULSD
• VMULSH
• VMULSS
• VORPD
• VORPS
• VPABSB
• VPABSD
• VPABSW
• VPACKSSDW
• VPACKSSWB
• VPACKUSDW
• VPACKUSWB
• VPADDB
• VPADDD
• VPADDQ
• VPADDSB
• VPADDSW
• VPADDUSB
• VPADDUSW
• VPADDW
• VPALIGNR
• VPANDD
• VPANDND
• VPANDNQ
• VPANDQ
• VPAVGB
• VPAVGW

All 686 AVX10.1 instructions