VPBROADCASTB

VPBROADCASTB copies one byte to every byte of a vector: 16, 32 or 64 copies of the low byte of an XMM register, of a byte in memory, or, in the EVEX forms of AVX512BW, of the low byte of a general-purpose register. VPBROADCASTD describes the family and how the sources differ by encoding; this page is about what bytes are broadcast for.

A character to compare against#

Byte-oriented vector code nearly always starts by spreading a run-time byte across a register: the character that memchr or strchr looks for, a delimiter to split on, a byte to count or to fill with. Each block of input is then compared with VPCMPEQB, and the matches leave the register with VPMOVMSKB or as a mask register. For _mm256_set1_epi8(c), GCC 14 and Clang 23 at -O2 emit VMOVD and VPBROADCASTB for -march=x86-64-v3, and VPBROADCASTB straight from the register for x86-64-v4. Clang vectorizes a loop counting the bytes equal to c this way, and GCC fills an array with c from a ZMM register made by vpbroadcastb zmm0, esi at x86-64-v4.

Before AVX2 there was no byte broadcast: SSSE3 code does it with PSHUFB and an all-zero selector, and for SSE2 both compilers need three shuffles after the MOVD, starting with PUNPCKLBW and ending with PSHUFD.

Not for constants#

Neither compiler uses VPBROADCASTB for a constant such as _mm256_set1_epi8(0x0f): since every byte is the same, any wider element is the same too. GCC builds 0F0F0F0Fh in a general-purpose register and broadcasts it with VPBROADCASTD; Clang broadcasts 8 bytes from memory with VBROADCASTSD, or applies them as an embedded broadcast {1to4} with AVX-512. The memory form of VPBROADCASTB is still useful when the byte comes from memory: it reads just that byte, with no alignment requirement, so it cannot touch the bytes around it.

AMD’s Zen 5 guide lists the forms with an XMM or memory source as one macro-op with a latency of one cycle, and the form with a general-purpose register source as two macro-ops.

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 // 8):
    if k1[i]:  # 1 without a write mask
        zmm1.b[i] = src2.b[0]
    elif zeroing:
        zmm1.b[i] = 0
    # else: zmm1.b[i] keeps its value
# 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.

VEX forms

avx-type-6

VPBROADCASTB xmm {k}{z}, xmm/m8 VPBROADCASTB ymm {k}{z}, xmm/m8 VPBROADCASTB zmm {k}{z}, xmm/m8

AVX512-E6

VPBROADCASTB xmm {k}{z}, r32 VPBROADCASTB ymm {k}{z}, r32 VPBROADCASTB zmm {k}{z}, r32

AVX512-E7NM

Attributes

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

EVEX forms

write masking

EVEX memory forms

memory fault suppression

Intrinsics

The C intrinsics of Clang 23.1.2 that compile to VPBROADCASTB, 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 AVX2 instructions

• VPACKSSDW
• VPACKSSWB
• VPACKUSDW
• VPACKUSWB
• VPADDB
• VPADDD
• VPADDQ
• VPADDSB
• VPADDSW
• VPADDUSB
• VPADDUSW
• VPADDW
• VPALIGNR
• VPAND
• VPANDN
• VPAVGB
• VPAVGW
• VPBLENDD
• VPBLENDVB
• VPBLENDW
• VPBROADCASTD
• VPBROADCASTQ
• VPBROADCASTW
• VPCMPEQB
• VPCMPEQD
• VPCMPEQQ
• VPCMPEQW
• VPCMPGTB
• VPCMPGTD
• VPCMPGTQ
• VPCMPGTW
• VPERM2I128
• VPERMD
• VPERMPD
• VPERMPS
• VPERMQ
• VPGATHERDD
• VPGATHERDQ
• VPGATHERQD
• VPGATHERQQ

All 138 AVX2 instructions

More AVX10.1 instructions

• VPACKUSWB
• VPADDB
• VPADDD
• VPADDQ
• VPADDSB
• VPADDSW
• VPADDUSB
• VPADDUSW
• VPADDW
• VPALIGNR
• VPANDD
• VPANDND
• VPANDNQ
• VPANDQ
• VPAVGB
• VPAVGW
• VPBLENDMB
• VPBLENDMD
• VPBLENDMQ
• VPBLENDMW
• VPBROADCASTD
• VPBROADCASTMB2Q
• VPBROADCASTMW2D
• VPBROADCASTQ
• VPBROADCASTW
• VPCLMULQDQ
• VPCMPB
• VPCMPD
• VPCMPEQB
• VPCMPEQD
• VPCMPEQQ
• VPCMPEQW
• VPCMPGTB
• VPCMPGTD
• VPCMPGTQ
• VPCMPGTW
• VPCMPQ
• VPCMPUB
• VPCMPUD
• VPCMPUQ

All 686 AVX10.1 instructions