PSHUFB — Packed Shuffle Bytes

NOTES:

1. See note in Section 2.4, “AVX and SSE Instruction Exception Specification” in the Intel® 64 and IA-32 Architectures Software Developer’s Manual, Volume 2A and Section 22.25.3, “Exception Conditions of Legacy SIMD Instructions Operating on MMX Registers” in the Intel® 64 and IA-32 Architectures Software Developer’s Manual, Volume 3A.

Instruction Operand Encoding

Description

PSHUFB performs in-place shuffles of bytes in the destination operand (the first operand) according to the shuffle control mask in the source operand (the second operand). The instruction permutes the data in the destination operand, leaving the shuffle mask unaffected. If the most significant bit (bit[7]) of each byte of the shuffle control mask is set, then constant zero is written in the result byte. Each byte in the shuffle control mask forms an index to permute the corresponding byte in the destination operand. The value of each index is the least significant 4 bits (128-bit operation) or 3 bits (64-bit operation) of the shuffle control byte. When the source operand is a 128-bit memory operand, the operand must be aligned on a 16-byte boundary or a general-protection exception (#GP) will be generated.

In 64-bit mode and not encoded with VEX/EVEX, use the REX prefix to access XMM8-XMM15 registers.

Legacy SSE version 64-bit operand: Both operands can be MMX registers.

128-bit Legacy SSE version: The first source operand and the destination operand are the same. Bits (VLMAX-1:128) of the corresponding YMM destination register remain unchanged.

VEX.128 encoded version: The destination operand is the first operand, the first source operand is the second operand, the second source operand is the third operand. Bits (VLMAX-1:128) of the destination YMM register are zeroed.

VEX.256 encoded version: Bits (255:128) of the destination YMM register stores the 16-byte shuffle result of the upper 16 bytes of the first source operand, using the upper 16-bytes of the second source operand as control mask. The value of each index is for the high 128-bit lane is the least significant 4 bits of the respective shuffle control byte. The index value selects a source data element within each 128-bit lane.

EVEX encoded version: The second source operand is an ZMM/YMM/XMM register or an 512/256/128-bit memory location. The first source operand and destination operands are ZMM/YMM/XMM registers. The destination is condi-tionally updated with writemask k1.

EVEX and VEX encoded version: Four/two in-lane 128-bit shuffles.

Operation

PSHUFB (with 64 bit operands)

TEMP ← DEST
for i = 0 to 7 {
    if (SRC[(i * 8)+7] = 1 ) then
         DEST[(i*8)+7...(i*8)+0] ← 0;
    else
         index[2..0] ← SRC[(i*8)+2 .. (i*8)+0];
         DEST[(i*8)+7...(i*8)+0] ← TEMP[(index*8+7)..(index*8+0)];
    endif;
}

PSHUFB (with 128 bit operands)

TEMP ← DEST
for i = 0 to 15 {
    if (SRC[(i * 8)+7] = 1 ) then
         DEST[(i*8)+7..(i*8)+0] ← 0;
     else
         index[3..0] ← SRC[(i*8)+3 .. (i*8)+0]; DEST[(i*8)+7..(i*8)+0] ← TEMP[(index*8+7)..(index*8+0)];
    endif
}

VPSHUFB (VEX.128 encoded version)

for i = 0 to 15 {
    if (SRC2[(i * 8)+7] = 1) then
         DEST[(i*8)+7..(i*8)+0] (cid:197) 0;
         else
         index[3..0] (cid:197) SRC2[(i*8)+3 .. (i*8)+0];
         DEST[(i*8)+7..(i*8)+0] (cid:197) SRC1[(index*8+7)..(index*8+0)];
    endif
}
DEST[VLMAX-1:128] (cid:197) 0

VPSHUFB (VEX.256 encoded version)

for i = 0 to 15 {
    if (SRC2[(i * 8)+7] == 1 ) then
         DEST[(i*8)+7..(i*8)+0] (cid:197) 0;
         else
         index[3..0] (cid:197) SRC2[(i*8)+3 .. (i*8)+0];
         DEST[(i*8)+7..(i*8)+0] (cid:197) SRC1[(index*8+7)..(index*8+0)];
    endif
    if (SRC2[128 + (i * 8)+7] == 1 ) then
         DEST[128 + (i*8)+7..(i*8)+0] (cid:197) 0;
         else
         index[3..0] (cid:197) SRC2[128 + (i*8)+3 .. (i*8)+0];
         DEST[128 + (i*8)+7..(i*8)+0] (cid:197) SRC1[128 + (index*8+7)..(index*8+0)];
    endif
} VPSHUFB (EVEX encoded versions)
(KL, VL) = (16, 128), (32, 256), (64, 512)
jmask (cid:197) (KL-1) & ~0xF
                                                                    // 0x00, 0x10, 0x30 depending on the VL
FOR j = 0 TO KL-1
                                                                    // dest
    IF kl[ i ] or no_masking
         index (cid:197) src.byte[ j ];
         IF index & 0x80
              Dest.byte[ j ] (cid:197) 0;
         ELSE
              index (cid:197) (index & 0xF) + (j & jmask);
                                                                    // 16-element in-lane lookup
              Dest.byte[ j ] (cid:197) src.byte[ index ];
    ELSE if zeroing
         Dest.byte[ j ] (cid:197) 0;
DEST[MAX_VL-1:VL] (cid:197) 0;

Figure 4-15. PSHUFB with 64-Bit Operands

Intel C/C++ Compiler Intrinsic Equivalent

VPSHUFB __m512i _mm512_shuffle_epi8(__m512i a, __m512i b);

VPSHUFB __m512i _mm512_mask_shuffle_epi8(__m512i s, __mmask64 k, __m512i a, __m512i b);

VPSHUFB __m512i _mm512_maskz_shuffle_epi8( __mmask64 k, __m512i a, __m512i b);

VPSHUFB __m256i _mm256_mask_shuffle_epi8(__m256i s, __mmask32 k, __m256i a, __m256i b);

VPSHUFB __m256i _mm256_maskz_shuffle_epi8( __mmask32 k, __m256i a, __m256i b);

VPSHUFB __m128i _mm_mask_shuffle_epi8(__m128i s, __mmask16 k, __m128i a, __m128i b);

VPSHUFB __m128i _mm_maskz_shuffle_epi8( __mmask16 k, __m128i a, __m128i b);

PSHUFB: __m64 _mm_shuffle_pi8 (__m64 a, __m64 b)

(V)PSHUFB: __m128i _mm_shuffle_epi8 (__m128i a, __m128i b)

VPSHUFB:__m256i _mm256_shuffle_epi8(__m256i a, __m256i b)

SIMD Floating-Point Exceptions

None.

Other Exceptions

Non-EVEX-encoded instruction, see Exceptions Type 4.

EVEX-encoded instruction, see Exceptions Type E4NF.nb.