VEXTRACTI128, VEXTRACTI32x4, VEXTRACTI64x2, VEXTRACTI32x8, VEXTRACTI64x4s (Intel x86/64 assembly instruction)

모두의 코드
VEXTRACTI128, VEXTRACTI32x4, VEXTRACTI64x2, VEXTRACTI32x8, VEXTRACTI64x4s (Intel x86/64 assembly instruction)

작성일 : 2020-09-01 이 글은 642 번 읽혔습니다.

VEXTRACTI128, VEXTRACTI32x4, VEXTRACTI64x2, VEXTRACTI32x8, VEXTRACTI64x4

Extract packed Integer Values

참고 사항

아래 표를 해석하는 방법은 x86-64 명령어 레퍼런스 읽는 법 글을 참조하시기 바랍니다.

Opcode/ Instruction	Op / En	64/32 bit Mode Support	CPUID Feature Flag	Description
`VEX.256.66.0F3A.W0 39 /r ib` VEXTRACTI128 xmm1/m128 ymm2 imm8	RMI	V/V	AVX2	Extract 128 bits of integer data from ymm2 and store results in xmm1/m128.
`EVEX.256.66.0F3A.W0 39 /r ibVEXTRACTI32X4 xmm1/m128 {k1}{z}, ymm2, imm8`	T4	V/V	AVX512VL AVX512F	Extract 128 bits of double-word integer values from ymm2 and store results in xmm1/m128 subject to writemask k1.
`EVEX.512.66.0F3A.W0 39 /r ib` VEXTRACTI32x4 xmm1/m128 {k1}{z} zmm2 imm8	T4	V/V	AVX512F	Extract 128 bits of double-word integer values from zmm2 and store results in xmm1/m128 subject to writemask k1.
`EVEX.256.66.0F3A.W1 39 /r ibVEXTRACTI64X2 xmm1/m128 {k1}{z}, ymm2, imm8`	T2	V/V	AVX512VL AVX512DQ	Extract 128 bits of quad-word integer values from ymm2 and store results in xmm1/m128 subject to writemask k1.
`EVEX.512.66.0F3A.W1 39 /r ibVEXTRACTI64X2 xmm1/m128 {k1}{z}, zmm2, imm8`	T2	V/V	AVX512DQ	Extract 128 bits of quad-word integer values from zmm2 and store results in xmm1/m128 subject to writemask k1.
`EVEX.512.66.0F3A.W0 3B /r ibVEXTRACTI32X8 ymm1/m256 {k1}{z}, zmm2, imm8`	T8	V/V	AVX512DQ	Extract 256 bits of double-word integer values from zmm2 and store results in ymm1/m256 subject to writemask k1.
`EVEX.512.66.0F3A.W1 3B /r ib` VEXTRACTI64x4 ymm1/m256 {k1}{z} zmm2 imm8	T4	V/V	AVX512F	Extract 256 bits of quad-word integer values from zmm2 and store results in ymm1/m256 subject to writemask k1.

Instruction Operand Encoding

Op/En	Operand 1	Operand 2	Operand 3	Operand 4
RMI	ModRM:r/m (w)	ModRM:reg (r)	Imm8	NA
T2, T4, T8	ModRM:r/m (w)	ModRM:reg (r)	Imm8	NA

Description

VEXTRACTI128/VEXTRACTI32x4 and VEXTRACTI64x2 extract 128-bits of doubleword integer values from the source operand (the second operand) and store to the low 128-bit of the destination operand (the first operand). The 128-bit data extraction occurs at an 128-bit granular offset specified by imm8[0] (256-bit) or imm8[1:0] as the multiply factor. The destination may be either a vector register or an 128-bit memory location.

VEXTRACTI32x4: The low 128-bit of the destination operand is updated at 32-bit granularity according to the writemask.

VEXTRACTI64x2: The low 128-bit of the destination operand is updated at 64-bit granularity according to the writemask.

VEXTRACTI32x8 and VEXTRACTI64x4 extract 256-bits of quadword integer values from the source operand (the second operand) and store to the low 256-bit of the destination operand (the first operand). The 256-bit data extraction occurs at an 256-bit granular offset specified by imm8[0] (256-bit) or imm8[0] as the multiply factor The destination may be either a vector register or a 256-bit memory location.

VEXTRACTI32x8: The low 256-bit of the destination operand is updated at 32-bit granularity according to the writemask.

VEXTRACTI64x4: The low 256-bit of the destination operand is updated at 64-bit granularity according to the writemask.

VEX.vvvv and EVEX.vvvv are reserved and must be 1111b otherwise instructions will #UD.

The high 7 bits (6 bits in EVEX.512) of the immediate are ignored.

If VEXTRACTI128 is encoded with VEX.L= 0, an attempt to execute the instruction encoded with VEX.L= 0 will cause an #UD exception.

Operation

VEXTRACTI32x4 (EVEX encoded versions) when destination is a register

VL = 256, 512
IF VL = 256
    CASE (imm8[0]) OF
          0: TMP_DEST[127:0] <-  SRC1[127:0]
          1: TMP_DEST[127:0] <-  SRC1[255:128]
    ESAC.
FI;
IF VL = 512 
    CASE (imm8[1:0]) OF
          00: TMP_DEST[127:0] <-  SRC1[127:0]
          01: TMP_DEST[127:0] <-  SRC1[255:128]
          10: TMP_DEST[127:0] <-  SRC1[383:256]
          11: TMP_DEST[127:0] <-  SRC1[511:384]
    ESAC.
FI;
FOR j <-  0 TO 3
    i <-  j * 32
    IF k1[j] OR *no writemask*
          THEN DEST[i+31:i] <-  TMP_DEST[i+31:i]
          ELSE 
                IF *merging-masking* ; merging-masking
                      THEN *DEST[i+31:i] remains unchanged*
                      ELSE *zeroing-masking* ; zeroing-masking
                            DEST[i+31:i] <-  0
                FI
    FI;
ENDFOR
DEST[MAX_VL-1:128] <-  0

VEXTRACTI32x4 (EVEX encoded versions) when destination is memory

VL = 256, 512
IF VL = 256
    CASE (imm8[0]) OF
          0: TMP_DEST[127:0] <-  SRC1[127:0]
          1: TMP_DEST[127:0] <-  SRC1[255:128]
    ESAC.
FI;
IF VL = 512 
    CASE (imm8[1:0]) OF
          00: TMP_DEST[127:0] <-  SRC1[127:0]
          01: TMP_DEST[127:0] <-  SRC1[255:128]
          10: TMP_DEST[127:0] <-  SRC1[383:256]
          11: TMP_DEST[127:0] <-  SRC1[511:384]
    ESAC.
FI;
FOR j <-  0 TO 3
    i <-  j * 32
    IF k1[j] OR *no writemask*
          THEN DEST[i+31:i] <-  TMP_DEST[i+31:i]
          ELSE *DEST[i+31:i] remains unchanged* ; merging-masking
    FI;
ENDFOR

VEXTRACTI64x2 (EVEX encoded versions) when destination is a register

VL = 256, 512
IF VL = 256
    CASE (imm8[0]) OF
          0: TMP_DEST[127:0] <-  SRC1[127:0]
          1: TMP_DEST[127:0] <-  SRC1[255:128]
    ESAC.
FI;
IF VL = 512 
    CASE (imm8[1:0]) OF
          00: TMP_DEST[127:0] <-  SRC1[127:0]
          01: TMP_DEST[127:0] <-  SRC1[255:128]
          10: TMP_DEST[127:0] <-  SRC1[383:256]
          11: TMP_DEST[127:0] <-  SRC1[511:384]
    ESAC.
FI;
FOR j <-  0 TO 1
    i <-  j * 64
    IF k1[j] OR *no writemask*
          THEN DEST[i+63:i] <-  TMP_DEST[i+63:i]
          ELSE 
                IF *merging-masking* ; merging-masking
                      THEN *DEST[i+63:i] remains unchanged*
                      ELSE *zeroing-masking* ; zeroing-masking
                            DEST[i+63:i] <-  0
                FI
    FI;
ENDFOR
DEST[MAX_VL-1:128] <-  0

VEXTRACTI64x2 (EVEX encoded versions) when destination is memory

VL = 256, 512
IF VL = 256
    CASE (imm8[0]) OF
          0: TMP_DEST[127:0] <-  SRC1[127:0]
          1: TMP_DEST[127:0] <-  SRC1[255:128]
    ESAC.
FI;
IF VL = 512 
    CASE (imm8[1:0]) OF
          00: TMP_DEST[127:0] <-  SRC1[127:0]
          01: TMP_DEST[127:0] <-  SRC1[255:128]
          10: TMP_DEST[127:0] <-  SRC1[383:256]
          11: TMP_DEST[127:0] <-  SRC1[511:384]
    ESAC.
FI;
FOR j <-  0 TO 1
    i <-  j * 64
    IF k1[j] OR *no writemask*
          THEN DEST[i+63:i] <-  TMP_DEST[i+63:i]
          ELSE *DEST[i+63:i] remains unchanged* ; merging-masking
    FI;
ENDFOR

VEXTRACTI32x8 (EVEX.U1.512 encoded version) when destination is a register

VL = 512
CASE (imm8[0]) OF
    0: TMP_DEST[255:0] <-  SRC1[255:0]
    1: TMP_DEST[255:0] <-  SRC1[511:256]
ESAC.
FOR j <-  0 TO 7
    i <-  j * 32
    IF k1[j] OR *no writemask*
          THEN DEST[i+31:i] <-  TMP_DEST[i+31:i]
          ELSE 
                IF *merging-masking* ; merging-masking
                      THEN *DEST[i+31:i] remains unchanged*
                      ELSE *zeroing-masking* ; zeroing-masking
                            DEST[i+31:i] <-  0
                FI
    FI;
ENDFOR
DEST[MAX_VL-1:256] <-  0

VEXTRACTI32x8 (EVEX.U1.512 encoded version) when destination is memory

CASE (imm8[0]) OF
    0: TMP_DEST[255:0] <-  SRC1[255:0]
    1: TMP_DEST[255:0] <-  SRC1[511:256]
ESAC.
FOR j <-  0 TO 7
    i <-  j * 32
    IF k1[j] OR *no writemask*
          THEN DEST[i+31:i] <-  TMP_DEST[i+31:i]
          ELSE *DEST[i+31:i] remains unchanged* ; merging-masking
    FI;
ENDFOR

VEXTRACTI64x4 (EVEX.512 encoded version) when destination is a register

VL = 512
CASE (imm8[0]) OF
    0: TMP_DEST[255:0] <-  SRC1[255:0]
    1: TMP_DEST[255:0] <-  SRC1[511:256]
ESAC.
FOR j <-  0 TO 3
    i <-  j * 64
    IF k1[j] OR *no writemask*
          THEN DEST[i+63:i] <-  TMP_DEST[i+63:i]
          ELSE 
                IF *merging-masking* ; merging-masking
                      THEN *DEST[i+63:i] remains unchanged*
                      ELSE *zeroing-masking* ; zeroing-masking
                            DEST[i+63:i] <-  0
                FI
    FI;
ENDFOR
DEST[MAX_VL-1:256] <-  0

VEXTRACTI64x4 (EVEX.512 encoded version) when destination is memory

CASE (imm8[0]) OF
    0: TMP_DEST[255:0] <-  SRC1[255:0]
    1: TMP_DEST[255:0] <-  SRC1[511:256]
ESAC.
FOR j <-  0 TO 3
    i <-  j * 64
    IF k1[j] OR *no writemask*
          THEN DEST[i+63:i] <-  TMP_DEST[i+63:i]
          ELSE *DEST[i+63:i] remains unchanged* ; merging-masking
    FI;
ENDFOR

VEXTRACTI128 (memory destination form)

CASE (imm8[0]) OF
    0: DEST[127:0] <- SRC1[127:0]
    1: DEST[127:0] <- SRC1[255:128]
ESAC.

VEXTRACTI128 (register destination form)

CASE (imm8[0]) OF
    0: DEST[127:0] <- SRC1[127:0]
    1: DEST[127:0] <- SRC1[255:128]
ESAC.
DEST[MAX_VL-1:128] <- 0

Intel C/C++ Compiler Intrinsic Equivalent

VEXTRACTI32x4 __m128i _mm512_extracti32x4_epi32(__m512i a, const int nidx);
VEXTRACTI32x4 __m128i _mm512_mask_extracti32x4_epi32(__m128i s, __mmask8 k,
                                                     __m512i a, const int nidx);
VEXTRACTI32x4 __m128i _mm512_maskz_extracti32x4_epi32(__mmask8 k, __m512i a,
                                                      const int nidx);
VEXTRACTI32x4 __m128i _mm256_extracti32x4_epi32(__m256i a, const int nidx);
VEXTRACTI32x4 __m128i _mm256_mask_extracti32x4_epi32(__m128i s, __mmask8 k,
                                                     __m256i a, const int nidx);
VEXTRACTI32x4 __m128i _mm256_maskz_extracti32x4_epi32(__mmask8 k, __m256i a,
                                                      const int nidx);
VEXTRACTI32x8 __m256i _mm512_extracti32x8_epi32(__m512i a, const int nidx);
VEXTRACTI32x8 __m256i _mm512_mask_extracti32x8_epi32(__m256i s, __mmask8 k,
                                                     __m512i a, const int nidx);
VEXTRACTI32x8 __m256i _mm512_maskz_extracti32x8_epi32(__mmask8 k, __m512i a,
                                                      const int nidx);
VEXTRACTI64x2 __m128i _mm512_extracti64x2_epi64(__m512i a, const int nidx);
VEXTRACTI64x2 __m128i _mm512_mask_extracti64x2_epi64(__m128i s, __mmask8 k,
                                                     __m512i a, const int nidx);
VEXTRACTI64x2 __m128i _mm512_maskz_extracti64x2_epi64(__mmask8 k, __m512i a,
                                                      const int nidx);
VEXTRACTI64x2 __m128i _mm256_extracti64x2_epi64(__m256i a, const int nidx);
VEXTRACTI64x2 __m128i _mm256_mask_extracti64x2_epi64(__m128i s, __mmask8 k,
                                                     __m256i a, const int nidx);
VEXTRACTI64x2 __m128i _mm256_maskz_extracti64x2_epi64(__mmask8 k, __m256i a,
                                                      const int nidx);
VEXTRACTI64x4 __m256i _mm512_extracti64x4_epi64(__m512i a, const int nidx);
VEXTRACTI64x4 __m256i _mm512_mask_extracti64x4_epi64(__m256i s, __mmask8 k,
                                                     __m512i a, const int nidx);
VEXTRACTI64x4 __m256i _mm512_maskz_extracti64x4_epi64(__mmask8 k, __m512i a,
                                                      const int nidx);
VEXTRACTI128 __m128i _mm256_extracti128_si256(__m256i a, int offset);

SIMD Floating-Point Exceptions

None

Other Exceptions

VEX-encoded instructions, see Exceptions Type 6;

EVEX-encoded instructions, see Exceptions Type E6NF.

#UD IF VEX.L = 0.

#UD If VEX.vvvv != 1111B or EVEX.vvvv != 1111B.

첫 댓글을 달아주세요!

강좌에 관련 없이 궁금한 내용은 여기를 사용해주세요

또는 직접 입력하세요 (댓글 수정시 비밀번호가 필요합니다)

댓글을 불러오는 중입니다..

모두의 코드 VEXTRACTI128, VEXTRACTI32x4, VEXTRACTI64x2, VEXTRACTI32x8, VEXTRACTI64x4s (Intel x86/64 assembly instruction)