Vesta
English
Source code

VDIVPH

Divide Packed FP16 Values

stableVMJITAOTinstruction

Encodings

OpcodeInstructionOp/En64-bitCompat/LegacyDescription
EVEX.128.NP.MAP5.W0 5E /rVDIVPH xmm1{k1}{z}, xmm2, xmm3/m128/m16bcstAValidValidDivide packed FP16 values in xmm2 by packed AND AVX512VL) FP16 values in xmm3/m128/m16bcst, and store OR AVX10.1 the result in xmm1 subject to writemask k1.
EVEX.256.NP.MAP5.W0 5E /rVDIVPH ymm1{k1}{z}, ymm2, ymm3/m256/m16bcstAValidValidDivide packed FP16 values in ymm2 by packed AND AVX512VL) FP16 values in ymm3/m256/m16bcst, and store OR AVX10.1 the result in ymm1 subject to writemask k1.
EVEX.512.NP.MAP5.W0 5E /rVDIVPH zmm1{k1}{z}, zmm2, zmm3/m512/m16bcst {er}AValidValidDivide packed FP16 values in zmm2 by packed OR AVX10.1 FP16 values in zmm3/m512/m16bcst, and store the result in zmm1 subject to writemask k1.

Operand encoding

Each mode is a value of the Op/En column above. It says which field of the encoded instruction carries each operand, in the order they are written, and whether the instruction reads it, writes it or both.

A

  1. modrm.reg escrituraModRM byte, reg field (bits 5-3)
  2. vex.vvvv lecturaVEX prefix, vvvv field (inverted)
  3. modrm.rm lecturaModRM byte, r/m field (bits 2-0); with the SIB byte and the displacement when the mod field asks for them

Tupla: Full

Measured cost

Loading measurements from arch-data...

Description

This instruction divides packed FP16 values from the first source operand by the corresponding elements in the second source operand, storing the packed FP16 result in the destination operand. The destination elements are updated according to the writemask.

Operation

VDIVPH (EVEX Encoded Versions) When SRC2 Operand is a Register
VL = 128, 256 or 512
KL := VL/16

IF (VL = 512) AND (EVEX.b = 1):
    SET_RM(EVEX.RC)

ELSE
    SET_RM(MXCSR.RC)

FOR j := 0 TO KL-1:
    IF k1[j] OR *no writemask*:
          DEST.fp16[j] := SRC1.fp16[j] / SRC2.fp16[j]
    ELSE IF *zeroing*:
          DEST.fp16[j] := 0
    // else dest.fp16[j] remains unchanged

DEST[MAXVL-1:VL] := 0


VDIVPH (EVEX Encoded Versions) When SRC2 Operand is a Memory Source
VL = 128, 256 or 512
KL := VL/16

FOR j := 0 TO KL-1:
    IF k1[j] OR *no writemask*:
          IF EVEX.b = 1:
                DEST.fp16[j] := SRC1.fp16[j] / SRC2.fp16[0]
          ELSE:
                DEST.fp16[j] := SRC1.fp16[j] / SRC2.fp16[j]
    ELSE IF *zeroing*:
          DEST.fp16[j] := 0
    // else dest.fp16[j] remains unchanged

DEST[MAXVL-1:VL] := 0

Intel C/C++ compiler intrinsics

VDIVPH __m128h _mm_div_ph (__m128h a, __m128h b);
VDIVPH __m128h _mm_mask_div_ph (__m128h src, __mmask8 k, __m128h a, __m128h b);
VDIVPH __m128h _mm_maskz_div_ph (__mmask8 k, __m128h a, __m128h b);
VDIVPH __m256h _mm256_div_ph (__m256h a, __m256h b);
VDIVPH __m256h _mm256_mask_div_ph (__m256h src, __mmask16 k, __m256h a, __m256h b);
VDIVPH __m256h _mm256_maskz_div_ph (__mmask16 k, __m256h a, __m256h b);
VDIVPH __m512h _mm512_div_ph (__m512h a, __m512h b);
VDIVPH __m512h _mm512_mask_div_ph (__m512h src, __mmask32 k, __m512h a, __m512h b);
VDIVPH __m512h _mm512_maskz_div_ph (__mmask32 k, __m512h a, __m512h b);
VDIVPH __m512h _mm512_div_round_ph (__m512h a, __m512h b, int rounding);
VDIVPH __m512h _mm512_mask_div_round_ph (__m512h src, __mmask32 k, __m512h a, __m512h b, int rounding);
VDIVPH __m512h _mm512_maskz_div_round_ph (__mmask32 k, __m512h a, __m512h b, int rounding);

SIMD Floating-Point Exceptions

Invalid, Underflow, Overflow, Precision, Denormal, Zero.

Other Exceptions

EVEX-encoded instructions, see Table 2-48, "Type E2 Class Exception Conditions."

Sources