Vesta
English
Source code

VCVTW2PH

Convert Packed Signed Word Integers to FP16 Values

stableVMJITAOTinstruction

Encodings

OpcodeInstructionOp/En64-bitCompat/LegacyDescription
EVEX.128.F3.MAP5.W0 7D /rVCVTW2PH xmm1{k1}{z}, xmm2/m128/m16bcstAValidValidConvert eight packed signed word integers from AND AVX512VL) xmm2/m128/m16bcst to FP16 values, and store OR AVX10.1 the result in xmm1 subject to writemask k1.
EVEX.256.F3.MAP5.W0 7D /rVCVTW2PH ymm1{k1}{z}, ymm2/m256/m16bcstAValidValidConvert sixteen packed signed word integers AND AVX512VL) from ymm2/m256/m16bcst to FP16 values, and OR AVX10.1 store the result in ymm1 subject to writemask k1.
EVEX.512.F3.MAP5.W0 7D /rVCVTW2PH zmm1{k1}{z}, zmm2/m512/m16bcst {er}AValidValidConvert thirty-two packed signed word integers OR AVX10.1 from zmm2/m512/m16bcst to FP16 values, 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. 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 converts packed signed word integers in the source operand to FP16 values in the destination operand. When conversion is inexact, the value returned is rounded according to the rounding control bits in the MXCSR register or embedded rounding controls.

The destination elements are updated according to the writemask.

Operation

VCVTW2PH dest, src
VL = 128, 256 or 512
KL := VL / 16

IF *SRC is a register* and (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*:
          IF *SRC is memory* and EVEX.b = 1:
                tsrc := SRC.word[0]
          ELSE
                tsrc := SRC.word[j]
          DEST.fp16[j] := Convert_integer16_to_fp16(tsrc)
    ELSE IF *zeroing*:
          DEST.fp16[j] := 0
    // else dest.fp16[j] remains unchanged

DEST[MAXVL-1:VL] := 0

Intel C/C++ compiler intrinsics

VCVTW2PH __m512h _mm512_cvt_roundepi16_ph (__m512i a, int rounding);
VCVTW2PH __m512h _mm512_mask_cvt_roundepi16_ph (__m512h src, __mmask32 k, __m512i a, int rounding);
VCVTW2PH __m512h _mm512_maskz_cvt_roundepi16_ph (__mmask32 k, __m512i a, int rounding);
VCVTW2PH __m128h _mm_cvtepi16_ph (__m128i a);
VCVTW2PH __m128h _mm_mask_cvtepi16_ph (__m128h src, __mmask8 k, __m128i a);
VCVTW2PH __m128h _mm_maskz_cvtepi16_ph (__mmask8 k, __m128i a);
VCVTW2PH __m256h _mm256_cvtepi16_ph (__m256i a);
VCVTW2PH __m256h _mm256_mask_cvtepi16_ph (__m256h src, __mmask16 k, __m256i a);
VCVTW2PH __m256h _mm256_maskz_cvtepi16_ph (__mmask16 k, __m256i a);
VCVTW2PH __m512h _mm512_cvtepi16_ph (__m512i a);
VCVTW2PH __m512h _mm512_mask_cvtepi16_ph (__m512h src, __mmask32 k, __m512i a);
VCVTW2PH __m512h _mm512_maskz_cvtepi16_ph (__mmask32 k, __m512i a);

SIMD Floating-Point Exceptions

Precision.

Other Exceptions

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

Sources