Vesta
English
Source code

VCVTTPH2W

Convert Packed FP16 Values to Signed Word Integers

stableVMJITAOTinstruction

Encodings

OpcodeInstructionOp/En64-bitCompat/LegacyDescription
EVEX.128.66.MAP5.W0 7C /rVCVTTPH2W xmm1{k1}{z}, xmm2/m128/m16bcstAValidValidConvert eight packed FP16 values in AND AVX512VL) xmm2/m128/m16bcst to eight signed word OR AVX10.1 integers, and store the result in xmm1 using truncation subject to writemask k1.
EVEX.256.66.MAP5.W0 7C /rVCVTTPH2W ymm1{k1}{z}, ymm2/m256/m16bcstAValidValidConvert sixteen packed FP16 values in AND AVX512VL) ymm2/m256/m16bcst to sixteen signed word OR AVX10.1 integers, and store the result in ymm1 using truncation subject to writemask k1.
EVEX.512.66.MAP5.W0 7C /rVCVTTPH2W zmm1{k1}{z}, zmm2/m512/m16bcst {sae}AValidValidConvert thirty-two packed FP16 values in OR AVX10.1 zmm2/m512/m16bcst to thirty-two signed word integers, and store the result in zmm1 using truncation 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 FP16 values in the source operand to signed word integers in the destination operand.

When a conversion is inexact, a truncated (round toward zero) value is returned. If a converted result cannot be represented in the destination format, the floating-point invalid exception is raised, and if this exception is masked, the integer indefinite value 8000H is returned.

The destination elements are updated according to the writemask.

Operation

VCVTTPH2W dest, src
VL = 128, 256 or 512
KL := VL / 16

FOR j := 0 TO KL-1:
    IF k1[j] OR *no writemask*:
          IF *SRC is memory* and EVEX.b = 1:
                tsrc := SRC.fp16[0]
          ELSE
                tsrc := SRC.fp16[j]
          DEST.word[j] := Convert_fp16_to_integer16_truncate(tsrc)
    ELSE IF *zeroing*:
          DEST.word[j] := 0
    // else dest.word[j] remains unchanged

DEST[MAXVL-1:VL] := 0

Intel C/C++ compiler intrinsics

VCVTTPH2W __m512i _mm512_cvtt_roundph_epi16 (__m512h a, int sae);
VCVTTPH2W __m512i _mm512_mask_cvtt_roundph_epi16 (__m512i src, __mmask32 k, __m512h a, int sae);
VCVTTPH2W __m512i _mm512_maskz_cvtt_roundph_epi16 (__mmask32 k, __m512h a, int sae);
VCVTTPH2W __m128i _mm_cvttph_epi16 (__m128h a);
VCVTTPH2W __m128i _mm_mask_cvttph_epi16 (__m128i src, __mmask8 k, __m128h a);
VCVTTPH2W __m128i _mm_maskz_cvttph_epi16 (__mmask8 k, __m128h a);
VCVTTPH2W __m256i _mm256_cvttph_epi16 (__m256h a);
VCVTTPH2W __m256i _mm256_mask_cvttph_epi16 (__m256i src, __mmask16 k, __m256h a);
VCVTTPH2W __m256i _mm256_maskz_cvttph_epi16 (__mmask16 k, __m256h a);
VCVTTPH2W __m512i _mm512_cvttph_epi16 (__m512h a);
VCVTTPH2W __m512i _mm512_mask_cvttph_epi16 (__m512i src, __mmask32 k, __m512h a);
VCVTTPH2W __m512i _mm512_maskz_cvttph_epi16 (__mmask32 k, __m512h a);

SIMD Floating-Point Exceptions

Invalid, Precision.

Other Exceptions

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

Sources