vcvtps2dq

Convert Packed Single to Packed Doubleword

VCVTPS2DQ ymm1, ymm2/m256

Converts four floats to 32-bit integers (Rounded).

Details

Converts four packed single-precision floating-point values to four packed 32-bit signed doubleword integers using the current MXCSR rounding mode. Overflowing results saturate to INT32_MIN or INT32_MAX; inexact results may set the PE flag in MXCSR. Available in 128-bit (XMM) and 256-bit (YMM) variants; the 256-bit form converts eight floats. No legacy x86 flags are modified.

Pseudocode Operation

if VEX.L = 0:
  for i in 0 to 1:
    xmm1[i] ← convert_float_to_dword_rounded(xmm2[i])
else:
  for i in 0 to 3:
    ymm1[i] ← convert_float_to_dword_rounded(ymm2[i])
update_MXCSR_flags()

Example

VCVTPS2DQ ymm1, ymm2/m256

Encoding

Binary Layout

VEX

Format VEX

Opcode VEX.256.66.0F.WIG 5B /r

Extension AVX

Operands

dest
256-bit YMM AVX register
src
256-bit YMM AVX register or Memory operand

Reference (Intel® SDM)

Instruction Forms

Opcode	Instruction	Op/En	64/32-bit Mode	CPUID	Description
66 0F 5B /r	CVTPS2DQ xmm1, xmm2/m128	A	V/V	SSE2	Convert four packed single precision floating-point values from xmm2/mem to four packed signed doubleword values in xmm1.
VEX.128.66.0F.WIG 5B /r	VCVTPS2DQ xmm1, xmm2/m128	A	V/V	AVX	Convert four packed single precision floating-point values from xmm2/mem to four packed signed doubleword values in xmm1.
VEX.256.66.0F.WIG 5B /r	VCVTPS2DQ ymm1, ymm2/m256	A	V/V	AVX	Convert eight packed single precision floating-point values from ymm2/mem to eight packed signed doubleword values in ymm1.
EVEX.128.66.0F.W0 5B /r	VCVTPS2DQ xmm1 {k1}{z}, xmm2/m128/m32bcst	B	V/V	(AVX512VL AND AVX512F) OR AVX10.1	Convert four packed single precision floating-point values from xmm2/m128/m32bcst to four packed signed doubleword values in xmm1 subject to writemask k1.
EVEX.256.66.0F.W0 5B /r	VCVTPS2DQ ymm1 {k1}{z}, ymm2/m256/m32bcst	B	V/V	(AVX512VL AND AVX512F) OR AVX10.1	Convert eight packed single precision floating-point values from ymm2/m256/m32bcst to eight packed signed doubleword values in ymm1 subject to writemask k1.
EVEX.512.66.0F.W0 5B /r	VCVTPS2DQ zmm1 {k1}{z}, zmm2/m512/m32bcst {er}	B	V/V	AVX512F OR AVX10.1	Convert sixteen packed single precision floating-point values from zmm2/m512/m32bcst to sixteen packed signed doubleword values in zmm1 subject to writemask k1.

Instruction Operand Encoding

Op/En	Tuple Type	Operand 1	Operand 2	Operand 3	Operand 4
A	N/A	ModRM:reg (w)	ModRM:r/m (r)	N/A	N/A
B	Full	ModRM:reg (w)	ModRM:r/m (r)	N/A	N/A

Description

Converts four, eight or sixteen packed single precision floating-point values in the source operand to four, eight or sixteen signed doubleword integers in the destination operand. When a conversion is inexact, the value returned is rounded according to the rounding control bits in the MXCSR register or the embedded rounding control bits. 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 indefinite integer value 80000000H is returned. EVEX encoded versions: The source operand is a ZMM register, a 512-bit memory location or a 512-bit vector broadcasted from a 32-bit memory location. The destination operand is a ZMM register conditionally updated with writemask k1. VEX.256 encoded version: The source operand is a YMM register or 256- bit memory location. The destination operand is a YMM register. The upper bits (MAXVL-1:256) of the corresponding ZMM register destination are zeroed. VEX.128 encoded version: The source operand is an XMM register or 128- bit memory location. The destination operand is a XMM register. The upper bits (MAXVL-1:128) of the corresponding ZMM register destination are zeroed. 128-bit Legacy SSE version: The source operand is an XMM register or 128- bit memory location. The destination operand is an XMM register. The upper bits (MAXVL-1:128) of the corresponding ZMM register destination are unmodified. CVTPS2DQ—Convert Packed Single Precision Floating-Point Values to Packed Signed Doubleword Integer Values Vol. 2A 3-225 VEX.vvvv and EVEX.vvvv are reserved and must be 1111b otherwise instructions will #UD.

Operation

VCVTPS2DQ (Encoded Versions) When SRC Operand is a Register
(KL, VL) = (4, 128), (8, 256), (16, 512)
IF (VL = 512) AND (EVEX.b = 1)
THEN
SET_ROUNDING_MODE_FOR_THIS_INSTRUCTION(EVEX.RC);
ELSE
SET_ROUNDING_MODE_FOR_THIS_INSTRUCTION(MXCSR.RC);
FI;

FOR j := 0 TO KL-1
i := j * 32
IF k1[j] OR *no writemask*
THEN DEST[i+31:i] :=
Convert_Single_Precision_Floating_Point_To_Integer(SRC[i+31:i])
ELSE
IF *merging-masking*                                 ; merging-masking
THEN *DEST[i+31:i] remains unchanged*
ELSE                                                         ; zeroing-masking
DEST[i+31:i] := 0
FI
FI;
ENDFOR
DEST[MAXVL-1:VL] := 0

VCVTPS2DQ (EVEX Encoded Versions) When SRC Operand is a Memory Source
(KL, VL) = (4, 128), (8, 256), (16, 512)

FOR j := 0 TO 15
i := j * 32
IF k1[j] OR *no writemask*
THEN
IF (EVEX.b = 1)
THEN
DEST[i+31:i] :=
Convert_Single_Precision_Floating_Point_To_Integer(SRC[31:0])
ELSE
DEST[i+31:i] :=
Convert_Single_Precision_Floating_Point_To_Integer(SRC[i+31:i])
FI;
ELSE
IF *merging-masking*                                 ; merging-masking
THEN *DEST[i+31:i] remains unchanged*
ELSE                                                         ; zeroing-masking
DEST[i+31:i] := 0
FI
FI;
ENDFOR
DEST[MAXVL-1:VL] := 0

VCVTPS2DQ (VEX.256 Encoded Version)



CVTPS2DQ—Convert Packed Single Precision Floating-Point Values to Packed Signed Doubleword Integer Values                           Vol. 2A 3-226
DEST[31:0] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[31:0])
DEST[63:32] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[63:32])
DEST[95:64] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[95:64])
DEST[127:96] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[127:96)
DEST[159:128] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[159:128])
DEST[191:160] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[191:160])
DEST[223:192] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[223:192])
DEST[255:224] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[255:224])

VCVTPS2DQ (VEX.128 Encoded Version)
DEST[31:0] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[31:0])
DEST[63:32] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[63:32])
DEST[95:64] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[95:64])
DEST[127:96] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[127:96])
DEST[MAXVL-1:128] := 0

CVTPS2DQ (128-bit Legacy SSE Version)
DEST[31:0] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[31:0])
DEST[63:32] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[63:32])
DEST[95:64] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[95:64])
DEST[127:96] := Convert_Single_Precision_Floating_Point_To_Integer(SRC[127:96])
DEST[MAXVL-1:128] (unmodified)

Intel C/C++ Compiler Intrinsic Equivalent

VCVTPS2DQ __m512i _mm512_cvtps_epi32( __m512 a);
VCVTPS2DQ __m512i _mm512_mask_cvtps_epi32( __m512i s, __mmask16 k, __m512 a);
VCVTPS2DQ __m512i _mm512_maskz_cvtps_epi32( __mmask16 k, __m512 a);
VCVTPS2DQ __m512i _mm512_cvt_roundps_epi32( __m512 a, int r);
VCVTPS2DQ __m512i _mm512_mask_cvt_roundps_epi32( __m512i s, __mmask16 k, __m512 a, int r);
VCVTPS2DQ __m512i _mm512_maskz_cvt_roundps_epi32( __mmask16 k, __m512 a, int r);
VCVTPS2DQ __m256i _mm256_mask_cvtps_epi32( __m256i s, __mmask8 k, __m256 a);
VCVTPS2DQ __m256i _mm256_maskz_cvtps_epi32( __mmask8 k, __m256 a);
VCVTPS2DQ __m128i _mm_mask_cvtps_epi32( __m128i s, __mmask8 k, __m128 a);
VCVTPS2DQ __m128i _mm_maskz_cvtps_epi32( __mmask8 k, __m128 a);
VCVTPS2DQ __ m256i _mm256_cvtps_epi32 (__m256 a)
CVTPS2DQ __m128i _mm_cvtps_epi32 (__m128 a)

Exceptions

SIMD Floating-Point Exceptions

Invalid, Precision.

Other Exceptions

VEX-encoded instructions, see Table 2-19, “Type 2 Class Exception Conditions.” EVEX-encoded instructions, see Table 2-48, “Type E2 Class Exception Conditions.” Additionally: #UD If VEX.vvvv != 1111B or EVEX.vvvv != 1111B. CVTPS2DQ—Convert Packed Single Precision Floating-Point Values to Packed Signed Doubleword Integer Values Vol. 2A 3-227