Vesta
English
Source code

VREDUCEPS

Perform Reduction Transformation on Packed Float32 Values

stableVMJITAOTinstruction

Encodings

OpcodeInstructionOp/En64-bitCompat/LegacyDescription
EVEX.128.66.0F3A.W0 56 /r ibVREDUCEPS xmm1 {k1}{z}, xmm2/m128/m32bcst, imm8AValidValidPerform reduction transformation on packed single- AND AVX512DQ) precision floating-point values in OR AVX10.1 xmm2/m128/m32bcst by subtracting a number of fraction bits specified by the imm8 field. Stores the result in xmm1 register under writemask k1.
EVEX.256.66.0F3A.W0 56 /r ibVREDUCEPS ymm1 {k1}{z}, ymm2/m256/m32bcst, imm8AValidValidPerform reduction transformation on packed single- AND AVX512DQ) precision floating-point values in OR AVX10.1 ymm2/m256/m32bcst by subtracting a number of fraction bits specified by the imm8 field. Stores the result in ymm1 register under writemask k1.
EVEX.512.66.0F3A.W0 56 /r ibVREDUCEPS zmm1 {k1}{z}, zmm2/m512/m32bcst{sae}, imm8AValidValidPerform reduction transformation on packed single- OR AVX10.1 precision floating-point values in zmm2/m512/m32bcst by subtracting a number of fraction bits specified by the imm8 field. Stores the result in zmm1 register under 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
  3. imm8immediate byte after the instruction

Tupla: Full

Measured cost

Loading measurements from arch-data...

Description

Perform reduction transformation of the packed binary encoded single precision floating-point values in the source operand (the second operand) and store the reduced results in binary floating-point format to the destination operand (the first operand) under the writemask k1.

The reduction transformation subtracts the integer part and the leading M fractional bits from the binary floatingpoint source value, where M is a unsigned integer specified by imm8[7:4], see Figure 5-28. Specifically, the reduction transformation can be expressed as: dest = src (ROUND(2Msrc))2-M; where "Round()" treats "src", "2M", and their product as binary floating-point numbers with normalized significand and biased exponents. The magnitude of the reduced result can be expressed by considering src= 2p*man2, where man2' is the normalized significand and p' is the unbiased exponent Then if RC = RNE: 0<=|Reduced Result|<=2p-M-1 Then if RC RNE: 0<=|Reduced Result|<2p-M

This instruction might end up with a precision exception set. However, in case of SPE set (i.e., Suppress Precision Exception, which is imm8[3]=1), no precision exception is reported.

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

Handling of special case of input values are listed in Table 5-27.

Operation

ReduceArgumentSP(SRC[31:0], imm8[7:0])
{

    // Check for NaN
    IF (SRC [31:0] = NAN) THEN

          RETURN (Convert SRC[31:0] to QNaN); FI
    M := imm8[7:4]; // Number of fraction bits of the normalized significand to be subtracted
    RC := imm8[1:0];// Round Control for ROUND() operation
    RC source := imm[2];
    SPE := imm[3];// Suppress Precision Exception
    TMP[31:0] := 2-M *{ROUND(2M*SRC[31:0], SPE, RC_source, RC)}; // ROUND() treats SRC and 2M as standard binary FP values
    TMP[31:0] := SRC[31:0]  TMP[31:0]; // subtraction under the same RC,SPE controls
RETURN TMP[31:0]; // binary encoded FP with biased exponent and normalized significand
}

VREDUCEPS

(KL, VL) = (4, 128), (8, 256), (16, 512)

FOR j := 0 TO KL-1

i := j * 32

IF k1[j] OR *no writemask* THEN

             IF (EVEX.b == 1) AND (SRC *is memory*)

                  THEN DEST[i+31:i] := ReduceArgumentSP(SRC[31:0], imm8[7:0]);

                  ELSE DEST[i+31:i] := ReduceArgumentSP(SRC[i+31:i], imm8[7:0]);

             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

Intel C/C++ compiler intrinsics

VREDUCEPS __m512 _mm512_mask_reduce_ps( __m512 a, int imm, int sae) VREDUCEPS __m512 _mm512_mask_reduce_ps(__m512 s, __mmask16 k, __m512 a, int imm, int sae) VREDUCEPS __m512 _mm512_maskz_reduce_ps(__mmask16 k, __m512 a, int imm, int sae) VREDUCEPS __m256 _mm256_mask_reduce_ps( __m256 a, int imm) VREDUCEPS __m256 _mm256_mask_reduce_ps(__m256 s, __mmask8 k, __m256 a, int imm) VREDUCEPS __m256 _mm256_maskz_reduce_ps(__mmask8 k, __m256 a, int imm) VREDUCEPS __m128 _mm_mask_reduce_ps( __m128 a, int imm) VREDUCEPS __m128 _mm_mask_reduce_ps(__m128 s, __mmask8 k, __m128 a, int imm) VREDUCEPS __m128 _mm_maskz_reduce_ps(__mmask8 k, __m128 a, int imm);

SIMD Floating-Point Exceptions

Invalid, Precision. If SPE is enabled, precision exception is not reported (regardless of MXCSR exception mask).

Other Exceptions

See Table 2-48, "Type E2 Class Exception Conditions"; additionally:

#UD                    If EVEX.vvvv != 1111B.

Sources