Floating-point complex multiply-add by indexed values with rotate
Multiply the duplicated real components for rotations 0 and 180, or imaginary components for rotations 90 and 270, of the floating-point complex numbers in each 128-bit segment of the first source vector by the specified complex number in the corresponding the second source vector segment rotated by 0, 90, 180 or 270 degrees in the direction from the positive real axis towards the positive imaginary axis, when considered in polar representation.
Then destructively add the products to the corresponding components of the complex numbers in the addend and destination vector, without intermediate rounding.
These transformations permit the creation of a variety of multiply-add and multiply-subtract operations on complex numbers by combining two of these instructions with the same vector operands but with rotations that are 90 degrees apart.
Each complex number is represented in a vector register as an even/odd pair of elements with the real part in the even-numbered element and the imaginary part in the odd-numbered element.
The complex numbers within the second source vector are specified using an immediate index which selects the same complex number position within each 128-bit vector segment. The index range is from 0 to one less than the number of complex numbers per 128-bit segment, encoded in 1 to 2 bits depending on the size of the complex number. This instruction is unpredicated.
It has encodings from 2 classes: Half-precision and Single-precision
31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
0 | 1 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | 1 | i2 | Zm | 0 | 0 | 0 | 1 | rot | Zn | Zda | ||||||||||||
size<1> | size<0> |
if !IsFeatureImplemented(FEAT_SVE) && !IsFeatureImplemented(FEAT_SME) then UNDEFINED; constant integer esize = 16; constant integer index = UInt(i2); constant integer n = UInt(Zn); constant integer m = UInt(Zm); constant integer da = UInt(Zda); constant integer sel_a = UInt(rot<0>); constant integer sel_b = UInt(NOT(rot<0>)); constant boolean neg_i = (rot<1> == '1'); constant boolean neg_r = (rot<0> != rot<1>);
31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |
0 | 1 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 1 | 1 | i1 | Zm | 0 | 0 | 0 | 1 | rot | Zn | Zda | ||||||||||||
size<1> | size<0> |
if !IsFeatureImplemented(FEAT_SVE) && !IsFeatureImplemented(FEAT_SME) then UNDEFINED; constant integer esize = 32; constant integer index = UInt(i1); constant integer n = UInt(Zn); constant integer m = UInt(Zm); constant integer da = UInt(Zda); constant integer sel_a = UInt(rot<0>); constant integer sel_b = UInt(NOT(rot<0>)); constant boolean neg_i = (rot<1> == '1'); constant boolean neg_r = (rot<0> != rot<1>);
<Zda> |
Is the name of the third source and destination scalable vector register, encoded in the "Zda" field. |
<Zn> |
Is the name of the first source scalable vector register, encoded in the "Zn" field. |
<const> |
Is the const specifier,
encoded in
|
CheckSVEEnabled(); constant integer VL = CurrentVL; constant integer PL = VL DIV 8; constant integer pairs = VL DIV (2 * esize); constant integer pairspersegment = 128 DIV (2 * esize); constant bits(VL) operand1 = Z[n, VL]; constant bits(VL) operand2 = Z[m, VL]; constant bits(VL) operand3 = Z[da, VL]; bits(VL) result; for p = 0 to pairs-1 segmentbase = p - (p MOD pairspersegment); s = segmentbase + index; addend_r = Elem[operand3, 2 * p + 0, esize]; addend_i = Elem[operand3, 2 * p + 1, esize]; elt1_a = Elem[operand1, 2 * p + sel_a, esize]; elt2_a = Elem[operand2, 2 * s + sel_a, esize]; elt2_b = Elem[operand2, 2 * s + sel_b, esize]; if neg_r then elt2_a = FPNeg(elt2_a, FPCR); if neg_i then elt2_b = FPNeg(elt2_b, FPCR); addend_r = FPMulAdd(addend_r, elt1_a, elt2_a, FPCR); addend_i = FPMulAdd(addend_i, elt1_a, elt2_b, FPCR); Elem[result, 2 * p + 0, esize] = addend_r; Elem[result, 2 * p + 1, esize] = addend_i; Z[da, VL] = result;
This instruction might be immediately preceded in program order by a MOVPRFX instruction. The MOVPRFX must conform to all of the following requirements, otherwise the behavior of the MOVPRFX and this instruction is constrained unpredictable:
Internal version only: aarchmrs v2024-03_relA, pseudocode v2024-03_rel, sve v2024-03_rel ; Build timestamp: 2024-03-26T09:45
Copyright © 2010-2024 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.