Corrected comment in add sv_expf_inline.h

[ClaudioMartino]

In sv_expf_inline.h, lines 55 to 63:

/* y = exp(r) - 1 ~= r + C0 r^2 + C1 r^3 + C2 r^4 + C3 r^5 + C4 r^6.  */
svfloat32_t p12 = svmla_lane (sv_f32 (d->c1), r, lane_consts, 2);
svfloat32_t p34 = svmla_lane (sv_f32 (d->c3), r, lane_consts, 3);
svfloat32_t r2 = svmul_x (svptrue_b32 (), r, r);
svfloat32_t p14 = svmla_x (pg, p12, p34, r2);
svfloat32_t p0 = svmul_lane (r, lane_consts, 1);
svfloat32_t poly = svmla_x (pg, p0, r2, p14);
return svmla_x (pg, scale, scale, poly);

lane_consts contains ln2_lo, c0, c2, c4, (lane_consts = svld1rq (svptrue_b32 (), &d->ln2_lo)), hence:

• p12 = svmla_lane (sv_f32 (d->c1), r, lane_consts, 2) is computing c1 + r c2
• p34 = svmla_lane (sv_f32 (d->c3), r, lane_consts, 3) is computing c3 + r c4
• p0 = svmul_lane (r, lane_consts, 1) is computing r c0

In p14 there is p12 + r^2 p34 = c1 + r c2 + r^2 (c3 + r c4) = c1 + r c2 + r^2 c3 + r^3 c4.
poly will contain the values for p0 + r^2 p14 = r c0 + r^2 (c1 + r c2 + r^2 c3 + r^3 c4) = r c0 + r^2 c1 + r^3 c2 + r^4 c3 + r^5 c4
Finally the returned value is scale + scale * poly = scale (1 + poly) = scale * exp(r).

Hence poly is a polynomial of degree 5, not 6 as written in the comment, and the coefficients have to be changed accordingly (c0 with r, c1 with r^2, etc.).
Since on line 42 the final result has been defined as exp(x) = 2^n (1 + poly(r)) and y is not used elsewhere, I would also change the y on line 55 with poly(r).

The error may come from exp.c where c0 is the coefficient of r^2:

/* y = exp(r) - 1 ~= r + C0 r^2 + C1 r^3 + C2 r^4 + C3 r^5.  */
svfloat64_t r2 = svmul_x (pg, r, r);
svfloat64_t p01 = svmla_x (pg, C (0), C (1), r);
svfloat64_t p23 = svmla_x (pg, C (2), C (3), r);
svfloat64_t p04 = svmla_x (pg, p01, p23, r2);
svfloat64_t y = svmla_x (pg, r, p04, r2);
svfloat64_t s = svexpa (u);
return svmla_x (pg, s, s, y)

[blapie]

Hello, Thanks for the contribution, as agreed we will follow the Arm-internal review process.

[ClaudioMartino]

Ok, thank you