A digression around VA's SRC #012b

[Albert Chan]

Hi, C.Ret

Below give some idea of time it take, for a physical HP71B.

I am basing numbers from my "fastest" Pentium 3 866Mhz, 512MB XP machine, running Emu71/DOS
If Emu71/DOS can run on modern machine without DOS emulator, I expected another 10X speedup.

FNQ(X) version: 9.84 + 23.17 + 26.72 = 59.13

From your 3:12:26 and 3:27:03 timings, Emulator speed at 195X to 210X. Let's say 200X

Y=X+1 version: 9.84 + 50.51 + 58.00 = 118.35 --> HP71B take 6½ hours.

We could speed up a bit, assuming Q = IP(Q) + FP(Q) = q + r

70 FOR I=N TO 0 STEP -1 @ R=(Q+R)*Y-R @ Q=D(I)-Q+IP(R) @ R=FP(R) @ NEXT I

IP-FP version: 9.85 + 45.31 + 51.87 = 107.03 --> HP71B take 6 hours.

All 3 versions produce the same real roots.

---

My Toshiba laptop, Intel i3 M370 2.40 GHz 4 GB, running original FNQ(x) code.
DosBox SVN-lfn can pushed up to 105% (weird, but it does)

DosBox, max 50%: 102.86 + 177.95 + 208.74 = 489.55 --> 25X (HP71B speed)
DosBox, max 105%: 65.05 + 111.32 + 130.74 = 307.11 --> 40X (HP71B speed)

For other programs, my laptop run at 8X my old Pentium Box.
But laptop at max speed, Emu71/DOS + DosBox still run 5 times slower.

--> Cost of DOS emulation layer ≈ 8*5 = 40X

This matched J-F Garnier numbers.

(06-04-2020 08:18 AM)J-F Garnier Wrote:  As a comparison, Emu71/DOS runs at about x300 speed natively on my W2K 32-bit system, and at x8 speed in DOSbox on my W10 64-bit system, so a ratio of 40 due to the DOSbox emulation layer.