#3015 Extremely long compile times for big arrays

[Gabriel Ravier]

char s[2048] = "a";

This takes 28 seconds to compile on my system, on version SDCC : mcs51/z80/z180/r2k/r3ka/gbz80/tlcs90/ez80_z80/z80n/ds390/pic16/pic14/TININative/ds400/hc08/s08/stm8/pdk13/pdk14/pdk15 4.0.2 #0 (Linux) (built from commit 11643).

[Philipp Klaus Krause]

Weirdly, this apparently depends on the backend. I tried a few on my system:

• stm8: 0s
• pdk14: 22s
• z80: 0s
• mcs51: 23s
• hc08: 0s

[Erik Petrich]

The backends that initialize data by doing a bulk copy from code memory compile quickly. The others generate the code equivalent of

s[0]=97;
s[1]=0;
s[2]=0;
s[3]=0;
...
s[2047]=0;

This in turn seems to be choking GCSE and dataflow analysis with several thousand iTemps.

Since data memory should already be zeroed, couldn't it just initialize the non-zero elements (assuming global or static)?

[Philipp Klaus Krause]

RFE [feature-requests:#342] looks related.
Currently SDCC only makes a difference between explicitly initalized vs. implicitly initialized, not checking for zeores that could be omitted in DATA.

A more complicated case would be INITALIZED. There, AFAIK currently is no zero-initalization and everything will be initialized via a copy (0 or not).
Then there is also RFE [feature-requests:#455] and RFE [feature-requests:#380].

[Konstantin Kim]

Philipp,

your RFE links wrongly points to bugs
https://sourceforge.net/p/sdcc/bugs/342/
instead of
https://sourceforge.net/p/sdcc/feature-requests/342/