6502/CLK
1
0
Fork 0
mirror of https://github.com/TomHarte/CLK.git synced 2025-01-12 00:30:31 +00:00
Code Issues Projects Releases Wiki Activity

Tweak per empirical results.

Browse Source
This commit is contained in:
Thomas Harte 2022-05-27 15:37:40 -04:00
parent e11990e453
commit e8dd8215ba
1 changed files with 11 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
Processors/68000Mk2/Implementation/68000Mk2Implementation.hpp
View File

@ -2576,22 +2576,26 @@ template <bool did_overflow> void ProcessorBase::did_divs(int32_t dividend, int3
return;
}
// It's either five or six microcycles to get into the main loop, depending
// It's either six or seven microcycles to get into the main loop, depending
// on dividend sign.
dynamic_instruction_length_ = 5 + (dividend < 0);
dynamic_instruction_length_ = 6 + (dividend < 0);
if(did_overflow) {
return;
}
// There's always a cost of four microcycles per bit, plus an additional
// one for each that is non-zero.
// There's a fixed cost per bit, plus an additional one for each that is zero.
//
// The sign bit does not count here; it's the low fifteen bits that matter
// The sign bit does not count here; it's the high fifteen bits that matter
// only, in the unsigned version of the result.
dynamic_instruction_length_ += 60;
//
// Disclaimer: per the flowchart it looks to me like this constant should be 60
// rather than 49 — four microcycles per bit. But the number 49 makes this
// algorithm exactly fit the stated minimum and maximum costs. Possibly the
// undefined difference between a nop cycle an an idle wait is relevant here?
dynamic_instruction_length_ += 49;
int result_bits = abs(dividend / divisor) & 0x7fff;
int result_bits = ~abs(dividend / divisor) & 0xfffe;
convert_to_bit_count_16(result_bits);
dynamic_instruction_length_ += result_bits;