68000 and 680x instruction sets

[Jonathan de Boyne Pollard]

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a single
bit negating the sense of the branch test, there was a branch-never
instruction, BRN, too. The same great taste as a branch, but zero calories.

[Nick Spalding]

Jonathan de Boyne Pollard wrote, in

Post by Jonathan de Boyne Pollard

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a single
bit negating the sense of the branch test, there was a branch-never
instruction, BRN, too. The same great taste as a branch, but zero calories.

Like the 1410/7010 move instruction where what got moved, how much of
it, and in what direction depended on a 6-bit modifier which when all
three 'what' bits were zero moved nothing but left the address registers
as they would have been if something had moved.

[jmfbahciv]

Post by Jonathan de Boyne Pollard

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a single
bit negating the sense of the branch test, there was a branch-never
instruction, BRN, too. The same great taste as a branch, but zero calories.

People used NOOPs because they were NOOPs.

/BAH

[Jonathan de Boyne Pollard]

Post by jmfbahciv

Post by Jonathan de Boyne Pollard

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a
single bit negating the sense of the branch test, there was a
branch-never instruction, BRN, too. The same great taste as a
branch, but zero calories.

People used NOOPs because they were NOOPs.

Actually, people used branch never to generate longer delays than a
NOP. As I said, BRN has the same great taste as a branch. It expends
all of the cycles that any other type of branch would (in the
branch-not-taken case). A 1-byte NOP takes 2 cycles. A 2-byte BRN
takes 3 cycles. (And a 4-byte LBRN takes 5 cycles.)

[Charles Richmond]

Post by Jonathan de Boyne Pollard

Post by jmfbahciv

Post by Jonathan de Boyne Pollard

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a
single bit negating the sense of the branch test, there was a
branch-never instruction, BRN, too. The same great taste as a
branch, but zero calories.

People used NOOPs because they were NOOPs.

Actually, people used branch never to generate longer delays than a
NOP. As I said, BRN has the same great taste as a branch. It expends
all of the cycles that any other type of branch would (in the
branch-not-taken case). A 1-byte NOP takes 2 cycles. A 2-byte BRN
takes 3 cycles. (And a 4-byte LBRN takes 5 cycles.)

And for *really* long delays, use something like this:


LABEL JMP LABEL


;-)

[jmfbahciv]

Post by Jonathan de Boyne Pollard

Post by jmfbahciv

Post by Jonathan de Boyne Pollard

There also was an unconditional branch, BRA, [...]

... and because of the way that the opcodes were encoded, with a
single bit negating the sense of the branch test, there was a
branch-never instruction, BRN, too. The same great taste as a
branch, but zero calories.

People used NOOPs because they were NOOPs.

Actually, people used branch never to generate longer delays than a
NOP. As I said, BRN has the same great taste as a branch. It expends
all of the cycles that any other type of branch would (in the
branch-not-taken case). A 1-byte NOP takes 2 cycles. A 2-byte BRN
takes 3 cycles. (And a 4-byte LBRN takes 5 cycles.)

Which instruction used depended on the reason it was used; some
reasons needed core to be touched others not. I used a NOOP as
the instruction of an error return whenever I wanted to ignore
it.

/BAH