Difference between revisions of "The code generator"
From Lazarus wiki
Jump to navigationJump to search| Line 1: | Line 1: | ||
With the version 1.9.x of FPC the code generator has been completly rewritten to fit the needs of a compiler with multiple processor target support. | With the version 1.9.x of FPC the code generator has been completly rewritten to fit the needs of a compiler with multiple processor target support. | ||
| + | |||
| + | |||
| + | == MM support == | ||
| + | |||
| + | === The shuffle operand === | ||
| + | |||
| + | { this record describes shuffle operations for mm operations; if a pointer a shuffle record | ||
| + | passed to an mm operation is nil, it means that the whole location is moved } | ||
| + | tmmshuffle = record | ||
| + | { describes how many shuffles are actually described, if len=0 then | ||
| + | moving the scalar with index 0 to the scalar with index 0 is meant } | ||
| + | len : byte; | ||
| + | { lower nibble of each entry of this array describes index of the source data index while | ||
| + | the upper nibble describes the destination index } | ||
| + | shuffles : array[1..1] of byte; | ||
| + | end; | ||
Revision as of 23:12, 20 December 2003
With the version 1.9.x of FPC the code generator has been completly rewritten to fit the needs of a compiler with multiple processor target support.
MM support
The shuffle operand
{ this record describes shuffle operations for mm operations; if a pointer a shuffle record
passed to an mm operation is nil, it means that the whole location is moved }
tmmshuffle = record
{ describes how many shuffles are actually described, if len=0 then
moving the scalar with index 0 to the scalar with index 0 is meant }
len : byte;
{ lower nibble of each entry of this array describes index of the source data index while
the upper nibble describes the destination index }
shuffles : array[1..1] of byte;
end;
