mirrors/kitty
1
0
Fork 0
mirror of https://github.com/kovidgoyal/kitty synced 2026-07-19 15:04:50 +02:00
Code Issues Packages Projects Releases Wiki Activity

Dont precalculate mask in loop body

Browse Source 
No need since we dont shift. Avoids the extra mask instructions for the
not found case.
This commit is contained in:
Kovid Goyal
2024-02-11 06:41:41 +05:30
parent a32e1aafa6
commit b5edd9ad57
1 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tools/simdstring/generate.go
View File

@@ -1368,7 +1368,7 @@ func (s *State) index_func(f *Function, test_bytes func(bytes_to_test, test_ans
f.MaskForCountDestructive(test_ans, mask)
f.Comment("We need to shift out the possible extra bytes at the start of the string caused by the unaligned read")
f.ShiftMaskRightDestructive(mask, extra_bytes)
f.JumpIfNonZero(mask, "byte_found")
f.JumpIfNonZero(mask, "byte_found_in_mask")
}()
f.Comment("Now loop over aligned blocks")
@@ -1383,12 +1383,13 @@ func (s *State) index_func(f *Function, test_bytes func(bytes_to_test, test_ans
f.Label("loop_body")
f.LoadPointerAligned(pos, bytes_to_test)
test_bytes(bytes_to_test, test_ans)
f.MaskForCountDestructive(test_ans, mask)
f.JumpIfNonZero(mask, "byte_found")
f.JumpIfNonZero(test_ans, "byte_found_in_vec")
f.JumpTo("loop_start")
f.Label("byte_found_in_vec")
f.MaskForCountDestructive(test_ans, mask)
f.Comment("Get the result from", mask, "and return it")
f.Label("byte_found")
f.Label("byte_found_in_mask")
f.CountLeadingZeroBytesInMask(mask, mask)
f.AddToSelf(mask, pos)
f.JumpIfLessThan(mask, limit, "result_in_bounds")