mapleFU commented on code in PR #43763:
URL: https://github.com/apache/arrow/pull/43763#discussion_r1738056046
##########
cpp/src/arrow/compute/row/row_encoder_internal.cc:
##########
@@ -145,41 +145,35 @@ void FixedWidthKeyEncoder::AddLengthNull(int32_t* length)
{
Status FixedWidthKeyEncoder::Encode(const ExecValue& data, int64_t
batch_length,
uint8_t** encoded_bytes) {
+ auto handle_next_valid_value = [&](std::string_view bytes) {
Review Comment:
Emm let me post a blog in chinese:
https://www.ykiko.me/en/articles/692886292/
> The efficiency of passing structures smaller than 16 bytes by value is
higher than by reference
This statement has been around for a long time, but I have never found the
basis for it. Finally, when I was studying the calling convention recently, I
found the reason. First of all, if the structure size is less than or equal to
8 bytes, it can be directly stuffed into a 64-bit register for parameter
passing. Passing parameters through registers requires fewer memory accesses
than passing parameters through references, which is more efficient and no
problem. What about 16 bytes? System V ABI allows a 16-byte structure to be
split into two 8-byte parts, and then passed using registers separately. In
this case, passing by value is indeed higher than passing by reference.
In general, they would be same, and passing by value is same. And 16b can be
more lightweight in some abi.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
