pitrou commented on code in PR #47994:
URL: https://github.com/apache/arrow/pull/47994#discussion_r2783602424
##########
cpp/src/arrow/util/bpacking_dispatch_internal.h:
##########
@@ -188,297 +202,431 @@ void unpack_width(const uint8_t* in, UnpackedUInt* out,
int batch_size, int bit_
return unpack_full(in, out, batch_size);
} else {
using UnpackerForWidth = Unpacker<UnpackedUInt, kPackedBitWidth>;
+ // Number of values extracted by one iteration of the kernel
constexpr auto kValuesUnpacked = UnpackerForWidth::kValuesUnpacked;
+ // Number of bytes read, but not necessarily unpacked, by one iteration
of the
+ // kernel. This constant prevent reading past buffer end.
+ constexpr auto kBytesRead = UnpackerForWidth::kBytesRead;
+
+ if constexpr (kValuesUnpacked > 0) {
+ const uint8_t* in_last = in_end - kBytesRead;
+ // Running the optimized kernel for batch extraction
+ while ((batch_size >= kValuesUnpacked) && (in <= in_last)) {
Review Comment:
I have no idea :) It could make a difference if the compiler decides not to
inline the inner function perhaps.
In any case, let's ignore this for now, perhaps adding a TODO for later
experimentation.
--
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]
