Work buffer length considerations

The work buffer length has the same considerations for both 31-bit and 64-bit. There are two main considerations, both of them are related to the collation level you specify. Following are the two possibilities:

Case 1 - CUNBOPRM_Coll_Level = 1, 2 or 3. For this case, you must consider at least twice the value of the source length (CUNBOPRM_SrcX_Buf_Len * 2), where X could be 1 or 2.
Case 2 - CUNBOPRM_Coll_Level = 4. For this level, you must require at least three times the value of the source (SrcX_Buf_Len * 3), where X could be 1 or 2.

For UCA version UCA400R1 and higher, the following table shows the size of the work buffers for most common UTF-16BE Code Points:

Table 1. Size of the work buffers for UTF-16BE Code Points
Collation Level / Strength	Work Buffer length per Code Point in Source buffer
1	4 - Bytes
2	7 - Bytes
3	9 - Bytes
4	12 - Bytes
5	15 - Bytes

Note:

Most common UTF-16BE Code Points require 2-bytes in Source buffer. Non-normal CP's are expansions, contractions, surrogates, surrogates expansions and surrogates contractions.

IBM® recommends allocating the same bytes for work buffer as for target buffer, see Target buffer length considerations. If Collation returns with RC = CUN_RC_USER_ERR, RS = CUN_RS_WRK_EXHAUSTED by following this recommendation (Wrk Buffer Len = Target buffer length), it is recommended to multiply failed work buffer length by 2 and so on.