Lines Matching +full:16 +full:- +full:byte
8 * of the Xtensa link-time HAL, and those files may differ per
11 * could lose the special functionality for accessing user-space
31 * do 16B chunks with a loop, and then finish up with
37 * This code tries to use fall-through braches for the common
76 srli a7, a4, 4 # number of loop iterations with 16B
80 __ssa8 a3 # set shift amount from byte offset
89 .Ldst1mod2: # dst is only byte aligned
90 bltui a4, 7, .Lbytecopy # do short copies byte by byte
92 # copy 1 byte
97 addi a4, a4, -1
100 .Ldst2mod4: # dst 16-bit aligned
102 bltui a4, 6, .Lbytecopy # do short copies byte by byte
109 addi a4, a4, -2
113 * Byte by byte copy
116 .byte 0 # 1 mod 4 alignment for LOOPNEZ
138 * Destination and source are word-aligned.
140 # copy 16 bytes per iteration for word-aligned dst and word-aligned src
142 .byte 0 # (0 mod 4 alignment for LBEG)
149 add a8, a8, a3 # a8 = end of last 16B source chunk
159 addi a3, a3, 16
161 addi a5, a5, 16
190 # copy 1 byte
202 .byte 0 # 1 mod 4 alignement for LOOPNEZ
205 # copy 16 bytes per iteration for word-aligned dst and unaligned src
216 add a10, a10, a3 # a10 = end of last 16B source chunk
219 add a12, a12, a3 # a12 = end of last 16B source chunk
230 EX(10f) l32i a6, a3, 16
233 addi a3, a3, 16
236 addi a5, a5, 16
278 # copy 1 byte
292 * bytes_copied = a5 - a2
293 * retval = bytes_not_copied = original len - bytes_copied
294 * retval = a11 - (a5 - a2)
299 sub a2, a5, a2 /* a2 <-- bytes copied */
300 sub a2, a11, a2 /* a2 <-- bytes not copied */