xtensa assembler is capable of representing register loads with either movi + addmi, l32r or const16, depending on the core configuration. Don't use '.literal' and 'l32r' directly in the code, use 'movi' and let the assembler relax them.
Signed-off-by: Max Filippov <[email protected]> --- libc/string/xtensa/strcmp.S | 17 +++++------------ 1 file changed, 5 insertions(+), 12 deletions(-) diff --git a/libc/string/xtensa/strcmp.S b/libc/string/xtensa/strcmp.S index a16da5da2e92..a106bee675f5 100644 --- a/libc/string/xtensa/strcmp.S +++ b/libc/string/xtensa/strcmp.S @@ -37,11 +37,6 @@ .text .align 4 .literal_position - .literal .Lmask0, MASK0 - .literal .Lmask1, MASK1 - .literal .Lmask2, MASK2 - .literal .Lmask3, MASK3 - .literal .Lmask4, MASK4 ENTRY (strcmp) /* a2 = s1, a3 = s2 */ @@ -130,10 +125,9 @@ ENTRY (strcmp) .align 4 #if XCHAL_HAVE_LOOPS .Laligned: - .begin no-transform movi a11, 0 - l32r a4, .Lmask0 /* mask for byte 0 */ - l32r a7, .Lmask4 + movi a4, MASK0 /* mask for byte 0 */ + movi a7, MASK4 loop a11, .Laligned_done /* Loop forever. */ /* First unrolled loop body. */ @@ -184,10 +178,10 @@ ENTRY (strcmp) If not, loop over the rest of string using normal algorithm. */ bnone a8, a4, .Leq /* if byte 0 is zero */ - l32r a5, .Lmask1 /* mask for byte 1 */ - l32r a6, .Lmask2 /* mask for byte 2 */ + movi a5, MASK1 /* mask for byte 1 */ + movi a6, MASK2 /* mask for byte 2 */ bnone a8, a5, .Leq /* if byte 1 is zero */ - l32r a7, .Lmask3 /* mask for byte 3 */ + movi a7, MASK3 /* mask for byte 3 */ bnone a8, a6, .Leq /* if byte 2 is zero */ bnone a8, a7, .Leq /* if byte 3 is zero */ addi.n a2, a2, 4 /* advance s1 pointer */ @@ -196,7 +190,6 @@ ENTRY (strcmp) /* align (1 mod 4) */ loop a11, .Leq /* loop forever */ - .end no-transform l32i a8, a2, 0 /* get word from s1 */ l32i a9, a3, 0 /* get word from s2 */ -- 2.11.0 _______________________________________________ devel mailing list [email protected] https://mailman.uclibc-ng.org/cgi-bin/mailman/listinfo/devel
