procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

[RISC-V] Avoid unnecessary sign extension after memcmp

Browse source 
Similar to the str[n]cmp work, this adjusts the block compare expansion to do
its work in X mode with an appropriate lowpart extraction of the results at the
end of the sequence.

This has gone through my tester on rv32 and rv64, but that's it. Waiting on
pre-commit testing before moving forward.

gcc/

	* config/riscv/riscv-string.cc (emit_memcmp_scalar_load_and_compare):
	Set RESULT directly rather than using a temporary.
	(emit_memcmp_scalar_result_calculation): Similarly.
	(riscv_expand_block_compare_scalar): Use CONST0_RTX rather than
	generating new RTL.
	* config/riscv/riscv.md (cmpmemsi): Pass an X mode temporary to the
	expansion routines.  If necessary extract low part of the word to store
	in final result location.
This commit is contained in:
Jeff Law 2024-07-12 07:53:41 -06:00
parent 13757e50ff
commit ae829a2778
2 changed files with 18 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
gcc/config/riscv/riscv-string.cc
View file

@ -663,9 +663,7 @@ emit_memcmp_scalar_load_and_compare (rtx result, rtx src1, rtx src2,
/* Fast-path for a single byte. */
if (cmp_bytes == 1)
{
rtx tmp = gen_reg_rtx (Xmode);
do_sub3 (tmp, data1, data2);
emit_insn (gen_movsi (result, gen_lowpart (SImode, tmp)));
do_sub3 (result, data1, data2);
emit_jump_insn (gen_jump (final_label));
emit_barrier (); /* No fall-through. */
return;
@ -702,12 +700,11 @@ emit_memcmp_scalar_result_calculation (rtx result, rtx data1, rtx data2)
/* Get bytes in big-endian order and compare as words. */
do_bswap2 (data1, data1);
do_bswap2 (data2, data2);
/* Synthesize (data1 >= data2) ? 1 : -1 in a branchless sequence. */
rtx tmp = gen_reg_rtx (Xmode);
emit_insn (gen_slt_3 (LTU, Xmode, Xmode, tmp, data1, data2));
do_neg2 (tmp, tmp);
do_ior3 (tmp, tmp, const1_rtx);
emit_insn (gen_movsi (result, gen_lowpart (SImode, tmp)));
emit_insn (gen_slt_3 (LTU, Xmode, Xmode, result, data1, data2));
do_neg2 (result, result);
do_ior3 (result, result, const1_rtx);
}
/* Expand memcmp using scalar instructions (incl. Zbb).
@ -773,7 +770,7 @@ riscv_expand_block_compare_scalar (rtx result, rtx src1, rtx src2, rtx nbytes)
data1, data2,
diff_label, final_label);
emit_insn (gen_rtx_SET (result, gen_rtx_CONST_INT (SImode, 0)));
emit_move_insn (result, CONST0_RTX (GET_MODE (result)));
emit_jump_insn (gen_jump (final_label));
emit_barrier (); /* No fall-through. */

14
gcc/config/riscv/riscv.md
View file

@ -2675,9 +2675,19 @@
operands[2], operands[3]))
DONE;
if (riscv_expand_block_compare (operands[0], operands[1], operands[2],
rtx temp = gen_reg_rtx (word_mode);
if (riscv_expand_block_compare (temp, operands[1], operands[2],
operands[3]))
DONE;
{
if (TARGET_64BIT)
{
temp = gen_lowpart (SImode, temp);
SUBREG_PROMOTED_VAR_P (temp) = 1;
SUBREG_PROMOTED_SET (temp, SRP_SIGNED);
}
emit_move_insn (operands[0], temp);
DONE;
}
else
FAIL;
})