[gcc(refs/vendors/riscv/heads/gcc-15-with-riscv-opts)] [to-be-committed][RISC-V] Handle 32bit operands in condition for conditional moves

[Jeff Law]

https://gcc.gnu.org/g:0c20acfa89617aec33000027503f0810a91be5b2

commit 0c20acfa89617aec33000027503f0810a91be5b2
Author: Jeff Law <jlaw@ventanamicro.com>
Date:   Sat Jun 7 07:48:46 2025 -0600

    [to-be-committed][RISC-V] Handle 32bit operands in condition for conditional moves
    
    So here's the next chunk of conditional move work from Shreya.
    
    It's been a long standing wart that the conditional move expander does not
    support sub-word operands in the comparison.  Particularly since we have
    support routines to handle the necessary extensions for that case.
    
    This patch adjusts the expander to use riscv_extend_comparands rather than fail
    for that case.  I've built spec2017 before/after this and we definitely get
    more conditional moves and they look sensible from a performance standpoint.
    None are likely hitting terribly hot code, so I wouldn't expect any performance
    jumps.
    
    Waiting on pre-commit testing to do its thing.
    
            * config/riscv/riscv.cc (riscv_expand_conditional_move): Use
            riscv_extend_comparands to extend sub-word comparison arguments.
    
                Co-authored-by: Jeff Law  <jlaw@ventanamicro.com>
    
    (cherry picked from commit 59a3da733a79f621700dd9ddc11a0efc07237c3a)

Diff:
---
 gcc/config/riscv/riscv.cc | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
index 3b7e5cbf5ac4..8aa02110b682 100644
--- a/gcc/config/riscv/riscv.cc
+++ b/gcc/config/riscv/riscv.cc
@@ -5437,13 +5437,18 @@ riscv_expand_conditional_move (rtx dest, rtx op, rtx cons, rtx alt)
       machine_mode mode0 = GET_MODE (op0);
       machine_mode mode1 = GET_MODE (op1);
 
-      /* An integer comparison must be comparing WORD_MODE objects.  We
-	 must enforce that so that we don't strip away a sign_extension
-	 thinking it is unnecessary.  We might consider using
-	 riscv_extend_operands if they are not already properly extended.  */
+      /* An integer comparison must be comparing WORD_MODE objects.
+	 Extend the comparison arguments as necessary.  */
       if ((INTEGRAL_MODE_P (mode0) && mode0 != word_mode)
 	  || (INTEGRAL_MODE_P (mode1) && mode1 != word_mode))
-	return false;
+	riscv_extend_comparands (code, &op0, &op1);
+
+      /* We might have been handed back a SUBREG.  Just to make things
+	 easy, force it into a REG.  */
+      if (!REG_P (op0) && !CONST_INT_P (op0))
+	op0 = force_reg (word_mode, op0);
+      if (!REG_P (op1) && !CONST_INT_P (op1))
+	op1 = force_reg (word_mode, op1);
 
       /* In the fallback generic case use MODE rather than WORD_MODE for
 	 the output of the SCC instruction, to match the mode of the NEG