diff mbox series

[committed,PR,target/109508] Adjust conditional move expansion for SFB

Message ID 9a3d9170-40de-597a-250b-6bec3445c4db@ventanamicro.com
State Committed
Commit a647198fcf7463a42c8e035a429200e7998735dc
Headers
Series [committed,PR,target/109508] Adjust conditional move expansion for SFB |

Commit Message

Jeff Law April 16, 2023, 3:57 p.m. UTC 
  Recently the conditional move expander's predicates were loosened for 
the benefit of the THEAD processors.  In particular one operand that was 
previously "register_operand" is now "reg_or_0_operand".  That's fine 
for THEAD, but breaks for SFB which requires a register for that operand.

This results in an ICE when compiling the testcase an SFB target such as 
the sifive s76.

This change adjusts the expansion code slightly to copy the value into a 
register for SFB.

Bootstrapped and regression tested (c,c++,fortran only) with a toolchain 
configured to enable SFB by default.

Installing on the trunk momentarily.

Jeff
commit a647198fcf7463a42c8e035a429200e7998735dc
Author: Jeff Law <jlaw@ventanamicro>
Date:   Sun Apr 16 09:55:32 2023 -0600

    [committed] [PR target/109508] Adjust conditional move expansion for SFB
    
    Recently the conditional move expander's predicates were loosened for the
    benefit of the THEAD processors.  In particular one operand that was
    previously "register_operand" is now "reg_or_0_operand".  That's fine for
    THEAD, but breaks for SFB which requires a register for that operand.
    
    This results in an ICE when compiling the testcase an SFB target such as
    the sifive s76.
    
    This change adjusts the expansion code slightly to copy the value into
    a register for SFB.
    
    Bootstrapped and regression tested (c,c++,fortran only) with a toolchain
    configured to enable SFB by default.
    
            PR target/109508
    gcc/
    
            * config/riscv/riscv.cc (riscv_expand_conditional_move): For
            TARGET_SFB_ALU, force the true arm into a register.
    
    gcc/testsuite
            * gcc.target/riscv/pr109508.c: New test.
diff mbox series

Patch

diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
index dc47434fac4..e88fa2d6337 100644
--- a/gcc/config/riscv/riscv.cc
+++ b/gcc/config/riscv/riscv.cc
@@ -3419,6 +3419,12 @@  riscv_expand_conditional_move (rtx dest, rtx op, rtx cons, rtx alt)
     {
       riscv_emit_int_compare (&code, &op0, &op1);
       rtx cond = gen_rtx_fmt_ee (code, GET_MODE (op0), op0, op1);
+
+      /* The expander allows (const_int 0) for CONS for the benefit of
+	 TARGET_XTHEADCONDMOV, but that case isn't supported for
+	 TARGET_SFB_ALU.  So force that operand into a register if
+	 necessary.  */
+      cons = force_reg (GET_MODE (dest), cons);
       emit_insn (gen_rtx_SET (dest, gen_rtx_IF_THEN_ELSE (GET_MODE (dest),
 							  cond, cons, alt)));
       return true;
diff --git a/gcc/testsuite/gcc.target/riscv/pr109508.c b/gcc/testsuite/gcc.target/riscv/pr109508.c
new file mode 100644
index 00000000000..65f291e17ed
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/pr109508.c
@@ -0,0 +1,12 @@ 
+/* { dg-do compile } */
+/* { dg-options "-mcpu=sifive-s76" } */
+
+typedef char __attribute__((__vector_size__ (1))) V;
+
+V v;
+
+void
+foo (void)
+{
+  (char) __builtin_shuffle (0 % v, (V){6}, v);
+}