arcv: implement bit-extract fusion

The bitfield zero_extract operation is normally expanded into an srai
followed by an andi.  (With the ZBS extension enabled, the special case
of 1-bit zero-extract is implemented with the bexti insn.)  However,
since the RHX core can execute a shift-left and a shift-right of the
same register in 1 cycle, we would prefer to emit those two instructions
instead, and schedule them together so that macro fusion can take place.

The required steps to achieve this are:

(1) Create an insn_and_split that handles the zero_extract RTX;
(2) Tell the combiner to use that split by lowering the cost of the
zero_extract RTX when the target is the RHX core;
(3) Designate the resulting slli + srli pair as fusable by the scheduler.

Attached is a small testcase demonstrating the split, and that the bexti
insn still takes priority over the shift pair.

Signed-off-by: Artemiy Volkov <artemiy@synopsys.com>

Author: artemiy-volkov

gcc/config/riscv/arcv-rhx100.md

@@ -47,6 +47,11 @@
        (eq_attr "type" "imul_fused"))
   "(arcv_rhx100_issueA_fuse0 + arcv_rhx100_issueA_fuse1 + arcv_rhx100_ALU_A_fuse0_early + arcv_rhx100_ALU_A_fuse1_early + arcv_rhx100_MPY32), nothing*3")
 
+(define_insn_reservation "arcv_rhx100_alu_fused" 1
+   (and (eq_attr "tune" "arcv_rhx100")
+       (eq_attr "type" "alu_fused"))
+  "(arcv_rhx100_issueA_fuse0 + arcv_rhx100_issueA_fuse1 + arcv_rhx100_ALU_A_fuse0_early + arcv_rhx100_ALU_A_fuse1_early) | (arcv_rhx100_issueB_fuse0 + arcv_rhx100_issueB_fuse1 + arcv_rhx100_ALU_B_fuse0_early + arcv_rhx100_ALU_B_fuse1_early)")
+
 (define_insn_reservation "arcv_rhx100_jmp_insn" 1
   (and (eq_attr "tune" "arcv_rhx100")
        (eq_attr "type" "branch,jump,call,jalr,ret,trap"))

gcc/config/riscv/riscv.cc

@@ -4337,7 +4337,8 @@ riscv_rtx_costs (rtx x, machine_mode mode, int outer_code, int opno ATTRIBUTE_UN
 	}
       gcc_fallthrough ();
     case SIGN_EXTRACT:
-      if (TARGET_XTHEADBB && outer_code == SET
+      if ((riscv_is_micro_arch (arcv_rhx100) || TARGET_XTHEADBB)
+	  && outer_code == SET
 	  && CONST_INT_P (XEXP (x, 1))
 	  && CONST_INT_P (XEXP (x, 2)))
 	{
@@ -11198,6 +11199,13 @@ arcv_macro_fusion_pair_p (rtx_insn *prev, rtx_insn *curr)
 	|| REGNO (SET_DEST (prev_set)) == REGNO (XEXP (SET_SRC (curr_set), 1))))
     return true;
 
+  /* Fuse logical shift left with logical shift right (bit-extract pattern). */
+  if (GET_CODE (SET_SRC (prev_set)) == ASHIFT
+      && GET_CODE (SET_SRC (curr_set)) == LSHIFTRT
+      && REGNO (SET_DEST (prev_set)) == REGNO (SET_DEST (curr_set))
+      && REGNO (SET_DEST (prev_set)) == REGNO (XEXP (SET_SRC (curr_set), 0)))
+    return true;
+
   return false;
 }
 

gcc/config/riscv/riscv.md

@@ -518,7 +518,7 @@
    vslideup,vslidedown,vislide1up,vislide1down,vfslide1up,vfslide1down,
    vgather,vcompress,vmov,vector,vandn,vbrev,vbrev8,vrev8,vclz,vctz,vcpop,vrol,vror,vwsll,
    vclmul,vclmulh,vghsh,vgmul,vaesef,vaesem,vaesdf,vaesdm,vaeskf1,vaeskf2,vaesz,
-   vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,vsm3me,vsm3c,vfncvtbf16,vfwcvtbf16,vfwmaccbf16,imul_fused,
+   vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,vsm3me,vsm3c,vfncvtbf16,vfwcvtbf16,vfwmaccbf16,imul_fused,alu_fused,
    sf_vc,sf_vc_se"
   (cond [(eq_attr "got" "load") (const_string "load")
 
@@ -4564,6 +4564,23 @@
   [(set_attr "type" "imul_fused")]
 )
 
+(define_insn "*zero_extract_fused"
+  [(set (match_operand:SI 0 "register_operand" "=r")
+	(zero_extract:SI (match_operand:SI 1 "register_operand" "r")
+			 (match_operand 2 "const_int_operand")
+			 (match_operand 3 "const_int_operand")))]
+  "riscv_is_micro_arch (arcv_rhx100) && !TARGET_64BIT
+     && (INTVAL (operands[2]) > 1 || !TARGET_ZBS)"
+  {
+     int amount = INTVAL (operands[2]);
+     int end = INTVAL (operands[3]) + amount;
+     operands[2] = GEN_INT (BITS_PER_WORD - end);
+     operands[3] = GEN_INT (BITS_PER_WORD - amount);
+     return "slli\t%0,%1,%2\n\tsrli\t%0,%0,%3";
+  }
+  [(set_attr "type" "alu_fused")]
+)
+
 ;; String compare with length insn.
 ;; Argument 0 is the target (result)
 ;; Argument 1 is the source1

gcc/testsuite/gcc.target/riscv/arcv-fusion-xbfu.c

@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-require-effective-target rv32 } */
+/* { dg-skip-if "" { *-*-* } { "-g" "-flto" "-O0" "-Oz" "-Os" } } */
+/* { dg-options "-mtune=arc-v-rhx-100-series -march=rv32im_zbs -mabi=ilp32" } */
+
+#define bit_extract(x,start,amt) (((x)>>(start)) & (~(0xffffffff << (amt))))
+
+int
+f (int x)
+{
+  return bit_extract(x,10,14) + bit_extract(x,1,1);
+}
+
+/* { dg-final { scan-assembler {\sslli\s([ast][0-9]+),a0,8\n\ssrli\s([ast][0-9]+),\1,18\n\sbexti\sa0,a0,1\n\sadd\sa0,\2,a0\n} } } */