[PATCH v2] RISC-V: Optimize vsetvl AVL for VLS VLMAX auto-vectorization

[pan2.li@intel.com]

From: Pan Li <pan2.li@intel.com>

This patch is optimizing the AVL for VLS auto-vectorzation.

Given below sample code:

typedef int8_t vnx2qi __attribute__ ((vector_size (2)));

__attribute__ ((noipa)) void
f_vnx2qi (int8_t a, int8_t b, int8_t *out)
{
  vnx2qi v = {a, b};
  *(vnx2qi *) out = v;
}

Before this patch:
f_vnx2qi:
        vsetvli a5,zero,e8,mf8,ta,ma
        vmv.v.x v1,a0
        vslide1down.vx  v1,v1,a1
        vse8.v  v1,0(a2)
        ret

After this patch:
f_vnx2qi:
        vsetivli        zero,2,e8,mf8,ta,ma
        vmv.v.x v1,a0
        vslide1down.vx  v1,v1,a1
        vse8.v  v1,0(a2)
        ret

Signed-off-by: Pan Li <pan2.li@intel.com>
Co-authored-by: Juzhe-Zhong <juzhe.zhong@rivai.ai>
Co-authored-by: kito-cheng <kito.cheng@sifive.com>

gcc/ChangeLog:

	* config/riscv/riscv-v.cc (const_vlmax_p): New function for
	deciding the mode is constant or not.
	(set_len_and_policy): Optimize VLS-VLMAX code gen to vsetivli.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/base/vf_avl-1.c: New test.
---
 gcc/config/riscv/riscv-v.cc                   | 26 ++++++++++++++++---
 .../gcc.target/riscv/rvv/base/vf_avl-1.c      | 15 +++++++++++
 2 files changed, 38 insertions(+), 3 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/vf_avl-1.c

diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc
index b8dc333f54e..d65e7300303 100644
--- a/gcc/config/riscv/riscv-v.cc
+++ b/gcc/config/riscv/riscv-v.cc
@@ -50,6 +50,19 @@ using namespace riscv_vector;
 
 namespace riscv_vector {
 
+/* Return true if vlmax is constant value and can be used in vsetivl.  */
+static bool
+const_vlmax_p (machine_mode mode)
+{
+  poly_uint64 nuints = GET_MODE_NUNITS (mode);
+
+  return nuints.is_constant ()
+    /* The vsetivli can only hold register 0~31.  */
+    ? (IN_RANGE (nuints.to_constant (), 0, 31))
+    /* Only allowed in VLS-VLMAX mode.  */
+    : false;
+}
+
 template <int MAX_OPERANDS> class insn_expander
 {
 public:
@@ -101,12 +114,19 @@ public:
 
   void set_len_and_policy (rtx len, bool force_vlmax = false)
     {
-      bool vlmax_p = force_vlmax;
+      bool vlmax_p = force_vlmax || !len;
       gcc_assert (has_dest);
 
-      if (!len)
+      if (vlmax_p && const_vlmax_p (dest_mode))
+	{
+	  /* Optimize VLS-VLMAX code gen, we can use vsetivli instead of the
+	     vsetvli to obtain the value of vlmax.  */
+	  poly_uint64 nunits = GET_MODE_NUNITS (dest_mode);
+	  len = gen_int_mode (nunits, Pmode);
+	  vlmax_p = false; /* It has became NONVLMAX now.  */
+	}
+      else if (!len)
 	{
-	  vlmax_p = true;
 	  len = gen_reg_rtx (Pmode);
 	  emit_vlmax_vsetvl (dest_mode, len);
 	}
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/vf_avl-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/vf_avl-1.c
new file mode 100644
index 00000000000..11adf6bc611
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/vf_avl-1.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-options "-O3 -march=rv32gcv -mabi=ilp32d --param riscv-autovec-preference=fixed-vlmax" } */
+
+#include <stdint-gcc.h>
+
+typedef int8_t vnx2qi __attribute__ ((vector_size (2)));
+
+__attribute__ ((noipa)) void
+f_vnx2qi (int8_t a, int8_t b, int8_t *out)
+{
+  vnx2qi v = {a, b};
+  *(vnx2qi *) out = v;
+}
+
+/* { dg-final { scan-assembler-times {vsetivli\s+zero,\s*2,\s*e8,\s*mf8,\s*t[au],\s*m[au]} 1 } } */
-- 
2.34.1