[VSETVL PASS] RISC-V: Optimize local AVL propagation

[Juzhe-Zhong <juzhe.zhong@rivai.ai>]

I recently noticed that current VSETVL pass has a unnecessary restriction on local
AVL propgation.

Consider this following case:

+                      insn 1: vsetvli a5,a3,e8,mf4,ta,mu
+                      insn 2: vsetvli zero,a5,e32,m1,ta,ma
+                      ...
+                      vle32.v v1,0(a1)
+                      vsetvli a2,zero,e32,m1,ta,ma
+                      vadd.vv v1,v1,v1
+                      vsetvli zero,a5,e32,m1,ta,ma
+                      vse32.v v1,0(a0)
+                      ...
+                      insn 3: sub     a3,a3,a5
+                      ...

We failed to elide insn 2 (vsetvl insn) since insn 3 is modifying "a3" AVL.
Actually, we don't really care about insn 3 since we should only check and make sure
there is no insn between insn 1 and insn 2 that modifies "a3" AVL. Then, we can propgate
AVL "a3" from insn 1 to insn 2. Finally, insn 2 is eliminated.

After this patch:

+                      insn 1: vsetvli a5,a3,e8,mf4,ta,ma
+                      ...
+                      vle32.v v1,0(a1)
+                      vsetvli a2,zero,e32,m1,ta,ma
+                      vadd.vv v1,v1,v1
+                      vsetvli zero,a5,e32,m1,ta,ma
+                      vse32.v v1,0(a0)
+                      ...
+                      insn 3: sub     a3,a3,a5
+                      ...

gcc/ChangeLog:

        * config/riscv/riscv-vsetvl.cc (vector_insn_info::parse_insn): Add early break.

gcc/testsuite/ChangeLog:

        * gcc.target/riscv/rvv/vsetvl/avl_prop-1.c: New test.

---
 gcc/config/riscv/riscv-vsetvl.cc              | 22 +++++++++++++++++++
 .../gcc.target/riscv/rvv/vsetvl/avl_prop-1.c  | 21 ++++++++++++++++++
 2 files changed, 43 insertions(+)
 create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_prop-1.c

diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc
index 2d576e8d5c1..ab47901e23f 100644
--- a/gcc/config/riscv/riscv-vsetvl.cc
+++ b/gcc/config/riscv/riscv-vsetvl.cc
@@ -2025,6 +2025,28 @@ vector_insn_info::parse_insn (insn_info *insn)
 		   real_insn_and_same_bb_p (i, get_insn ()->bb ());
 		   i = i->next_nondebug_insn ())
 		{
+		  /* Consider this following sequence:
+
+		       insn 1: vsetvli a5,a3,e8,mf4,ta,mu
+		       insn 2: vsetvli zero,a5,e32,m1,ta,ma
+		       ...
+		       vle32.v v1,0(a1)
+		       vsetvli a2,zero,e32,m1,ta,ma
+		       vadd.vv v1,v1,v1
+		       vsetvli zero,a5,e32,m1,ta,ma
+		       vse32.v v1,0(a0)
+		       ...
+		       insn 3: sub     a3,a3,a5
+		       ...
+
+		       We can local AVL propagate "a3" from insn 1 to insn 2
+		       if no insns between insn 1 and insn 2 modify "a3 even
+		       though insn 3 modifies "a3".
+		       Otherwise, we can't perform local AVL propagation.
+
+		       Early break if we reach the insn 2.  */
+		  if (!before_p (i, insn))
+		    break;
 		  if (find_access (i->defs (), REGNO (new_info.get_avl ())))
 		    {
 		      modified_p = true;
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_prop-1.c b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_prop-1.c
new file mode 100644
index 00000000000..19ea0f14df5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_prop-1.c
@@ -0,0 +1,21 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv32gcv -mabi=ilp32 -fno-schedule-insns -fno-schedule-insns2" } */
+
+#include "riscv_vector.h"
+
+void
+foo (void *a, void *b, void *c, size_t n)
+{
+  for (size_t vl; n > 0; n -= vl, a += vl, b += vl * 4, c += vl)
+    {
+      vl = __riscv_vsetvl_e8mf4 (n);
+      vint32m1_t vec_b = __riscv_vle32_v_i32m1 (b, vl);
+      vint32m1_t vec_a = __riscv_vadd_vv_i32m1 (vec_b, vec_b, __riscv_vsetvlmax_e32m1 ());
+      __riscv_vse32_v_i32m1 (a, vec_a, vl);
+    }
+}
+
+/* { dg-final { scan-assembler-times {vsetvli} 3 { target { no-opts "-O0" no-opts "-Os" no-opts "-Oz" no-opts "-O1" no-opts "-g" no-opts "-funroll-loops" } } } } */
+/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*[a-x0-9]+,\s*e8,\s*mf4,\s*t[au],\s*m[au]} 1 { target { no-opts "-O0" no-opts "-Os" no-opts "-Oz" no-opts "-O1" no-opts "-g" no-opts "-funroll-loops" } } } } */
+/* { dg-final { scan-assembler-times {vsetvli\s+zero,\s*[a-x0-9]+,\s*e32,\s*m1,\s*t[au],\s*m[au]} 1 { target { no-opts "-O0" no-opts "-Os" no-opts "-Oz" no-opts "-O1" no-opts "-g" no-opts "-funroll-loops" } } } } */
+/* { dg-final { scan-assembler-times {vsetvli\s+[a-x0-9]+,\s*zero,\s*e32,\s*m1,\s*t[au],\s*m[au]} 1 { target { no-opts "-O0" no-opts "-Os" no-opts "-Oz" no-opts "-O1" no-opts "-g" no-opts "-funroll-loops" } } } } */
-- 
2.36.3