From 416aafa85d493d1add1e80da99b698c5cb496a33 Mon Sep 17 00:00:00 2001
From: qwe661234 <qwe661234@gmail.com>
Date: Mon, 12 Dec 2022 20:38:06 +0800
Subject: [PATCH] Use TCO to accelerate funciton emulate

We need to refactor the function emulate to a recursive version for
meeting the requirement of tail-call optimization(TCO). To achieve this,
I add a variable is_tail to the struct rv_insn_t to help us determine
whether the basic block is terminate or not. As a result, we can use
this variable to rewrite function emulate into a self-recursive
function.

Running coremark and dhrystone benchmark now produces faster results
than it did previously.
---
 src/decode.h  |  1 +
 src/emulate.c | 28 +++++-----------------------
 2 files changed, 6 insertions(+), 23 deletions(-)

diff --git a/src/decode.h b/src/decode.h
index 63f134944..b5022c1f6 100644
--- a/src/decode.h
+++ b/src/decode.h
@@ -240,6 +240,7 @@ typedef struct {
 
     /* instruction length */
     uint8_t insn_len;
+    uint8_t is_tail;
 } rv_insn_t;
 
 /* translated basic block */
diff --git a/src/emulate.c b/src/emulate.c
index 995e3b6f1..99d37a8ba 100644
--- a/src/emulate.c
+++ b/src/emulate.c
@@ -1147,7 +1147,9 @@ static bool emulate(riscv_t *rv, const rv_insn_t *ir)
 
     /* step over instruction */
     rv->PC += ir->insn_len;
-    return true;
+    if(ir->is_tail)
+        return true;
+    return emulate(rv, ir + 1);
 }
 
 static bool insn_is_branch(uint8_t opcode)
@@ -1240,27 +1242,6 @@ static block_t *block_find(const block_map_t *map, const uint32_t addr)
     return NULL;
 }
 
-/* execute a basic block */
-static bool block_emulate(riscv_t *rv, const block_t *block)
-{
-    const uint32_t n_insn = block->n_insn;
-    const rv_insn_t *ir = block->ir;
-
-    /* execute the block */
-    for (uint32_t i = 0; i < n_insn; i++) {
-        /* enforce zero register */
-        rv->X[rv_reg_zero] = 0;
-
-        /* execute the instruction */
-        if (!emulate(rv, ir + i))
-            return false;
-
-        /* increment the cycles csr */
-        rv->csr_cycle++;
-    }
-    return true;
-}
-
 static void block_translate(riscv_t *rv, block_t *block)
 {
     block->pc_start = block->pc_end = rv->PC;
@@ -1288,6 +1269,7 @@ static void block_translate(riscv_t *rv, block_t *block)
         if (insn_is_branch(ir->opcode))
             break;
     }
+    (block->ir + block->n_insn - 1)->is_tail = true;
 }
 
 static block_t *block_find_or_translate(riscv_t *rv, block_t *prev)
@@ -1350,7 +1332,7 @@ void rv_step(riscv_t *rv, int32_t cycles)
         assert(block);
 
         /* execute the block */
-        if (!block_emulate(rv, block))
+        if (!emulate(rv, block->ir))
             break;
 
         prev = block;