From c629f5dc2661dab0dfe74076ba5188b4d6a7c866 Mon Sep 17 00:00:00 2001
From: Stephen Dolan <mu@netsoc.tcd.ie>
Date: Wed, 2 Jul 2014 21:54:48 +0100
Subject: [PATCH] Move tail call optimisations to compile.c.

Needs more testing.
---
 compile.c | 21 +++++++++++++---
 execute.c | 72 -------------------------------------------------------
 2 files changed, 18 insertions(+), 75 deletions(-)

diff --git a/compile.c b/compile.c
index 597fa6f145..48527675b2 100644
--- a/compile.c
+++ b/compile.c
@@ -74,6 +74,7 @@ static inst* inst_new(opcode op) {
   i->subfn = gen_noop();
   i->arglist = gen_noop();
   i->source = UNKNOWN_LOCATION;
+  i->compiled = 0;
   return i;
 }
 
@@ -474,6 +475,12 @@ static int count_cfunctions(block b) {
 }
 
 
+static int returns_immediately(inst* i) {
+  while (i->op == JUMP)
+    i = i->imm.target;
+  return i->op == RET;
+}
+
 // Expands call instructions into a calling sequence
 static int expand_call_arglist(struct locfile* locations, block* b) {
   int errors = 0;
@@ -498,6 +505,12 @@ static int expand_call_arglist(struct locfile* locations, block* b) {
       case CLOSURE_CREATE: 
       case CLOSURE_PARAM: {
         block callargs = gen_noop();
+        /* A call is a tail call if:
+           a) the next instruction is a RET or a chain of unconditional JUMPs
+              that ends in a RET, and
+           b) none of the closures -callee included- are in the current bytecode block. */
+        int is_tail_call = b->first && returns_immediately(b->first);
+        if (!curr->bound_by->compiled) is_tail_call = 0;
         for (inst* i; (i = block_take(&curr->arglist));) {
           assert(opcode_describe(i->op)->flags & OP_IS_CALL_PSEUDO);
           block b = inst_block(i);
@@ -511,10 +524,12 @@ static int expand_call_arglist(struct locfile* locations, block* b) {
             block_append(&callargs, gen_op_bound(CLOSURE_REF, b));
             break;
           }
+          if (!i->bound_by->compiled) is_tail_call = 0;
           actual_args++;
         }
         curr->imm.intval = actual_args;
         curr->arglist = callargs;
+        if (is_tail_call) curr->op = TAIL_CALL_JQ;
 
         if (curr->bound_by->op == CLOSURE_CREATE) {
           for (inst* i = curr->bound_by->arglist.first; i; i = i->next) {
@@ -559,13 +574,13 @@ static int compile(struct locfile* locations, struct bytecode* bc, block b) {
   int pos = 0;
   int var_frame_idx = 0;
   bc->nsubfunctions = 0;
-  errors += expand_call_arglist(locations, &b);
   b = BLOCK(b, gen_op_simple(RET));
+  errors += expand_call_arglist(locations, &b);
   jv localnames = jv_array();
   for (inst* curr = b.first; curr; curr = curr->next) {
     if (!curr->next) assert(curr == b.last);
     int length = opcode_describe(curr->op)->length;
-    if (curr->op == CALL_JQ) {
+    if (curr->op == CALL_JQ || curr->op == TAIL_CALL_JQ) {
       for (inst* arg = curr->arglist.first; arg; arg = arg->next) {
         length += 2;
       }
@@ -639,7 +654,7 @@ static int compile(struct locfile* locations, struct bytecode* bc, block b) {
       assert(!curr->arglist.first);
       code[pos++] = (uint16_t)curr->imm.intval;
       code[pos++] = curr->bound_by->imm.intval;
-    } else if (curr->op == CALL_JQ) {
+    } else if (curr->op == CALL_JQ || curr->op == TAIL_CALL_JQ) {
       assert(curr->bound_by->op == CLOSURE_CREATE ||
              curr->bound_by->op == CLOSURE_PARAM);
       code[pos++] = (uint16_t)curr->imm.intval;
diff --git a/execute.c b/execute.c
index 860e46b790..f0700ba35f 100644
--- a/execute.c
+++ b/execute.c
@@ -796,76 +796,6 @@ void jq_teardown(jq_state **jq) {
   jv_mem_free(old_jq);
 }
 
-static int ret_follows(uint16_t *pc) {
-  if (*pc == RET)
-    return 1;
-  if (*pc++ != JUMP)
-    return 0;
-  return ret_follows(pc + *pc + 1); // FIXME, might be ironic
-}
-
-/*
- * Look for tail calls that can be optimized: tail calls with no
- * references left to the current frame.
- *
- * We're staring at this bytecode layout:
- *
- *   CALL_JQ
- *   <nclosures>
- *   <callee closure>       (2 units)
- *   <nclosures closures>   (2 units each)
- *   <next instruction>
- *
- * A closure is:
- *
- *   <level>    (a relative frame count chased via the current frame's env)
- *   <index>    (an index of a subfunction or closure in that frame)
- *
- * We're looking for:
- *
- * a) the next instruction is a RET or a chain of unconditional JUMPs
- * that ends in a RET, and
- *
- * b) none of the closures -callee included- have level == 0.
- */
-static uint16_t tail_call_analyze(uint16_t *pc) {
-  assert(*pc == CALL_JQ);
-  pc++;
-  // + 1 for the callee closure
-  for (uint16_t nclosures = *pc++ + 1; nclosures > 0; pc++, nclosures--) {
-    if (*pc++ == 0)
-      return CALL_JQ;
-  }
-  if (ret_follows(pc))
-    return TAIL_CALL_JQ;
-  return CALL_JQ;
-}
-
-static struct bytecode *optimize_code(struct bytecode *bc) {
-  uint16_t *pc = bc->code;
-  // FIXME: Don't mutate bc->code...
-  while (pc < bc->code + bc->codelen) {
-    switch (*pc) {
-    case CALL_JQ:
-      *pc = tail_call_analyze(pc);
-      break;
-
-    // Other bytecode optimizations here.  A peephole optimizer would
-    // fit right in.
-    default: break;
-    }
-    pc += bytecode_operation_length(pc);
-  }
-  return bc;
-}
-
-static struct bytecode *optimize(struct bytecode *bc) {
-  for (int i=0; i<bc->nsubfunctions; i++) {
-    bc->subfunctions[i] = optimize(bc->subfunctions[i]);
-  }
-  return optimize_code(bc);
-}
-
 int jq_compile_args(jq_state *jq, const char* str, jv args) {
   jv_nomem_handler(jq->nomem_handler, jq->nomem_handler_data);
   assert(jv_get_kind(args) == JV_KIND_ARRAY);
@@ -903,8 +833,6 @@ int jq_compile_args(jq_state *jq, const char* str, jv args) {
       fprintf(stderr, "%s\n", jv_string_value(s));
     jv_free(s);
   }
-  if (jq->bc)
-    jq->bc = optimize(jq->bc);
   locfile_free(&locations);
   return jq->bc != NULL;
 }