4 files changed, 37 insertions, 8 deletions

diff --git a/src/lj_asm.c b/src/lj_asm.c
index a550d3b3..e847b8c9 100644
--- a/src/lj_asm.c
+++ b/src/lj_asm.c
@@ -1058,14 +1058,22 @@ static void asm_bufhdr(ASMState *as, IRIns *ir)
 static void asm_bufput(ASMState *as, IRIns *ir)
 {
   const CCallInfo *ci = &lj_ir_callinfo[IRCALL_lj_buf_putstr];
-  IRRef args[2];
+  IRRef args[3];
   IRIns *irs;
+  int kchar = -1;
   if (!ra_used(ir)) return;
   args[0] = ir->op1;  /* SBuf * */
   args[1] = ir->op2;  /* GCstr * */
   irs = IR(ir->op2);
   lua_assert(irt_isstr(irs->t));
-  if (mayfuse(as, ir->op2) && ra_noreg(irs->r)) {
+  if (irs->o == IR_KGC) {
+    GCstr *s = ir_kstr(irs);
+    if (s->len == 1) {  /* Optimize put of single-char string constant. */
+      kchar = strdata(s)[0];
+      ci = &lj_ir_callinfo[IRCALL_lj_buf_putchar];
+      args[1] = ASMREF_TMP1;  /* int, truncated to char */
+    }
+  } else if (mayfuse(as, ir->op2) && ra_noreg(irs->r)) {
     if (irs->o == IR_TOSTR) {  /* Fuse number to string conversions. */
       if (LJ_SOFTFP ? (irs+1)->o == IR_HIOP : irt_isnum(IR(irs->op1)->t)) {
         ci = &lj_ir_callinfo[IRCALL_lj_buf_putnum];
@@ -1075,12 +1083,21 @@ static void asm_bufput(ASMState *as, IRIns *ir)
         ci = &lj_ir_callinfo[IRCALL_lj_buf_putint];
         args[1] = irs->op1;  /* int */
       }
+    } else if (irs->o == IR_SNEW) {  /* Fuse string allocation. */
+      ci = &lj_ir_callinfo[IRCALL_lj_buf_putmem];
+      args[1] = irs->op1;  /* const void * */
+      args[2] = irs->op2;  /* MSize */
     }
   }
   asm_setupresult(as, ir, ci);  /* SBuf * */
   asm_gencall(as, ci, args);
-  if (args[1] == ASMREF_TMP1)
-    asm_tvptr(as, ra_releasetmp(as, ASMREF_TMP1), irs->op1);
+  if (args[1] == ASMREF_TMP1) {
+    Reg tmp = ra_releasetmp(as, ASMREF_TMP1);
+    if (kchar == -1)
+      asm_tvptr(as, tmp, irs->op1);
+    else
+      ra_allockreg(as, kchar, tmp);
+  }
 }
 
 static void asm_bufstr(ASMState *as, IRIns *ir)
@@ -2015,11 +2032,11 @@ static void asm_setup_regsp(ASMState *as)
       /* fallthrough */
 #endif
     /* C calls evict all scratch regs and return results in RID_RET. */
-    case IR_SNEW: case IR_XSNEW: case IR_NEWREF:
+    case IR_SNEW: case IR_XSNEW: case IR_NEWREF: case IR_BUFPUT:
       if (REGARG_NUMGPR < 3 && as->evenspill < 3)
         as->evenspill = 3;  /* lj_str_new and lj_tab_newkey need 3 args. */
     case IR_TNEW: case IR_TDUP: case IR_CNEW: case IR_CNEWI: case IR_TOSTR:
-    case IR_BUFPUT: case IR_BUFSTR:
+    case IR_BUFSTR:
       ir->prev = REGSP_HINT(RID_RET);
       if (inloop)
         as->modset = RSET_SCRATCH;
diff --git a/src/lj_buf.c b/src/lj_buf.c
index ef48b580..67525694 100644
--- a/src/lj_buf.c
+++ b/src/lj_buf.c
@@ -58,11 +58,12 @@ char *lj_buf_wmem(char *p, const void *q, MSize len)
   return p;
 }
 
-void lj_buf_putmem(SBuf *sb, const void *q, MSize len)
+SBuf * lj_buf_putmem(SBuf *sb, const void *q, MSize len)
 {
   char *p = lj_buf_more(sb, len);
   p = lj_buf_wmem(p, q, len);
   setsbufP(sb, p);
+  return sb;
 }
 
 #if LJ_HASJIT
@@ -75,6 +76,14 @@ SBuf * LJ_FASTCALL lj_buf_putstr(SBuf *sb, GCstr *s)
   return sb;
 }
 
+SBuf * LJ_FASTCALL lj_buf_putchar(SBuf *sb, int c)
+{
+  char *p = lj_buf_more(sb, 1);
+  *p++ = (char)c;
+  setsbufP(sb, p);
+  return sb;
+}
+
 SBuf * LJ_FASTCALL lj_buf_putint(SBuf *sb, int32_t k)
 {
   setsbufP(sb, lj_str_bufint(lj_buf_more(sb, LJ_STR_INTBUF), k));
diff --git a/src/lj_buf.h b/src/lj_buf.h
index e028a434..426ec720 100644
--- a/src/lj_buf.h
+++ b/src/lj_buf.h
@@ -25,9 +25,10 @@ LJ_FUNC void LJ_FASTCALL lj_buf_grow(SBuf *sb, char *en);
 LJ_FUNC void LJ_FASTCALL lj_buf_shrink(lua_State *L, SBuf *sb);
 
 LJ_FUNC char *lj_buf_wmem(char *p, const void *q, MSize len);
-LJ_FUNC void lj_buf_putmem(SBuf *sb, const void *q, MSize len);
+LJ_FUNC SBuf * lj_buf_putmem(SBuf *sb, const void *q, MSize len);
 #if LJ_HASJIT
 LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putstr(SBuf *sb, GCstr *s);
+LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putchar(SBuf *sb, int c);
 LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putint(SBuf *sb, int32_t k);
 LJ_FUNC SBuf * LJ_FASTCALL lj_buf_putnum(SBuf *sb, cTValue *o);
 LJ_FUNC GCstr * LJ_FASTCALL lj_buf_tostr(SBuf *sb);
diff --git a/src/lj_ircall.h b/src/lj_ircall.h
index 3b1cc928..46bb54fe 100644
--- a/src/lj_ircall.h
+++ b/src/lj_ircall.h
@@ -105,7 +105,9 @@ typedef struct CCallInfo {
   _(ANY,        lj_strscan_num,         2,  FN, INT, 0) \
   _(ANY,        lj_str_fromint,         2,  FN, STR, CCI_L) \
   _(ANY,        lj_str_fromnum,         2,  FN, STR, CCI_L) \
+  _(ANY,        lj_buf_putmem,          3,   S, P32, 0) \
   _(ANY,        lj_buf_putstr,          2,  FS, P32, 0) \
+  _(ANY,        lj_buf_putchar,         2,  FS, P32, 0) \
   _(ANY,        lj_buf_putint,          2,  FS, P32, 0) \
   _(ANY,        lj_buf_putnum,          2,  FS, P32, 0) \
   _(ANY,        lj_buf_tostr,           1,  FL, STR, 0) \