First rough version of the JIT.

It can successfully parse SpeedMessage1.
Preliminary results: 750MB/s on Core2 2.4GHz.
This number is 2.5x proto2.
This isn't apples-to-apples, because
proto2 is parsing to a struct and we are
just doing stream parsing, but for apps
that are currently using proto2, this is the
improvement they would see if they could
move to stream-based processing.

Unfortunately perf-regression-test.py is
broken, and I'm not 100% sure why.  It would
be nice to fix it first (to ensure that
there are no performance regressions for
the table-based decoder) but I'm really
impatient to get the JIT checked in.

1 files changed, 16 insertions, 3 deletions

diff --git a/src/upb_decoder.h b/src/upb_decoder.h
index bb54930..1be31c4 100644
--- a/src/upb_decoder.h
+++ b/src/upb_decoder.h
@@ -27,13 +27,12 @@ extern "C" {
 
 /* upb_decoder *****************************************************************/
 
+struct dasm_State;
+
 struct _upb_decoder {
   // Bytesrc from which we pull serialized data.
   upb_bytesrc *bytesrc;
 
-  // Dispatcher to which we push parsed data.
-  upb_dispatcher dispatcher;
-
   // String to hold our input buffer; is only active if d->buf != NULL.
   upb_string *bufstr;
 
@@ -48,6 +47,7 @@ struct _upb_decoder {
 
   // End of this buffer, relative to *ptr.
   const char *end;
+  const char *jit_end;
 
   // Members which may also be written by the JIT:
 
@@ -57,8 +57,21 @@ struct _upb_decoder {
   // End of this submessage, relative to *ptr.
   const char *submsg_end;
 
+  // MIN(end, submsg_end)
+  const char *effective_end;
+
   // Where we will store any errors that occur.
   upb_status *status;
+
+  // Dispatcher to which we push parsed data.
+  upb_dispatcher dispatcher;
+
+  // JIT-generated machine code (else NULL).
+  char *jit_code;
+  size_t jit_size;
+  char *debug_info;
+
+  struct dasm_State *dynasm;
 };
 
 // A upb_decoder decodes the binary protocol buffer format, writing the data it