runtime: simpler and faster GC

Implement the design described in:
https://docs.google.com/document/d/1v4Oqa0WwHunqlb8C3ObL_uNQw3DfSY-ztoA-4wWbKcg/pub

Summary of the changes:
GC uses "2-bits per word" pointer type info embed directly into bitmap.
Scanning of stacks/data/heap is unified.
The old spans types go away.
Compiler generates "sparse" 4-bits type info for GC (directly for GC bitmap).
Linker generates "dense" 2-bits type info for data/bss (the same as stacks use).

Summary of results:
-1680 lines of code total (-1000+ in mgc0.c only)
-25% memory consumption
-3-7% binary size
-15% GC pause reduction
-7% run time reduction

LGTM=khr
R=golang-codereviews, rsc, christoph, khr
CC=golang-codereviews, rlh
https://codereview.appspot.com/106260045

1 files changed, 24 insertions, 4 deletions

diff --git a/src/cmd/ld/decodesym.c b/src/cmd/ld/decodesym.c
index 1773387f5..b5fe47ce9 100644
--- a/src/cmd/ld/decodesym.c
+++ b/src/cmd/ld/decodesym.c
@@ -70,14 +70,28 @@ decode_inuxi(uchar* p, int sz)
 static int
 commonsize(void)
 {
-	return 7*PtrSize + 8;
+	return 8*PtrSize + 8;
 }
 
 // Type.commonType.kind
 uint8
 decodetype_kind(LSym *s)
 {
-	return s->p[1*PtrSize + 7] & ~KindNoPointers;	//  0x13 / 0x1f
+	return s->p[1*PtrSize + 7] & KindMask;	//  0x13 / 0x1f
+}
+
+// Type.commonType.kind
+uint8
+decodetype_noptr(LSym *s)
+{
+	return s->p[1*PtrSize + 7] & KindNoPointers;	//  0x13 / 0x1f
+}
+
+// Type.commonType.kind
+uint8
+decodetype_usegcprog(LSym *s)
+{
+	return s->p[1*PtrSize + 7] & KindGCProg;	//  0x13 / 0x1f
 }
 
 // Type.commonType.size
@@ -89,9 +103,15 @@ decodetype_size(LSym *s)
 
 // Type.commonType.gc
 LSym*
-decodetype_gc(LSym *s)
+decodetype_gcprog(LSym *s)
+{
+	return decode_reloc_sym(s, 1*PtrSize + 8 + 2*PtrSize);
+}
+
+uint8*
+decodetype_gcmask(LSym *s)
 {
-	return decode_reloc_sym(s, 1*PtrSize + 8 + 1*PtrSize);
+	return (uint8*)(s->p + 1*PtrSize + 8 + 1*PtrSize);
 }
 
 // Type.ArrayType.elem and Type.SliceType.Elem