diff --git a/codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt b/codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt
index 14de30269..6da28dbcd 100644
--- a/codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt
+++ b/codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt
@@ -173,7 +173,11 @@ internal class ProgramAndVarsGen(
     }
 
     private fun footer() {
-        asmgen.out("; bss sections\n  .dsection BSS\n  .dsection slabs_BSS")
+        asmgen.out("; bss sections")
+        asmgen.out("prog8_bss_section_start")
+        asmgen.out("  .dsection BSS")
+        asmgen.out("prog8_bss_section_size = * - prog8_bss_section_start")
+        asmgen.out("  .dsection slabs_BSS")
         asmgen.out("prog8_program_end\t; end of program label for progend()")
     }
 
@@ -205,7 +209,9 @@ internal class ProgramAndVarsGen(
         if (initializers.isNotEmpty()) {
             asmgen.out("prog8_init_vars\t.block")
             initializers.forEach { assign ->
-                asmgen.translate(assign)
+                if((assign.value as? PtNumber)?.number != 0.0 || allocator.isZpVar(assign.target.identifier!!.name))
+                    asmgen.translate(assign)
+                // the other variables that should be set to zero are done so as part of the BSS section.
             }
             asmgen.out("  rts\n  .bend")
         }
@@ -363,7 +369,21 @@ internal class ProgramAndVarsGen(
 
     private fun entrypointInitialization() {
         asmgen.out("; program startup initialization")
-        asmgen.out("  cld")
+        asmgen.out("  cld |  tsx |  stx  prog8_lib.orig_stackpointer    ; required for sys.exit()")
+        // set full BSS area to zero
+        asmgen.out("""
+    .if  prog8_bss_section_size>0
+    ; reset all variables in BSS section to zero
+	lda  #<prog8_bss_section_start
+	ldy  #>prog8_bss_section_start
+	sta  P8ZP_SCRATCH_W1
+	sty  P8ZP_SCRATCH_W1+1
+	ldx  #<prog8_bss_section_size
+	ldy  #>prog8_bss_section_size
+	lda  #0
+	jsr  prog8_lib.memset
+    .endif""")
+
         blockVariableInitializers.forEach {
             if (it.value.isNotEmpty())
                 asmgen.out("  jsr  ${it.key.name}.prog8_init_vars")
@@ -414,8 +434,7 @@ internal class ProgramAndVarsGen(
             arrayVariable2asm(varname, it.alloc.dt, it.value, null)
         }
 
-        asmgen.out("""+       tsx
-                    stx  prog8_lib.orig_stackpointer    ; required for sys.exit()                    
+        asmgen.out("""+        
                     ldx  #255       ; init estack ptr
                     clv
                     clc""")
diff --git a/docs/source/programming.rst b/docs/source/programming.rst
index 86b40731f..2a0a50453 100644
--- a/docs/source/programming.rst
+++ b/docs/source/programming.rst
@@ -890,7 +890,8 @@ sizeof(name)
 
 memory(name, size, alignment)
     Returns the address of the first location of a statically "reserved" block of memory of the given size in bytes,
-    with the given name.  If you specify an alignment value >1, it means the block of memory will
+    with the given name. The block is uninitialized memory, it is *not* set to zero!
+    If you specify an alignment value >1, it means the block of memory will
     be aligned to such a dividable address in memory, for instance an alignment of $100 means the
     memory block is aligned on a page boundary, and $2 means word aligned (even addresses).
     Requesting the address of such a named memory block again later with
diff --git a/docs/source/todo.rst b/docs/source/todo.rst
index ec03f3669..b570348be 100644
--- a/docs/source/todo.rst
+++ b/docs/source/todo.rst
@@ -3,8 +3,7 @@ TODO
 
 For next minor release
 ^^^^^^^^^^^^^^^^^^^^^^
-- BSS: initialize BSS block (without slabs!) to zeros on start, using 1 loop instead of all those initialization assignments
-- BSS subroutine parameters don't have to be set to 0. But meh, the loop should be super fast anyway.
+- subroutine parameters don't have to be initialized to 0 in prog8_init_vars()
 - option to put BSS in specific upper memory block ($a000-$bfff, $c000-$cfff on C64) add a .cerror check for overflow!
 - document bss stuff in the manual
 
diff --git a/examples/test.p8 b/examples/test.p8
index 59ced2ac0..bf6e22760 100644
--- a/examples/test.p8
+++ b/examples/test.p8
@@ -6,8 +6,8 @@
 main {
 
     uword b_wordvar
-    uword b_initwordvar = 12345     ; TODO FIX THIS INIT VALUE FOR noreinit=true
-    ubyte b_bb =123      ; TODO FIX THIS INIT VALUE FOR noreinit=true
+    uword b_initwordvar = 12345
+    ubyte b_bb =123
     float b_fl
     ubyte[10] b_emptyarray
     ubyte[10] b_filledarray = [1,2,3,4,5,6,7,8,9,10]