cpu.x86: use push and pop when emitting %prologue and %epilogue

push/pop REG has the same effect as sub/add rsp, 8 so use them instead
when applicable to generate a little shorter code

basis/cpu/x86/x86-docs.factor

@@ -61,10 +61,13 @@ HELP: (%slot)
   }
 } ;
 
-
 HELP: decr-stack-reg
 { $values { "n" number } }
-{ $description "Emits an instruction for decrementing the stack register the given number of bytes." } ;
+{ $description "Emits an instruction for decrementing the stack register the given number of bytes. If n is equal to the " { $link cell } " size, then a " { $link PUSH } " instruction is emitted instead because it has a shorter encoding." } ;
+
+HELP: incr-stack-reg
+{ $values { "n" number } }
+{ $description "Emits an instruction for incrementing the stack register the given number of bytes. If n is equal to the " { $link cell } " size, then a " { $link POP } " instruction is emitted instead because it has a shorter encoding." } ;
 
 HELP: load-zone-offset
 { $values { "nursery-ptr" "a register symbol" } }
@@ -106,3 +109,11 @@ HELP: copy-register*
     "0000000533c61fe0: 0f28ca  movaps xmm1, xmm2"
   }
 } ;
+
+ARTICLE: "cpu.x86" "x86 compiler backend"
+"Implementation of " { $vocab-link "cpu.architecture" } " for x86 machines."
+$nl
+{ $link ADD } " and " { $link SUB } " variants:"
+{ $subsections (%inc) decr-stack-reg incr-stack-reg } ;
+
+ABOUT: "cpu.x86"

basis/cpu/x86/x86-tests.factor

@@ -1,6 +1,7 @@
 USING: compiler.cfg.debugger compiler.cfg.instructions
-compiler.codegen.gc-maps compiler.codegen.relocation compiler.cfg.registers
-cpu.architecture cpu.x86.features kernel kernel.private make math math.libm
+compiler.cfg.registers compiler.codegen.gc-maps
+compiler.codegen.relocation cpu.architecture cpu.x86 cpu.x86.assembler
+cpu.x86.features kernel kernel.private layouts make math math.libm
 namespaces sequences system tools.test ;
 IN: cpu.x86.tests
 
@@ -37,3 +38,14 @@ IN: cpu.x86.tests
     [ D 0 %clear ] B{ } make
     cpu x86.32? B{ 199 6 144 18 0 0 } B{ 73 199 6 144 18 0 0 } ? =
 ] unit-test
+
+! %prologue
+{ t } [
+    [ 2 cells %prologue ] B{ } make
+    [ pic-tail-reg PUSH ] B{ } make =
+] unit-test
+
+{ t } [
+    [ 8 cells %prologue ] B{ } make
+    [ stack-reg 7 cells SUB ] B{ } make =
+] unit-test

basis/cpu/x86/x86.factor

@@ -30,6 +30,8 @@ HOOK: stack-reg cpu ( -- reg )
 
 HOOK: reserved-stack-space cpu ( -- n )
 
+HOOK: pic-tail-reg cpu ( -- reg )
+
 : stack@ ( n -- op ) stack-reg swap [+] ;
 
 : special-offset ( m -- n )
@@ -37,11 +39,17 @@ HOOK: reserved-stack-space cpu ( -- n )
 
 : spill@ ( n -- op ) spill-offset special-offset stack@ ;
 
+: (%inc) ( n reg -- ) swap cells dup 0 > [ ADD ] [ neg SUB ] if ; inline
+
 : decr-stack-reg ( n -- )
-    [ stack-reg swap SUB ] unless-zero ;
+    [
+        dup cell = [ drop pic-tail-reg PUSH ] [ stack-reg swap SUB ] if
+    ] unless-zero ;
 
 : incr-stack-reg ( n -- )
-    [ stack-reg swap ADD ] unless-zero ;
+    [
+        dup cell = [ drop pic-tail-reg POP ] [ stack-reg swap ADD ] if
+    ] unless-zero ;
 
 : align-stack ( n -- n' ) 16 align ;
 
@@ -51,8 +59,6 @@ M: x86 stack-frame-size ( stack-frame -- i )
     cell +
     align-stack ;
 
-HOOK: pic-tail-reg cpu ( -- reg )
-
 M: x86 complex-addressing? t ;
 
 M: x86 fused-unboxing? t ;
@@ -93,8 +99,6 @@ M: x86 %replace-imm
 M: x86 %clear ( loc -- )
     297 swap %replace-imm ;
 
-: (%inc) ( n reg -- ) swap cells dup 0 > [ ADD ] [ neg SUB ] if ; inline
-
 M: x86 %inc ( loc -- )
     [ n>> ] [ ds-loc? ds-reg rs-reg ? ] bi (%inc) ;