refactor: move bumping algorithm to its own module for testability an…

…d readability
bluurryy · Aug 24, 2024 · 89a241d · 89a241d
1 parent 02aed02
commit 89a241d
Show file tree

Hide file tree

Showing 7 changed files with 329 additions and 218 deletions.
diff --git a/README.md b/README.md
@@ -65,14 +65,14 @@ let mut bump: Bump = Bump::new();
 bump.scoped(|mut bump| {
     let hello = bump.alloc_str("hello");
     assert_eq!(bump.stats().allocated(), 5);
-    
+
     bump.scoped(|bump| {
         let world = bump.alloc_str("world");
 
         println!("{hello} and {world} are both live");
         assert_eq!(bump.stats().allocated(), 10);
     });
-    
+
     println!("{hello} is still live");
     assert_eq!(bump.stats().allocated(), 5);
 });
@@ -86,7 +86,7 @@ assert_eq!(bump.stats().allocated(), 0);
 
     let hello = bump.alloc_str("hello");
     assert_eq!(bump.stats().allocated(), 5);
-    
+
     {
         let mut guard = bump.scope_guard();
         let bump = guard.scope();
@@ -96,7 +96,7 @@ assert_eq!(bump.stats().allocated(), 0);
         println!("{hello} and {world} are both live");
         assert_eq!(bump.stats().allocated(), 10);
     }
-    
+
     println!("{hello} is still live");
     assert_eq!(bump.stats().allocated(), 5);
 }

diff --git a/crates/inspect-asm/out/x86-64/alloc_iter_u32/exact_down.asm b/crates/inspect-asm/out/x86-64/alloc_iter_u32/exact_down.asm
@@ -1,53 +1,54 @@
 inspect_asm::alloc_iter_u32::exact_down:
-	push r15
 	push r14
 	push rbx
+	push rax
 	mov rax, rdx
 	shr rax, 61
 	jne .LBB0_7
-	lea rbx, [4*rdx]
+	lea r8, [4*rdx]
 	mov rcx, qword ptr [rdi]
 	mov rax, qword ptr [rcx]
-	mov r8, rax
-	sub r8, qword ptr [rcx + 8]
-	cmp rbx, r8
+	mov r9, rax
+	sub r9, qword ptr [rcx + 8]
+	cmp r8, r9
 	ja .LBB0_5
-	sub rax, rbx
+	sub rax, r8
 	and rax, -4
 	mov qword ptr [rcx], rax
 	test rdx, rdx
 	je .LBB0_4
 .LBB0_0:
-	add rbx, -4
-	shr rbx, 2
-	cmp rdx, rbx
-	cmovb rbx, rdx
+	lea rcx, [rdx - 1]
+	movabs r8, 4611686018427387903
+	and r8, rcx
+	cmp rdx, r8
+	cmovb r8, rdx
 	mov rcx, rsi
 	mov rdi, rax
-	cmp rbx, 8
+	cmp r8, 8
 	jb .LBB0_2
-	mov r8, rax
-	sub r8, rsi
+	mov r9, rax
+	sub r9, rsi
 	mov rcx, rsi
 	mov rdi, rax
-	cmp r8, 31
+	cmp r9, 31
 	jbe .LBB0_2
-	inc rbx
-	mov ecx, ebx
+	inc r8
+	mov ecx, r8d
 	and ecx, 7
 	mov edi, 8
 	cmovne rdi, rcx
-	sub rbx, rdi
-	lea rcx, [rsi + 4*rbx]
-	lea rdi, [rax + 4*rbx]
-	xor r8d, r8d
+	sub r8, rdi
+	lea rcx, [rsi + 4*r8]
+	lea rdi, [rax + 4*r8]
+	xor r9d, r9d
 .LBB0_1:
-	movups xmm0, xmmword ptr [rsi + 4*r8]
-	movups xmm1, xmmword ptr [rsi + 4*r8 + 16]
-	movups xmmword ptr [rax + 4*r8], xmm0
-	movups xmmword ptr [rax + 4*r8 + 16], xmm1
-	add r8, 8
-	cmp rbx, r8
+	movups xmm0, xmmword ptr [rsi + 4*r9]
+	movups xmm1, xmmword ptr [rsi + 4*r9 + 16]
+	movups xmmword ptr [rax + 4*r9], xmm0
+	movups xmmword ptr [rax + 4*r9 + 16], xmm1
+	add r9, 8
+	cmp r8, r9
 	jne .LBB0_1
 .LBB0_2:
 	lea rsi, [rsi + 4*rdx]
@@ -62,17 +63,17 @@ inspect_asm::alloc_iter_u32::exact_down:
 	cmp rdi, r8
 	jne .LBB0_3
 .LBB0_4:
+	add rsp, 8
 	pop rbx
 	pop r14
-	pop r15
 	ret
 .LBB0_5:
-	mov r14, rsi
+	mov rbx, rsi
 	mov rsi, rdx
-	mov r15, rdx
+	mov r14, rdx
 	call bump_scope::bump_scope::BumpScope<A,_,_,_>::do_alloc_slice_in_another_chunk
-	mov rsi, r14
-	mov rdx, r15
+	mov rsi, rbx
+	mov rdx, r14
 	test rdx, rdx
 	jne .LBB0_0
 	jmp .LBB0_4

diff --git a/crates/inspect-asm/out/x86-64/alloc_iter_u32/exact_down_a.asm b/crates/inspect-asm/out/x86-64/alloc_iter_u32/exact_down_a.asm
@@ -1,52 +1,53 @@
 inspect_asm::alloc_iter_u32::exact_down_a:
-	push r15
 	push r14
 	push rbx
+	push rax
 	mov rax, rdx
 	shr rax, 61
 	jne .LBB0_7
-	lea rbx, [4*rdx]
-	mov rcx, qword ptr [rdi]
-	mov rax, qword ptr [rcx]
-	mov r8, rax
-	sub r8, qword ptr [rcx + 8]
-	cmp rbx, r8
+	lea rcx, [4*rdx]
+	mov r8, qword ptr [rdi]
+	mov rax, qword ptr [r8]
+	mov r9, rax
+	sub r9, qword ptr [r8 + 8]
+	cmp rcx, r9
 	ja .LBB0_5
-	sub rax, rbx
-	mov qword ptr [rcx], rax
+	sub rax, rcx
+	mov qword ptr [r8], rax
 	test rdx, rdx
 	je .LBB0_4
 .LBB0_0:
-	add rbx, -4
-	shr rbx, 2
-	cmp rdx, rbx
-	cmovb rbx, rdx
+	lea rcx, [rdx - 1]
+	movabs r8, 4611686018427387903
+	and r8, rcx
+	cmp rdx, r8
+	cmovb r8, rdx
 	mov rcx, rsi
 	mov rdi, rax
-	cmp rbx, 8
+	cmp r8, 8
 	jb .LBB0_2
-	mov r8, rax
-	sub r8, rsi
+	mov r9, rax
+	sub r9, rsi
 	mov rcx, rsi
 	mov rdi, rax
-	cmp r8, 31
+	cmp r9, 31
 	jbe .LBB0_2
-	inc rbx
-	mov ecx, ebx
+	inc r8
+	mov ecx, r8d
 	and ecx, 7
 	mov edi, 8
 	cmovne rdi, rcx
-	sub rbx, rdi
-	lea rcx, [rsi + 4*rbx]
-	lea rdi, [rax + 4*rbx]
-	xor r8d, r8d
+	sub r8, rdi
+	lea rcx, [rsi + 4*r8]
+	lea rdi, [rax + 4*r8]
+	xor r9d, r9d
 .LBB0_1:
-	movups xmm0, xmmword ptr [rsi + 4*r8]
-	movups xmm1, xmmword ptr [rsi + 4*r8 + 16]
-	movups xmmword ptr [rax + 4*r8], xmm0
-	movups xmmword ptr [rax + 4*r8 + 16], xmm1
-	add r8, 8
-	cmp rbx, r8
+	movups xmm0, xmmword ptr [rsi + 4*r9]
+	movups xmm1, xmmword ptr [rsi + 4*r9 + 16]
+	movups xmmword ptr [rax + 4*r9], xmm0
+	movups xmmword ptr [rax + 4*r9 + 16], xmm1
+	add r9, 8
+	cmp r8, r9
 	jne .LBB0_1
 .LBB0_2:
 	lea rsi, [rsi + 4*rdx]
@@ -61,17 +62,17 @@ inspect_asm::alloc_iter_u32::exact_down_a:
 	cmp rdi, r8
 	jne .LBB0_3
 .LBB0_4:
+	add rsp, 8
 	pop rbx
 	pop r14
-	pop r15
 	ret
 .LBB0_5:
-	mov r14, rsi
+	mov rbx, rsi
 	mov rsi, rdx
-	mov r15, rdx
+	mov r14, rdx
 	call bump_scope::bump_scope::BumpScope<A,_,_,_>::do_alloc_slice_in_another_chunk
-	mov rsi, r14
-	mov rdx, r15
+	mov rsi, rbx
+	mov rdx, r14
 	test rdx, rdx
 	jne .LBB0_0
 	jmp .LBB0_4