Compare commits

...

2 Commits

Author SHA1 Message Date
Ed_
81328819c6 watl.v0.llvm.lottes.c: Did FArena 2025-11-04 02:07:38 -05:00
Ed_
f437be32e2 WIP: watl.v0.llvm.lottes.c fleshing out 2025-11-03 23:14:40 -05:00
2 changed files with 816 additions and 3 deletions

View File

@@ -4,8 +4,11 @@ Version: 0 (From Scratch, 1-Stage Compilation, LLVM & WinAPI Only, Win CRT Mul
Host: Windows 11 (x86-64)
Toolchain: LLVM (2025-08-30), C-Stanard: 11
Following strictly: Neokineogfx - Fixing C
Following strictly (mostly): Neokineogfx - Fixing C
https://youtu.be/RrL7121MOeA
Unlike lottes_hybrid this file will be entirely untyped for any pointer addressing.
Win CRT imports will also be typeless signatures.
*/
#pragma clang diagnostic ignored "-Wunused-const-variable"
@@ -43,14 +46,624 @@ https://youtu.be/RrL7121MOeA
#define S_ static
#define I_ static inline __attribute__((always_inline))
#define N_ static __attribute__((noinline))
#define R_ __restrict
#define V_ volatile
#define R_ __restrict
#define V_ volatile
#define W_ __attribute((__stdcall__)) __attribute__((__force_align_arg_pointer__))
#define reg register
#define glue_impl(A, B) A ## B
#define glue(A, B) glue_impl(A, B)
#define stringify_impl(S) #S
#define stringify(S) stringify_impl(S)
#define tmpl(prefix, type) prefix ## _ ## type
#define local_persist static
#define global static
#define internal static
#define static_assert _Static_assert
#define typeof __typeof__
#define typeof_ptr(ptr) typeof(ptr[0])
#define typeof_same(a, b) _Generic((a), typeof((b)): 1, default: 0)
#define def_R_(type) type*restrict type ## _R
#define def_V_(type) type*volatile type ## _V
#define def_ptr_set(type) def_R_(type); typedef def_V_(type)
#define def_tset(type) type; typedef def_ptr_set(type)
/* Deviation from Lottes's Convention: Using byte-width for the with a single letter to indicating underlying type or intent.
U1: B1
U2: W1
U4: I1
U8: L1
S1: SB1
S2: SW1
S4: SI1
S8: SL1
F4: F1
F8: D1
F4_4: F4
*/
typedef __UINT8_TYPE__ def_tset(U1); typedef __UINT16_TYPE__ def_tset(U2); typedef __UINT32_TYPE__ def_tset(U4); typedef __UINT64_TYPE__ def_tset(U8);
typedef __INT8_TYPE__ def_tset(S1); typedef __INT16_TYPE__ def_tset(S2); typedef __INT32_TYPE__ def_tset(S4); typedef __INT64_TYPE__ def_tset(S8);
typedef unsigned char def_tset(B1); typedef __UINT16_TYPE__ def_tset(B2); typedef __UINT32_TYPE__ def_tset(B4); typedef __UINT64_TYPE__ def_tset(B8);
typedef float def_tset(F4);
typedef double def_tset(F8);
typedef float F4_4 __attribute__((vector_size(16))); typedef def_ptr_set(F4_4);
enum { false = 0, true = 1, true_overflow, };
#define u1_r(value) cast(U1_R, value)
#define u2_r(value) cast(U2_R, value)
#define u4_r(value) cast(U4_R, value)
#define u8_r(value) cast(U8_R, value)
#define u1_v(value) cast(U1_V, value)
#define u2_v(value) cast(U2_V, value)
#define u4_v(value) cast(U4_V, value)
#define u8_v(value) cast(U8_V, value)
#define u1_(value) cast(U1, value)
#define u2_(value) cast(U2, value)
#define u4_(value) cast(U4, value)
#define u8_(value) cast(U8, value)
#define s1_(value) cast(S1, value)
#define s2_(value) cast(S2, value)
#define s4_(value) cast(S4, value)
#define s8_(value) cast(S8, value)
#define f4_(value) cast(F4, value)
#define f8_(value) cast(F8, value)
#define uvar(Type, sym) B1 sym[sizeof(Type)]
#define farray_len(array) (U8)sizeof(array) / size_of( typeof((array)[0]))
#define farray_init(type, ...) (type[]){__VA_ARGS__}
#define def_farray_sym(_type, _len) A ## _len ## _ ## _type
#define def_farray_impl(_type, _len) _type def_farray_sym(_type, _len)[_len]; typedef def_ptr_set(def_farray_sym(_type, _len))
#define def_farray(type, len) def_farray_impl(type, len)
#define def_enum(underlying_type, symbol) underlying_type def_tset(symbol); enum symbol
#define def_struct(symbol) struct symbol def_tset(symbol); struct symbol
#define def_union(symbol) union symbol def_tset(symbol); union symbol
#define def_proc(symbol) symbol
#define opt_args(symbol, ...) &(symbol){__VA_ARGS__}
#define alignas _Alignas
#define alignof _Alignof
#define cast(type, data) ((type)(data))
#define pcast(type, data) cast(type*, & (data))[0]
#define nullptr cast(void*, 0)
#define null cast(U8, 0)
#define offset_of(type, member) cast(U8, & (((type*) 0)->member))
#define size_of(data) cast(U8, sizeof(data))
#define r_(ptr) cast(typeof_ptr(ptr)*R_, ptr)
#define v_(ptr) cast(typeof_ptr(ptr)*V_, ptr)
#define tr_(type, ptr) cast(type*R_, ptr)
#define tv_(type, ptr) cast(type*V_, ptr)
#define kilo(n) (cast(U8, n) << 10)
#define mega(n) (cast(U8, n) << 20)
#define giga(n) (cast(U8, n) << 30)
#define tera(n) (cast(U8, n) << 40)
// Deviation from Lottes's Convention: Using lower snake case for the naming.
#define sop_1(op, a, b) cast(U1, s1_(a) op s1_(b))
#define sop_2(op, a, b) cast(U2, s2_(a) op s2_(b))
#define sop_4(op, a, b) cast(U4, s4_(a) op s4_(b))
#define sop_8(op, a, b) cast(U8, s8_(a) op s8_(b))
#define def_signed_op(id, op, width) I_ U ## width id ## _s ## width(U ## width a, U ## width b) {return sop_ ## width(op, a, b); }
#define def_signed_ops(id, op) def_signed_op(id, op, 1) def_signed_op(id, op, 2) def_signed_op(id, op, 4) def_signed_op(id, op, 8)
def_signed_ops(add, +) def_signed_ops(sub, -)
def_signed_ops(mut, *) def_signed_ops(div, /)
def_signed_ops(gt, >) def_signed_ops(lt, <)
def_signed_ops(ge, >=) def_signed_ops(le, <=)
#define def_generic_sop(op, a, ...) _Generic((a), U1: op ## _s1, U2: op ## _s2, U4: op ## _s4, U8: op ## _s8) (a, __VA_ARGS__)
#define add_s(a,b) def_generic_sop(add,a,b)
#define sub_s(a,b) def_generic_sop(sub,a,b)
#define mut_s(a,b) def_generic_sop(mut,a,b)
#define gt_s(a,b) def_generic_sop(gt, a,b)
#define lt_s(a,b) def_generic_sop(lt, a,b)
#define ge_s(a,b) def_generic_sop(ge, a,b)
#define le_s(a,b) def_generic_sop(le, a,b)
I_ U4 atm_add_u4 (U4_R a, U4 v){__asm__ volatile("lock xaddl %0,%1":"=r"(v),"=m"(*a):"0"(v),"m"(*a):"memory","cc");return v;}
I_ U8 atm_add_u8 (U8_R a, U8 v){__asm__ volatile("lock xaddq %0,%1":"=r"(v),"=m"(*a):"0"(v),"m"(*a):"memory","cc");return v;}
I_ U4 atm_swap_u4(U4_R a, U4 v){__asm__ volatile("lock xchgl %0,%1":"=r"(v),"=m"(*a):"0"(v),"m"(*a):"memory","cc");return v;}
I_ U8 atm_swap_u8(U8_R a, U8 v){__asm__ volatile("lock xchgq %0,%1":"=r"(v),"=m"(*a):"0"(v),"m"(*a):"memory","cc");return v;}
I_ void barrier_compiler(void){__asm__ volatile("::""memory");} // Compiler Barrier
I_ void barrier_memory (void){__builtin_ia32_mfence();} // Memory Barrier
I_ void barrier_read (void){__builtin_ia32_lfence();} // Read Barrier
I_ void barrier_write (void){__builtin_ia32_sfence();} // Write Barrier
I_ U8 clock(void){U8 aa,dd;__asm__ volatile("rdtsc":"=a"(aa),"=d"(dd));return aa;}
I_ void pause(void){__asm__ volatile("pause":::"memory");}
#pragma endregion DSL
#pragma region Strings
typedef unsigned char def_tset(UTF8);
typedef def_struct(Str8) { U8 ptr; U8 len; }; typedef Str8 def_tset(Slice_UTF8);
typedef def_struct(Slice_Str8) { U8 ptr; U8 len; };
#define lit(string_literal) (Str8){ u8_(string_literal), size_of(string_literal) - 1 }
#pragma endregion Strings
#pragma region Debug
#ifdef BUILD_DEBUG
#define debug_trap() __debugbreak()
#define assert_trap(cond) do { if (cond) __debug_trap(); } while(0)
#define assert(cond) assert_msg(cond, nullptr)
#define assert_msg(cond, msg, ...) do { \
if (! (cond)) \
{ \
assert_handler( \
stringify(cond), \
__FILE__, \
__func__, \
cast(S4, __LINE__), \
msg, \
## __VA_ARGS__); \
debug_trap(); \
} \
} while(0)
// Deviation from Lottes's Convention: Don't want to mess with passing in typeless strings to the assert handler.
void assert_handler(UTF8*R_ condition, UTF8*R_ file, UTF8*R_ function, S4 line, UTF8*R_ msg, ... );
#else
#define debug_trap()
#define assert_trap(cond)
#define assert(cond)
#define assert_msg(cond, msg, ...)
#endif
#pragma endregion Debug
#pragma region Memory
typedef def_farray(B1, 1);
typedef def_farray(B1, 2);
typedef def_farray(B1, 4);
typedef def_farray(B1, 8);
I_ U8 mem_copy (U8 dest, U8 src, U8 len) { return (U8)(__builtin_memcpy ((void*)dest, (void const*)src, len)); }
I_ U8 mem_copy_overlapping(U8 dest, U8 src, U8 len) { return (U8)(__builtin_memmove((void*)dest, (void const*)src, len)); }
I_ U8 mem_fill (U8 dest, U8 value, U8 len) { return (U8)(__builtin_memset ((void*)dest, (int) value, len)); }
I_ B4 mem_zero (U8 dest, U8 len) { if (dest == 0) return false; mem_fill(dest, 0, len); return true; }
#define struct_copy(type, dest, src) mem_copy(dest, src, sizeof(type))
#define struct_zero(type, dest) mem_zero(dest, sizeof(type))
I_ U8 align_pow2(U8 x, U8 b) {
assert(b != 0);
assert((b & (b - 1)) == 0); // Check power of 2
return ((x + b - 1) & (~(b - 1)));
}
#define align_struct(type_width) ((U8)(((type_width) + 7) / 8 * 8))
#define assert_bounds(point, start, end) do { \
assert(start <= point); \
assert(point <= end); \
} while(0)
#define check_nil(nil, p) ((p) == 0 || (p) == nil)
#define set_nil(nil, p) ((p) = nil)
#define sll_stack_push_n(f, n, next) do { (n)->next = (f); (f) = (n); } while(0)
#define sll_queue_push_nz(nil, f, l, n, next) \
( \
check_nil(nil, f) ? ( \
(f) = (l) = (n), \
set_nil(nil, (n)->next) \
) \
: ( \
(l)->next=(n), \
(l) = (n), \
set_nil(nil,(n)->next) \
) \
)
#define sll_queue_push_n(f, l, n, next) sll_queue_push_nz(0, f, l, n, next)
#define def_Slice(type) def_struct(tmpl(Slice,type)) { type* ptr; U8 len; }; typedef def_ptr_set(tmpl(Slice,type))
#define slice_assert(slice) do { assert((slice).ptr != 0); assert((slice).len > 0); } while(0)
#define slice_end(slice) ((slice).ptr + (slice).len)
#define size_of_slice_type(slice) size_of( (slice).ptr[0] )
typedef def_struct(Slice_Mem) { U8 ptr; U8 len; };
#define slice_mem(ptr, len) ((Slice_Mem){u8_(ptr), u8_(len)})
#define slice_mem_s(slice) ((Slice_Mem){u8_((slice).ptr), (slice).len * size_of_slice_type(slice) })
typedef def_Slice(void);
typedef def_Slice(B1);
#define slice_to_bytes(slice) ((Slice_B1){cast(B1*, (slice).ptr), (slice).len * size_of_slice_type(slice)})
#define slice_fmem(mem) slice_mem(u8_(mem), size_of(mem))
I_ void slice__zero(Slice_B1 mem, U8 typewidth) { slice_assert(mem); memory_zero(u8_(mem.ptr), mem.len); }
#define slice_zero(slice) slice__zero(slice_mem_s(slice), size_of_slice_type(slice))
I_ void slice__copy(Slice_B1 dest, U8 dest_typewidth, Slice_B1 src, U8 src_typewidth) {
assert(dest.len >= src.len);
slice_assert(dest);
slice_assert(src);
mem_copy(u8_(dest.ptr), u8_(src.ptr), src.len);
}
#define slice_copy(dest, src) do { \
static_assert(typeof_same(dest, src)); \
slice__copy(slice_to_bytes(dest), size_of_slice_type(dest), slice_to_bytes(src), size_of_slice_type(src)); \
} while (0)
#define slice_iter(container, iter) (typeof((container).ptr) iter = (container).ptr; iter != slice_end(container); ++ iter)
#define slice_arg_from_array(type, ...) & (tmpl(Slice,type)) { .ptr = farray_init(type, __VA_ARGS__), .len = farray_len( farray_init(type, __VA_ARGS__)) }
#define span_iter(type, iter, m_begin, op, m_end) \
( \
tmpl(Iter_Span,type) iter = { \
.r = {(m_begin), (m_end)}, \
.cursor = (m_begin) }; \
iter.cursor op iter.r.end; \
++ iter.cursor \
)
#define def_span(type) \
def_struct(tmpl( Span,type)) { type begin; type end; }; \
typedef def_struct(tmpl(Iter_Span,type)) { tmpl(Span,type) r; type cursor; }
typedef def_span(B1);
typedef def_span(U4);
typedef def_span(U8);
#pragma endregion Memory
#pragma region Math
#define min(A, B) (((A) < (B)) ? (A) : (B))
#define max(A, B) (((A) > (B)) ? (A) : (B))
#define clamp_bot(X, B) max(X, B)
#pragma endregion Math
#pragma region Allocator Interface
typedef def_enum(U4, AllocatorOp) {
AllocatorOp_Alloc_NoZero = 0, // If Alloc exist, so must No_Zero
AllocatorOp_Alloc,
AllocatorOp_Free,
AllocatorOp_Reset,
AllocatorOp_Grow_NoZero,
AllocatorOp_Grow,
AllocatorOp_Shrink,
AllocatorOp_Rewind,
AllocatorOp_SavePoint,
AllocatorOp_Query, // Must always be implemented
};
typedef def_enum(U4, AllocatorQueryFlags) {
AllocatorQuery_Alloc = (1 << 0),
AllocatorQuery_Free = (1 << 1),
// Wipe the allocator's state
AllocatorQuery_Reset = (1 << 2),
// Supports both grow and shrink
AllocatorQuery_Shrink = (1 << 4),
AllocatorQuery_Grow = (1 << 5),
AllocatorQuery_Resize = AllocatorQuery_Grow | AllocatorQuery_Shrink,
// Ability to rewind to a save point (ex: arenas, stack), must also be able to save such a point
AllocatorQuery_Rewind = (1 << 6),
};
typedef struct AllocatorProc_In def_tset(AllocatorProc_In);
typedef struct AllocatorProc_Out def_tset(AllocatorProc_Out);
typedef struct AllocatorSP AllocatorSP;
typedef void def_proc(AllocatorProc) (U8 data, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out);
struct AllocatorSP {
AllocatorProc* type_sig;
U8 slot;
};
struct AllocatorProc_In {
U8 data;
U8 requested_size;
U8 alignment;
union {
Slice_Mem old_allocation;
AllocatorSP save_point;
};
AllocatorOp op;
A4_B1 _PAD_;
};
struct AllocatorProc_Out {
union {
Slice_Mem allocation;
AllocatorSP save_point;
};
AllocatorQueryFlags features;
A4_B1 _PAD_;
U8 left; // Contiguous memory left
U8 max_alloc;
U8 min_alloc;
A4_B1 _PAD_2;
};
typedef def_struct(AllocatorInfo) {
AllocatorProc* proc;
U8 data;
};
static_assert(size_of(AllocatorSP) <= size_of(Slice_Mem));
typedef def_struct(AllocatorQueryInfo) {
AllocatorSP save_point;
AllocatorQueryFlags features;
A4_B1 _PAD_;
U8 left; // Contiguous memory left
U8 max_alloc;
U8 min_alloc;
A4_B1 _PAD_2;
};
static_assert(size_of(AllocatorProc_Out) == size_of(AllocatorQueryInfo));
#define MEMORY_ALIGNMENT_DEFAULT (2 * size_of(void*))
I_ void allocator_query__u(U8 ainfo_proc, U8 ainfo_data, U8 allocator_query_info);
I_ void mem_free__u (U8 proc, U8 data, U8 mem_ptr, U8 mem_len);
I_ void mem_reset__u (U8 proc, U8 data);
I_ void mem_rewind__u (U8 proc, U8 data, U8 sp_type_sig, U8 sp_slot);
I_ void mem_save_point__u(U8 proc, U8 data, U8 sp);
I_ AllocatorQueryInfo allocator_query(AllocatorInfo ainfo);
I_ void mem_free (AllocatorInfo ainfo, Slice_Mem mem);
I_ void mem_reset (AllocatorInfo ainfo);
I_ void mem_rewind (AllocatorInfo ainfo, AllocatorSP save_point);
I_ AllocatorSP mem_save_point(AllocatorInfo ainfo);
I_ void mem__alloc__u (U8 out_mem, U8 proc, U8 data, U8 size, U8 alignemnt, B4 no_zero);
I_ void mem__grow__u (U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_actual);
I_ void mem__resize__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_actual);
I_ void mem__shrink__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment);
typedef def_struct(Opts_mem_alloc) { U8 alignment; B4 no_zero; A4_B1 _PAD_; };
typedef def_struct(Opts_mem_grow) { U8 alignment; B4 no_zero; B4 give_actual; };
typedef def_struct(Opts_mem_resize) { U8 alignment; B4 no_zero; B4 give_actual; };
typedef def_struct(Opts_mem_shrink) { U8 alignment; };
I_ Slice_Mem mem__alloc (AllocatorInfo ainfo, U8 size, Opts_mem_alloc_R opts);
I_ Slice_Mem mem__grow (AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_grow_R opts);
I_ Slice_Mem mem__resize(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_resize_R opts);
I_ Slice_Mem mem__shrink(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_shrink_R opts);
#define mem_alloc(ainfo, size, ...) mem__alloc (ainfo, size, opt_args(Opts_mem_alloc, __VA_ARGS__))
#define mem_grow(ainfo, mem, size, ...) mem__grow (ainfo, mem, size, opt_args(Opts_mem_grow, __VA_ARGS__))
#define mem_resize(ainfo, mem, size, ...) mem__resize(ainfo, mem, size, opt_args(Opts_mem_resize, __VA_ARGS__))
#define mem_shrink(ainfo, mem, size, ...) mem__shrink(ainfo, mem, size, opt_args(Opts_mem_shrink, __VA_ARGS__))
#define alloc_type(ainfo, type, ...) (type*) mem__alloc(ainfo, size_of(type), opt_args(Opts_mem_alloc, __VA_ARGS__)).ptr
#define alloc_slice(ainfo, type, num, ...) (tmpl(Slice,type)){ (type*)mem__alloc(ainfo, size_of(type) * num, opt_args(Opts_mem_alloc, __VA_ARGS__)).ptr, num }
#pragma endregion Allocator Interface
#pragma region FArena (Fixed-Sized Arena)
typedef def_struct(Opts_farena) {
U8 alignment;
};
typedef def_struct(FArena) {
U8 start;
U8 capacity;
U8 used;
};
I_ void farena_init__u (U8 arena, U8 mem_ptr, U8 mem_len);
void farena__push__u (U8 arena, U8 amount, U8 type_width, U8 alignment, U8 slice_mem);
I_ void farena_reset__u (U8 arena);
I_ void farena_rewind__u(U8 arena, U8 sp_type_sig, U8 sp_slot);
I_ void farena_save__u (U8 arena, U8 sp);
I_ FArena farena_make (Slice_Mem mem);
I_ void farena_init (FArena_R arena, Slice_Mem byte);
I_ Slice_Mem farena__push (FArena_R arena, U8 amount, U8 type_width, Opts_farena*R_ opts);
I_ void farena_reset (FArena_R arena);
I_ void farena_rewind(FArena_R arena, AllocatorSP save_point);
I_ AllocatorSP farena_save (FArena arena);
void farena_allocator_proc(U8 data, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out);
#define ainfo_farena(arena) (AllocatorInfo){ .proc = farena_allocator_proc, .data = u8_(& arena) }
#define farena_push_mem(arena, amount, ...) farena__push(arena, amount, 1, opt_args(Opts_farena, lit(stringify(B1)), __VA_ARGS__))
#define farena_push(arena, type, ...) \
cast(type*, farena__push(arena, size_of(type), 1, opt_args(Opts_farena, __VA_ARGS__))).ptr
#define farena_push_array(arena, type, amount, ...) \
(Slice ## type){ farena__push(arena, size_of(type), amount, opt_args(Opts_farena, __VA_ARGS__)).ptr, amount }
#pragma endregion FArena
#pragma endregion Header
#pragma region Implementation
#pragma region Allocator Interface
I_ void allocator_query__u(U8 ainfo_proc, U8 ainfo_data, U8 allocator_query_info) {
assert(ainfo_proc != null);
cast(AllocatorProc*, ainfo_proc)(ainfo_data, 0, 0, 0, 0, AllocatorOp_Query, allocator_query_info);
}
I_ void mem_free__u(U8 proc, U8 data, U8 mem_ptr, U8 mem_len) {
assert(proc != null);
cast(AllocatorProc*, proc)(data, 0, 0, mem_ptr, mem_len, AllocatorOp_Free, 0);
}
I_ void mem_reset__u(U8 proc, U8 data) {
assert(proc != null);
cast(AllocatorProc*, proc)(data, 0, 0, 0, 0, AllocatorOp_Reset, 0);
}
I_ void mem_rewind__u(U8 proc, U8 data, U8 sp_type_sig, U8 sp_slot) {
assert(proc != null);
cast(AllocatorProc*, proc)(data, 0, 0, sp_type_sig, sp_slot, AllocatorOp_Rewind, 0);
}
I_ void mem_save_point__u(U8 proc, U8 data, U8 sp) {
assert(proc != null);
uvar(AllocatorProc_Out, out) = {0};
cast(AllocatorProc*, proc)(data, 0, 0, 0, 0, AllocatorOp_SavePoint, u8_(out));
struct_copy(AllocatorSP, sp, (U8) out + offset_of(AllocatorProc_Out, save_point));
}
I_ void mem__alloc__u(U8 out_mem, U8 proc, U8 data, U8 size, U8 alignment, B4 no_zero) {
assert(proc != null);
uvar(AllocatorProc_Out, out) = {0};
cast(AllocatorProc*, proc)(data, size, alignment, 0, 0, no_zero ? AllocatorOp_Alloc_NoZero : AllocatorOp_Alloc, u8_(out));
struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
}
I_ void mem__grow__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_actual) {
assert(proc != null);
uvar(AllocatorProc_Out, out) = {0};
cast(AllocatorProc*, proc)(data, size, alignment, old_ptr, old_len, no_zero ? AllocatorOp_Grow_NoZero : AllocatorOp_Grow, u8_(out));
if (give_actual == false) { u8_r(out + offset_of(AllocatorProc_Out, allocation) + offset_of(Slice_Mem, len))[0] = size; }
struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
}
I_ void mem__shrink__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment) {
assert(proc != null);
uvar(AllocatorProc_Out, out) = {0};
cast(AllocatorProc*, proc)(data, size, alignment, old_ptr, old_len, AllocatorOp_Shrink, u8_(out));
struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
}
I_ void mem__resize__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_acutal) {
if (old_len == size) { struct_copy(Slice_Mem, out_mem, (U8)& slice_mem(old_ptr, old_len)); }
if (old_len < size) { mem__grow__u (out_mem, proc, data, old_ptr, old_len, size, alignment, no_zero, give_acutal); }
else { mem__shrink__u(out_mem, proc, data, old_ptr, old_len, size, alignment); }
}
I_ AllocatorQueryInfo allocator_query(AllocatorInfo ainfo) { AllocatorQueryInfo out; allocator_query__u(u8_(ainfo.proc), ainfo.data, u8_(& out)); return out; }
I_ void mem_free (AllocatorInfo ainfo, Slice_Mem mem) { mem_free__u (u8_(ainfo.proc), ainfo.data, mem.ptr, mem.len); }
I_ void mem_reset (AllocatorInfo ainfo) { mem_reset__u (u8_(ainfo.proc), ainfo.data); }
I_ void mem_rewind(AllocatorInfo ainfo, AllocatorSP save_point) { mem_rewind__u(u8_(ainfo.proc), ainfo.data, u8_(save_point.type_sig), save_point.slot); }
I_ AllocatorSP mem_save_point(AllocatorInfo ainfo) { AllocatorSP sp; mem_save_point__u(u8_(ainfo.proc), ainfo.data, u8_(& sp)); return sp; }
I_ Slice_Mem mem__alloc(AllocatorInfo ainfo, U8 size, Opts_mem_alloc_R opts) {
assert(opts != nullptr); Slice_Mem result;
mem__alloc__u(u8_(& result), u8_(ainfo.proc), ainfo.data, size, opts->alignment, opts->no_zero);
return result;
}
I_ Slice_Mem mem__grow(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_grow_R opts) {
assert(opts != nullptr);
Slice_Mem out; mem__grow__u(u8_(& out), u8_(ainfo.proc), ainfo.data, mem.ptr, mem.len, size, opts->alignment, opts->no_zero, opts->give_actual);
if (!opts->give_actual) { out.len = size; }
return out;
}
I_ Slice_Mem mem__resize(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_resize_R opts) {
assert(opts != nullptr);
Slice_Mem out; mem__resize__u(u8_(& out), u8_(ainfo.proc), ainfo.data, mem.ptr, mem.len, size, opts->alignment, opts->no_zero, opts->give_actual);
return out;
}
I_ Slice_Mem mem__shrink(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_shrink_R opts) {
assert(opts != nullptr);
Slice_Mem out; mem__shrink__u(u8_(& out), u8_(ainfo.proc), ainfo.data, mem.ptr, mem.len, size, opts->alignment);
return out;
}
#pragma endregion Allocator Interface
#pragma region FArena (Fixed-Sized Arena)
I_ void farena_init__u(U8 arena, U8 mem_ptr, U8 mem_len) {
assert(arena != null);
u8_r(arena + offset_of(FArena, start) )[0] = mem_ptr;
u8_r(arena + offset_of(FArena, capacity))[0] = mem_len;
u8_r(arena + offset_of(FArena, used) )[0] = 0;
}
inline void farena__push__u(U8 arena, U8 amount, U8 type_width, U8 alignment, U8 result) {
if (amount == 0) { struct_zero(Slice_Mem, result); }
U8 reg desired = type_width * amount;
U8 reg to_commit = align_pow2(desired, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
U8_R reg used = u8_r(arena + offset_of(FArena, used));
U8 reg unused = u8_r(arena + offset_of(FArena, capacity))[0] - used[0]; assert(to_commit <= unused);
U8 reg ptr = u8_r(arena + offset_of(FArena, start) )[0] + used[0];
used[0] += to_commit;
struct_copy(Slice_Mem, result, (U8)& slice_mem(ptr, desired));
}
inline void farena__grow__u(U8 result, U8 arena, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment, B4 should_zero) {
assert(result != null);
assert(arena != null);
U8_R reg used = u8_r(arena + offset_of(FArena, used));
/*Check if the allocation is at the end of the arena*/{
U8 reg alloc_end = old_ptr + old_len;
U8 reg arena_end = u8_r(arena + offset_of(FArena, start))[0] + used[0];
if (alloc_end != arena_end) {
// Not at the end, can't grow in place
struct_zero(Slice_Mem, result);
return;
}
}
// Calculate growth
U8 reg grow_amount = requested_size - old_len;
U8 reg aligned_grow = align_pow2(grow_amount, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
U8 reg unused = u8_r(arena + offset_of(FArena, capacity))[0] - used[0];
if (aligned_grow > unused) {
// Not enough space
struct_zero(Slice_Mem, result);
return;
}
used[0] += aligned_grow;
struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, aligned_grow + requested_size));
memory_zero(old_ptr + old_len, grow_amount * cast(U8, should_zero));
}
inline void farena__shrink__u(U8 result, U8 arena, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment) {
assert(result != null);
assert(arena != null);
U8_R reg used = u8_r(arena + offset_of(FArena, used));
/*Check if the allocation is at the end of the arena*/ {
U8 reg alloc_end = old_ptr + old_len;
U8 reg arena_end = u8_r(arena + offset_of(FArena, start))[0] + used[0];
if (alloc_end != arena_end) {
// Not at the end, can't shrink but return adjusted size
struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, requested_size));
return;
}
}
U8 reg aligned_original = align_pow2(old_len, MEMORY_ALIGNMENT_DEFAULT);
U8 reg aligned_new = align_pow2(requested_size, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
used[0] -= (aligned_original - aligned_new);
struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, requested_size));
}
I_ void farena_reset__u(U8 arena) { u8_r(arena + offset_of(FArena, used))[0] = 0; }
I_ void farena_rewind__u(U8 arena, U8 sp_type_sig, U8 sp_slot) {
assert(sp_type_sig == (U8)& farena_allocator_proc);
U8 reg start = u8_r(arena + offset_of(FArena, start))[0];
U8_R reg used = u8_r(arena + offset_of(FArena, used));
U8 reg end = start + used[0]; assert_bounds(sp_slot, start, end);
used[0] -= sp_slot - start;
}
I_ void farena_save__u(U8 arena, U8 sp) {
u8_r(sp + offset_of(AllocatorSP, type_sig))[0] = (U8)& farena_allocator_proc;
u8_r(sp + offset_of(AllocatorSP, slot ))[0] = u8_r(arena + offset_of(FArena, used))[0];
}
void farena_allocator_proc(U8 arena, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out)
{
assert(out != null);
assert(arena != null);
U8 reg allocation = arena + offset_of(AllocatorProc_Out, allocation);
switch (op)
{
case AllocatorOp_Alloc:
case AllocatorOp_Alloc_NoZero:
farena__push__u(arena, requested_size, 1, alignment, allocation);
memory_zero(u8_r(allocation + offset_of(Slice_Mem, ptr))[0], u8_r(allocation + offset_of(Slice_Mem, len))[0] * op);
break;
case AllocatorOp_Free: break;
case AllocatorOp_Reset: farena_reset__u(arena); break;
case AllocatorOp_Grow:
case AllocatorOp_Grow_NoZero:
farena__grow__u(allocation, arena, old_ptr, old_len, requested_size, alignment, op - AllocatorOp_Grow_NoZero);
break;
case AllocatorOp_Shrink:
farena__shrink__u(allocation, arena, old_ptr, old_len, requested_size, alignment);
break;
case AllocatorOp_Rewind: farena_rewind__u(arena, old_ptr, old_len); break;
case AllocatorOp_SavePoint: farena_save__u(arena, allocation); break;
case AllocatorOp_Query:
u4_r(out + offset_of(AllocatorQueryInfo, features))[0] =
AllocatorQuery_Alloc
| AllocatorQuery_Reset
| AllocatorQuery_Resize
| AllocatorQuery_Rewind
;
U8 reg max_alloc = u8_r(arena + offset_of(FArena, capacity))[0] - u8_r(arena + offset_of(FArena, used))[0];
u8_r(out + offset_of(AllocatorQueryInfo, max_alloc))[0] = max_alloc;
u8_r(out + offset_of(AllocatorQueryInfo, min_alloc))[0] = 0;
u8_r(out + offset_of(AllocatorQueryInfo, left ))[0] = max_alloc;
farena_save__u(arena, out + offset_of(AllocatorQueryInfo, save_point));
break;
}
return;
}
#pragma endregion FArena
#pragma endregion Implementation
int main(void)

200
scripts/build.c_lottes.ps1 Normal file
View File

@@ -0,0 +1,200 @@
$misc = Join-Path $PSScriptRoot 'helpers/misc.psm1'
import-module $misc
# This script now uses the LLVM toolchain (clang-cl, lld-link).
# Ensure these tools are available in your PATH.
# The original call to the MSVC devshell has been removed.
# & (join-path $PSScriptRoot 'helpers/devshell.ps1') -arch amd64
$path_root = Get-ScriptRepoRoot
$path_root = split-path -Path $PSScriptRoot -Parent
$path_toolchain = join-path $path_root 'toolchain'
$path_rad = join-path $path_toolchain 'rad'
# https://learn.microsoft.com/en-us/cpp/build/reference/compiler-options-listed-by-category?view=msvc-170
# Most cl.exe flags are compatible with clang-cl.exe
$flag_all_c = '/TC'
$flag_c11 = '/std:c11'
$flag_c23 = '/std:c23'
$flag_all_cpp = '/TP'
$flag_compile = '/c'
$flag_charset_utf8 = '/utf-8'
$flag_debug = '/Zi'
$flag_define = '/D'
$flag_exceptions_disabled = '/EHsc-'
$flag_RTTI_disabled = '/GR-'
$flag_include = '/I'
$flag_full_src_path = '/FC'
$flag_asm_listing_file = '/FAs'
$flag_nologo = '/nologo'
$flag_dll = '/LD'
$flag_dll_debug = '/LDd'
$flag_linker = '/link'
# $flag_link_lib = '/lib'
$flag_link_dll = '/DLL'
$flag_link_no_incremental = '/INCREMENTAL:NO'
$flag_link_mapfile = '/MAP:'
$flag_link_optimize_references = '/OPT:REF'
$flag_link_win_debug = '/DEBUG'
$flag_link_win_pdb = '/PDB:'
$flag_link_win_machine_32 = '/MACHINE:X86'
$flag_link_win_machine_64 = '/MACHINE:X64'
$flag_link_win_path_output = '/OUT:'
$flag_link_win_rt_dll = '/MD'
$flag_link_win_rt_dll_debug = '/MDd'
$flag_link_win_rt_static = '/MT'
$flag_link_win_rt_static_debug = '/MTd'
$flag_link_win_subsystem_console = '/SUBSYSTEM:CONSOLE'
$flag_link_win_subsystem_windows = '/SUBSYSTEM:WINDOWS'
$flag_no_optimization = '/Od'
$flag_optimize_speed_max = '/Ox'
$flag_optimize_fast = '/O2'
$flag_optimize_size = '/O1'
$flag_optimize_intrinsics = '/Oi'
$flag_optimized_debug_forceinline = '/d2Obforceinline'
$flag_optimized_debug = '/Zo'
$flag_preprocess_to_file = '/P'
$flag_preprocess_preserve_comments = '/C'
# $flag_out_name = '/OUT:'
$flag_path_interm = '/Fo'
$flag_path_debug = '/Fd'
$flag_path_output = '/Fe'
$flag_preprocess_conform = '/Zc:preprocessor'
$flag_sanitize_address = '/fsanitize=address'
$flag_updated_cpp_macro = "/Zc:__cplusplus"
$flag_set_stack_size = '/F'
$flag_syntax_only = '/Zs'
$flag_wall = '/Wall'
$flag_warnings_as_errors = '/WX'
$flag_lib_list = '/LIST'
$archiver = 'llvm-lib'
$compiler = 'clang-cl'
$linker = 'lld-link'
$radbin = join-path $path_rad 'radbin.exe'
$radlink = join-path $path_rad 'radlink.exe'
$path_build = join-path $path_root 'build'
if ( -not(test-path -Path $path_build) ) {
new-item -ItemType Directory -Path $path_build
}
push-location $path_build
write-host "Compiling with clang-cl"
$compiler_args = @()
$compiler_args += $flag_nologo
# Constraints on interpeting all files as C code
$compiler_args += $flag_all_c
$compiler_args += $flag_c11
# Constraints on C program code-gen
$compiler_args += $flag_exceptions_disabled
$compiler_args += $flag_RTTI_disabled
# $compiler_args += $flag_preprocess_conform
# $compiler_args += $flag_sanitize_address
$compiler_args += $flag_wall
# Set charset encoding for both execution and source to UTF-8
$compiler_args += $flag_charset_utf8
# Specifing output pathing
$compiler_args += ( $flag_path_interm + $path_build + '\' )
# $compiler_args += ( $flag_path_output + $path_build + '\' )
# Dump preprocess file
if ($false) {
$compiler_args += $flag_preprocess_to_file
$compiler_args += $flag_preprocess_preserve_comments
}
# Diagnostic logging
$compiler_args += $flag_full_src_path
# $compiler_args += $flag_asm_listing_file
# $compiler_args += $flag_optimize_speed_max
# $compiler_args += $flag_optimize_fast
# $compiler_args += $flag_optimize_size
# $compiler_args += $flag_optimize_intrinsics
$compiler_args += $flag_no_optimization
# Debug setup
$compiler_args += ($flag_define + 'BUILD_DEBUG')
$compiler_args += $flag_debug
$compiler_args += ( $flag_path_debug + $path_build + '\' )
# Use the static, multithreaded, debug runtime library
# $compiler_args += $flag_link_win_rt_static_debug
# Include setup
$compiler_args += ($flag_include + $path_root)
$unit_name = "watl.v0.llvm.lottes"
# Specify unit to compile
$unit = join-path $path_root "C\$unit_name.c"
$compiler_args += $flag_compile, $unit
# Diagnoistc print for the args
$compiler_args | ForEach-Object { Write-Host $_ }
# Compile the unit
$compilation_time = Measure-Command {
& $compiler $compiler_args
}
write-host "Compilation took $($compilation_time.TotalMilliseconds)ms"
write-host
$binary = join-path $path_build "$unit_name.exe"
$object = join-path $path_build "$unit_name.obj"
$pdb = join-path $path_build "$unit_name.pdb"
$map = join-path $path_build "$unit_name.map"
$rdi = join-path $path_build "$unit_name.rdi"
$rdi_listing = join-path $path_build "$unit_name.rdi.list"
if ($true) {
write-host "Linking with lld-link"
$linker_args = @()
$linker_args += $flag_nologo
$linker_args += $flag_link_win_machine_64
$linker_args += $flag_link_no_incremental
$linker_args += ($flag_link_win_path_output + $binary)
$linker_args += "$flag_link_win_debug"
$linker_args += $flag_link_win_pdb + $pdb
$linker_args += $flag_link_mapfile + $map
$linker_args += $flag_link_win_subsystem_console
$linker_args += $object
# Add necessary libraries for a basic Windows application
$linker_args += "kernel32.lib", "user32.lib", "gdi32.lib"
# Diagnoistc print for the args
$linker_args | ForEach-Object { Write-Host $_ }
$linking_time = Measure-Command { & $linker $linker_args }
# & $radlink $linker_args
write-host "Linking took $($linking_time.TotalMilliseconds)ms"
write-host
}
if ($false) {
write-host "Dumping Debug Info"
$rbin_out = '--out:'
$rbin_dump = '--dump'
$nargs = @($pdb, ($rbin_out + $rdi))
& $radbin $nargs
$nargs = @($rbin_dump, $rdi)
$dump = & $radbin $nargs
$dump > $rdi_listing
}
Pop-Location