Generate handle_cycle_limit at runtime so it can use the generated save/load_context functions. Since the hand written versions of save/load are no longer used they have been removed.

3 files changed, 41 insertions, 74 deletions

diff --git a/m68k_to_x86.c b/m68k_to_x86.c
index 916a9b1..54ec824 100644
--- a/m68k_to_x86.c
+++ b/m68k_to_x86.c
@@ -30,12 +30,10 @@ char disasm_buf[1024];
 
 m68k_context * sync_components(m68k_context * context, uint32_t address);
 
-void handle_cycle_limit();
 void m68k_invalid();
 void set_sr();
 void set_ccr();
 void get_sr();
-void do_sync();
 void bcd_add();
 void bcd_sub();
 
@@ -56,12 +54,12 @@ uint8_t * check_cycles_int(uint8_t * dst, uint32_t address, x86_68k_options * op
 	return dst;
 }
 
-uint8_t * check_cycles(uint8_t * dst)
+uint8_t * check_cycles(uint8_t * dst, x86_68k_options * opts)
 {
 	dst = cmp_rr(dst, CYCLES, LIMIT, SZ_D);
 	uint8_t * jmp_off = dst+1;
 	dst = jcc(dst, CC_NC, dst + 7);
-	dst = call(dst, (uint8_t *)handle_cycle_limit);
+	dst = call(dst, opts->handle_cycle_limit);
 	*jmp_off = dst - (jmp_off+1);
 	return dst;
 }
@@ -2850,7 +2848,7 @@ uint8_t * translate_m68k(uint8_t * dst, m68kinst * inst, x86_68k_options * opts)
 				dst = mov_rrdisp8(dst, SCRATCH1, CONTEXT, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_B);
 			}
 			if (inst->src.params.immed & 0x700) {
-				dst = call(dst, (uint8_t *)do_sync);
+				dst = call(dst, opts->do_sync);
 			}
 		}
 		break;
@@ -3166,7 +3164,7 @@ uint8_t * translate_m68k(uint8_t * dst, m68kinst * inst, x86_68k_options * opts)
 		if (inst->op == M68K_ORI_SR) {
 			dst = xor_irdisp8(dst, inst->src.params.immed >> 8, CONTEXT, offsetof(m68k_context, status), SZ_B);
 			if (inst->src.params.immed & 0x700) {
-				dst = call(dst, (uint8_t *)do_sync);
+				dst = call(dst, opts->do_sync);
 			}
 		}
 		break;
@@ -3225,7 +3223,7 @@ uint8_t * translate_m68k(uint8_t * dst, m68kinst * inst, x86_68k_options * opts)
 					dst = mov_rdisp8r(dst, CONTEXT, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->aregs[7], SZ_D);
 					dst = mov_rrdisp8(dst, SCRATCH1, CONTEXT, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D);
 				}
-				dst = call(dst, (uint8_t *)do_sync);
+				dst = call(dst, opts->do_sync);
 			}
 			dst = cycles(dst, 12);
 		} else {
@@ -3414,7 +3412,7 @@ uint8_t * translate_m68k(uint8_t * dst, m68kinst * inst, x86_68k_options * opts)
 		if (inst->op == M68K_ORI_SR) {
 			dst = or_irdisp8(dst, inst->src.params.immed >> 8, CONTEXT, offsetof(m68k_context, status), SZ_B);
 			if (inst->src.params.immed & 0x700) {
-				dst = call(dst, (uint8_t *)do_sync);
+				dst = call(dst, opts->do_sync);
 			}
 		}
 		break;
@@ -3714,7 +3712,7 @@ uint8_t * translate_m68k(uint8_t * dst, m68kinst * inst, x86_68k_options * opts)
 			dst = mov_rrdisp8(dst, SCRATCH1, CONTEXT, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D);
 		}
 		uint8_t * loop_top = dst;
-		dst = call(dst, (uint8_t *)do_sync);
+		dst = call(dst, opts->do_sync);
     dst = cmp_rr(dst, LIMIT, CYCLES, SZ_D);
     uint8_t * normal_cycle_up = dst + 1;
     dst = jcc(dst, CC_A, dst+2);
@@ -4116,7 +4114,7 @@ uint8_t * gen_mem_fun(x86_68k_options * opts, memmap_chunk * memmap, uint32_t nu
 {
 	uint8_t * dst = opts->cur_code;
 	uint8_t * start = dst;
-	dst = check_cycles(dst);
+	dst = check_cycles(dst, opts);
 	dst = cycles(dst, BUS);
 	dst = and_ir(dst, 0xFFFFFF, SCRATCH1, SZ_D);
 	uint8_t *lb_jcc = NULL, *ub_jcc = NULL;
@@ -4436,6 +4434,34 @@ void init_x86_68k_opts(x86_68k_options * opts, memmap_chunk * memmap, uint32_t n
 	dst = call(dst, opts->load_context);
 	dst = retn(dst);
 
+	opts->handle_cycle_limit = dst;
+	dst = cmp_rdisp8r(dst, CONTEXT, offsetof(m68k_context, sync_cycle), CYCLES, SZ_D);
+	uint8_t * skip_sync = dst+1;
+	dst = jcc(dst, CC_C, dst+2);
+	opts->do_sync = dst;
+	dst = push_r(dst, SCRATCH1);
+	dst = push_r(dst, SCRATCH2);
+	dst = call(dst, opts->save_context);
+	dst = mov_rr(dst, CONTEXT, RDI, SZ_Q);
+	dst = xor_rr(dst, RSI, RSI, SZ_D);
+	dst = test_ir(dst, 8, RSP, SZ_D);
+	uint8_t *adjust_rsp = dst+1;
+	dst = jcc(dst, CC_NZ, dst+2);
+	dst = call(dst, (uint8_t *)sync_components);
+	uint8_t *no_adjust = dst+1;
+	dst = jmp(dst, dst+2);
+	*adjust_rsp = dst - (adjust_rsp + 1);
+	dst = sub_ir(dst, 8, RSP, SZ_Q);
+	dst = call(dst, (uint8_t *)sync_components);
+	dst = add_ir(dst, 8, RSP, SZ_Q);
+	*no_adjust = dst - (no_adjust+1);
+	dst = mov_rr(dst, RAX, CONTEXT, SZ_Q);
+	dst = call(dst, opts->load_context);
+	dst = pop_r(dst, SCRATCH2);
+	dst = pop_r(dst, SCRATCH1);
+	*skip_sync = dst - (skip_sync+1);
+	dst = retn(dst);
+
 	opts->cur_code = dst;
 
 	opts->read_16 = gen_mem_fun(opts, memmap, num_chunks, READ_16);
@@ -4484,16 +4510,16 @@ void init_x86_68k_opts(x86_68k_options * opts, memmap_chunk * memmap, uint32_t n
 	uint8_t * do_int = dst+1;
 	dst = jcc(dst, CC_NC, dst+2);
 	dst = cmp_rdisp8r(dst, CONTEXT, offsetof(m68k_context, sync_cycle), CYCLES, SZ_D);
-	uint8_t * skip_sync = dst+1;
+	skip_sync = dst+1;
 	dst = jcc(dst, CC_C, dst+2);
 	dst = call(dst, opts->save_context);
 	dst = mov_rr(dst, CONTEXT, RDI, SZ_Q);
 	dst = mov_rr(dst, SCRATCH1, RSI, SZ_D);
 	dst = test_ir(dst, 8, RSP, SZ_D);
-	uint8_t *adjust_rsp = dst+1;
+	adjust_rsp = dst+1;
 	dst = jcc(dst, CC_NZ, dst+2);
 	dst = call(dst, (uint8_t *)sync_components);
-	uint8_t *no_adjust = dst+1;
+	no_adjust = dst+1;
 	dst = jmp(dst, dst+2);
 	*adjust_rsp = dst - (adjust_rsp + 1);
 	dst = sub_ir(dst, 8, RSP, SZ_Q);
diff --git a/m68k_to_x86.h b/m68k_to_x86.h
index 4b64c53..c3a50d7 100644
--- a/m68k_to_x86.h
+++ b/m68k_to_x86.h
@@ -36,7 +36,9 @@ typedef struct {
 	uint8_t         *read_32;
 	uint8_t         *write_32_lowfirst;
 	uint8_t         *write_32_highfirst;
+	uint8_t         *handle_cycle_limit;
 	uint8_t         *handle_cycle_limit_int;
+	uint8_t         *do_sync;
 	uint8_t         *trap;
 	uint8_t			*save_context;
 	uint8_t			*load_context;
diff --git a/runtime.S b/runtime.S
index b93a46e..ded2c2e 100644
--- a/runtime.S
+++ b/runtime.S
@@ -1,30 +1,4 @@
 
-	.global handle_cycle_limit
-	.global do_sync
-handle_cycle_limit:
-	cmp 84(%rsi), %eax
-	jb skip_sync
-do_sync:
-	push %rcx
-	push %rdi
-	call m68k_save_context
-	mov %rsi, %rdi
-	xor %esi, %esi
-	test $8, %esp
-	jnz adjust_rsp
-	call sync_components
-	jmp done_adjust
-adjust_rsp:
-	sub $8, %rsp
-	call sync_components
-	add $8, %rsp
-done_adjust:
-	mov %rax, %rsi
-	call m68k_load_context
-	pop %rdi
-	pop %rcx
-skip_sync:
-	ret
 
 invalid_msg:
 	.asciz "Invalid instruction at %X\n"
@@ -149,38 +123,3 @@ set_ccr:
 	mov %cl, (%rsi)
 	ret
 
-	.global m68k_save_context
-m68k_save_context:
-	mov %bl, 1(%rsi) /* N Flag */
-	mov %dl, 2(%rsi) /* Z flag */
-	mov %bh, 3(%rsi) /* V flag */
-	mov %dh, 4(%rsi) /* C flag */
-	mov %r10d, 8(%rsi) /* d0 */
-	mov %r11d, 12(%rsi) /* d1 */
-	mov %r12d, 16(%rsi) /* d2 */
-	mov %r8d,  20(%rsi) /* d3 */
-	mov %r13d, 40(%rsi) /* a0 */
-	mov %r14d, 44(%rsi) /* a1 */
-	mov %r9d,  48(%rsi) /* a2 */
-	mov %r15d, 68(%rsi) /* a7 */
-	mov %eax, 80(%rsi) /* current cycle count */
-	ret
-
-	.global m68k_load_context
-m68k_load_context:
-	mov 1(%rsi), %bl /* N Flag */
-	mov 2(%rsi), %dl /* Z flag */
-	mov 3(%rsi), %bh /* V flag */
-	mov 4(%rsi), %dh /* C flag */
-	mov 8(%rsi), %r10d /* d0 */
-	mov 12(%rsi), %r11d /* d1 */
-	mov 16(%rsi), %r12d /* d2 */
-	mov 20(%rsi), %r8d  /* d3 */
-	mov 40(%rsi), %r13d /* a0 */
-	mov 44(%rsi), %r14d /* a1 */
-	mov 48(%rsi), %r9d  /* a2 */
-	mov 68(%rsi), %r15d /* a7 */
-	mov 76(%rsi), %ebp /* target cycle count */
-	mov 80(%rsi), %eax /* current cycle count */
-	ret
-