aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMike Pall <mike>2015-01-04 23:01:16 +0100
committerMike Pall <mike>2015-01-04 23:01:16 +0100
commitdfe84fd8c545b6f2da721e3f86f9b00f31711abd (patch)
treeaec798be428f542a17d36990686f83a0552ae70d
parent968725c4d57851f46d5aa51ad8db8880e323300b (diff)
downloadluajit-dfe84fd8c545b6f2da721e3f86f9b00f31711abd.tar.gz
luajit-dfe84fd8c545b6f2da721e3f86f9b00f31711abd.tar.bz2
luajit-dfe84fd8c545b6f2da721e3f86f9b00f31711abd.zip
PPC/e500: Drop support for this architecture.
Diffstat
-rw-r--r--doc/ext_jit.html2
-rw-r--r--doc/install.html9
-rw-r--r--doc/luajit.html2
-rw-r--r--src/Makefile5
-rw-r--r--src/host/buildvm.c4
-rw-r--r--src/host/buildvm_asm.c7
-rw-r--r--src/jit/bcsave.lua6
-rw-r--r--src/lib_jit.c2
-rw-r--r--src/lj_arch.h37
-rw-r--r--src/lj_ccall.c36
-rw-r--r--src/lj_ccall.h10
-rw-r--r--src/lj_ctype.h2
-rw-r--r--src/lj_frame.h9
-rw-r--r--src/vm_ppcspe.dasc3685
14 files changed, 16 insertions, 3800 deletions
diff --git a/doc/ext_jit.html b/doc/ext_jit.html
index ddc8337d..7be5c99f 100644
--- a/doc/ext_jit.html
+++ b/doc/ext_jit.html
@@ -153,7 +153,7 @@ Contains the target OS name:
153<h3 id="jit_arch"><tt>jit.arch</tt></h3> 153<h3 id="jit_arch"><tt>jit.arch</tt></h3>
154<p> 154<p>
155Contains the target architecture name: 155Contains the target architecture name:
156"x86", "x64", "arm", "ppc", "ppcspe", or "mips". 156"x86", "x64", "arm", "ppc", or "mips".
157</p> 157</p>
158 158
159<h2 id="jit_opt"><tt>jit.opt.*</tt> &mdash; JIT compiler optimization control</h2> 159<h2 id="jit_opt"><tt>jit.opt.*</tt> &mdash; JIT compiler optimization control</h2>
diff --git a/doc/install.html b/doc/install.html
index 2ad60c32..4d4b5ae3 100644
--- a/doc/install.html
+++ b/doc/install.html
@@ -141,13 +141,6 @@ operating systems, CPUs and compilers:
141<td class="compatos">XEDK (<a href="#xbox360">Xbox 360</a>)</td> 141<td class="compatos">XEDK (<a href="#xbox360">Xbox 360</a>)</td>
142</tr> 142</tr>
143<tr class="odd"> 143<tr class="odd">
144<td class="compatcpu"><a href="#cross2">PPC/e500v2</a></td>
145<td class="compatos">GCC 4.3+</td>
146<td class="compatos">GCC 4.3+</td>
147<td class="compatos compatno">&nbsp;</td>
148<td class="compatos compatno">&nbsp;</td>
149</tr>
150<tr class="even">
151<td class="compatcpu"><a href="#cross2">MIPS</a></td> 144<td class="compatcpu"><a href="#cross2">MIPS</a></td>
152<td class="compatos">GCC 4.3+</td> 145<td class="compatos">GCC 4.3+</td>
153<td class="compatos">GCC 4.3+</td> 146<td class="compatos">GCC 4.3+</td>
@@ -387,8 +380,6 @@ make HOST_CC="gcc -m32" CROSS=arm-linux-gnueabihf-
387 380
388# PPC 381# PPC
389make HOST_CC="gcc -m32" CROSS=powerpc-linux-gnu- 382make HOST_CC="gcc -m32" CROSS=powerpc-linux-gnu-
390# PPC/e500v2 (fast interpreter only)
391make HOST_CC="gcc -m32" CROSS=powerpc-e500v2-linux-gnuspe-
392 383
393# MIPS big-endian 384# MIPS big-endian
394make HOST_CC="gcc -m32" CROSS=mips-linux- 385make HOST_CC="gcc -m32" CROSS=mips-linux-
diff --git a/doc/luajit.html b/doc/luajit.html
index 1a85f033..b9984f4b 100644
--- a/doc/luajit.html
+++ b/doc/luajit.html
@@ -172,7 +172,7 @@ LuaJIT is Copyright &copy; 2005-2014 Mike Pall, released under the
172<tr><td>GCC</td><td>CLANG<br>LLVM</td><td>MSVC</td></tr> 172<tr><td>GCC</td><td>CLANG<br>LLVM</td><td>MSVC</td></tr>
173</table> 173</table>
174<table class="feature cpu"> 174<table class="feature cpu">
175<tr><td>x86</td><td>x64</td><td>ARM</td><td>PPC</td><td>e500</td><td>MIPS</td></tr> 175<tr><td>x86</td><td>x64</td><td>ARM</td><td>PPC</td><td>MIPS</td></tr>
176</table> 176</table>
177<table class="feature fcompat"> 177<table class="feature fcompat">
178<tr><td>Lua&nbsp;5.1<br>API+ABI</td><td>+&nbsp;JIT</td><td>+&nbsp;BitOp</td><td>+&nbsp;FFI</td><td>Drop-in<br>DLL/.so</td></tr> 178<tr><td>Lua&nbsp;5.1<br>API+ABI</td><td>+&nbsp;JIT</td><td>+&nbsp;BitOp</td><td>+&nbsp;FFI</td><td>Drop-in<br>DLL/.so</td></tr>
diff --git a/src/Makefile b/src/Makefile
index 88dc8166..90cb28ce 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -50,7 +50,6 @@ CCOPT_x64=
50CCOPT_arm= 50CCOPT_arm=
51CCOPT_arm64= 51CCOPT_arm64=
52CCOPT_ppc= 52CCOPT_ppc=
53CCOPT_ppcspe=
54CCOPT_mips= 53CCOPT_mips=
55# 54#
56CCDEBUG= 55CCDEBUG=
@@ -224,9 +223,6 @@ else
224ifneq (,$(findstring LJ_TARGET_PPC ,$(TARGET_TESTARCH))) 223ifneq (,$(findstring LJ_TARGET_PPC ,$(TARGET_TESTARCH)))
225 TARGET_LJARCH= ppc 224 TARGET_LJARCH= ppc
226else 225else
227ifneq (,$(findstring LJ_TARGET_PPCSPE ,$(TARGET_TESTARCH)))
228 TARGET_LJARCH= ppcspe
229else
230ifneq (,$(findstring LJ_TARGET_MIPS ,$(TARGET_TESTARCH))) 226ifneq (,$(findstring LJ_TARGET_MIPS ,$(TARGET_TESTARCH)))
231 ifneq (,$(findstring MIPSEL ,$(TARGET_TESTARCH))) 227 ifneq (,$(findstring MIPSEL ,$(TARGET_TESTARCH)))
232 TARGET_ARCH= -D__MIPSEL__=1 228 TARGET_ARCH= -D__MIPSEL__=1
@@ -240,7 +236,6 @@ endif
240endif 236endif
241endif 237endif
242endif 238endif
243endif
244 239
245ifneq (,$(findstring LJ_TARGET_PS3 1,$(TARGET_TESTARCH))) 240ifneq (,$(findstring LJ_TARGET_PS3 1,$(TARGET_TESTARCH)))
246 TARGET_SYS= PS3 241 TARGET_SYS= PS3
diff --git a/src/host/buildvm.c b/src/host/buildvm.c
index d162086e..865236d7 100644
--- a/src/host/buildvm.c
+++ b/src/host/buildvm.c
@@ -63,8 +63,6 @@ static int collect_reloc(BuildCtx *ctx, uint8_t *addr, int idx, int type);
63#include "../dynasm/dasm_arm64.h" 63#include "../dynasm/dasm_arm64.h"
64#elif LJ_TARGET_PPC 64#elif LJ_TARGET_PPC
65#include "../dynasm/dasm_ppc.h" 65#include "../dynasm/dasm_ppc.h"
66#elif LJ_TARGET_PPCSPE
67#include "../dynasm/dasm_ppc.h"
68#elif LJ_TARGET_MIPS 66#elif LJ_TARGET_MIPS
69#include "../dynasm/dasm_mips.h" 67#include "../dynasm/dasm_mips.h"
70#else 68#else
@@ -115,7 +113,7 @@ static const char *sym_decorate(BuildCtx *ctx,
115 name[0] = '@'; 113 name[0] = '@';
116 else 114 else
117 *p = '\0'; 115 *p = '\0';
118#elif (LJ_TARGET_PPC || LJ_TARGET_PPCSPE) && !LJ_TARGET_CONSOLE 116#elif LJ_TARGET_PPC && !LJ_TARGET_CONSOLE
119 /* Keep @plt. */ 117 /* Keep @plt. */
120#else 118#else
121 *p = '\0'; 119 *p = '\0';
diff --git a/src/host/buildvm_asm.c b/src/host/buildvm_asm.c
index 7bea4d17..d1f5224f 100644
--- a/src/host/buildvm_asm.c
+++ b/src/host/buildvm_asm.c
@@ -130,7 +130,7 @@ static void emit_asm_wordreloc(BuildCtx *ctx, uint8_t *p, int n,
130 ins, sym); 130 ins, sym);
131 exit(1); 131 exit(1);
132 } 132 }
133#elif LJ_TARGET_PPC || LJ_TARGET_PPCSPE 133#elif LJ_TARGET_PPC
134#if LJ_TARGET_PS3 134#if LJ_TARGET_PS3
135#define TOCPREFIX "." 135#define TOCPREFIX "."
136#else 136#else
@@ -313,10 +313,7 @@ void emit_asm(BuildCtx *ctx)
313#if !(LJ_TARGET_PS3 || LJ_TARGET_PSVITA) 313#if !(LJ_TARGET_PS3 || LJ_TARGET_PSVITA)
314 fprintf(ctx->fp, "\t.section .note.GNU-stack,\"\"," ELFASM_PX "progbits\n"); 314 fprintf(ctx->fp, "\t.section .note.GNU-stack,\"\"," ELFASM_PX "progbits\n");
315#endif 315#endif
316#if LJ_TARGET_PPCSPE 316#if LJ_TARGET_PPC && !LJ_TARGET_PS3
317 /* Soft-float ABI + SPE. */
318 fprintf(ctx->fp, "\t.gnu_attribute 4, 2\n\t.gnu_attribute 8, 3\n");
319#elif LJ_TARGET_PPC && !LJ_TARGET_PS3
320 /* Hard-float ABI. */ 317 /* Hard-float ABI. */
321 fprintf(ctx->fp, "\t.gnu_attribute 4, 1\n"); 318 fprintf(ctx->fp, "\t.gnu_attribute 4, 1\n");
322#endif 319#endif
diff --git a/src/jit/bcsave.lua b/src/jit/bcsave.lua
index 8303263d..05bc78d6 100644
--- a/src/jit/bcsave.lua
+++ b/src/jit/bcsave.lua
@@ -63,7 +63,7 @@ local map_type = {
63} 63}
64 64
65local map_arch = { 65local map_arch = {
66 x86 = true, x64 = true, arm = true, ppc = true, ppcspe = true, 66 x86 = true, x64 = true, arm = true, ppc = true,
67 mips = true, mipsel = true, 67 mips = true, mipsel = true,
68} 68}
69 69
@@ -202,7 +202,7 @@ typedef struct {
202 local is64, isbe = false, false 202 local is64, isbe = false, false
203 if ctx.arch == "x64" then 203 if ctx.arch == "x64" then
204 is64 = true 204 is64 = true
205 elseif ctx.arch == "ppc" or ctx.arch == "ppcspe" or ctx.arch == "mips" then 205 elseif ctx.arch == "ppc" or ctx.arch == "mips" then
206 isbe = true 206 isbe = true
207 end 207 end
208 208
@@ -237,7 +237,7 @@ typedef struct {
237 hdr.eendian = isbe and 2 or 1 237 hdr.eendian = isbe and 2 or 1
238 hdr.eversion = 1 238 hdr.eversion = 1
239 hdr.type = f16(1) 239 hdr.type = f16(1)
240 hdr.machine = f16(({ x86=3, x64=62, arm=40, ppc=20, ppcspe=20, mips=8, mipsel=8 })[ctx.arch]) 240 hdr.machine = f16(({ x86=3, x64=62, arm=40, ppc=20, mips=8, mipsel=8 })[ctx.arch])
241 if ctx.arch == "mips" or ctx.arch == "mipsel" then 241 if ctx.arch == "mips" or ctx.arch == "mipsel" then
242 hdr.flags = 0x50001006 242 hdr.flags = 0x50001006
243 end 243 end
diff --git a/src/lib_jit.c b/src/lib_jit.c
index 7a810ae0..4c9e5119 100644
--- a/src/lib_jit.c
+++ b/src/lib_jit.c
@@ -706,8 +706,6 @@ static uint32_t jit_cpudetect(lua_State *L)
706 flags |= JIT_F_ROUND; 706 flags |= JIT_F_ROUND;
707#endif 707#endif
708#endif 708#endif
709#elif LJ_TARGET_PPCSPE
710 /* Nothing to do. */
711#elif LJ_TARGET_MIPS 709#elif LJ_TARGET_MIPS
712#if LJ_HASJIT 710#if LJ_HASJIT
713 /* Compile-time MIPS CPU detection. */ 711 /* Compile-time MIPS CPU detection. */
diff --git a/src/lj_arch.h b/src/lj_arch.h
index 0d69d528..202fc5cb 100644
--- a/src/lj_arch.h
+++ b/src/lj_arch.h
@@ -22,11 +22,9 @@
22#define LUAJIT_ARCH_ARM64 4 22#define LUAJIT_ARCH_ARM64 4
23#define LUAJIT_ARCH_arm64 4 23#define LUAJIT_ARCH_arm64 4
24#define LUAJIT_ARCH_PPC 5 24#define LUAJIT_ARCH_PPC 5
25#define LUAJIT_ARCH_ppc 6 25#define LUAJIT_ARCH_ppc 5
26#define LUAJIT_ARCH_PPCSPE 6 26#define LUAJIT_ARCH_MIPS 6
27#define LUAJIT_ARCH_ppcspe 6 27#define LUAJIT_ARCH_mips 6
28#define LUAJIT_ARCH_MIPS 7
29#define LUAJIT_ARCH_mips 7
30 28
31/* Target OS. */ 29/* Target OS. */
32#define LUAJIT_OS_OTHER 0 30#define LUAJIT_OS_OTHER 0
@@ -48,11 +46,7 @@
48#elif defined(__aarch64__) 46#elif defined(__aarch64__)
49#define LUAJIT_TARGET LUAJIT_ARCH_ARM64 47#define LUAJIT_TARGET LUAJIT_ARCH_ARM64
50#elif defined(__ppc__) || defined(__ppc) || defined(__PPC__) || defined(__PPC) || defined(__powerpc__) || defined(__powerpc) || defined(__POWERPC__) || defined(__POWERPC) || defined(_M_PPC) 48#elif defined(__ppc__) || defined(__ppc) || defined(__PPC__) || defined(__PPC) || defined(__powerpc__) || defined(__powerpc) || defined(__POWERPC__) || defined(__POWERPC) || defined(_M_PPC)
51#ifdef __NO_FPRS__
52#define LUAJIT_TARGET LUAJIT_ARCH_PPCSPE
53#else
54#define LUAJIT_TARGET LUAJIT_ARCH_PPC 49#define LUAJIT_TARGET LUAJIT_ARCH_PPC
55#endif
56#elif defined(__mips__) || defined(__mips) || defined(__MIPS__) || defined(__MIPS) 50#elif defined(__mips__) || defined(__mips) || defined(__MIPS__) || defined(__MIPS)
57#define LUAJIT_TARGET LUAJIT_ARCH_MIPS 51#define LUAJIT_TARGET LUAJIT_ARCH_MIPS
58#else 52#else
@@ -260,26 +254,6 @@
260#define LJ_ARCH_XENON 1 254#define LJ_ARCH_XENON 1
261#endif 255#endif
262 256
263#elif LUAJIT_TARGET == LUAJIT_ARCH_PPCSPE
264
265#error "The PPC/e500 port is broken and will be abandoned with LuaJIT 2.1"
266#define LJ_ARCH_NAME "ppcspe"
267#define LJ_ARCH_BITS 32
268#define LJ_ARCH_ENDIAN LUAJIT_BE
269#ifndef LJ_ABI_SOFTFP
270#define LJ_ABI_SOFTFP 1
271#endif
272#define LJ_ABI_EABI 1
273#define LJ_TARGET_PPCSPE 1
274#define LJ_TARGET_EHRETREG 3
275#define LJ_TARGET_JUMPRANGE 25 /* +-2^25 = +-32MB */
276#define LJ_TARGET_MASKSHIFT 0
277#define LJ_TARGET_MASKROT 1
278#define LJ_TARGET_UNIFYROT 1 /* Want only IR_BROL. */
279#define LJ_ARCH_NUMMODE LJ_NUMMODE_SINGLE
280#define LJ_ARCH_NOFFI 1 /* NYI: comparisons, calls. */
281#define LJ_ARCH_NOJIT 1
282
283#elif LUAJIT_TARGET == LUAJIT_ARCH_MIPS 257#elif LUAJIT_TARGET == LUAJIT_ARCH_MIPS
284 258
285#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) 259#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL)
@@ -356,7 +330,7 @@
356#if defined(_ILP32) 330#if defined(_ILP32)
357#error "No support for ILP32 model on ARM64" 331#error "No support for ILP32 model on ARM64"
358#endif 332#endif
359#elif LJ_TARGET_PPC || LJ_TARGET_PPCSPE 333#elif LJ_TARGET_PPC
360#if defined(_SOFT_FLOAT) || defined(_SOFT_DOUBLE) 334#if defined(_SOFT_FLOAT) || defined(_SOFT_DOUBLE)
361#error "No support for PowerPC CPUs without double-precision FPU" 335#error "No support for PowerPC CPUs without double-precision FPU"
362#endif 336#endif
@@ -366,6 +340,9 @@
366#if defined(_LP64) 340#if defined(_LP64)
367#error "No support for PowerPC 64 bit mode" 341#error "No support for PowerPC 64 bit mode"
368#endif 342#endif
343#ifdef __NO_FPRS__
344#error "No support for PPC/e500 anymore (use LuaJIT 2.0)"
345#endif
369#elif LJ_TARGET_MIPS 346#elif LJ_TARGET_MIPS
370#if defined(__mips_soft_float) 347#if defined(__mips_soft_float)
371#error "No support for MIPS CPUs without FPU" 348#error "No support for MIPS CPUs without FPU"
diff --git a/src/lj_ccall.c b/src/lj_ccall.c
index 1011a1ad..fd9683dc 100644
--- a/src/lj_ccall.c
+++ b/src/lj_ccall.c
@@ -338,42 +338,6 @@
338 if (ctype_isfp(ctr->info) && ctr->size == sizeof(float)) \ 338 if (ctype_isfp(ctr->info) && ctr->size == sizeof(float)) \
339 ctr = ctype_get(cts, CTID_DOUBLE); /* FPRs always hold doubles. */ 339 ctr = ctype_get(cts, CTID_DOUBLE); /* FPRs always hold doubles. */
340 340
341#elif LJ_TARGET_PPCSPE
342/* -- PPC/SPE calling conventions ----------------------------------------- */
343
344#define CCALL_HANDLE_STRUCTRET \
345 cc->retref = 1; /* Return all structs by reference. */ \
346 cc->gpr[ngpr++] = (GPRArg)dp;
347
348#define CCALL_HANDLE_COMPLEXRET \
349 /* Complex values are returned in 2 or 4 GPRs. */ \
350 cc->retref = 0;
351
352#define CCALL_HANDLE_COMPLEXRET2 \
353 memcpy(dp, sp, ctr->size); /* Copy complex from GPRs. */
354
355#define CCALL_HANDLE_STRUCTARG \
356 rp = cdataptr(lj_cdata_new(cts, did, sz)); \
357 sz = CTSIZE_PTR; /* Pass all structs by reference. */
358
359#define CCALL_HANDLE_COMPLEXARG \
360 /* Pass complex by value in 2 or 4 GPRs. */
361
362/* PPC/SPE has a softfp ABI. */
363#define CCALL_HANDLE_REGARG \
364 if (n > 1) { /* Doesn't fit in a single GPR? */ \
365 lua_assert(n == 2 || n == 4); /* int64_t, double or complex (float). */ \
366 if (n == 2) \
367 ngpr = (ngpr + 1u) & ~1u; /* Only align 64 bit value to regpair. */ \
368 else if (ngpr + n > maxgpr) \
369 ngpr = maxgpr; /* Prevent reordering. */ \
370 } \
371 if (ngpr + n <= maxgpr) { \
372 dp = &cc->gpr[ngpr]; \
373 ngpr += n; \
374 goto done; \
375 }
376
377#elif LJ_TARGET_MIPS 341#elif LJ_TARGET_MIPS
378/* -- MIPS calling conventions -------------------------------------------- */ 342/* -- MIPS calling conventions -------------------------------------------- */
379 343
diff --git a/src/lj_ccall.h b/src/lj_ccall.h
index c3ea9e6f..81d2bb39 100644
--- a/src/lj_ccall.h
+++ b/src/lj_ccall.h
@@ -80,16 +80,6 @@ typedef union FPRArg {
80typedef intptr_t GPRArg; 80typedef intptr_t GPRArg;
81typedef double FPRArg; 81typedef double FPRArg;
82 82
83#elif LJ_TARGET_PPCSPE
84
85#define CCALL_NARG_GPR 8
86#define CCALL_NARG_FPR 0
87#define CCALL_NRET_GPR 4 /* For softfp complex double. */
88#define CCALL_NRET_FPR 0
89#define CCALL_SPS_FREE 0 /* NYI */
90
91typedef intptr_t GPRArg;
92
93#elif LJ_TARGET_MIPS 83#elif LJ_TARGET_MIPS
94 84
95#define CCALL_NARG_GPR 4 85#define CCALL_NARG_GPR 4
diff --git a/src/lj_ctype.h b/src/lj_ctype.h
index ff8ee069..c9919d4a 100644
--- a/src/lj_ctype.h
+++ b/src/lj_ctype.h
@@ -263,7 +263,7 @@ typedef struct CTState {
263/* -- Predefined types ---------------------------------------------------- */ 263/* -- Predefined types ---------------------------------------------------- */
264 264
265/* Target-dependent types. */ 265/* Target-dependent types. */
266#if LJ_TARGET_PPC || LJ_TARGET_PPCSPE 266#if LJ_TARGET_PPC
267#define CTTYDEFP(_) \ 267#define CTTYDEFP(_) \
268 _(LINT32, 4, CT_NUM, CTF_LONG|CTALIGN(2)) 268 _(LINT32, 4, CT_NUM, CTF_LONG|CTALIGN(2))
269#else 269#else
diff --git a/src/lj_frame.h b/src/lj_frame.h
index f024c43f..4a412fe1 100644
--- a/src/lj_frame.h
+++ b/src/lj_frame.h
@@ -197,15 +197,6 @@ enum { LJ_CONT_TAILCALL, LJ_CONT_FFI_CALLBACK }; /* Special continuations. */
197#define CFRAME_SIZE 272 197#define CFRAME_SIZE 272
198#define CFRAME_SHIFT_MULTRES 3 198#define CFRAME_SHIFT_MULTRES 3
199#endif 199#endif
200#elif LJ_TARGET_PPCSPE
201#define CFRAME_OFS_ERRF 28
202#define CFRAME_OFS_NRES 24
203#define CFRAME_OFS_PREV 20
204#define CFRAME_OFS_L 16
205#define CFRAME_OFS_PC 12
206#define CFRAME_OFS_MULTRES 8
207#define CFRAME_SIZE 184
208#define CFRAME_SHIFT_MULTRES 3
209#elif LJ_TARGET_MIPS 200#elif LJ_TARGET_MIPS
210#define CFRAME_OFS_ERRF 124 201#define CFRAME_OFS_ERRF 124
211#define CFRAME_OFS_NRES 120 202#define CFRAME_OFS_NRES 120
diff --git a/src/vm_ppcspe.dasc b/src/vm_ppcspe.dasc
deleted file mode 100644
index ee394031..00000000
--- a/src/vm_ppcspe.dasc
+++ /dev/null
@@ -1,3685 +0,0 @@
1|// Low-level VM code for PowerPC/e500 CPUs.
2|// Bytecode interpreter, fast functions and helper functions.
3|// Copyright (C) 2005-2014 Mike Pall. See Copyright Notice in luajit.h
4|
5|.arch ppc
6|.section code_op, code_sub
7|
8|.actionlist build_actionlist
9|.globals GLOB_
10|.globalnames globnames
11|.externnames extnames
12|
13|// Note: The ragged indentation of the instructions is intentional.
14|// The starting columns indicate data dependencies.
15|
16|//-----------------------------------------------------------------------
17|
18|// Fixed register assignments for the interpreter.
19|// Don't use: r1 = sp, r2 and r13 = reserved and/or small data area ptr
20|
21|// The following must be C callee-save (but BASE is often refetched).
22|.define BASE, r14 // Base of current Lua stack frame.
23|.define KBASE, r15 // Constants of current Lua function.
24|.define PC, r16 // Next PC.
25|.define DISPATCH, r17 // Opcode dispatch table.
26|.define LREG, r18 // Register holding lua_State (also in SAVE_L).
27|.define MULTRES, r19 // Size of multi-result: (nresults+1)*8.
28|
29|// Constants for vectorized type-comparisons (hi+low GPR). C callee-save.
30|.define TISNUM, r22
31|.define TISSTR, r23
32|.define TISTAB, r24
33|.define TISFUNC, r25
34|.define TISNIL, r26
35|.define TOBIT, r27
36|.define ZERO, TOBIT // Zero in lo word.
37|
38|// The following temporaries are not saved across C calls, except for RA.
39|.define RA, r20 // Callee-save.
40|.define RB, r10
41|.define RC, r11
42|.define RD, r12
43|.define INS, r7 // Overlaps CARG5.
44|
45|.define TMP0, r0
46|.define TMP1, r8
47|.define TMP2, r9
48|.define TMP3, r6 // Overlaps CARG4.
49|
50|// Saved temporaries.
51|.define SAVE0, r21
52|
53|// Calling conventions.
54|.define CARG1, r3
55|.define CARG2, r4
56|.define CARG3, r5
57|.define CARG4, r6 // Overlaps TMP3.
58|.define CARG5, r7 // Overlaps INS.
59|
60|.define CRET1, r3
61|.define CRET2, r4
62|
63|// Stack layout while in interpreter. Must match with lj_frame.h.
64|.define SAVE_LR, 188(sp)
65|.define CFRAME_SPACE, 184 // Delta for sp.
66|// Back chain for sp: 184(sp) <-- sp entering interpreter
67|.define SAVE_r31, 176(sp) // 64 bit register saves.
68|.define SAVE_r30, 168(sp)
69|.define SAVE_r29, 160(sp)
70|.define SAVE_r28, 152(sp)
71|.define SAVE_r27, 144(sp)
72|.define SAVE_r26, 136(sp)
73|.define SAVE_r25, 128(sp)
74|.define SAVE_r24, 120(sp)
75|.define SAVE_r23, 112(sp)
76|.define SAVE_r22, 104(sp)
77|.define SAVE_r21, 96(sp)
78|.define SAVE_r20, 88(sp)
79|.define SAVE_r19, 80(sp)
80|.define SAVE_r18, 72(sp)
81|.define SAVE_r17, 64(sp)
82|.define SAVE_r16, 56(sp)
83|.define SAVE_r15, 48(sp)
84|.define SAVE_r14, 40(sp)
85|.define SAVE_CR, 36(sp)
86|.define UNUSED1, 32(sp)
87|.define SAVE_ERRF, 28(sp) // 32 bit C frame info.
88|.define SAVE_NRES, 24(sp)
89|.define SAVE_CFRAME, 20(sp)
90|.define SAVE_L, 16(sp)
91|.define SAVE_PC, 12(sp)
92|.define SAVE_MULTRES, 8(sp)
93|// Next frame lr: 4(sp)
94|// Back chain for sp: 0(sp) <-- sp while in interpreter
95|
96|.macro save_, reg; evstdd reg, SAVE_..reg; .endmacro
97|.macro rest_, reg; evldd reg, SAVE_..reg; .endmacro
98|
99|.macro saveregs
100| stwu sp, -CFRAME_SPACE(sp)
101| save_ r14; save_ r15; save_ r16; save_ r17; save_ r18; save_ r19
102| mflr r0; mfcr r12
103| save_ r20; save_ r21; save_ r22; save_ r23; save_ r24; save_ r25
104| stw r0, SAVE_LR; stw r12, SAVE_CR
105| save_ r26; save_ r27; save_ r28; save_ r29; save_ r30; save_ r31
106|.endmacro
107|
108|.macro restoreregs
109| lwz r0, SAVE_LR; lwz r12, SAVE_CR
110| rest_ r14; rest_ r15; rest_ r16; rest_ r17; rest_ r18; rest_ r19
111| mtlr r0; mtcrf 0x38, r12
112| rest_ r20; rest_ r21; rest_ r22; rest_ r23; rest_ r24; rest_ r25
113| rest_ r26; rest_ r27; rest_ r28; rest_ r29; rest_ r30; rest_ r31
114| addi sp, sp, CFRAME_SPACE
115|.endmacro
116|
117|// Type definitions. Some of these are only used for documentation.
118|.type L, lua_State, LREG
119|.type GL, global_State
120|.type TVALUE, TValue
121|.type GCOBJ, GCobj
122|.type STR, GCstr
123|.type TAB, GCtab
124|.type LFUNC, GCfuncL
125|.type CFUNC, GCfuncC
126|.type PROTO, GCproto
127|.type UPVAL, GCupval
128|.type NODE, Node
129|.type NARGS8, int
130|.type TRACE, GCtrace
131|
132|//-----------------------------------------------------------------------
133|
134|// These basic macros should really be part of DynASM.
135|.macro srwi, rx, ry, n; rlwinm rx, ry, 32-n, n, 31; .endmacro
136|.macro slwi, rx, ry, n; rlwinm rx, ry, n, 0, 31-n; .endmacro
137|.macro rotlwi, rx, ry, n; rlwinm rx, ry, n, 0, 31; .endmacro
138|.macro rotlw, rx, ry, rn; rlwnm rx, ry, rn, 0, 31; .endmacro
139|.macro subi, rx, ry, i; addi rx, ry, -i; .endmacro
140|
141|// Trap for not-yet-implemented parts.
142|.macro NYI; tw 4, sp, sp; .endmacro
143|
144|//-----------------------------------------------------------------------
145|
146|// Access to frame relative to BASE.
147|.define FRAME_PC, -8
148|.define FRAME_FUNC, -4
149|
150|// Instruction decode.
151|.macro decode_OP4, dst, ins; rlwinm dst, ins, 2, 22, 29; .endmacro
152|.macro decode_RA8, dst, ins; rlwinm dst, ins, 27, 21, 28; .endmacro
153|.macro decode_RB8, dst, ins; rlwinm dst, ins, 11, 21, 28; .endmacro
154|.macro decode_RC8, dst, ins; rlwinm dst, ins, 19, 21, 28; .endmacro
155|.macro decode_RD8, dst, ins; rlwinm dst, ins, 19, 13, 28; .endmacro
156|
157|.macro decode_OP1, dst, ins; rlwinm dst, ins, 0, 24, 31; .endmacro
158|.macro decode_RD4, dst, ins; rlwinm dst, ins, 18, 14, 29; .endmacro
159|
160|// Instruction fetch.
161|.macro ins_NEXT1
162| lwz INS, 0(PC)
163| addi PC, PC, 4
164|.endmacro
165|// Instruction decode+dispatch.
166|.macro ins_NEXT2
167| decode_OP4 TMP1, INS
168| decode_RB8 RB, INS
169| decode_RD8 RD, INS
170| lwzx TMP0, DISPATCH, TMP1
171| decode_RA8 RA, INS
172| decode_RC8 RC, INS
173| mtctr TMP0
174| bctr
175|.endmacro
176|.macro ins_NEXT
177| ins_NEXT1
178| ins_NEXT2
179|.endmacro
180|
181|// Instruction footer.
182|.if 1
183| // Replicated dispatch. Less unpredictable branches, but higher I-Cache use.
184| .define ins_next, ins_NEXT
185| .define ins_next_, ins_NEXT
186| .define ins_next1, ins_NEXT1
187| .define ins_next2, ins_NEXT2
188|.else
189| // Common dispatch. Lower I-Cache use, only one (very) unpredictable branch.
190| // Affects only certain kinds of benchmarks (and only with -j off).
191| .macro ins_next
192| b ->ins_next
193| .endmacro
194| .macro ins_next1
195| .endmacro
196| .macro ins_next2
197| b ->ins_next
198| .endmacro
199| .macro ins_next_
200| ->ins_next:
201| ins_NEXT
202| .endmacro
203|.endif
204|
205|// Call decode and dispatch.
206|.macro ins_callt
207| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
208| lwz PC, LFUNC:RB->pc
209| lwz INS, 0(PC)
210| addi PC, PC, 4
211| decode_OP4 TMP1, INS
212| decode_RA8 RA, INS
213| lwzx TMP0, DISPATCH, TMP1
214| add RA, RA, BASE
215| mtctr TMP0
216| bctr
217|.endmacro
218|
219|.macro ins_call
220| // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, PC = caller PC
221| stw PC, FRAME_PC(BASE)
222| ins_callt
223|.endmacro
224|
225|//-----------------------------------------------------------------------
226|
227|// Macros to test operand types.
228|.macro checknum, reg; evcmpltu reg, TISNUM; .endmacro
229|.macro checkstr, reg; evcmpeq reg, TISSTR; .endmacro
230|.macro checktab, reg; evcmpeq reg, TISTAB; .endmacro
231|.macro checkfunc, reg; evcmpeq reg, TISFUNC; .endmacro
232|.macro checknil, reg; evcmpeq reg, TISNIL; .endmacro
233|.macro checkok, label; blt label; .endmacro
234|.macro checkfail, label; bge label; .endmacro
235|.macro checkanyfail, label; bns label; .endmacro
236|.macro checkallok, label; bso label; .endmacro
237|
238|.macro branch_RD
239| srwi TMP0, RD, 1
240| add PC, PC, TMP0
241| addis PC, PC, -(BCBIAS_J*4 >> 16)
242|.endmacro
243|
244|// Assumes DISPATCH is relative to GL.
245#define DISPATCH_GL(field) (GG_DISP2G + (int)offsetof(global_State, field))
246#define DISPATCH_J(field) (GG_DISP2J + (int)offsetof(jit_State, field))
247|
248#define PC2PROTO(field) ((int)offsetof(GCproto, field)-(int)sizeof(GCproto))
249|
250|.macro hotloop
251| NYI
252|.endmacro
253|
254|.macro hotcall
255| NYI
256|.endmacro
257|
258|// Set current VM state. Uses TMP0.
259|.macro li_vmstate, st; li TMP0, ~LJ_VMST_..st; .endmacro
260|.macro st_vmstate; stw TMP0, DISPATCH_GL(vmstate)(DISPATCH); .endmacro
261|
262|// Move table write barrier back. Overwrites mark and tmp.
263|.macro barrierback, tab, mark, tmp
264| lwz tmp, DISPATCH_GL(gc.grayagain)(DISPATCH)
265| // Assumes LJ_GC_BLACK is 0x04.
266| rlwinm mark, mark, 0, 30, 28 // black2gray(tab)
267| stw tab, DISPATCH_GL(gc.grayagain)(DISPATCH)
268| stb mark, tab->marked
269| stw tmp, tab->gclist
270|.endmacro
271|
272|//-----------------------------------------------------------------------
273
274/* Generate subroutines used by opcodes and other parts of the VM. */
275/* The .code_sub section should be last to help static branch prediction. */
276static void build_subroutines(BuildCtx *ctx)
277{
278 |.code_sub
279 |
280 |//-----------------------------------------------------------------------
281 |//-- Return handling ----------------------------------------------------
282 |//-----------------------------------------------------------------------
283 |
284 |->vm_returnp:
285 | // See vm_return. Also: TMP2 = previous base.
286 | andi. TMP0, PC, FRAME_P
287 | evsplati TMP1, LJ_TTRUE
288 | beq ->cont_dispatch
289 |
290 | // Return from pcall or xpcall fast func.
291 | lwz PC, FRAME_PC(TMP2) // Fetch PC of previous frame.
292 | mr BASE, TMP2 // Restore caller base.
293 | // Prepending may overwrite the pcall frame, so do it at the end.
294 | stwu TMP1, FRAME_PC(RA) // Prepend true to results.
295 |
296 |->vm_returnc:
297 | addi RD, RD, 8 // RD = (nresults+1)*8.
298 | andi. TMP0, PC, FRAME_TYPE
299 | cmpwi cr1, RD, 0
300 | li CRET1, LUA_YIELD
301 | beq cr1, ->vm_unwind_c_eh
302 | mr MULTRES, RD
303 | beq ->BC_RET_Z // Handle regular return to Lua.
304 |
305 |->vm_return:
306 | // BASE = base, RA = resultptr, RD/MULTRES = (nresults+1)*8, PC = return
307 | // TMP0 = PC & FRAME_TYPE
308 | cmpwi TMP0, FRAME_C
309 | rlwinm TMP2, PC, 0, 0, 28
310 | li_vmstate C
311 | sub TMP2, BASE, TMP2 // TMP2 = previous base.
312 | bne ->vm_returnp
313 |
314 | addic. TMP1, RD, -8
315 | stw TMP2, L->base
316 | lwz TMP2, SAVE_NRES
317 | subi BASE, BASE, 8
318 | st_vmstate
319 | slwi TMP2, TMP2, 3
320 | beq >2
321 |1:
322 | addic. TMP1, TMP1, -8
323 | evldd TMP0, 0(RA)
324 | addi RA, RA, 8
325 | evstdd TMP0, 0(BASE)
326 | addi BASE, BASE, 8
327 | bne <1
328 |
329 |2:
330 | cmpw TMP2, RD // More/less results wanted?
331 | bne >6
332 |3:
333 | stw BASE, L->top // Store new top.
334 |
335 |->vm_leave_cp:
336 | lwz TMP0, SAVE_CFRAME // Restore previous C frame.
337 | li CRET1, 0 // Ok return status for vm_pcall.
338 | stw TMP0, L->cframe
339 |
340 |->vm_leave_unw:
341 | restoreregs
342 | blr
343 |
344 |6:
345 | ble >7 // Less results wanted?
346 | // More results wanted. Check stack size and fill up results with nil.
347 | lwz TMP1, L->maxstack
348 | cmplw BASE, TMP1
349 | bge >8
350 | evstdd TISNIL, 0(BASE)
351 | addi RD, RD, 8
352 | addi BASE, BASE, 8
353 | b <2
354 |
355 |7: // Less results wanted.
356 | sub TMP0, RD, TMP2
357 | cmpwi TMP2, 0 // LUA_MULTRET+1 case?
358 | sub TMP0, BASE, TMP0 // Subtract the difference.
359 | iseleq BASE, BASE, TMP0 // Either keep top or shrink it.
360 | b <3
361 |
362 |8: // Corner case: need to grow stack for filling up results.
363 | // This can happen if:
364 | // - A C function grows the stack (a lot).
365 | // - The GC shrinks the stack in between.
366 | // - A return back from a lua_call() with (high) nresults adjustment.
367 | stw BASE, L->top // Save current top held in BASE (yes).
368 | mr SAVE0, RD
369 | mr CARG2, TMP2
370 | mr CARG1, L
371 | bl extern lj_state_growstack // (lua_State *L, int n)
372 | lwz TMP2, SAVE_NRES
373 | mr RD, SAVE0
374 | slwi TMP2, TMP2, 3
375 | lwz BASE, L->top // Need the (realloced) L->top in BASE.
376 | b <2
377 |
378 |->vm_unwind_c: // Unwind C stack, return from vm_pcall.
379 | // (void *cframe, int errcode)
380 | mr sp, CARG1
381 | mr CRET1, CARG2
382 |->vm_unwind_c_eh: // Landing pad for external unwinder.
383 | lwz L, SAVE_L
384 | li TMP0, ~LJ_VMST_C
385 | lwz GL:TMP1, L->glref
386 | stw TMP0, GL:TMP1->vmstate
387 | b ->vm_leave_unw
388 |
389 |->vm_unwind_ff: // Unwind C stack, return from ff pcall.
390 | // (void *cframe)
391 | rlwinm sp, CARG1, 0, 0, 29
392 |->vm_unwind_ff_eh: // Landing pad for external unwinder.
393 | lwz L, SAVE_L
394 | evsplati TISNUM, LJ_TISNUM+1 // Setup type comparison constants.
395 | evsplati TISFUNC, LJ_TFUNC
396 | lus TOBIT, 0x4338
397 | evsplati TISTAB, LJ_TTAB
398 | li TMP0, 0
399 | lwz BASE, L->base
400 | evmergelo TOBIT, TOBIT, TMP0
401 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
402 | evsplati TISSTR, LJ_TSTR
403 | li TMP1, LJ_TFALSE
404 | evsplati TISNIL, LJ_TNIL
405 | li_vmstate INTERP
406 | lwz PC, FRAME_PC(BASE) // Fetch PC of previous frame.
407 | la RA, -8(BASE) // Results start at BASE-8.
408 | addi DISPATCH, DISPATCH, GG_G2DISP
409 | stw TMP1, 0(RA) // Prepend false to error message.
410 | li RD, 16 // 2 results: false + error message.
411 | st_vmstate
412 | b ->vm_returnc
413 |
414 |//-----------------------------------------------------------------------
415 |//-- Grow stack for calls -----------------------------------------------
416 |//-----------------------------------------------------------------------
417 |
418 |->vm_growstack_c: // Grow stack for C function.
419 | li CARG2, LUA_MINSTACK
420 | b >2
421 |
422 |->vm_growstack_l: // Grow stack for Lua function.
423 | // BASE = new base, RA = BASE+framesize*8, RC = nargs*8, PC = first PC
424 | add RC, BASE, RC
425 | sub RA, RA, BASE
426 | stw BASE, L->base
427 | addi PC, PC, 4 // Must point after first instruction.
428 | stw RC, L->top
429 | srwi CARG2, RA, 3
430 |2:
431 | // L->base = new base, L->top = top
432 | stw PC, SAVE_PC
433 | mr CARG1, L
434 | bl extern lj_state_growstack // (lua_State *L, int n)
435 | lwz BASE, L->base
436 | lwz RC, L->top
437 | lwz LFUNC:RB, FRAME_FUNC(BASE)
438 | sub RC, RC, BASE
439 | // BASE = new base, RB = LFUNC/CFUNC, RC = nargs*8, FRAME_PC(BASE) = PC
440 | ins_callt // Just retry the call.
441 |
442 |//-----------------------------------------------------------------------
443 |//-- Entry points into the assembler VM ---------------------------------
444 |//-----------------------------------------------------------------------
445 |
446 |->vm_resume: // Setup C frame and resume thread.
447 | // (lua_State *L, TValue *base, int nres1 = 0, ptrdiff_t ef = 0)
448 | saveregs
449 | mr L, CARG1
450 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
451 | mr BASE, CARG2
452 | lbz TMP1, L->status
453 | stw L, SAVE_L
454 | li PC, FRAME_CP
455 | addi TMP0, sp, CFRAME_RESUME
456 | addi DISPATCH, DISPATCH, GG_G2DISP
457 | stw CARG3, SAVE_NRES
458 | cmplwi TMP1, 0
459 | stw CARG3, SAVE_ERRF
460 | stw TMP0, L->cframe
461 | stw CARG3, SAVE_CFRAME
462 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
463 | beq >3
464 |
465 | // Resume after yield (like a return).
466 | mr RA, BASE
467 | lwz BASE, L->base
468 | evsplati TISNUM, LJ_TISNUM+1 // Setup type comparison constants.
469 | lwz TMP1, L->top
470 | evsplati TISFUNC, LJ_TFUNC
471 | lus TOBIT, 0x4338
472 | evsplati TISTAB, LJ_TTAB
473 | lwz PC, FRAME_PC(BASE)
474 | li TMP2, 0
475 | evsplati TISSTR, LJ_TSTR
476 | sub RD, TMP1, BASE
477 | evmergelo TOBIT, TOBIT, TMP2
478 | stb CARG3, L->status
479 | andi. TMP0, PC, FRAME_TYPE
480 | li_vmstate INTERP
481 | addi RD, RD, 8
482 | evsplati TISNIL, LJ_TNIL
483 | mr MULTRES, RD
484 | st_vmstate
485 | beq ->BC_RET_Z
486 | b ->vm_return
487 |
488 |->vm_pcall: // Setup protected C frame and enter VM.
489 | // (lua_State *L, TValue *base, int nres1, ptrdiff_t ef)
490 | saveregs
491 | li PC, FRAME_CP
492 | stw CARG4, SAVE_ERRF
493 | b >1
494 |
495 |->vm_call: // Setup C frame and enter VM.
496 | // (lua_State *L, TValue *base, int nres1)
497 | saveregs
498 | li PC, FRAME_C
499 |
500 |1: // Entry point for vm_pcall above (PC = ftype).
501 | lwz TMP1, L:CARG1->cframe
502 | stw CARG3, SAVE_NRES
503 | mr L, CARG1
504 | stw CARG1, SAVE_L
505 | mr BASE, CARG2
506 | stw sp, L->cframe // Add our C frame to cframe chain.
507 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
508 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
509 | stw TMP1, SAVE_CFRAME
510 | addi DISPATCH, DISPATCH, GG_G2DISP
511 |
512 |3: // Entry point for vm_cpcall/vm_resume (BASE = base, PC = ftype).
513 | lwz TMP2, L->base // TMP2 = old base (used in vmeta_call).
514 | evsplati TISNUM, LJ_TISNUM+1 // Setup type comparison constants.
515 | lwz TMP1, L->top
516 | evsplati TISFUNC, LJ_TFUNC
517 | add PC, PC, BASE
518 | evsplati TISTAB, LJ_TTAB
519 | lus TOBIT, 0x4338
520 | li TMP0, 0
521 | sub PC, PC, TMP2 // PC = frame delta + frame type
522 | evsplati TISSTR, LJ_TSTR
523 | sub NARGS8:RC, TMP1, BASE
524 | evmergelo TOBIT, TOBIT, TMP0
525 | li_vmstate INTERP
526 | evsplati TISNIL, LJ_TNIL
527 | st_vmstate
528 |
529 |->vm_call_dispatch:
530 | // TMP2 = old base, BASE = new base, RC = nargs*8, PC = caller PC
531 | li TMP0, -8
532 | evlddx LFUNC:RB, BASE, TMP0
533 | checkfunc LFUNC:RB
534 | checkfail ->vmeta_call
535 |
536 |->vm_call_dispatch_f:
537 | ins_call
538 | // BASE = new base, RB = func, RC = nargs*8, PC = caller PC
539 |
540 |->vm_cpcall: // Setup protected C frame, call C.
541 | // (lua_State *L, lua_CFunction func, void *ud, lua_CPFunction cp)
542 | saveregs
543 | mr L, CARG1
544 | lwz TMP0, L:CARG1->stack
545 | stw CARG1, SAVE_L
546 | lwz TMP1, L->top
547 | stw CARG1, SAVE_PC // Any value outside of bytecode is ok.
548 | sub TMP0, TMP0, TMP1 // Compute -savestack(L, L->top).
549 | lwz TMP1, L->cframe
550 | stw sp, L->cframe // Add our C frame to cframe chain.
551 | li TMP2, 0
552 | stw TMP0, SAVE_NRES // Neg. delta means cframe w/o frame.
553 | stw TMP2, SAVE_ERRF // No error function.
554 | stw TMP1, SAVE_CFRAME
555 | mtctr CARG4
556 | bctrl // (lua_State *L, lua_CFunction func, void *ud)
557 | mr. BASE, CRET1
558 | lwz DISPATCH, L->glref // Setup pointer to dispatch table.
559 | li PC, FRAME_CP
560 | addi DISPATCH, DISPATCH, GG_G2DISP
561 | bne <3 // Else continue with the call.
562 | b ->vm_leave_cp // No base? Just remove C frame.
563 |
564 |//-----------------------------------------------------------------------
565 |//-- Metamethod handling ------------------------------------------------
566 |//-----------------------------------------------------------------------
567 |
568 |// The lj_meta_* functions (except for lj_meta_cat) don't reallocate the
569 |// stack, so BASE doesn't need to be reloaded across these calls.
570 |
571 |//-- Continuation dispatch ----------------------------------------------
572 |
573 |->cont_dispatch:
574 | // BASE = meta base, RA = resultptr, RD = (nresults+1)*8
575 | lwz TMP0, -12(BASE) // Continuation.
576 | mr RB, BASE
577 | mr BASE, TMP2 // Restore caller BASE.
578 | lwz LFUNC:TMP1, FRAME_FUNC(TMP2)
579 | cmplwi TMP0, 0
580 | lwz PC, -16(RB) // Restore PC from [cont|PC].
581 | beq >1
582 | subi TMP2, RD, 8
583 | lwz TMP1, LFUNC:TMP1->pc
584 | evstddx TISNIL, RA, TMP2 // Ensure one valid arg.
585 | lwz KBASE, PC2PROTO(k)(TMP1)
586 | // BASE = base, RA = resultptr, RB = meta base
587 | mtctr TMP0
588 | bctr // Jump to continuation.
589 |
590 |1: // Tail call from C function.
591 | subi TMP1, RB, 16
592 | sub RC, TMP1, BASE
593 | b ->vm_call_tail
594 |
595 |->cont_cat: // RA = resultptr, RB = meta base
596 | lwz INS, -4(PC)
597 | subi CARG2, RB, 16
598 | decode_RB8 SAVE0, INS
599 | evldd TMP0, 0(RA)
600 | add TMP1, BASE, SAVE0
601 | stw BASE, L->base
602 | cmplw TMP1, CARG2
603 | sub CARG3, CARG2, TMP1
604 | decode_RA8 RA, INS
605 | evstdd TMP0, 0(CARG2)
606 | bne ->BC_CAT_Z
607 | evstddx TMP0, BASE, RA
608 | b ->cont_nop
609 |
610 |//-- Table indexing metamethods -----------------------------------------
611 |
612 |->vmeta_tgets1:
613 | evmergelo STR:RC, TISSTR, STR:RC
614 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
615 | decode_RB8 RB, INS
616 | evstdd STR:RC, 0(CARG3)
617 | add CARG2, BASE, RB
618 | b >1
619 |
620 |->vmeta_tgets:
621 | evmergelo TAB:RB, TISTAB, TAB:RB
622 | la CARG2, DISPATCH_GL(tmptv)(DISPATCH)
623 | evmergelo STR:RC, TISSTR, STR:RC
624 | evstdd TAB:RB, 0(CARG2)
625 | la CARG3, DISPATCH_GL(tmptv2)(DISPATCH)
626 | evstdd STR:RC, 0(CARG3)
627 | b >1
628 |
629 |->vmeta_tgetb: // TMP0 = index
630 | efdcfsi TMP0, TMP0
631 | decode_RB8 RB, INS
632 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
633 | add CARG2, BASE, RB
634 | evstdd TMP0, 0(CARG3)
635 | b >1
636 |
637 |->vmeta_tgetv:
638 | decode_RB8 RB, INS
639 | decode_RC8 RC, INS
640 | add CARG2, BASE, RB
641 | add CARG3, BASE, RC
642 |1:
643 | stw BASE, L->base
644 | mr CARG1, L
645 | stw PC, SAVE_PC
646 | bl extern lj_meta_tget // (lua_State *L, TValue *o, TValue *k)
647 | // Returns TValue * (finished) or NULL (metamethod).
648 | cmplwi CRET1, 0
649 | beq >3
650 | evldd TMP0, 0(CRET1)
651 | evstddx TMP0, BASE, RA
652 | ins_next
653 |
654 |3: // Call __index metamethod.
655 | // BASE = base, L->top = new base, stack = cont/func/t/k
656 | subfic TMP1, BASE, FRAME_CONT
657 | lwz BASE, L->top
658 | stw PC, -16(BASE) // [cont|PC]
659 | add PC, TMP1, BASE
660 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
661 | li NARGS8:RC, 16 // 2 args for func(t, k).
662 | b ->vm_call_dispatch_f
663 |
664 |//-----------------------------------------------------------------------
665 |
666 |->vmeta_tsets1:
667 | evmergelo STR:RC, TISSTR, STR:RC
668 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
669 | decode_RB8 RB, INS
670 | evstdd STR:RC, 0(CARG3)
671 | add CARG2, BASE, RB
672 | b >1
673 |
674 |->vmeta_tsets:
675 | evmergelo TAB:RB, TISTAB, TAB:RB
676 | la CARG2, DISPATCH_GL(tmptv)(DISPATCH)
677 | evmergelo STR:RC, TISSTR, STR:RC
678 | evstdd TAB:RB, 0(CARG2)
679 | la CARG3, DISPATCH_GL(tmptv2)(DISPATCH)
680 | evstdd STR:RC, 0(CARG3)
681 | b >1
682 |
683 |->vmeta_tsetb: // TMP0 = index
684 | efdcfsi TMP0, TMP0
685 | decode_RB8 RB, INS
686 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
687 | add CARG2, BASE, RB
688 | evstdd TMP0, 0(CARG3)
689 | b >1
690 |
691 |->vmeta_tsetv:
692 | decode_RB8 RB, INS
693 | decode_RC8 RC, INS
694 | add CARG2, BASE, RB
695 | add CARG3, BASE, RC
696 |1:
697 | stw BASE, L->base
698 | mr CARG1, L
699 | stw PC, SAVE_PC
700 | bl extern lj_meta_tset // (lua_State *L, TValue *o, TValue *k)
701 | // Returns TValue * (finished) or NULL (metamethod).
702 | cmplwi CRET1, 0
703 | evlddx TMP0, BASE, RA
704 | beq >3
705 | // NOBARRIER: lj_meta_tset ensures the table is not black.
706 | evstdd TMP0, 0(CRET1)
707 | ins_next
708 |
709 |3: // Call __newindex metamethod.
710 | // BASE = base, L->top = new base, stack = cont/func/t/k/(v)
711 | subfic TMP1, BASE, FRAME_CONT
712 | lwz BASE, L->top
713 | stw PC, -16(BASE) // [cont|PC]
714 | add PC, TMP1, BASE
715 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
716 | li NARGS8:RC, 24 // 3 args for func(t, k, v)
717 | evstdd TMP0, 16(BASE) // Copy value to third argument.
718 | b ->vm_call_dispatch_f
719 |
720 |//-- Comparison metamethods ---------------------------------------------
721 |
722 |->vmeta_comp:
723 | mr CARG1, L
724 | subi PC, PC, 4
725 | add CARG2, BASE, RA
726 | stw PC, SAVE_PC
727 | add CARG3, BASE, RD
728 | stw BASE, L->base
729 | decode_OP1 CARG4, INS
730 | bl extern lj_meta_comp // (lua_State *L, TValue *o1, *o2, int op)
731 | // Returns 0/1 or TValue * (metamethod).
732 |3:
733 | cmplwi CRET1, 1
734 | bgt ->vmeta_binop
735 |4:
736 | lwz INS, 0(PC)
737 | addi PC, PC, 4
738 | decode_RD4 TMP2, INS
739 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
740 | add TMP2, TMP2, TMP3
741 | isellt PC, PC, TMP2
742 |->cont_nop:
743 | ins_next
744 |
745 |->cont_ra: // RA = resultptr
746 | lwz INS, -4(PC)
747 | evldd TMP0, 0(RA)
748 | decode_RA8 TMP1, INS
749 | evstddx TMP0, BASE, TMP1
750 | b ->cont_nop
751 |
752 |->cont_condt: // RA = resultptr
753 | lwz TMP0, 0(RA)
754 | li TMP1, LJ_TTRUE
755 | cmplw TMP1, TMP0 // Branch if result is true.
756 | b <4
757 |
758 |->cont_condf: // RA = resultptr
759 | lwz TMP0, 0(RA)
760 | li TMP1, LJ_TFALSE
761 | cmplw TMP0, TMP1 // Branch if result is false.
762 | b <4
763 |
764 |->vmeta_equal:
765 | // CARG2, CARG3, CARG4 are already set by BC_ISEQV/BC_ISNEV.
766 | subi PC, PC, 4
767 | stw BASE, L->base
768 | mr CARG1, L
769 | stw PC, SAVE_PC
770 | bl extern lj_meta_equal // (lua_State *L, GCobj *o1, *o2, int ne)
771 | // Returns 0/1 or TValue * (metamethod).
772 | b <3
773 |
774 |//-- Arithmetic metamethods ---------------------------------------------
775 |
776 |->vmeta_arith_vn:
777 | add CARG3, BASE, RB
778 | add CARG4, KBASE, RC
779 | b >1
780 |
781 |->vmeta_arith_nv:
782 | add CARG3, KBASE, RC
783 | add CARG4, BASE, RB
784 | b >1
785 |
786 |->vmeta_unm:
787 | add CARG3, BASE, RD
788 | mr CARG4, CARG3
789 | b >1
790 |
791 |->vmeta_arith_vv:
792 | add CARG3, BASE, RB
793 | add CARG4, BASE, RC
794 |1:
795 | add CARG2, BASE, RA
796 | stw BASE, L->base
797 | mr CARG1, L
798 | stw PC, SAVE_PC
799 | decode_OP1 CARG5, INS // Caveat: CARG5 overlaps INS.
800 | bl extern lj_meta_arith // (lua_State *L, TValue *ra,*rb,*rc, BCReg op)
801 | // Returns NULL (finished) or TValue * (metamethod).
802 | cmplwi CRET1, 0
803 | beq ->cont_nop
804 |
805 | // Call metamethod for binary op.
806 |->vmeta_binop:
807 | // BASE = old base, CRET1 = new base, stack = cont/func/o1/o2
808 | sub TMP1, CRET1, BASE
809 | stw PC, -16(CRET1) // [cont|PC]
810 | mr TMP2, BASE
811 | addi PC, TMP1, FRAME_CONT
812 | mr BASE, CRET1
813 | li NARGS8:RC, 16 // 2 args for func(o1, o2).
814 | b ->vm_call_dispatch
815 |
816 |->vmeta_len:
817#if LJ_52
818 | mr SAVE0, CARG1
819#endif
820 | add CARG2, BASE, RD
821 | stw BASE, L->base
822 | mr CARG1, L
823 | stw PC, SAVE_PC
824 | bl extern lj_meta_len // (lua_State *L, TValue *o)
825 | // Returns NULL (retry) or TValue * (metamethod base).
826#if LJ_52
827 | cmplwi CRET1, 0
828 | bne ->vmeta_binop // Binop call for compatibility.
829 | mr CARG1, SAVE0
830 | b ->BC_LEN_Z
831#else
832 | b ->vmeta_binop // Binop call for compatibility.
833#endif
834 |
835 |//-- Call metamethod ----------------------------------------------------
836 |
837 |->vmeta_call: // Resolve and call __call metamethod.
838 | // TMP2 = old base, BASE = new base, RC = nargs*8
839 | mr CARG1, L
840 | stw TMP2, L->base // This is the callers base!
841 | subi CARG2, BASE, 8
842 | stw PC, SAVE_PC
843 | add CARG3, BASE, RC
844 | mr SAVE0, NARGS8:RC
845 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
846 | lwz LFUNC:RB, FRAME_FUNC(BASE) // Guaranteed to be a function here.
847 | addi NARGS8:RC, SAVE0, 8 // Got one more argument now.
848 | ins_call
849 |
850 |->vmeta_callt: // Resolve __call for BC_CALLT.
851 | // BASE = old base, RA = new base, RC = nargs*8
852 | mr CARG1, L
853 | stw BASE, L->base
854 | subi CARG2, RA, 8
855 | stw PC, SAVE_PC
856 | add CARG3, RA, RC
857 | mr SAVE0, NARGS8:RC
858 | bl extern lj_meta_call // (lua_State *L, TValue *func, TValue *top)
859 | lwz TMP1, FRAME_PC(BASE)
860 | addi NARGS8:RC, SAVE0, 8 // Got one more argument now.
861 | lwz LFUNC:RB, FRAME_FUNC(RA) // Guaranteed to be a function here.
862 | b ->BC_CALLT_Z
863 |
864 |//-- Argument coercion for 'for' statement ------------------------------
865 |
866 |->vmeta_for:
867 | mr CARG1, L
868 | stw BASE, L->base
869 | mr CARG2, RA
870 | stw PC, SAVE_PC
871 | mr SAVE0, INS
872 | bl extern lj_meta_for // (lua_State *L, TValue *base)
873 |.if JIT
874 | decode_OP1 TMP0, SAVE0
875 |.endif
876 | decode_RA8 RA, SAVE0
877 |.if JIT
878 | cmpwi TMP0, BC_JFORI
879 |.endif
880 | decode_RD8 RD, SAVE0
881 |.if JIT
882 | beq =>BC_JFORI
883 |.endif
884 | b =>BC_FORI
885 |
886 |//-----------------------------------------------------------------------
887 |//-- Fast functions -----------------------------------------------------
888 |//-----------------------------------------------------------------------
889 |
890 |.macro .ffunc, name
891 |->ff_ .. name:
892 |.endmacro
893 |
894 |.macro .ffunc_1, name
895 |->ff_ .. name:
896 | cmplwi NARGS8:RC, 8
897 | evldd CARG1, 0(BASE)
898 | blt ->fff_fallback
899 |.endmacro
900 |
901 |.macro .ffunc_2, name
902 |->ff_ .. name:
903 | cmplwi NARGS8:RC, 16
904 | evldd CARG1, 0(BASE)
905 | evldd CARG2, 8(BASE)
906 | blt ->fff_fallback
907 |.endmacro
908 |
909 |.macro .ffunc_n, name
910 | .ffunc_1 name
911 | checknum CARG1
912 | checkfail ->fff_fallback
913 |.endmacro
914 |
915 |.macro .ffunc_nn, name
916 | .ffunc_2 name
917 | evmergehi TMP0, CARG1, CARG2
918 | checknum TMP0
919 | checkanyfail ->fff_fallback
920 |.endmacro
921 |
922 |// Inlined GC threshold check. Caveat: uses TMP0 and TMP1.
923 |.macro ffgccheck
924 | lwz TMP0, DISPATCH_GL(gc.total)(DISPATCH)
925 | lwz TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
926 | cmplw TMP0, TMP1
927 | bgel ->fff_gcstep
928 |.endmacro
929 |
930 |//-- Base library: checks -----------------------------------------------
931 |
932 |.ffunc assert
933 | cmplwi NARGS8:RC, 8
934 | evldd TMP0, 0(BASE)
935 | blt ->fff_fallback
936 | evaddw TMP1, TISNIL, TISNIL // Synthesize LJ_TFALSE.
937 | la RA, -8(BASE)
938 | evcmpltu cr1, TMP0, TMP1
939 | lwz PC, FRAME_PC(BASE)
940 | bge cr1, ->fff_fallback
941 | evstdd TMP0, 0(RA)
942 | addi RD, NARGS8:RC, 8 // Compute (nresults+1)*8.
943 | beq ->fff_res // Done if exactly 1 argument.
944 | li TMP1, 8
945 | subi RC, RC, 8
946 |1:
947 | cmplw TMP1, RC
948 | evlddx TMP0, BASE, TMP1
949 | evstddx TMP0, RA, TMP1
950 | addi TMP1, TMP1, 8
951 | bne <1
952 | b ->fff_res
953 |
954 |.ffunc type
955 | cmplwi NARGS8:RC, 8
956 | lwz CARG1, 0(BASE)
957 | blt ->fff_fallback
958 | li TMP2, ~LJ_TNUMX
959 | cmplw CARG1, TISNUM
960 | not TMP1, CARG1
961 | isellt TMP1, TMP2, TMP1
962 | slwi TMP1, TMP1, 3
963 | la TMP2, CFUNC:RB->upvalue
964 | evlddx STR:CRET1, TMP2, TMP1
965 | b ->fff_restv
966 |
967 |//-- Base library: getters and setters ---------------------------------
968 |
969 |.ffunc_1 getmetatable
970 | checktab CARG1
971 | evmergehi TMP1, CARG1, CARG1
972 | checkfail >6
973 |1: // Field metatable must be at same offset for GCtab and GCudata!
974 | lwz TAB:RB, TAB:CARG1->metatable
975 |2:
976 | evmr CRET1, TISNIL
977 | cmplwi TAB:RB, 0
978 | lwz STR:RC, DISPATCH_GL(gcroot[GCROOT_MMNAME+MM_metatable])(DISPATCH)
979 | beq ->fff_restv
980 | lwz TMP0, TAB:RB->hmask
981 | evmergelo CRET1, TISTAB, TAB:RB // Use metatable as default result.
982 | lwz TMP1, STR:RC->hash
983 | lwz NODE:TMP2, TAB:RB->node
984 | evmergelo STR:RC, TISSTR, STR:RC
985 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
986 | slwi TMP0, TMP1, 5
987 | slwi TMP1, TMP1, 3
988 | sub TMP1, TMP0, TMP1
989 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
990 |3: // Rearranged logic, because we expect _not_ to find the key.
991 | evldd TMP0, NODE:TMP2->key
992 | evldd TMP1, NODE:TMP2->val
993 | evcmpeq TMP0, STR:RC
994 | lwz NODE:TMP2, NODE:TMP2->next
995 | checkallok >5
996 | cmplwi NODE:TMP2, 0
997 | beq ->fff_restv // Not found, keep default result.
998 | b <3
999 |5:
1000 | checknil TMP1
1001 | checkok ->fff_restv // Ditto for nil value.
1002 | evmr CRET1, TMP1 // Return value of mt.__metatable.
1003 | b ->fff_restv
1004 |
1005 |6:
1006 | cmpwi TMP1, LJ_TUDATA
1007 | not TMP1, TMP1
1008 | beq <1
1009 | checknum CARG1
1010 | slwi TMP1, TMP1, 2
1011 | li TMP2, 4*~LJ_TNUMX
1012 | isellt TMP1, TMP2, TMP1
1013 | la TMP2, DISPATCH_GL(gcroot[GCROOT_BASEMT])(DISPATCH)
1014 | lwzx TAB:RB, TMP2, TMP1
1015 | b <2
1016 |
1017 |.ffunc_2 setmetatable
1018 | // Fast path: no mt for table yet and not clearing the mt.
1019 | evmergehi TMP0, TAB:CARG1, TAB:CARG2
1020 | checktab TMP0
1021 | checkanyfail ->fff_fallback
1022 | lwz TAB:TMP1, TAB:CARG1->metatable
1023 | cmplwi TAB:TMP1, 0
1024 | lbz TMP3, TAB:CARG1->marked
1025 | bne ->fff_fallback
1026 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
1027 | stw TAB:CARG2, TAB:CARG1->metatable
1028 | beq ->fff_restv
1029 | barrierback TAB:CARG1, TMP3, TMP0
1030 | b ->fff_restv
1031 |
1032 |.ffunc rawget
1033 | cmplwi NARGS8:RC, 16
1034 | evldd CARG2, 0(BASE)
1035 | blt ->fff_fallback
1036 | checktab CARG2
1037 | la CARG3, 8(BASE)
1038 | checkfail ->fff_fallback
1039 | mr CARG1, L
1040 | bl extern lj_tab_get // (lua_State *L, GCtab *t, cTValue *key)
1041 | // Returns cTValue *.
1042 | evldd CRET1, 0(CRET1)
1043 | b ->fff_restv
1044 |
1045 |//-- Base library: conversions ------------------------------------------
1046 |
1047 |.ffunc tonumber
1048 | // Only handles the number case inline (without a base argument).
1049 | cmplwi NARGS8:RC, 8
1050 | evldd CARG1, 0(BASE)
1051 | bne ->fff_fallback // Exactly one argument.
1052 | checknum CARG1
1053 | checkok ->fff_restv
1054 | b ->fff_fallback
1055 |
1056 |.ffunc_1 tostring
1057 | // Only handles the string or number case inline.
1058 | checkstr CARG1
1059 | // A __tostring method in the string base metatable is ignored.
1060 | checkok ->fff_restv // String key?
1061 | // Handle numbers inline, unless a number base metatable is present.
1062 | lwz TMP0, DISPATCH_GL(gcroot[GCROOT_BASEMT_NUM])(DISPATCH)
1063 | checknum CARG1
1064 | cmplwi cr1, TMP0, 0
1065 | stw BASE, L->base // Add frame since C call can throw.
1066 | crand 4*cr0+eq, 4*cr0+lt, 4*cr1+eq
1067 | stw PC, SAVE_PC // Redundant (but a defined value).
1068 | bne ->fff_fallback
1069 | ffgccheck
1070 | mr CARG1, L
1071 | mr CARG2, BASE
1072 | bl extern lj_str_fromnum // (lua_State *L, lua_Number *np)
1073 | // Returns GCstr *.
1074 | evmergelo STR:CRET1, TISSTR, STR:CRET1
1075 | b ->fff_restv
1076 |
1077 |//-- Base library: iterators -------------------------------------------
1078 |
1079 |.ffunc next
1080 | cmplwi NARGS8:RC, 8
1081 | evldd CARG2, 0(BASE)
1082 | blt ->fff_fallback
1083 | evstddx TISNIL, BASE, NARGS8:RC // Set missing 2nd arg to nil.
1084 | checktab TAB:CARG2
1085 | lwz PC, FRAME_PC(BASE)
1086 | checkfail ->fff_fallback
1087 | stw BASE, L->base // Add frame since C call can throw.
1088 | mr CARG1, L
1089 | stw BASE, L->top // Dummy frame length is ok.
1090 | la CARG3, 8(BASE)
1091 | stw PC, SAVE_PC
1092 | bl extern lj_tab_next // (lua_State *L, GCtab *t, TValue *key)
1093 | // Returns 0 at end of traversal.
1094 | cmplwi CRET1, 0
1095 | evmr CRET1, TISNIL
1096 | beq ->fff_restv // End of traversal: return nil.
1097 | evldd TMP0, 8(BASE) // Copy key and value to results.
1098 | la RA, -8(BASE)
1099 | evldd TMP1, 16(BASE)
1100 | evstdd TMP0, 0(RA)
1101 | li RD, (2+1)*8
1102 | evstdd TMP1, 8(RA)
1103 | b ->fff_res
1104 |
1105 |.ffunc_1 pairs
1106 | checktab TAB:CARG1
1107 | lwz PC, FRAME_PC(BASE)
1108 | checkfail ->fff_fallback
1109#if LJ_52
1110 | lwz TAB:TMP2, TAB:CARG1->metatable
1111 | evldd CFUNC:TMP0, CFUNC:RB->upvalue[0]
1112 | cmplwi TAB:TMP2, 0
1113 | la RA, -8(BASE)
1114 | bne ->fff_fallback
1115#else
1116 | evldd CFUNC:TMP0, CFUNC:RB->upvalue[0]
1117 | la RA, -8(BASE)
1118#endif
1119 | evstdd TISNIL, 8(BASE)
1120 | li RD, (3+1)*8
1121 | evstdd CFUNC:TMP0, 0(RA)
1122 | b ->fff_res
1123 |
1124 |.ffunc_2 ipairs_aux
1125 | checktab TAB:CARG1
1126 | lwz PC, FRAME_PC(BASE)
1127 | checkfail ->fff_fallback
1128 | checknum CARG2
1129 | lus TMP3, 0x3ff0
1130 | checkfail ->fff_fallback
1131 | efdctsi TMP2, CARG2
1132 | lwz TMP0, TAB:CARG1->asize
1133 | evmergelo TMP3, TMP3, ZERO
1134 | lwz TMP1, TAB:CARG1->array
1135 | efdadd CARG2, CARG2, TMP3
1136 | addi TMP2, TMP2, 1
1137 | la RA, -8(BASE)
1138 | cmplw TMP0, TMP2
1139 | slwi TMP3, TMP2, 3
1140 | evstdd CARG2, 0(RA)
1141 | ble >2 // Not in array part?
1142 | evlddx TMP1, TMP1, TMP3
1143 |1:
1144 | checknil TMP1
1145 | li RD, (0+1)*8
1146 | checkok ->fff_res // End of iteration, return 0 results.
1147 | li RD, (2+1)*8
1148 | evstdd TMP1, 8(RA)
1149 | b ->fff_res
1150 |2: // Check for empty hash part first. Otherwise call C function.
1151 | lwz TMP0, TAB:CARG1->hmask
1152 | cmplwi TMP0, 0
1153 | li RD, (0+1)*8
1154 | beq ->fff_res
1155 | mr CARG2, TMP2
1156 | bl extern lj_tab_getinth // (GCtab *t, int32_t key)
1157 | // Returns cTValue * or NULL.
1158 | cmplwi CRET1, 0
1159 | li RD, (0+1)*8
1160 | beq ->fff_res
1161 | evldd TMP1, 0(CRET1)
1162 | b <1
1163 |
1164 |.ffunc_1 ipairs
1165 | checktab TAB:CARG1
1166 | lwz PC, FRAME_PC(BASE)
1167 | checkfail ->fff_fallback
1168#if LJ_52
1169 | lwz TAB:TMP2, TAB:CARG1->metatable
1170 | evldd CFUNC:TMP0, CFUNC:RB->upvalue[0]
1171 | cmplwi TAB:TMP2, 0
1172 | la RA, -8(BASE)
1173 | bne ->fff_fallback
1174#else
1175 | evldd CFUNC:TMP0, CFUNC:RB->upvalue[0]
1176 | la RA, -8(BASE)
1177#endif
1178 | evsplati TMP1, 0
1179 | li RD, (3+1)*8
1180 | evstdd TMP1, 8(BASE)
1181 | evstdd CFUNC:TMP0, 0(RA)
1182 | b ->fff_res
1183 |
1184 |//-- Base library: catch errors ----------------------------------------
1185 |
1186 |.ffunc pcall
1187 | cmplwi NARGS8:RC, 8
1188 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1189 | blt ->fff_fallback
1190 | mr TMP2, BASE
1191 | la BASE, 8(BASE)
1192 | // Remember active hook before pcall.
1193 | rlwinm TMP3, TMP3, 32-HOOK_ACTIVE_SHIFT, 31, 31
1194 | subi NARGS8:RC, NARGS8:RC, 8
1195 | addi PC, TMP3, 8+FRAME_PCALL
1196 | b ->vm_call_dispatch
1197 |
1198 |.ffunc_2 xpcall
1199 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1200 | mr TMP2, BASE
1201 | checkfunc CARG2 // Traceback must be a function.
1202 | checkfail ->fff_fallback
1203 | la BASE, 16(BASE)
1204 | // Remember active hook before pcall.
1205 | rlwinm TMP3, TMP3, 32-HOOK_ACTIVE_SHIFT, 31, 31
1206 | evstdd CARG2, 0(TMP2) // Swap function and traceback.
1207 | subi NARGS8:RC, NARGS8:RC, 16
1208 | evstdd CARG1, 8(TMP2)
1209 | addi PC, TMP3, 16+FRAME_PCALL
1210 | b ->vm_call_dispatch
1211 |
1212 |//-- Coroutine library --------------------------------------------------
1213 |
1214 |.macro coroutine_resume_wrap, resume
1215 |.if resume
1216 |.ffunc_1 coroutine_resume
1217 | evmergehi TMP0, L:CARG1, L:CARG1
1218 |.else
1219 |.ffunc coroutine_wrap_aux
1220 | lwz L:CARG1, CFUNC:RB->upvalue[0].gcr
1221 |.endif
1222 |.if resume
1223 | cmpwi TMP0, LJ_TTHREAD
1224 | bne ->fff_fallback
1225 |.endif
1226 | lbz TMP0, L:CARG1->status
1227 | lwz TMP1, L:CARG1->cframe
1228 | lwz CARG2, L:CARG1->top
1229 | cmplwi cr0, TMP0, LUA_YIELD
1230 | lwz TMP2, L:CARG1->base
1231 | cmplwi cr1, TMP1, 0
1232 | lwz TMP0, L:CARG1->maxstack
1233 | cmplw cr7, CARG2, TMP2
1234 | lwz PC, FRAME_PC(BASE)
1235 | crorc 4*cr6+lt, 4*cr0+gt, 4*cr1+eq // st>LUA_YIELD || cframe!=0
1236 | add TMP2, CARG2, NARGS8:RC
1237 | crandc 4*cr6+gt, 4*cr7+eq, 4*cr0+eq // base==top && st!=LUA_YIELD
1238 | cmplw cr1, TMP2, TMP0
1239 | cror 4*cr6+lt, 4*cr6+lt, 4*cr6+gt
1240 | stw PC, SAVE_PC
1241 | cror 4*cr6+lt, 4*cr6+lt, 4*cr1+gt // cond1 || cond2 || stackov
1242 | stw BASE, L->base
1243 | blt cr6, ->fff_fallback
1244 |1:
1245 |.if resume
1246 | addi BASE, BASE, 8 // Keep resumed thread in stack for GC.
1247 | subi NARGS8:RC, NARGS8:RC, 8
1248 | subi TMP2, TMP2, 8
1249 |.endif
1250 | stw TMP2, L:CARG1->top
1251 | li TMP1, 0
1252 | stw BASE, L->top
1253 |2: // Move args to coroutine.
1254 | cmpw TMP1, NARGS8:RC
1255 | evlddx TMP0, BASE, TMP1
1256 | beq >3
1257 | evstddx TMP0, CARG2, TMP1
1258 | addi TMP1, TMP1, 8
1259 | b <2
1260 |3:
1261 | li CARG3, 0
1262 | mr L:SAVE0, L:CARG1
1263 | li CARG4, 0
1264 | bl ->vm_resume // (lua_State *L, TValue *base, 0, 0)
1265 | // Returns thread status.
1266 |4:
1267 | lwz TMP2, L:SAVE0->base
1268 | cmplwi CRET1, LUA_YIELD
1269 | lwz TMP3, L:SAVE0->top
1270 | li_vmstate INTERP
1271 | lwz BASE, L->base
1272 | st_vmstate
1273 | bgt >8
1274 | sub RD, TMP3, TMP2
1275 | lwz TMP0, L->maxstack
1276 | cmplwi RD, 0
1277 | add TMP1, BASE, RD
1278 | beq >6 // No results?
1279 | cmplw TMP1, TMP0
1280 | li TMP1, 0
1281 | bgt >9 // Need to grow stack?
1282 |
1283 | subi TMP3, RD, 8
1284 | stw TMP2, L:SAVE0->top // Clear coroutine stack.
1285 |5: // Move results from coroutine.
1286 | cmplw TMP1, TMP3
1287 | evlddx TMP0, TMP2, TMP1
1288 | evstddx TMP0, BASE, TMP1
1289 | addi TMP1, TMP1, 8
1290 | bne <5
1291 |6:
1292 | andi. TMP0, PC, FRAME_TYPE
1293 |.if resume
1294 | li TMP1, LJ_TTRUE
1295 | la RA, -8(BASE)
1296 | stw TMP1, -8(BASE) // Prepend true to results.
1297 | addi RD, RD, 16
1298 |.else
1299 | mr RA, BASE
1300 | addi RD, RD, 8
1301 |.endif
1302 |7:
1303 | stw PC, SAVE_PC
1304 | mr MULTRES, RD
1305 | beq ->BC_RET_Z
1306 | b ->vm_return
1307 |
1308 |8: // Coroutine returned with error (at co->top-1).
1309 |.if resume
1310 | andi. TMP0, PC, FRAME_TYPE
1311 | la TMP3, -8(TMP3)
1312 | li TMP1, LJ_TFALSE
1313 | evldd TMP0, 0(TMP3)
1314 | stw TMP3, L:SAVE0->top // Remove error from coroutine stack.
1315 | li RD, (2+1)*8
1316 | stw TMP1, -8(BASE) // Prepend false to results.
1317 | la RA, -8(BASE)
1318 | evstdd TMP0, 0(BASE) // Copy error message.
1319 | b <7
1320 |.else
1321 | mr CARG1, L
1322 | mr CARG2, L:SAVE0
1323 | bl extern lj_ffh_coroutine_wrap_err // (lua_State *L, lua_State *co)
1324 |.endif
1325 |
1326 |9: // Handle stack expansion on return from yield.
1327 | mr CARG1, L
1328 | srwi CARG2, RD, 3
1329 | bl extern lj_state_growstack // (lua_State *L, int n)
1330 | li CRET1, 0
1331 | b <4
1332 |.endmacro
1333 |
1334 | coroutine_resume_wrap 1 // coroutine.resume
1335 | coroutine_resume_wrap 0 // coroutine.wrap
1336 |
1337 |.ffunc coroutine_yield
1338 | lwz TMP0, L->cframe
1339 | add TMP1, BASE, NARGS8:RC
1340 | stw BASE, L->base
1341 | andi. TMP0, TMP0, CFRAME_RESUME
1342 | stw TMP1, L->top
1343 | li CRET1, LUA_YIELD
1344 | beq ->fff_fallback
1345 | stw ZERO, L->cframe
1346 | stb CRET1, L->status
1347 | b ->vm_leave_unw
1348 |
1349 |//-- Math library -------------------------------------------------------
1350 |
1351 |.ffunc_n math_abs
1352 | efdabs CRET1, CARG1
1353 | // Fallthrough.
1354 |
1355 |->fff_restv:
1356 | // CRET1 = TValue result.
1357 | lwz PC, FRAME_PC(BASE)
1358 | la RA, -8(BASE)
1359 | evstdd CRET1, 0(RA)
1360 |->fff_res1:
1361 | // RA = results, PC = return.
1362 | li RD, (1+1)*8
1363 |->fff_res:
1364 | // RA = results, RD = (nresults+1)*8, PC = return.
1365 | andi. TMP0, PC, FRAME_TYPE
1366 | mr MULTRES, RD
1367 | bne ->vm_return
1368 | lwz INS, -4(PC)
1369 | decode_RB8 RB, INS
1370 |5:
1371 | cmplw RB, RD // More results expected?
1372 | decode_RA8 TMP0, INS
1373 | bgt >6
1374 | ins_next1
1375 | // Adjust BASE. KBASE is assumed to be set for the calling frame.
1376 | sub BASE, RA, TMP0
1377 | ins_next2
1378 |
1379 |6: // Fill up results with nil.
1380 | subi TMP1, RD, 8
1381 | addi RD, RD, 8
1382 | evstddx TISNIL, RA, TMP1
1383 | b <5
1384 |
1385 |.macro math_extern, func
1386 | .ffunc math_ .. func
1387 | cmplwi NARGS8:RC, 8
1388 | evldd CARG2, 0(BASE)
1389 | blt ->fff_fallback
1390 | checknum CARG2
1391 | evmergehi CARG1, CARG2, CARG2
1392 | checkfail ->fff_fallback
1393 | bl extern func@plt
1394 | evmergelo CRET1, CRET1, CRET2
1395 | b ->fff_restv
1396 |.endmacro
1397 |
1398 |.macro math_extern2, func
1399 | .ffunc math_ .. func
1400 | cmplwi NARGS8:RC, 16
1401 | evldd CARG2, 0(BASE)
1402 | evldd CARG4, 8(BASE)
1403 | blt ->fff_fallback
1404 | evmergehi CARG1, CARG4, CARG2
1405 | checknum CARG1
1406 | evmergehi CARG3, CARG4, CARG4
1407 | checkanyfail ->fff_fallback
1408 | bl extern func@plt
1409 | evmergelo CRET1, CRET1, CRET2
1410 | b ->fff_restv
1411 |.endmacro
1412 |
1413 |.macro math_round, func
1414 | .ffunc math_ .. func
1415 | cmplwi NARGS8:RC, 8
1416 | evldd CARG2, 0(BASE)
1417 | blt ->fff_fallback
1418 | checknum CARG2
1419 | evmergehi CARG1, CARG2, CARG2
1420 | checkfail ->fff_fallback
1421 | lwz PC, FRAME_PC(BASE)
1422 | bl ->vm_..func.._hilo;
1423 | la RA, -8(BASE)
1424 | evstdd CRET2, 0(RA)
1425 | b ->fff_res1
1426 |.endmacro
1427 |
1428 | math_round floor
1429 | math_round ceil
1430 |
1431 | math_extern sqrt
1432 |
1433 |.ffunc math_log
1434 | cmplwi NARGS8:RC, 8
1435 | evldd CARG2, 0(BASE)
1436 | bne ->fff_fallback // Need exactly 1 argument.
1437 | checknum CARG2
1438 | evmergehi CARG1, CARG2, CARG2
1439 | checkfail ->fff_fallback
1440 | bl extern log@plt
1441 | evmergelo CRET1, CRET1, CRET2
1442 | b ->fff_restv
1443 |
1444 | math_extern log10
1445 | math_extern exp
1446 | math_extern sin
1447 | math_extern cos
1448 | math_extern tan
1449 | math_extern asin
1450 | math_extern acos
1451 | math_extern atan
1452 | math_extern sinh
1453 | math_extern cosh
1454 | math_extern tanh
1455 | math_extern2 pow
1456 | math_extern2 atan2
1457 | math_extern2 fmod
1458 |
1459 |.ffunc math_ldexp
1460 | cmplwi NARGS8:RC, 16
1461 | evldd CARG2, 0(BASE)
1462 | evldd CARG4, 8(BASE)
1463 | blt ->fff_fallback
1464 | evmergehi CARG1, CARG4, CARG2
1465 | checknum CARG1
1466 | checkanyfail ->fff_fallback
1467 | efdctsi CARG3, CARG4
1468 | bl extern ldexp@plt
1469 | evmergelo CRET1, CRET1, CRET2
1470 | b ->fff_restv
1471 |
1472 |.ffunc math_frexp
1473 | cmplwi NARGS8:RC, 8
1474 | evldd CARG2, 0(BASE)
1475 | blt ->fff_fallback
1476 | checknum CARG2
1477 | evmergehi CARG1, CARG2, CARG2
1478 | checkfail ->fff_fallback
1479 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
1480 | lwz PC, FRAME_PC(BASE)
1481 | bl extern frexp@plt
1482 | lwz TMP1, DISPATCH_GL(tmptv)(DISPATCH)
1483 | evmergelo CRET1, CRET1, CRET2
1484 | efdcfsi CRET2, TMP1
1485 | la RA, -8(BASE)
1486 | evstdd CRET1, 0(RA)
1487 | li RD, (2+1)*8
1488 | evstdd CRET2, 8(RA)
1489 | b ->fff_res
1490 |
1491 |.ffunc math_modf
1492 | cmplwi NARGS8:RC, 8
1493 | evldd CARG2, 0(BASE)
1494 | blt ->fff_fallback
1495 | checknum CARG2
1496 | evmergehi CARG1, CARG2, CARG2
1497 | checkfail ->fff_fallback
1498 | la CARG3, -8(BASE)
1499 | lwz PC, FRAME_PC(BASE)
1500 | bl extern modf@plt
1501 | evmergelo CRET1, CRET1, CRET2
1502 | la RA, -8(BASE)
1503 | evstdd CRET1, 0(BASE)
1504 | li RD, (2+1)*8
1505 | b ->fff_res
1506 |
1507 |.macro math_minmax, name, cmpop
1508 | .ffunc_1 name
1509 | checknum CARG1
1510 | li TMP1, 8
1511 | checkfail ->fff_fallback
1512 |1:
1513 | evlddx CARG2, BASE, TMP1
1514 | cmplw cr1, TMP1, NARGS8:RC
1515 | checknum CARG2
1516 | bge cr1, ->fff_restv // Ok, since CRET1 = CARG1.
1517 | checkfail ->fff_fallback
1518 | cmpop CARG2, CARG1
1519 | addi TMP1, TMP1, 8
1520 | crmove 4*cr0+lt, 4*cr0+gt
1521 | evsel CARG1, CARG2, CARG1
1522 | b <1
1523 |.endmacro
1524 |
1525 | math_minmax math_min, efdtstlt
1526 | math_minmax math_max, efdtstgt
1527 |
1528 |//-- String library -----------------------------------------------------
1529 |
1530 |.ffunc_1 string_len
1531 | checkstr STR:CARG1
1532 | checkfail ->fff_fallback
1533 | lwz TMP0, STR:CARG1->len
1534 | efdcfsi CRET1, TMP0
1535 | b ->fff_restv
1536 |
1537 |.ffunc string_byte // Only handle the 1-arg case here.
1538 | cmplwi NARGS8:RC, 8
1539 | evldd STR:CARG1, 0(BASE)
1540 | bne ->fff_fallback // Need exactly 1 argument.
1541 | checkstr STR:CARG1
1542 | la RA, -8(BASE)
1543 | checkfail ->fff_fallback
1544 | lwz TMP0, STR:CARG1->len
1545 | li RD, (0+1)*8
1546 | lbz TMP1, STR:CARG1[1] // Access is always ok (NUL at end).
1547 | li TMP2, (1+1)*8
1548 | cmplwi TMP0, 0
1549 | lwz PC, FRAME_PC(BASE)
1550 | efdcfsi CRET1, TMP1
1551 | iseleq RD, RD, TMP2
1552 | evstdd CRET1, 0(RA)
1553 | b ->fff_res
1554 |
1555 |.ffunc string_char // Only handle the 1-arg case here.
1556 | ffgccheck
1557 | cmplwi NARGS8:RC, 8
1558 | evldd CARG1, 0(BASE)
1559 | bne ->fff_fallback // Exactly 1 argument.
1560 | checknum CARG1
1561 | la CARG2, DISPATCH_GL(tmptv)(DISPATCH)
1562 | checkfail ->fff_fallback
1563 | efdctsiz TMP0, CARG1
1564 | li CARG3, 1
1565 | cmplwi TMP0, 255
1566 | stb TMP0, 0(CARG2)
1567 | bgt ->fff_fallback
1568 |->fff_newstr:
1569 | mr CARG1, L
1570 | stw BASE, L->base
1571 | stw PC, SAVE_PC
1572 | bl extern lj_str_new // (lua_State *L, char *str, size_t l)
1573 | // Returns GCstr *.
1574 | lwz BASE, L->base
1575 | evmergelo STR:CRET1, TISSTR, STR:CRET1
1576 | b ->fff_restv
1577 |
1578 |.ffunc string_sub
1579 | ffgccheck
1580 | cmplwi NARGS8:RC, 16
1581 | evldd CARG3, 16(BASE)
1582 | evldd STR:CARG1, 0(BASE)
1583 | blt ->fff_fallback
1584 | evldd CARG2, 8(BASE)
1585 | li TMP2, -1
1586 | beq >1
1587 | checknum CARG3
1588 | checkfail ->fff_fallback
1589 | efdctsiz TMP2, CARG3
1590 |1:
1591 | checknum CARG2
1592 | checkfail ->fff_fallback
1593 | checkstr STR:CARG1
1594 | efdctsiz TMP1, CARG2
1595 | checkfail ->fff_fallback
1596 | lwz TMP0, STR:CARG1->len
1597 | cmplw TMP0, TMP2 // len < end? (unsigned compare)
1598 | add TMP3, TMP2, TMP0
1599 | blt >5
1600 |2:
1601 | cmpwi TMP1, 0 // start <= 0?
1602 | add TMP3, TMP1, TMP0
1603 | ble >7
1604 |3:
1605 | sub. CARG3, TMP2, TMP1
1606 | addi CARG2, STR:CARG1, #STR-1
1607 | addi CARG3, CARG3, 1
1608 | add CARG2, CARG2, TMP1
1609 | isellt CARG3, r0, CARG3
1610 | b ->fff_newstr
1611 |
1612 |5: // Negative end or overflow.
1613 | cmpw TMP0, TMP2
1614 | addi TMP3, TMP3, 1
1615 | iselgt TMP2, TMP3, TMP0 // end = end > len ? len : end+len+1
1616 | b <2
1617 |
1618 |7: // Negative start or underflow.
1619 | cmpwi cr1, TMP3, 0
1620 | iseleq TMP1, r0, TMP3
1621 | isel TMP1, r0, TMP1, 4*cr1+lt
1622 | addi TMP1, TMP1, 1 // start = 1 + (start ? start+len : 0)
1623 | b <3
1624 |
1625 |.ffunc string_rep // Only handle the 1-char case inline.
1626 | ffgccheck
1627 | cmplwi NARGS8:RC, 16
1628 | evldd CARG1, 0(BASE)
1629 | evldd CARG2, 8(BASE)
1630 | bne ->fff_fallback // Exactly 2 arguments.
1631 | checknum CARG2
1632 | checkfail ->fff_fallback
1633 | checkstr STR:CARG1
1634 | efdctsiz CARG3, CARG2
1635 | checkfail ->fff_fallback
1636 | lwz TMP0, STR:CARG1->len
1637 | cmpwi CARG3, 0
1638 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1639 | ble >2 // Count <= 0? (or non-int)
1640 | cmplwi TMP0, 1
1641 | subi TMP2, CARG3, 1
1642 | blt >2 // Zero length string?
1643 | cmplw cr1, TMP1, CARG3
1644 | bne ->fff_fallback // Fallback for > 1-char strings.
1645 | lbz TMP0, STR:CARG1[1]
1646 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1647 | blt cr1, ->fff_fallback
1648 |1: // Fill buffer with char. Yes, this is suboptimal code (do you care?).
1649 | cmplwi TMP2, 0
1650 | stbx TMP0, CARG2, TMP2
1651 | subi TMP2, TMP2, 1
1652 | bne <1
1653 | b ->fff_newstr
1654 |2: // Return empty string.
1655 | la STR:CRET1, DISPATCH_GL(strempty)(DISPATCH)
1656 | evmergelo CRET1, TISSTR, STR:CRET1
1657 | b ->fff_restv
1658 |
1659 |.ffunc string_reverse
1660 | ffgccheck
1661 | cmplwi NARGS8:RC, 8
1662 | evldd CARG1, 0(BASE)
1663 | blt ->fff_fallback
1664 | checkstr STR:CARG1
1665 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1666 | checkfail ->fff_fallback
1667 | lwz CARG3, STR:CARG1->len
1668 | la CARG1, #STR(STR:CARG1)
1669 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1670 | li TMP2, 0
1671 | cmplw TMP1, CARG3
1672 | subi TMP3, CARG3, 1
1673 | blt ->fff_fallback
1674 |1: // Reverse string copy.
1675 | cmpwi TMP3, 0
1676 | lbzx TMP1, CARG1, TMP2
1677 | blt ->fff_newstr
1678 | stbx TMP1, CARG2, TMP3
1679 | subi TMP3, TMP3, 1
1680 | addi TMP2, TMP2, 1
1681 | b <1
1682 |
1683 |.macro ffstring_case, name, lo
1684 | .ffunc name
1685 | ffgccheck
1686 | cmplwi NARGS8:RC, 8
1687 | evldd CARG1, 0(BASE)
1688 | blt ->fff_fallback
1689 | checkstr STR:CARG1
1690 | lwz TMP1, DISPATCH_GL(tmpbuf.sz)(DISPATCH)
1691 | checkfail ->fff_fallback
1692 | lwz CARG3, STR:CARG1->len
1693 | la CARG1, #STR(STR:CARG1)
1694 | lwz CARG2, DISPATCH_GL(tmpbuf.buf)(DISPATCH)
1695 | cmplw TMP1, CARG3
1696 | li TMP2, 0
1697 | blt ->fff_fallback
1698 |1: // ASCII case conversion.
1699 | cmplw TMP2, CARG3
1700 | lbzx TMP1, CARG1, TMP2
1701 | bge ->fff_newstr
1702 | subi TMP0, TMP1, lo
1703 | xori TMP3, TMP1, 0x20
1704 | cmplwi TMP0, 26
1705 | isellt TMP1, TMP3, TMP1
1706 | stbx TMP1, CARG2, TMP2
1707 | addi TMP2, TMP2, 1
1708 | b <1
1709 |.endmacro
1710 |
1711 |ffstring_case string_lower, 65
1712 |ffstring_case string_upper, 97
1713 |
1714 |//-- Table library ------------------------------------------------------
1715 |
1716 |.ffunc_1 table_getn
1717 | checktab CARG1
1718 | checkfail ->fff_fallback
1719 | bl extern lj_tab_len // (GCtab *t)
1720 | // Returns uint32_t (but less than 2^31).
1721 | efdcfsi CRET1, CRET1
1722 | b ->fff_restv
1723 |
1724 |//-- Bit library --------------------------------------------------------
1725 |
1726 |.macro .ffunc_bit, name
1727 | .ffunc_n bit_..name
1728 | efdadd CARG1, CARG1, TOBIT
1729 |.endmacro
1730 |
1731 |.ffunc_bit tobit
1732 |->fff_resbit:
1733 | efdcfsi CRET1, CARG1
1734 | b ->fff_restv
1735 |
1736 |.macro .ffunc_bit_op, name, ins
1737 | .ffunc_bit name
1738 | li TMP1, 8
1739 |1:
1740 | evlddx CARG2, BASE, TMP1
1741 | cmplw cr1, TMP1, NARGS8:RC
1742 | checknum CARG2
1743 | bge cr1, ->fff_resbit
1744 | checkfail ->fff_fallback
1745 | efdadd CARG2, CARG2, TOBIT
1746 | ins CARG1, CARG1, CARG2
1747 | addi TMP1, TMP1, 8
1748 | b <1
1749 |.endmacro
1750 |
1751 |.ffunc_bit_op band, and
1752 |.ffunc_bit_op bor, or
1753 |.ffunc_bit_op bxor, xor
1754 |
1755 |.ffunc_bit bswap
1756 | rotlwi TMP0, CARG1, 8
1757 | rlwimi TMP0, CARG1, 24, 0, 7
1758 | rlwimi TMP0, CARG1, 24, 16, 23
1759 | efdcfsi CRET1, TMP0
1760 | b ->fff_restv
1761 |
1762 |.ffunc_bit bnot
1763 | not TMP0, CARG1
1764 | efdcfsi CRET1, TMP0
1765 | b ->fff_restv
1766 |
1767 |.macro .ffunc_bit_sh, name, ins, shmod
1768 | .ffunc_nn bit_..name
1769 | efdadd CARG2, CARG2, TOBIT
1770 | efdadd CARG1, CARG1, TOBIT
1771 |.if shmod == 1
1772 | rlwinm CARG2, CARG2, 0, 27, 31
1773 |.elif shmod == 2
1774 | neg CARG2, CARG2
1775 |.endif
1776 | ins TMP0, CARG1, CARG2
1777 | efdcfsi CRET1, TMP0
1778 | b ->fff_restv
1779 |.endmacro
1780 |
1781 |.ffunc_bit_sh lshift, slw, 1
1782 |.ffunc_bit_sh rshift, srw, 1
1783 |.ffunc_bit_sh arshift, sraw, 1
1784 |.ffunc_bit_sh rol, rotlw, 0
1785 |.ffunc_bit_sh ror, rotlw, 2
1786 |
1787 |//-----------------------------------------------------------------------
1788 |
1789 |->fff_fallback: // Call fast function fallback handler.
1790 | // BASE = new base, RB = CFUNC, RC = nargs*8
1791 | lwz TMP3, CFUNC:RB->f
1792 | add TMP1, BASE, NARGS8:RC
1793 | lwz PC, FRAME_PC(BASE) // Fallback may overwrite PC.
1794 | addi TMP0, TMP1, 8*LUA_MINSTACK
1795 | lwz TMP2, L->maxstack
1796 | stw PC, SAVE_PC // Redundant (but a defined value).
1797 | cmplw TMP0, TMP2
1798 | stw BASE, L->base
1799 | stw TMP1, L->top
1800 | mr CARG1, L
1801 | bgt >5 // Need to grow stack.
1802 | mtctr TMP3
1803 | bctrl // (lua_State *L)
1804 | // Either throws an error, or recovers and returns -1, 0 or nresults+1.
1805 | lwz BASE, L->base
1806 | cmpwi CRET1, 0
1807 | slwi RD, CRET1, 3
1808 | la RA, -8(BASE)
1809 | bgt ->fff_res // Returned nresults+1?
1810 |1: // Returned 0 or -1: retry fast path.
1811 | lwz TMP0, L->top
1812 | lwz LFUNC:RB, FRAME_FUNC(BASE)
1813 | sub NARGS8:RC, TMP0, BASE
1814 | bne ->vm_call_tail // Returned -1?
1815 | ins_callt // Returned 0: retry fast path.
1816 |
1817 |// Reconstruct previous base for vmeta_call during tailcall.
1818 |->vm_call_tail:
1819 | andi. TMP0, PC, FRAME_TYPE
1820 | rlwinm TMP1, PC, 0, 0, 28
1821 | bne >3
1822 | lwz INS, -4(PC)
1823 | decode_RA8 TMP1, INS
1824 | addi TMP1, TMP1, 8
1825 |3:
1826 | sub TMP2, BASE, TMP1
1827 | b ->vm_call_dispatch // Resolve again for tailcall.
1828 |
1829 |5: // Grow stack for fallback handler.
1830 | li CARG2, LUA_MINSTACK
1831 | bl extern lj_state_growstack // (lua_State *L, int n)
1832 | lwz BASE, L->base
1833 | cmpw TMP0, TMP0 // Set 4*cr0+eq to force retry.
1834 | b <1
1835 |
1836 |->fff_gcstep: // Call GC step function.
1837 | // BASE = new base, RC = nargs*8
1838 | mflr SAVE0
1839 | stw BASE, L->base
1840 | add TMP0, BASE, NARGS8:RC
1841 | stw PC, SAVE_PC // Redundant (but a defined value).
1842 | stw TMP0, L->top
1843 | mr CARG1, L
1844 | bl extern lj_gc_step // (lua_State *L)
1845 | lwz BASE, L->base
1846 | mtlr SAVE0
1847 | lwz TMP0, L->top
1848 | sub NARGS8:RC, TMP0, BASE
1849 | lwz CFUNC:RB, FRAME_FUNC(BASE)
1850 | blr
1851 |
1852 |//-----------------------------------------------------------------------
1853 |//-- Special dispatch targets -------------------------------------------
1854 |//-----------------------------------------------------------------------
1855 |
1856 |->vm_record: // Dispatch target for recording phase.
1857 |.if JIT
1858 | NYI
1859 |.endif
1860 |
1861 |->vm_rethook: // Dispatch target for return hooks.
1862 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1863 | andi. TMP0, TMP3, HOOK_ACTIVE // Hook already active?
1864 | beq >1
1865 |5: // Re-dispatch to static ins.
1866 | addi TMP1, TMP1, GG_DISP2STATIC // Assumes decode_OP4 TMP1, INS.
1867 | lwzx TMP0, DISPATCH, TMP1
1868 | mtctr TMP0
1869 | bctr
1870 |
1871 |->vm_inshook: // Dispatch target for instr/line hooks.
1872 | lbz TMP3, DISPATCH_GL(hookmask)(DISPATCH)
1873 | lwz TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1874 | andi. TMP0, TMP3, HOOK_ACTIVE // Hook already active?
1875 | rlwinm TMP0, TMP3, 31-LUA_HOOKLINE, 31, 0
1876 | bne <5
1877 |
1878 | cmpwi cr1, TMP0, 0
1879 | addic. TMP2, TMP2, -1
1880 | beq cr1, <5
1881 | stw TMP2, DISPATCH_GL(hookcount)(DISPATCH)
1882 | beq >1
1883 | bge cr1, <5
1884 |1:
1885 | mr CARG1, L
1886 | stw MULTRES, SAVE_MULTRES
1887 | mr CARG2, PC
1888 | stw BASE, L->base
1889 | // SAVE_PC must hold the _previous_ PC. The callee updates it with PC.
1890 | bl extern lj_dispatch_ins // (lua_State *L, const BCIns *pc)
1891 |3:
1892 | lwz BASE, L->base
1893 |4: // Re-dispatch to static ins.
1894 | lwz INS, -4(PC)
1895 | decode_OP4 TMP1, INS
1896 | decode_RB8 RB, INS
1897 | addi TMP1, TMP1, GG_DISP2STATIC
1898 | decode_RD8 RD, INS
1899 | lwzx TMP0, DISPATCH, TMP1
1900 | decode_RA8 RA, INS
1901 | decode_RC8 RC, INS
1902 | mtctr TMP0
1903 | bctr
1904 |
1905 |->cont_hook: // Continue from hook yield.
1906 | addi PC, PC, 4
1907 | lwz MULTRES, -20(RB) // Restore MULTRES for *M ins.
1908 | b <4
1909 |
1910 |->vm_hotloop: // Hot loop counter underflow.
1911 |.if JIT
1912 | NYI
1913 |.endif
1914 |
1915 |->vm_callhook: // Dispatch target for call hooks.
1916 | mr CARG2, PC
1917 |.if JIT
1918 | b >1
1919 |.endif
1920 |
1921 |->vm_hotcall: // Hot call counter underflow.
1922 |.if JIT
1923 | ori CARG2, PC, 1
1924 |1:
1925 |.endif
1926 | add TMP0, BASE, RC
1927 | stw PC, SAVE_PC
1928 | mr CARG1, L
1929 | stw BASE, L->base
1930 | sub RA, RA, BASE
1931 | stw TMP0, L->top
1932 | bl extern lj_dispatch_call // (lua_State *L, const BCIns *pc)
1933 | // Returns ASMFunction.
1934 | lwz BASE, L->base
1935 | lwz TMP0, L->top
1936 | stw ZERO, SAVE_PC // Invalidate for subsequent line hook.
1937 | sub NARGS8:RC, TMP0, BASE
1938 | add RA, BASE, RA
1939 | lwz LFUNC:RB, FRAME_FUNC(BASE)
1940 | mtctr CRET1
1941 | bctr
1942 |
1943 |//-----------------------------------------------------------------------
1944 |//-- Trace exit handler -------------------------------------------------
1945 |//-----------------------------------------------------------------------
1946 |
1947 |->vm_exit_handler:
1948 |.if JIT
1949 | NYI
1950 |.endif
1951 |->vm_exit_interp:
1952 |.if JIT
1953 | NYI
1954 |.endif
1955 |
1956 |//-----------------------------------------------------------------------
1957 |//-- Math helper functions ----------------------------------------------
1958 |//-----------------------------------------------------------------------
1959 |
1960 |// FP value rounding. Called by math.floor/math.ceil fast functions
1961 |// and from JIT code.
1962 |//
1963 |// This can be inlined if the CPU has the frin/friz/frip/frim instructions.
1964 |// The alternative hard-float approaches have a deep dependency chain.
1965 |// The resulting latency is at least 3x-7x the double-precision FP latency
1966 |// (e500v2: 6cy, e600: 5cy, Cell: 10cy) or around 20-70 cycles.
1967 |//
1968 |// The soft-float approach is tedious, but much faster (e500v2: ~11cy/~6cy).
1969 |// However it relies on a fast way to transfer the FP value to GPRs
1970 |// (e500v2: 0cy for lo-word, 1cy for hi-word).
1971 |//
1972 |.macro vm_round, name, mode
1973 | // Used temporaries: TMP0, TMP1, TMP2, TMP3.
1974 |->name.._efd: // Input: CARG2, output: CRET2
1975 | evmergehi CARG1, CARG2, CARG2
1976 |->name.._hilo:
1977 | // Input: CARG1 (hi), CARG2 (hi, lo), output: CRET2
1978 | rlwinm TMP2, CARG1, 12, 21, 31
1979 | addic. TMP2, TMP2, -1023 // exp = exponent(x) - 1023
1980 | li TMP1, -1
1981 | cmplwi cr1, TMP2, 51 // 0 <= exp <= 51?
1982 | subfic TMP0, TMP2, 52
1983 | bgt cr1, >1
1984 | lus TMP3, 0xfff0
1985 | slw TMP0, TMP1, TMP0 // lomask = -1 << (52-exp)
1986 | sraw TMP1, TMP3, TMP2 // himask = (int32_t)0xfff00000 >> exp
1987 |.if mode == 2 // trunc(x):
1988 | evmergelo TMP0, TMP1, TMP0
1989 | evand CRET2, CARG2, TMP0 // hi &= himask, lo &= lomask
1990 |.else
1991 | andc TMP2, CARG2, TMP0
1992 | andc TMP3, CARG1, TMP1
1993 | or TMP2, TMP2, TMP3 // ztest = (hi&~himask) | (lo&~lomask)
1994 | srawi TMP3, CARG1, 31 // signmask = (int32_t)hi >> 31
1995 |.if mode == 0 // floor(x):
1996 | and. TMP2, TMP2, TMP3 // iszero = ((ztest & signmask) == 0)
1997 |.else // ceil(x):
1998 | andc. TMP2, TMP2, TMP3 // iszero = ((ztest & ~signmask) == 0)
1999 |.endif
2000 | and CARG2, CARG2, TMP0 // lo &= lomask
2001 | and CARG1, CARG1, TMP1 // hi &= himask
2002 | subc TMP0, CARG2, TMP0
2003 | iseleq TMP0, CARG2, TMP0 // lo = iszero ? lo : lo-lomask
2004 | sube TMP1, CARG1, TMP1
2005 | iseleq TMP1, CARG1, TMP1 // hi = iszero ? hi : hi-himask+carry
2006 | evmergelo CRET2, TMP1, TMP0
2007 |.endif
2008 | blr
2009 |1:
2010 | bgtlr // Already done if >=2^52, +-inf or nan.
2011 |.if mode == 2 // trunc(x):
2012 | rlwinm TMP1, CARG1, 0, 0, 0 // hi = sign(x)
2013 | li TMP0, 0
2014 | evmergelo CRET2, TMP1, TMP0
2015 |.else
2016 | rlwinm TMP2, CARG1, 0, 1, 31
2017 | srawi TMP0, CARG1, 31 // signmask = (int32_t)hi >> 31
2018 | or TMP2, TMP2, CARG2 // ztest = abs(hi) | lo
2019 | lus TMP1, 0x3ff0
2020 |.if mode == 0 // floor(x):
2021 | and. TMP2, TMP2, TMP0 // iszero = ((ztest & signmask) == 0)
2022 |.else // ceil(x):
2023 | andc. TMP2, TMP2, TMP0 // iszero = ((ztest & ~signmask) == 0)
2024 |.endif
2025 | li TMP0, 0
2026 | iseleq TMP1, r0, TMP1
2027 | rlwimi CARG1, TMP1, 0, 1, 31 // hi = sign(x) | (iszero ? 0.0 : 1.0)
2028 | evmergelo CRET2, CARG1, TMP0
2029 |.endif
2030 | blr
2031 |.endmacro
2032 |
2033 |->vm_floor:
2034 | mflr CARG3
2035 | evmergelo CARG2, CARG1, CARG2
2036 | bl ->vm_floor_hilo
2037 | mtlr CARG3
2038 | evmergehi CRET1, CRET2, CRET2
2039 | blr
2040 |
2041 | vm_round vm_floor, 0
2042 | vm_round vm_ceil, 1
2043 |.if JIT
2044 | vm_round vm_trunc, 2
2045 |.else
2046 |->vm_trunc_efd:
2047 |->vm_trunc_hilo:
2048 |.endif
2049 |
2050 |//-----------------------------------------------------------------------
2051 |//-- Miscellaneous functions --------------------------------------------
2052 |//-----------------------------------------------------------------------
2053 |
2054 |//-----------------------------------------------------------------------
2055 |//-- FFI helper functions -----------------------------------------------
2056 |//-----------------------------------------------------------------------
2057 |
2058 |->vm_ffi_call:
2059 |.if FFI
2060 | NYI
2061 |.endif
2062 |
2063 |//-----------------------------------------------------------------------
2064}
2065
2066/* Generate the code for a single instruction. */
2067static void build_ins(BuildCtx *ctx, BCOp op, int defop)
2068{
2069 int vk = 0;
2070 |=>defop:
2071
2072 switch (op) {
2073
2074 /* -- Comparison ops ---------------------------------------------------- */
2075
2076 /* Remember: all ops branch for a true comparison, fall through otherwise. */
2077
2078 case BC_ISLT: case BC_ISGE: case BC_ISLE: case BC_ISGT:
2079 | // RA = src1*8, RD = src2*8, JMP with RD = target
2080 | evlddx TMP0, BASE, RA
2081 | addi PC, PC, 4
2082 | evlddx TMP1, BASE, RD
2083 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2084 | lwz TMP2, -4(PC)
2085 | evmergehi RB, TMP0, TMP1
2086 | decode_RD4 TMP2, TMP2
2087 | checknum RB
2088 | add TMP2, TMP2, TMP3
2089 | checkanyfail ->vmeta_comp
2090 | efdcmplt TMP0, TMP1
2091 if (op == BC_ISLE || op == BC_ISGT) {
2092 | efdcmpeq cr1, TMP0, TMP1
2093 | cror 4*cr0+gt, 4*cr0+gt, 4*cr1+gt
2094 }
2095 if (op == BC_ISLT || op == BC_ISLE) {
2096 | iselgt PC, TMP2, PC
2097 } else {
2098 | iselgt PC, PC, TMP2
2099 }
2100 | ins_next
2101 break;
2102
2103 case BC_ISEQV: case BC_ISNEV:
2104 vk = op == BC_ISEQV;
2105 | // RA = src1*8, RD = src2*8, JMP with RD = target
2106 | evlddx CARG2, BASE, RA
2107 | addi PC, PC, 4
2108 | evlddx CARG3, BASE, RD
2109 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2110 | lwz TMP2, -4(PC)
2111 | evmergehi RB, CARG2, CARG3
2112 | decode_RD4 TMP2, TMP2
2113 | checknum RB
2114 | add TMP2, TMP2, TMP3
2115 | checkanyfail >5
2116 | efdcmpeq CARG2, CARG3
2117 if (vk) {
2118 | iselgt PC, TMP2, PC
2119 } else {
2120 | iselgt PC, PC, TMP2
2121 }
2122 |1:
2123 | ins_next
2124 |
2125 |5: // Either or both types are not numbers.
2126 | evcmpeq CARG2, CARG3
2127 | not TMP3, RB
2128 | cmplwi cr1, TMP3, ~LJ_TISPRI // Primitive?
2129 | crorc 4*cr7+lt, 4*cr0+so, 4*cr0+lt // 1: Same tv or different type.
2130 | cmplwi cr6, TMP3, ~LJ_TISTABUD // Table or userdata?
2131 | crandc 4*cr7+gt, 4*cr0+lt, 4*cr1+gt // 2: Same type and primitive.
2132 | mr SAVE0, PC
2133 if (vk) {
2134 | isel PC, TMP2, PC, 4*cr7+gt
2135 } else {
2136 | isel TMP2, PC, TMP2, 4*cr7+gt
2137 }
2138 | cror 4*cr7+lt, 4*cr7+lt, 4*cr7+gt // 1 or 2.
2139 if (vk) {
2140 | isel PC, TMP2, PC, 4*cr0+so
2141 } else {
2142 | isel PC, PC, TMP2, 4*cr0+so
2143 }
2144 | blt cr7, <1 // Done if 1 or 2.
2145 | blt cr6, <1 // Done if not tab/ud.
2146 |
2147 | // Different tables or userdatas. Need to check __eq metamethod.
2148 | // Field metatable must be at same offset for GCtab and GCudata!
2149 | lwz TAB:TMP2, TAB:CARG2->metatable
2150 | li CARG4, 1-vk // ne = 0 or 1.
2151 | cmplwi TAB:TMP2, 0
2152 | beq <1 // No metatable?
2153 | lbz TMP2, TAB:TMP2->nomm
2154 | andi. TMP2, TMP2, 1<<MM_eq
2155 | bne <1 // Or 'no __eq' flag set?
2156 | mr PC, SAVE0 // Restore old PC.
2157 | b ->vmeta_equal // Handle __eq metamethod.
2158 break;
2159
2160 case BC_ISEQS: case BC_ISNES:
2161 vk = op == BC_ISEQS;
2162 | // RA = src*8, RD = str_const*8 (~), JMP with RD = target
2163 | evlddx TMP0, BASE, RA
2164 | srwi RD, RD, 1
2165 | lwz INS, 0(PC)
2166 | subfic RD, RD, -4
2167 | addi PC, PC, 4
2168 | lwzx STR:TMP1, KBASE, RD // KBASE-4-str_const*4
2169 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2170 | decode_RD4 TMP2, INS
2171 | evmergelo STR:TMP1, TISSTR, STR:TMP1
2172 | add TMP2, TMP2, TMP3
2173 | evcmpeq TMP0, STR:TMP1
2174 if (vk) {
2175 | isel PC, TMP2, PC, 4*cr0+so
2176 } else {
2177 | isel PC, PC, TMP2, 4*cr0+so
2178 }
2179 | ins_next
2180 break;
2181
2182 case BC_ISEQN: case BC_ISNEN:
2183 vk = op == BC_ISEQN;
2184 | // RA = src*8, RD = num_const*8, JMP with RD = target
2185 | evlddx TMP0, BASE, RA
2186 | addi PC, PC, 4
2187 | evlddx TMP1, KBASE, RD
2188 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2189 | lwz INS, -4(PC)
2190 | checknum TMP0
2191 | checkfail >5
2192 | efdcmpeq TMP0, TMP1
2193 |1:
2194 | decode_RD4 TMP2, INS
2195 | add TMP2, TMP2, TMP3
2196 if (vk) {
2197 | iselgt PC, TMP2, PC
2198 |5:
2199 } else {
2200 | iselgt PC, PC, TMP2
2201 }
2202 |3:
2203 | ins_next
2204 if (!vk) {
2205 |5:
2206 | decode_RD4 TMP2, INS
2207 | add PC, TMP2, TMP3
2208 | b <3
2209 }
2210 break;
2211
2212 case BC_ISEQP: case BC_ISNEP:
2213 vk = op == BC_ISEQP;
2214 | // RA = src*8, RD = primitive_type*8 (~), JMP with RD = target
2215 | lwzx TMP0, BASE, RA
2216 | srwi TMP1, RD, 3
2217 | lwz INS, 0(PC)
2218 | addi PC, PC, 4
2219 | not TMP1, TMP1
2220 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2221 | cmplw TMP0, TMP1
2222 | decode_RD4 TMP2, INS
2223 | add TMP2, TMP2, TMP3
2224 if (vk) {
2225 | iseleq PC, TMP2, PC
2226 } else {
2227 | iseleq PC, PC, TMP2
2228 }
2229 | ins_next
2230 break;
2231
2232 /* -- Unary test and copy ops ------------------------------------------- */
2233
2234 case BC_ISTC: case BC_ISFC: case BC_IST: case BC_ISF:
2235 | // RA = dst*8 or unused, RD = src*8, JMP with RD = target
2236 | evlddx TMP0, BASE, RD
2237 | evaddw TMP1, TISNIL, TISNIL // Synthesize LJ_TFALSE.
2238 | lwz INS, 0(PC)
2239 | evcmpltu TMP0, TMP1
2240 | addi PC, PC, 4
2241 if (op == BC_IST || op == BC_ISF) {
2242 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
2243 | decode_RD4 TMP2, INS
2244 | add TMP2, TMP2, TMP3
2245 if (op == BC_IST) {
2246 | isellt PC, TMP2, PC
2247 } else {
2248 | isellt PC, PC, TMP2
2249 }
2250 } else {
2251 if (op == BC_ISTC) {
2252 | checkfail >1
2253 } else {
2254 | checkok >1
2255 }
2256 | addis PC, PC, -(BCBIAS_J*4 >> 16)
2257 | decode_RD4 TMP2, INS
2258 | evstddx TMP0, BASE, RA
2259 | add PC, PC, TMP2
2260 |1:
2261 }
2262 | ins_next
2263 break;
2264
2265 /* -- Unary ops --------------------------------------------------------- */
2266
2267 case BC_MOV:
2268 | // RA = dst*8, RD = src*8
2269 | ins_next1
2270 | evlddx TMP0, BASE, RD
2271 | evstddx TMP0, BASE, RA
2272 | ins_next2
2273 break;
2274 case BC_NOT:
2275 | // RA = dst*8, RD = src*8
2276 | ins_next1
2277 | lwzx TMP0, BASE, RD
2278 | subfic TMP1, TMP0, LJ_TTRUE
2279 | adde TMP0, TMP0, TMP1
2280 | stwx TMP0, BASE, RA
2281 | ins_next2
2282 break;
2283 case BC_UNM:
2284 | // RA = dst*8, RD = src*8
2285 | evlddx TMP0, BASE, RD
2286 | checknum TMP0
2287 | checkfail ->vmeta_unm
2288 | efdneg TMP0, TMP0
2289 | ins_next1
2290 | evstddx TMP0, BASE, RA
2291 | ins_next2
2292 break;
2293 case BC_LEN:
2294 | // RA = dst*8, RD = src*8
2295 | evlddx CARG1, BASE, RD
2296 | checkstr CARG1
2297 | checkfail >2
2298 | lwz CRET1, STR:CARG1->len
2299 |1:
2300 | ins_next1
2301 | efdcfsi TMP0, CRET1
2302 | evstddx TMP0, BASE, RA
2303 | ins_next2
2304 |2:
2305 | checktab CARG1
2306 | checkfail ->vmeta_len
2307#if LJ_52
2308 | lwz TAB:TMP2, TAB:CARG1->metatable
2309 | cmplwi TAB:TMP2, 0
2310 | bne >9
2311 |3:
2312#endif
2313 |->BC_LEN_Z:
2314 | bl extern lj_tab_len // (GCtab *t)
2315 | // Returns uint32_t (but less than 2^31).
2316 | b <1
2317#if LJ_52
2318 |9:
2319 | lbz TMP0, TAB:TMP2->nomm
2320 | andi. TMP0, TMP0, 1<<MM_len
2321 | bne <3 // 'no __len' flag set: done.
2322 | b ->vmeta_len
2323#endif
2324 break;
2325
2326 /* -- Binary ops -------------------------------------------------------- */
2327
2328 |.macro ins_arithpre, t0, t1
2329 | // RA = dst*8, RB = src1*8, RC = src2*8 | num_const*8
2330 ||vk = ((int)op - BC_ADDVN) / (BC_ADDNV-BC_ADDVN);
2331 ||switch (vk) {
2332 ||case 0:
2333 | evlddx t0, BASE, RB
2334 | checknum t0
2335 | evlddx t1, KBASE, RC
2336 | checkfail ->vmeta_arith_vn
2337 || break;
2338 ||case 1:
2339 | evlddx t1, BASE, RB
2340 | checknum t1
2341 | evlddx t0, KBASE, RC
2342 | checkfail ->vmeta_arith_nv
2343 || break;
2344 ||default:
2345 | evlddx t0, BASE, RB
2346 | evlddx t1, BASE, RC
2347 | evmergehi TMP2, t0, t1
2348 | checknum TMP2
2349 | checkanyfail ->vmeta_arith_vv
2350 || break;
2351 ||}
2352 |.endmacro
2353 |
2354 |.macro ins_arith, ins
2355 | ins_arithpre TMP0, TMP1
2356 | ins_next1
2357 | ins TMP0, TMP0, TMP1
2358 | evstddx TMP0, BASE, RA
2359 | ins_next2
2360 |.endmacro
2361
2362 case BC_ADDVN: case BC_ADDNV: case BC_ADDVV:
2363 | ins_arith efdadd
2364 break;
2365 case BC_SUBVN: case BC_SUBNV: case BC_SUBVV:
2366 | ins_arith efdsub
2367 break;
2368 case BC_MULVN: case BC_MULNV: case BC_MULVV:
2369 | ins_arith efdmul
2370 break;
2371 case BC_DIVVN: case BC_DIVNV: case BC_DIVVV:
2372 | ins_arith efddiv
2373 break;
2374 case BC_MODVN:
2375 | ins_arithpre RD, SAVE0
2376 |->BC_MODVN_Z:
2377 | efddiv CARG2, RD, SAVE0
2378 | bl ->vm_floor_efd // floor(b/c)
2379 | efdmul TMP0, CRET2, SAVE0
2380 | ins_next1
2381 | efdsub TMP0, RD, TMP0 // b - floor(b/c)*c
2382 | evstddx TMP0, BASE, RA
2383 | ins_next2
2384 break;
2385 case BC_MODNV: case BC_MODVV:
2386 | ins_arithpre RD, SAVE0
2387 | b ->BC_MODVN_Z // Avoid 3 copies. It's slow anyway.
2388 break;
2389 case BC_POW:
2390 | evlddx CARG2, BASE, RB
2391 | evlddx CARG4, BASE, RC
2392 | evmergehi CARG1, CARG4, CARG2
2393 | checknum CARG1
2394 | evmergehi CARG3, CARG4, CARG4
2395 | checkanyfail ->vmeta_arith_vv
2396 | bl extern pow@plt
2397 | evmergelo CRET2, CRET1, CRET2
2398 | evstddx CRET2, BASE, RA
2399 | ins_next
2400 break;
2401
2402 case BC_CAT:
2403 | // RA = dst*8, RB = src_start*8, RC = src_end*8
2404 | sub CARG3, RC, RB
2405 | stw BASE, L->base
2406 | add CARG2, BASE, RC
2407 | mr SAVE0, RB
2408 |->BC_CAT_Z:
2409 | stw PC, SAVE_PC
2410 | mr CARG1, L
2411 | srwi CARG3, CARG3, 3
2412 | bl extern lj_meta_cat // (lua_State *L, TValue *top, int left)
2413 | // Returns NULL (finished) or TValue * (metamethod).
2414 | cmplwi CRET1, 0
2415 | lwz BASE, L->base
2416 | bne ->vmeta_binop
2417 | evlddx TMP0, BASE, SAVE0 // Copy result from RB to RA.
2418 | evstddx TMP0, BASE, RA
2419 | ins_next
2420 break;
2421
2422 /* -- Constant ops ------------------------------------------------------ */
2423
2424 case BC_KSTR:
2425 | // RA = dst*8, RD = str_const*8 (~)
2426 | ins_next1
2427 | srwi TMP1, RD, 1
2428 | subfic TMP1, TMP1, -4
2429 | lwzx TMP0, KBASE, TMP1 // KBASE-4-str_const*4
2430 | evmergelo TMP0, TISSTR, TMP0
2431 | evstddx TMP0, BASE, RA
2432 | ins_next2
2433 break;
2434 case BC_KCDATA:
2435 |.if FFI
2436 | // RA = dst*8, RD = cdata_const*8 (~)
2437 | ins_next1
2438 | srwi TMP1, RD, 1
2439 | subfic TMP1, TMP1, -4
2440 | lwzx TMP0, KBASE, TMP1 // KBASE-4-cdata_const*4
2441 | li TMP2, LJ_TCDATA
2442 | evmergelo TMP0, TMP2, TMP0
2443 | evstddx TMP0, BASE, RA
2444 | ins_next2
2445 |.endif
2446 break;
2447 case BC_KSHORT:
2448 | // RA = dst*8, RD = int16_literal*8
2449 | srwi TMP1, RD, 3
2450 | extsh TMP1, TMP1
2451 | ins_next1
2452 | efdcfsi TMP0, TMP1
2453 | evstddx TMP0, BASE, RA
2454 | ins_next2
2455 break;
2456 case BC_KNUM:
2457 | // RA = dst*8, RD = num_const*8
2458 | evlddx TMP0, KBASE, RD
2459 | ins_next1
2460 | evstddx TMP0, BASE, RA
2461 | ins_next2
2462 break;
2463 case BC_KPRI:
2464 | // RA = dst*8, RD = primitive_type*8 (~)
2465 | srwi TMP1, RD, 3
2466 | not TMP0, TMP1
2467 | ins_next1
2468 | stwx TMP0, BASE, RA
2469 | ins_next2
2470 break;
2471 case BC_KNIL:
2472 | // RA = base*8, RD = end*8
2473 | evstddx TISNIL, BASE, RA
2474 | addi RA, RA, 8
2475 |1:
2476 | evstddx TISNIL, BASE, RA
2477 | cmpw RA, RD
2478 | addi RA, RA, 8
2479 | blt <1
2480 | ins_next_
2481 break;
2482
2483 /* -- Upvalue and function ops ------------------------------------------ */
2484
2485 case BC_UGET:
2486 | // RA = dst*8, RD = uvnum*8
2487 | ins_next1
2488 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2489 | srwi RD, RD, 1
2490 | addi RD, RD, offsetof(GCfuncL, uvptr)
2491 | lwzx UPVAL:RB, LFUNC:RB, RD
2492 | lwz TMP1, UPVAL:RB->v
2493 | evldd TMP0, 0(TMP1)
2494 | evstddx TMP0, BASE, RA
2495 | ins_next2
2496 break;
2497 case BC_USETV:
2498 | // RA = uvnum*8, RD = src*8
2499 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2500 | srwi RA, RA, 1
2501 | addi RA, RA, offsetof(GCfuncL, uvptr)
2502 | evlddx TMP1, BASE, RD
2503 | lwzx UPVAL:RB, LFUNC:RB, RA
2504 | lbz TMP3, UPVAL:RB->marked
2505 | lwz CARG2, UPVAL:RB->v
2506 | andi. TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
2507 | lbz TMP0, UPVAL:RB->closed
2508 | evmergehi TMP2, TMP1, TMP1
2509 | evstdd TMP1, 0(CARG2)
2510 | cmplwi cr1, TMP0, 0
2511 | cror 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
2512 | subi TMP2, TMP2, (LJ_TISNUM+1)
2513 | bne >2 // Upvalue is closed and black?
2514 |1:
2515 | ins_next
2516 |
2517 |2: // Check if new value is collectable.
2518 | cmplwi TMP2, LJ_TISGCV - (LJ_TISNUM+1)
2519 | bge <1 // tvisgcv(v)
2520 | lbz TMP3, GCOBJ:TMP1->gch.marked
2521 | andi. TMP3, TMP3, LJ_GC_WHITES // iswhite(v)
2522 | la CARG1, GG_DISP2G(DISPATCH)
2523 | // Crossed a write barrier. Move the barrier forward.
2524 | beq <1
2525 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
2526 | b <1
2527 break;
2528 case BC_USETS:
2529 | // RA = uvnum*8, RD = str_const*8 (~)
2530 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2531 | srwi TMP1, RD, 1
2532 | srwi RA, RA, 1
2533 | subfic TMP1, TMP1, -4
2534 | addi RA, RA, offsetof(GCfuncL, uvptr)
2535 | lwzx STR:TMP1, KBASE, TMP1 // KBASE-4-str_const*4
2536 | lwzx UPVAL:RB, LFUNC:RB, RA
2537 | evmergelo STR:TMP1, TISSTR, STR:TMP1
2538 | lbz TMP3, UPVAL:RB->marked
2539 | lwz CARG2, UPVAL:RB->v
2540 | andi. TMP3, TMP3, LJ_GC_BLACK // isblack(uv)
2541 | lbz TMP3, STR:TMP1->marked
2542 | lbz TMP2, UPVAL:RB->closed
2543 | evstdd STR:TMP1, 0(CARG2)
2544 | bne >2
2545 |1:
2546 | ins_next
2547 |
2548 |2: // Check if string is white and ensure upvalue is closed.
2549 | andi. TMP3, TMP3, LJ_GC_WHITES // iswhite(str)
2550 | cmplwi cr1, TMP2, 0
2551 | cror 4*cr0+eq, 4*cr0+eq, 4*cr1+eq
2552 | la CARG1, GG_DISP2G(DISPATCH)
2553 | // Crossed a write barrier. Move the barrier forward.
2554 | beq <1
2555 | bl extern lj_gc_barrieruv // (global_State *g, TValue *tv)
2556 | b <1
2557 break;
2558 case BC_USETN:
2559 | // RA = uvnum*8, RD = num_const*8
2560 | ins_next1
2561 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2562 | srwi RA, RA, 1
2563 | addi RA, RA, offsetof(GCfuncL, uvptr)
2564 | evlddx TMP0, KBASE, RD
2565 | lwzx UPVAL:RB, LFUNC:RB, RA
2566 | lwz TMP1, UPVAL:RB->v
2567 | evstdd TMP0, 0(TMP1)
2568 | ins_next2
2569 break;
2570 case BC_USETP:
2571 | // RA = uvnum*8, RD = primitive_type*8 (~)
2572 | ins_next1
2573 | lwz LFUNC:RB, FRAME_FUNC(BASE)
2574 | srwi RA, RA, 1
2575 | addi RA, RA, offsetof(GCfuncL, uvptr)
2576 | srwi TMP0, RD, 3
2577 | lwzx UPVAL:RB, LFUNC:RB, RA
2578 | not TMP0, TMP0
2579 | lwz TMP1, UPVAL:RB->v
2580 | stw TMP0, 0(TMP1)
2581 | ins_next2
2582 break;
2583
2584 case BC_UCLO:
2585 | // RA = level*8, RD = target
2586 | lwz TMP1, L->openupval
2587 | branch_RD // Do this first since RD is not saved.
2588 | stw BASE, L->base
2589 | cmplwi TMP1, 0
2590 | mr CARG1, L
2591 | beq >1
2592 | add CARG2, BASE, RA
2593 | bl extern lj_func_closeuv // (lua_State *L, TValue *level)
2594 | lwz BASE, L->base
2595 |1:
2596 | ins_next
2597 break;
2598
2599 case BC_FNEW:
2600 | // RA = dst*8, RD = proto_const*8 (~) (holding function prototype)
2601 | srwi TMP1, RD, 1
2602 | stw BASE, L->base
2603 | subfic TMP1, TMP1, -4
2604 | stw PC, SAVE_PC
2605 | lwzx CARG2, KBASE, TMP1 // KBASE-4-tab_const*4
2606 | mr CARG1, L
2607 | lwz CARG3, FRAME_FUNC(BASE)
2608 | // (lua_State *L, GCproto *pt, GCfuncL *parent)
2609 | bl extern lj_func_newL_gc
2610 | // Returns GCfuncL *.
2611 | lwz BASE, L->base
2612 | evmergelo LFUNC:CRET1, TISFUNC, LFUNC:CRET1
2613 | evstddx LFUNC:CRET1, BASE, RA
2614 | ins_next
2615 break;
2616
2617 /* -- Table ops --------------------------------------------------------- */
2618
2619 case BC_TNEW:
2620 case BC_TDUP:
2621 | // RA = dst*8, RD = (hbits|asize)*8 | tab_const*8 (~)
2622 | lwz TMP0, DISPATCH_GL(gc.total)(DISPATCH)
2623 | mr CARG1, L
2624 | lwz TMP1, DISPATCH_GL(gc.threshold)(DISPATCH)
2625 | stw BASE, L->base
2626 | cmplw TMP0, TMP1
2627 | stw PC, SAVE_PC
2628 | bge >5
2629 |1:
2630 if (op == BC_TNEW) {
2631 | rlwinm CARG2, RD, 29, 21, 31
2632 | rlwinm CARG3, RD, 18, 27, 31
2633 | cmpwi CARG2, 0x7ff
2634 | li TMP1, 0x801
2635 | iseleq CARG2, TMP1, CARG2
2636 | bl extern lj_tab_new // (lua_State *L, int32_t asize, uint32_t hbits)
2637 | // Returns Table *.
2638 } else {
2639 | srwi TMP1, RD, 1
2640 | subfic TMP1, TMP1, -4
2641 | lwzx CARG2, KBASE, TMP1 // KBASE-4-tab_const*4
2642 | bl extern lj_tab_dup // (lua_State *L, Table *kt)
2643 | // Returns Table *.
2644 }
2645 | lwz BASE, L->base
2646 | evmergelo TAB:CRET1, TISTAB, TAB:CRET1
2647 | evstddx TAB:CRET1, BASE, RA
2648 | ins_next
2649 |5:
2650 | mr SAVE0, RD
2651 | bl extern lj_gc_step_fixtop // (lua_State *L)
2652 | mr RD, SAVE0
2653 | mr CARG1, L
2654 | b <1
2655 break;
2656
2657 case BC_GGET:
2658 | // RA = dst*8, RD = str_const*8 (~)
2659 case BC_GSET:
2660 | // RA = src*8, RD = str_const*8 (~)
2661 | lwz LFUNC:TMP2, FRAME_FUNC(BASE)
2662 | srwi TMP1, RD, 1
2663 | lwz TAB:RB, LFUNC:TMP2->env
2664 | subfic TMP1, TMP1, -4
2665 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
2666 if (op == BC_GGET) {
2667 | b ->BC_TGETS_Z
2668 } else {
2669 | b ->BC_TSETS_Z
2670 }
2671 break;
2672
2673 case BC_TGETV:
2674 | // RA = dst*8, RB = table*8, RC = key*8
2675 | evlddx TAB:RB, BASE, RB
2676 | evlddx RC, BASE, RC
2677 | checktab TAB:RB
2678 | checkfail ->vmeta_tgetv
2679 | checknum RC
2680 | checkfail >5
2681 | // Convert number key to integer
2682 | efdctsi TMP2, RC
2683 | lwz TMP0, TAB:RB->asize
2684 | efdcfsi TMP1, TMP2
2685 | cmplw cr0, TMP0, TMP2
2686 | efdcmpeq cr1, RC, TMP1
2687 | lwz TMP1, TAB:RB->array
2688 | crand 4*cr0+gt, 4*cr0+gt, 4*cr1+gt
2689 | slwi TMP2, TMP2, 3
2690 | ble ->vmeta_tgetv // Integer key and in array part?
2691 | evlddx TMP1, TMP1, TMP2
2692 | checknil TMP1
2693 | checkok >2
2694 |1:
2695 | evstddx TMP1, BASE, RA
2696 | ins_next
2697 |
2698 |2: // Check for __index if table value is nil.
2699 | lwz TAB:TMP2, TAB:RB->metatable
2700 | cmplwi TAB:TMP2, 0
2701 | beq <1 // No metatable: done.
2702 | lbz TMP0, TAB:TMP2->nomm
2703 | andi. TMP0, TMP0, 1<<MM_index
2704 | bne <1 // 'no __index' flag set: done.
2705 | b ->vmeta_tgetv
2706 |
2707 |5:
2708 | checkstr STR:RC // String key?
2709 | checkok ->BC_TGETS_Z
2710 | b ->vmeta_tgetv
2711 break;
2712 case BC_TGETS:
2713 | // RA = dst*8, RB = table*8, RC = str_const*8 (~)
2714 | evlddx TAB:RB, BASE, RB
2715 | srwi TMP1, RC, 1
2716 | checktab TAB:RB
2717 | subfic TMP1, TMP1, -4
2718 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
2719 | checkfail ->vmeta_tgets1
2720 |->BC_TGETS_Z:
2721 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = dst*8
2722 | lwz TMP0, TAB:RB->hmask
2723 | lwz TMP1, STR:RC->hash
2724 | lwz NODE:TMP2, TAB:RB->node
2725 | evmergelo STR:RC, TISSTR, STR:RC
2726 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
2727 | slwi TMP0, TMP1, 5
2728 | slwi TMP1, TMP1, 3
2729 | sub TMP1, TMP0, TMP1
2730 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
2731 |1:
2732 | evldd TMP0, NODE:TMP2->key
2733 | evldd TMP1, NODE:TMP2->val
2734 | evcmpeq TMP0, STR:RC
2735 | checkanyfail >4
2736 | checknil TMP1
2737 | checkok >5 // Key found, but nil value?
2738 |3:
2739 | evstddx TMP1, BASE, RA
2740 | ins_next
2741 |
2742 |4: // Follow hash chain.
2743 | lwz NODE:TMP2, NODE:TMP2->next
2744 | cmplwi NODE:TMP2, 0
2745 | bne <1
2746 | // End of hash chain: key not found, nil result.
2747 | evmr TMP1, TISNIL
2748 |
2749 |5: // Check for __index if table value is nil.
2750 | lwz TAB:TMP2, TAB:RB->metatable
2751 | cmplwi TAB:TMP2, 0
2752 | beq <3 // No metatable: done.
2753 | lbz TMP0, TAB:TMP2->nomm
2754 | andi. TMP0, TMP0, 1<<MM_index
2755 | bne <3 // 'no __index' flag set: done.
2756 | b ->vmeta_tgets
2757 break;
2758 case BC_TGETB:
2759 | // RA = dst*8, RB = table*8, RC = index*8
2760 | evlddx TAB:RB, BASE, RB
2761 | srwi TMP0, RC, 3
2762 | checktab TAB:RB
2763 | checkfail ->vmeta_tgetb
2764 | lwz TMP1, TAB:RB->asize
2765 | lwz TMP2, TAB:RB->array
2766 | cmplw TMP0, TMP1
2767 | bge ->vmeta_tgetb
2768 | evlddx TMP1, TMP2, RC
2769 | checknil TMP1
2770 | checkok >5
2771 |1:
2772 | ins_next1
2773 | evstddx TMP1, BASE, RA
2774 | ins_next2
2775 |
2776 |5: // Check for __index if table value is nil.
2777 | lwz TAB:TMP2, TAB:RB->metatable
2778 | cmplwi TAB:TMP2, 0
2779 | beq <1 // No metatable: done.
2780 | lbz TMP2, TAB:TMP2->nomm
2781 | andi. TMP2, TMP2, 1<<MM_index
2782 | bne <1 // 'no __index' flag set: done.
2783 | b ->vmeta_tgetb // Caveat: preserve TMP0!
2784 break;
2785
2786 case BC_TSETV:
2787 | // RA = src*8, RB = table*8, RC = key*8
2788 | evlddx TAB:RB, BASE, RB
2789 | evlddx RC, BASE, RC
2790 | checktab TAB:RB
2791 | checkfail ->vmeta_tsetv
2792 | checknum RC
2793 | checkfail >5
2794 | // Convert number key to integer
2795 | efdctsi TMP2, RC
2796 | evlddx SAVE0, BASE, RA
2797 | lwz TMP0, TAB:RB->asize
2798 | efdcfsi TMP1, TMP2
2799 | cmplw cr0, TMP0, TMP2
2800 | efdcmpeq cr1, RC, TMP1
2801 | lwz TMP1, TAB:RB->array
2802 | crand 4*cr0+gt, 4*cr0+gt, 4*cr1+gt
2803 | slwi TMP0, TMP2, 3
2804 | ble ->vmeta_tsetv // Integer key and in array part?
2805 | lbz TMP3, TAB:RB->marked
2806 | evlddx TMP2, TMP1, TMP0
2807 | checknil TMP2
2808 | checkok >3
2809 |1:
2810 | andi. TMP2, TMP3, LJ_GC_BLACK // isblack(table)
2811 | evstddx SAVE0, TMP1, TMP0
2812 | bne >7
2813 |2:
2814 | ins_next
2815 |
2816 |3: // Check for __newindex if previous value is nil.
2817 | lwz TAB:TMP2, TAB:RB->metatable
2818 | cmplwi TAB:TMP2, 0
2819 | beq <1 // No metatable: done.
2820 | lbz TMP2, TAB:TMP2->nomm
2821 | andi. TMP2, TMP2, 1<<MM_newindex
2822 | bne <1 // 'no __newindex' flag set: done.
2823 | b ->vmeta_tsetv
2824 |
2825 |5:
2826 | checkstr STR:RC // String key?
2827 | checkok ->BC_TSETS_Z
2828 | b ->vmeta_tsetv
2829 |
2830 |7: // Possible table write barrier for the value. Skip valiswhite check.
2831 | barrierback TAB:RB, TMP3, TMP0
2832 | b <2
2833 break;
2834 case BC_TSETS:
2835 | // RA = src*8, RB = table*8, RC = str_const*8 (~)
2836 | evlddx TAB:RB, BASE, RB
2837 | srwi TMP1, RC, 1
2838 | checktab TAB:RB
2839 | subfic TMP1, TMP1, -4
2840 | lwzx STR:RC, KBASE, TMP1 // KBASE-4-str_const*4
2841 | checkfail ->vmeta_tsets1
2842 |->BC_TSETS_Z:
2843 | // TAB:RB = GCtab *, STR:RC = GCstr *, RA = src*8
2844 | lwz TMP0, TAB:RB->hmask
2845 | lwz TMP1, STR:RC->hash
2846 | lwz NODE:TMP2, TAB:RB->node
2847 | evmergelo STR:RC, TISSTR, STR:RC
2848 | stb ZERO, TAB:RB->nomm // Clear metamethod cache.
2849 | and TMP1, TMP1, TMP0 // idx = str->hash & tab->hmask
2850 | evlddx SAVE0, BASE, RA
2851 | slwi TMP0, TMP1, 5
2852 | slwi TMP1, TMP1, 3
2853 | sub TMP1, TMP0, TMP1
2854 | lbz TMP3, TAB:RB->marked
2855 | add NODE:TMP2, NODE:TMP2, TMP1 // node = tab->node + (idx*32-idx*8)
2856 |1:
2857 | evldd TMP0, NODE:TMP2->key
2858 | evldd TMP1, NODE:TMP2->val
2859 | evcmpeq TMP0, STR:RC
2860 | checkanyfail >5
2861 | checknil TMP1
2862 | checkok >4 // Key found, but nil value?
2863 |2:
2864 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
2865 | evstdd SAVE0, NODE:TMP2->val
2866 | bne >7
2867 |3:
2868 | ins_next
2869 |
2870 |4: // Check for __newindex if previous value is nil.
2871 | lwz TAB:TMP1, TAB:RB->metatable
2872 | cmplwi TAB:TMP1, 0
2873 | beq <2 // No metatable: done.
2874 | lbz TMP0, TAB:TMP1->nomm
2875 | andi. TMP0, TMP0, 1<<MM_newindex
2876 | bne <2 // 'no __newindex' flag set: done.
2877 | b ->vmeta_tsets
2878 |
2879 |5: // Follow hash chain.
2880 | lwz NODE:TMP2, NODE:TMP2->next
2881 | cmplwi NODE:TMP2, 0
2882 | bne <1
2883 | // End of hash chain: key not found, add a new one.
2884 |
2885 | // But check for __newindex first.
2886 | lwz TAB:TMP1, TAB:RB->metatable
2887 | la CARG3, DISPATCH_GL(tmptv)(DISPATCH)
2888 | stw PC, SAVE_PC
2889 | mr CARG1, L
2890 | cmplwi TAB:TMP1, 0
2891 | stw BASE, L->base
2892 | beq >6 // No metatable: continue.
2893 | lbz TMP0, TAB:TMP1->nomm
2894 | andi. TMP0, TMP0, 1<<MM_newindex
2895 | beq ->vmeta_tsets // 'no __newindex' flag NOT set: check.
2896 |6:
2897 | mr CARG2, TAB:RB
2898 | evstdd STR:RC, 0(CARG3)
2899 | bl extern lj_tab_newkey // (lua_State *L, GCtab *t, TValue *k)
2900 | // Returns TValue *.
2901 | lwz BASE, L->base
2902 | evstdd SAVE0, 0(CRET1)
2903 | b <3 // No 2nd write barrier needed.
2904 |
2905 |7: // Possible table write barrier for the value. Skip valiswhite check.
2906 | barrierback TAB:RB, TMP3, TMP0
2907 | b <3
2908 break;
2909 case BC_TSETB:
2910 | // RA = src*8, RB = table*8, RC = index*8
2911 | evlddx TAB:RB, BASE, RB
2912 | srwi TMP0, RC, 3
2913 | checktab TAB:RB
2914 | checkfail ->vmeta_tsetb
2915 | lwz TMP1, TAB:RB->asize
2916 | lwz TMP2, TAB:RB->array
2917 | lbz TMP3, TAB:RB->marked
2918 | cmplw TMP0, TMP1
2919 | evlddx SAVE0, BASE, RA
2920 | bge ->vmeta_tsetb
2921 | evlddx TMP1, TMP2, RC
2922 | checknil TMP1
2923 | checkok >5
2924 |1:
2925 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
2926 | evstddx SAVE0, TMP2, RC
2927 | bne >7
2928 |2:
2929 | ins_next
2930 |
2931 |5: // Check for __newindex if previous value is nil.
2932 | lwz TAB:TMP1, TAB:RB->metatable
2933 | cmplwi TAB:TMP1, 0
2934 | beq <1 // No metatable: done.
2935 | lbz TMP1, TAB:TMP1->nomm
2936 | andi. TMP1, TMP1, 1<<MM_newindex
2937 | bne <1 // 'no __newindex' flag set: done.
2938 | b ->vmeta_tsetb // Caveat: preserve TMP0!
2939 |
2940 |7: // Possible table write barrier for the value. Skip valiswhite check.
2941 | barrierback TAB:RB, TMP3, TMP0
2942 | b <2
2943 break;
2944
2945 case BC_TSETM:
2946 | // RA = base*8 (table at base-1), RD = num_const*8 (start index)
2947 | add RA, BASE, RA
2948 |1:
2949 | add TMP3, KBASE, RD
2950 | lwz TAB:CARG2, -4(RA) // Guaranteed to be a table.
2951 | addic. TMP0, MULTRES, -8
2952 | lwz TMP3, 4(TMP3) // Integer constant is in lo-word.
2953 | srwi CARG3, TMP0, 3
2954 | beq >4 // Nothing to copy?
2955 | add CARG3, CARG3, TMP3
2956 | lwz TMP2, TAB:CARG2->asize
2957 | slwi TMP1, TMP3, 3
2958 | lbz TMP3, TAB:CARG2->marked
2959 | cmplw CARG3, TMP2
2960 | add TMP2, RA, TMP0
2961 | lwz TMP0, TAB:CARG2->array
2962 | bgt >5
2963 | add TMP1, TMP1, TMP0
2964 | andi. TMP0, TMP3, LJ_GC_BLACK // isblack(table)
2965 |3: // Copy result slots to table.
2966 | evldd TMP0, 0(RA)
2967 | addi RA, RA, 8
2968 | cmpw cr1, RA, TMP2
2969 | evstdd TMP0, 0(TMP1)
2970 | addi TMP1, TMP1, 8
2971 | blt cr1, <3
2972 | bne >7
2973 |4:
2974 | ins_next
2975 |
2976 |5: // Need to resize array part.
2977 | stw BASE, L->base
2978 | mr CARG1, L
2979 | stw PC, SAVE_PC
2980 | mr SAVE0, RD
2981 | bl extern lj_tab_reasize // (lua_State *L, GCtab *t, int nasize)
2982 | // Must not reallocate the stack.
2983 | mr RD, SAVE0
2984 | b <1
2985 |
2986 |7: // Possible table write barrier for any value. Skip valiswhite check.
2987 | barrierback TAB:CARG2, TMP3, TMP0
2988 | b <4
2989 break;
2990
2991 /* -- Calls and vararg handling ----------------------------------------- */
2992
2993 case BC_CALLM:
2994 | // RA = base*8, (RB = (nresults+1)*8,) RC = extra_nargs*8
2995 | add NARGS8:RC, NARGS8:RC, MULTRES
2996 | // Fall through. Assumes BC_CALL follows.
2997 break;
2998 case BC_CALL:
2999 | // RA = base*8, (RB = (nresults+1)*8,) RC = (nargs+1)*8
3000 | evlddx LFUNC:RB, BASE, RA
3001 | mr TMP2, BASE
3002 | add BASE, BASE, RA
3003 | subi NARGS8:RC, NARGS8:RC, 8
3004 | checkfunc LFUNC:RB
3005 | addi BASE, BASE, 8
3006 | checkfail ->vmeta_call
3007 | ins_call
3008 break;
3009
3010 case BC_CALLMT:
3011 | // RA = base*8, (RB = 0,) RC = extra_nargs*8
3012 | add NARGS8:RC, NARGS8:RC, MULTRES
3013 | // Fall through. Assumes BC_CALLT follows.
3014 break;
3015 case BC_CALLT:
3016 | // RA = base*8, (RB = 0,) RC = (nargs+1)*8
3017 | evlddx LFUNC:RB, BASE, RA
3018 | add RA, BASE, RA
3019 | lwz TMP1, FRAME_PC(BASE)
3020 | subi NARGS8:RC, NARGS8:RC, 8
3021 | checkfunc LFUNC:RB
3022 | addi RA, RA, 8
3023 | checkfail ->vmeta_callt
3024 |->BC_CALLT_Z:
3025 | andi. TMP0, TMP1, FRAME_TYPE // Caveat: preserve cr0 until the crand.
3026 | lbz TMP3, LFUNC:RB->ffid
3027 | xori TMP2, TMP1, FRAME_VARG
3028 | cmplwi cr1, NARGS8:RC, 0
3029 | bne >7
3030 |1:
3031 | stw LFUNC:RB, FRAME_FUNC(BASE) // Copy function down, but keep PC.
3032 | li TMP2, 0
3033 | cmplwi cr7, TMP3, 1 // (> FF_C) Calling a fast function?
3034 | beq cr1, >3
3035 |2:
3036 | addi TMP3, TMP2, 8
3037 | evlddx TMP0, RA, TMP2
3038 | cmplw cr1, TMP3, NARGS8:RC
3039 | evstddx TMP0, BASE, TMP2
3040 | mr TMP2, TMP3
3041 | bne cr1, <2
3042 |3:
3043 | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+gt
3044 | beq >5
3045 |4:
3046 | ins_callt
3047 |
3048 |5: // Tailcall to a fast function with a Lua frame below.
3049 | lwz INS, -4(TMP1)
3050 | decode_RA8 RA, INS
3051 | sub TMP1, BASE, RA
3052 | lwz LFUNC:TMP1, FRAME_FUNC-8(TMP1)
3053 | lwz TMP1, LFUNC:TMP1->pc
3054 | lwz KBASE, PC2PROTO(k)(TMP1) // Need to prepare KBASE.
3055 | b <4
3056 |
3057 |7: // Tailcall from a vararg function.
3058 | andi. TMP0, TMP2, FRAME_TYPEP
3059 | bne <1 // Vararg frame below?
3060 | sub BASE, BASE, TMP2 // Relocate BASE down.
3061 | lwz TMP1, FRAME_PC(BASE)
3062 | andi. TMP0, TMP1, FRAME_TYPE
3063 | b <1
3064 break;
3065
3066 case BC_ITERC:
3067 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 ((2+1)*8))
3068 | subi RA, RA, 24 // evldd doesn't support neg. offsets.
3069 | mr TMP2, BASE
3070 | evlddx LFUNC:RB, BASE, RA
3071 | add BASE, BASE, RA
3072 | evldd TMP0, 8(BASE)
3073 | evldd TMP1, 16(BASE)
3074 | evstdd LFUNC:RB, 24(BASE) // Copy callable.
3075 | checkfunc LFUNC:RB
3076 | evstdd TMP0, 32(BASE) // Copy state.
3077 | li NARGS8:RC, 16 // Iterators get 2 arguments.
3078 | evstdd TMP1, 40(BASE) // Copy control var.
3079 | addi BASE, BASE, 32
3080 | checkfail ->vmeta_call
3081 | ins_call
3082 break;
3083
3084 case BC_ITERN:
3085 | // RA = base*8, (RB = (nresults+1)*8, RC = (nargs+1)*8 (2+1)*8)
3086 |.if JIT
3087 | // NYI: add hotloop, record BC_ITERN.
3088 |.endif
3089 | add RA, BASE, RA
3090 | lwz TAB:RB, -12(RA)
3091 | lwz RC, -4(RA) // Get index from control var.
3092 | lwz TMP0, TAB:RB->asize
3093 | lwz TMP1, TAB:RB->array
3094 | addi PC, PC, 4
3095 |1: // Traverse array part.
3096 | cmplw RC, TMP0
3097 | slwi TMP3, RC, 3
3098 | bge >5 // Index points after array part?
3099 | evlddx TMP2, TMP1, TMP3
3100 | checknil TMP2
3101 | lwz INS, -4(PC)
3102 | checkok >4
3103 | efdcfsi TMP0, RC
3104 | addi RC, RC, 1
3105 | addis TMP3, PC, -(BCBIAS_J*4 >> 16)
3106 | evstdd TMP2, 8(RA)
3107 | decode_RD4 TMP1, INS
3108 | stw RC, -4(RA) // Update control var.
3109 | add PC, TMP1, TMP3
3110 | evstdd TMP0, 0(RA)
3111 |3:
3112 | ins_next
3113 |
3114 |4: // Skip holes in array part.
3115 | addi RC, RC, 1
3116 | b <1
3117 |
3118 |5: // Traverse hash part.
3119 | lwz TMP1, TAB:RB->hmask
3120 | sub RC, RC, TMP0
3121 | lwz TMP2, TAB:RB->node
3122 |6:
3123 | cmplw RC, TMP1 // End of iteration? Branch to ITERL+1.
3124 | slwi TMP3, RC, 5
3125 | bgt <3
3126 | slwi RB, RC, 3
3127 | sub TMP3, TMP3, RB
3128 | evlddx RB, TMP2, TMP3
3129 | add NODE:TMP3, TMP2, TMP3
3130 | checknil RB
3131 | lwz INS, -4(PC)
3132 | checkok >7
3133 | evldd TMP3, NODE:TMP3->key
3134 | addis TMP2, PC, -(BCBIAS_J*4 >> 16)
3135 | evstdd RB, 8(RA)
3136 | add RC, RC, TMP0
3137 | decode_RD4 TMP1, INS
3138 | evstdd TMP3, 0(RA)
3139 | addi RC, RC, 1
3140 | add PC, TMP1, TMP2
3141 | stw RC, -4(RA) // Update control var.
3142 | b <3
3143 |
3144 |7: // Skip holes in hash part.
3145 | addi RC, RC, 1
3146 | b <6
3147 break;
3148
3149 case BC_ISNEXT:
3150 | // RA = base*8, RD = target (points to ITERN)
3151 | add RA, BASE, RA
3152 | li TMP2, -24
3153 | evlddx CFUNC:TMP1, RA, TMP2
3154 | lwz TMP2, -16(RA)
3155 | lwz TMP3, -8(RA)
3156 | evmergehi TMP0, CFUNC:TMP1, CFUNC:TMP1
3157 | cmpwi cr0, TMP2, LJ_TTAB
3158 | cmpwi cr1, TMP0, LJ_TFUNC
3159 | cmpwi cr6, TMP3, LJ_TNIL
3160 | bne cr1, >5
3161 | lbz TMP1, CFUNC:TMP1->ffid
3162 | crand 4*cr0+eq, 4*cr0+eq, 4*cr6+eq
3163 | cmpwi cr7, TMP1, FF_next_N
3164 | srwi TMP0, RD, 1
3165 | crand 4*cr0+eq, 4*cr0+eq, 4*cr7+eq
3166 | add TMP3, PC, TMP0
3167 | bne cr0, >5
3168 | lus TMP1, 0xfffe
3169 | ori TMP1, TMP1, 0x7fff
3170 | stw ZERO, -4(RA) // Initialize control var.
3171 | stw TMP1, -8(RA)
3172 | addis PC, TMP3, -(BCBIAS_J*4 >> 16)
3173 |1:
3174 | ins_next
3175 |5: // Despecialize bytecode if any of the checks fail.
3176 | li TMP0, BC_JMP
3177 | li TMP1, BC_ITERC
3178 | stb TMP0, -1(PC)
3179 | addis PC, TMP3, -(BCBIAS_J*4 >> 16)
3180 | stb TMP1, 3(PC)
3181 | b <1
3182 break;
3183
3184 case BC_VARG:
3185 | // RA = base*8, RB = (nresults+1)*8, RC = numparams*8
3186 | lwz TMP0, FRAME_PC(BASE)
3187 | add RC, BASE, RC
3188 | add RA, BASE, RA
3189 | addi RC, RC, FRAME_VARG
3190 | add TMP2, RA, RB
3191 | subi TMP3, BASE, 8 // TMP3 = vtop
3192 | sub RC, RC, TMP0 // RC = vbase
3193 | // Note: RC may now be even _above_ BASE if nargs was < numparams.
3194 | cmplwi cr1, RB, 0
3195 | sub. TMP1, TMP3, RC
3196 | beq cr1, >5 // Copy all varargs?
3197 | subi TMP2, TMP2, 16
3198 | ble >2 // No vararg slots?
3199 |1: // Copy vararg slots to destination slots.
3200 | evldd TMP0, 0(RC)
3201 | addi RC, RC, 8
3202 | evstdd TMP0, 0(RA)
3203 | cmplw RA, TMP2
3204 | cmplw cr1, RC, TMP3
3205 | bge >3 // All destination slots filled?
3206 | addi RA, RA, 8
3207 | blt cr1, <1 // More vararg slots?
3208 |2: // Fill up remainder with nil.
3209 | evstdd TISNIL, 0(RA)
3210 | cmplw RA, TMP2
3211 | addi RA, RA, 8
3212 | blt <2
3213 |3:
3214 | ins_next
3215 |
3216 |5: // Copy all varargs.
3217 | lwz TMP0, L->maxstack
3218 | li MULTRES, 8 // MULTRES = (0+1)*8
3219 | ble <3 // No vararg slots?
3220 | add TMP2, RA, TMP1
3221 | cmplw TMP2, TMP0
3222 | addi MULTRES, TMP1, 8
3223 | bgt >7
3224 |6:
3225 | evldd TMP0, 0(RC)
3226 | addi RC, RC, 8
3227 | evstdd TMP0, 0(RA)
3228 | cmplw RC, TMP3
3229 | addi RA, RA, 8
3230 | blt <6 // More vararg slots?
3231 | b <3
3232 |
3233 |7: // Grow stack for varargs.
3234 | mr CARG1, L
3235 | stw RA, L->top
3236 | sub SAVE0, RC, BASE // Need delta, because BASE may change.
3237 | stw BASE, L->base
3238 | sub RA, RA, BASE
3239 | stw PC, SAVE_PC
3240 | srwi CARG2, TMP1, 3
3241 | bl extern lj_state_growstack // (lua_State *L, int n)
3242 | lwz BASE, L->base
3243 | add RA, BASE, RA
3244 | add RC, BASE, SAVE0
3245 | subi TMP3, BASE, 8
3246 | b <6
3247 break;
3248
3249 /* -- Returns ----------------------------------------------------------- */
3250
3251 case BC_RETM:
3252 | // RA = results*8, RD = extra_nresults*8
3253 | add RD, RD, MULTRES // MULTRES >= 8, so RD >= 8.
3254 | // Fall through. Assumes BC_RET follows.
3255 break;
3256
3257 case BC_RET:
3258 | // RA = results*8, RD = (nresults+1)*8
3259 | lwz PC, FRAME_PC(BASE)
3260 | add RA, BASE, RA
3261 | mr MULTRES, RD
3262 |1:
3263 | andi. TMP0, PC, FRAME_TYPE
3264 | xori TMP1, PC, FRAME_VARG
3265 | bne ->BC_RETV_Z
3266 |
3267 |->BC_RET_Z:
3268 | // BASE = base, RA = resultptr, RD = (nresults+1)*8, PC = return
3269 | lwz INS, -4(PC)
3270 | cmpwi RD, 8
3271 | subi TMP2, BASE, 8
3272 | subi RC, RD, 8
3273 | decode_RB8 RB, INS
3274 | beq >3
3275 | li TMP1, 0
3276 |2:
3277 | addi TMP3, TMP1, 8
3278 | evlddx TMP0, RA, TMP1
3279 | cmpw TMP3, RC
3280 | evstddx TMP0, TMP2, TMP1
3281 | beq >3
3282 | addi TMP1, TMP3, 8
3283 | evlddx TMP0, RA, TMP3
3284 | cmpw TMP1, RC
3285 | evstddx TMP0, TMP2, TMP3
3286 | bne <2
3287 |3:
3288 |5:
3289 | cmplw RB, RD
3290 | decode_RA8 RA, INS
3291 | bgt >6
3292 | sub BASE, TMP2, RA
3293 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
3294 | ins_next1
3295 | lwz TMP1, LFUNC:TMP1->pc
3296 | lwz KBASE, PC2PROTO(k)(TMP1)
3297 | ins_next2
3298 |
3299 |6: // Fill up results with nil.
3300 | subi TMP1, RD, 8
3301 | addi RD, RD, 8
3302 | evstddx TISNIL, TMP2, TMP1
3303 | b <5
3304 |
3305 |->BC_RETV_Z: // Non-standard return case.
3306 | andi. TMP2, TMP1, FRAME_TYPEP
3307 | bne ->vm_return
3308 | // Return from vararg function: relocate BASE down.
3309 | sub BASE, BASE, TMP1
3310 | lwz PC, FRAME_PC(BASE)
3311 | b <1
3312 break;
3313
3314 case BC_RET0: case BC_RET1:
3315 | // RA = results*8, RD = (nresults+1)*8
3316 | lwz PC, FRAME_PC(BASE)
3317 | add RA, BASE, RA
3318 | mr MULTRES, RD
3319 | andi. TMP0, PC, FRAME_TYPE
3320 | xori TMP1, PC, FRAME_VARG
3321 | bne ->BC_RETV_Z
3322 |
3323 | lwz INS, -4(PC)
3324 | subi TMP2, BASE, 8
3325 | decode_RB8 RB, INS
3326 if (op == BC_RET1) {
3327 | evldd TMP0, 0(RA)
3328 | evstdd TMP0, 0(TMP2)
3329 }
3330 |5:
3331 | cmplw RB, RD
3332 | decode_RA8 RA, INS
3333 | bgt >6
3334 | sub BASE, TMP2, RA
3335 | lwz LFUNC:TMP1, FRAME_FUNC(BASE)
3336 | ins_next1
3337 | lwz TMP1, LFUNC:TMP1->pc
3338 | lwz KBASE, PC2PROTO(k)(TMP1)
3339 | ins_next2
3340 |
3341 |6: // Fill up results with nil.
3342 | subi TMP1, RD, 8
3343 | addi RD, RD, 8
3344 | evstddx TISNIL, TMP2, TMP1
3345 | b <5
3346 break;
3347
3348 /* -- Loops and branches ------------------------------------------------ */
3349
3350 case BC_FORL:
3351 |.if JIT
3352 | hotloop
3353 |.endif
3354 | // Fall through. Assumes BC_IFORL follows.
3355 break;
3356
3357 case BC_JFORI:
3358 case BC_JFORL:
3359#if !LJ_HASJIT
3360 break;
3361#endif
3362 case BC_FORI:
3363 case BC_IFORL:
3364 | // RA = base*8, RD = target (after end of loop or start of loop)
3365 vk = (op == BC_IFORL || op == BC_JFORL);
3366 | add RA, BASE, RA
3367 | evldd TMP1, FORL_IDX*8(RA)
3368 | evldd TMP3, FORL_STEP*8(RA)
3369 | evldd TMP2, FORL_STOP*8(RA)
3370 if (!vk) {
3371 | evcmpgtu cr0, TMP1, TISNUM
3372 | evcmpgtu cr7, TMP3, TISNUM
3373 | evcmpgtu cr1, TMP2, TISNUM
3374 | cror 4*cr0+lt, 4*cr0+lt, 4*cr7+lt
3375 | cror 4*cr0+lt, 4*cr0+lt, 4*cr1+lt
3376 | blt ->vmeta_for
3377 }
3378 if (vk) {
3379 | efdadd TMP1, TMP1, TMP3
3380 | evstdd TMP1, FORL_IDX*8(RA)
3381 }
3382 | evcmpgts TMP3, TISNIL
3383 | evstdd TMP1, FORL_EXT*8(RA)
3384 | bge >2
3385 | efdcmpgt TMP1, TMP2
3386 |1:
3387 if (op != BC_JFORL) {
3388 | srwi RD, RD, 1
3389 | add RD, PC, RD
3390 if (op == BC_JFORI) {
3391 | addis PC, RD, -(BCBIAS_J*4 >> 16)
3392 } else {
3393 | addis RD, RD, -(BCBIAS_J*4 >> 16)
3394 }
3395 }
3396 if (op == BC_FORI) {
3397 | iselgt PC, RD, PC
3398 } else if (op == BC_IFORL) {
3399 | iselgt PC, PC, RD
3400 } else {
3401 | ble =>BC_JLOOP
3402 }
3403 | ins_next
3404 |2:
3405 | efdcmpgt TMP2, TMP1
3406 | b <1
3407 break;
3408
3409 case BC_ITERL:
3410 |.if JIT
3411 | hotloop
3412 |.endif
3413 | // Fall through. Assumes BC_IITERL follows.
3414 break;
3415
3416 case BC_JITERL:
3417#if !LJ_HASJIT
3418 break;
3419#endif
3420 case BC_IITERL:
3421 | // RA = base*8, RD = target
3422 | evlddx TMP1, BASE, RA
3423 | subi RA, RA, 8
3424 | checknil TMP1
3425 | checkok >1 // Stop if iterator returned nil.
3426 if (op == BC_JITERL) {
3427 | NYI
3428 } else {
3429 | branch_RD // Otherwise save control var + branch.
3430 | evstddx TMP1, BASE, RA
3431 }
3432 |1:
3433 | ins_next
3434 break;
3435
3436 case BC_LOOP:
3437 | // RA = base*8, RD = target (loop extent)
3438 | // Note: RA/RD is only used by trace recorder to determine scope/extent
3439 | // This opcode does NOT jump, it's only purpose is to detect a hot loop.
3440 |.if JIT
3441 | hotloop
3442 |.endif
3443 | // Fall through. Assumes BC_ILOOP follows.
3444 break;
3445
3446 case BC_ILOOP:
3447 | // RA = base*8, RD = target (loop extent)
3448 | ins_next
3449 break;
3450
3451 case BC_JLOOP:
3452 |.if JIT
3453 | NYI
3454 |.endif
3455 break;
3456
3457 case BC_JMP:
3458 | // RA = base*8 (only used by trace recorder), RD = target
3459 | branch_RD
3460 | ins_next
3461 break;
3462
3463 /* -- Function headers -------------------------------------------------- */
3464
3465 case BC_FUNCF:
3466 |.if JIT
3467 | hotcall
3468 |.endif
3469 case BC_FUNCV: /* NYI: compiled vararg functions. */
3470 | // Fall through. Assumes BC_IFUNCF/BC_IFUNCV follow.
3471 break;
3472
3473 case BC_JFUNCF:
3474#if !LJ_HASJIT
3475 break;
3476#endif
3477 case BC_IFUNCF:
3478 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
3479 | lwz TMP2, L->maxstack
3480 | lbz TMP1, -4+PC2PROTO(numparams)(PC)
3481 | lwz KBASE, -4+PC2PROTO(k)(PC)
3482 | cmplw RA, TMP2
3483 | slwi TMP1, TMP1, 3
3484 | bgt ->vm_growstack_l
3485 | ins_next1
3486 |2:
3487 | cmplw NARGS8:RC, TMP1 // Check for missing parameters.
3488 | ble >3
3489 if (op == BC_JFUNCF) {
3490 | NYI
3491 } else {
3492 | ins_next2
3493 }
3494 |
3495 |3: // Clear missing parameters.
3496 | evstddx TISNIL, BASE, NARGS8:RC
3497 | addi NARGS8:RC, NARGS8:RC, 8
3498 | b <2
3499 break;
3500
3501 case BC_JFUNCV:
3502#if !LJ_HASJIT
3503 break;
3504#endif
3505 | NYI // NYI: compiled vararg functions
3506 break; /* NYI: compiled vararg functions. */
3507
3508 case BC_IFUNCV:
3509 | // BASE = new base, RA = BASE+framesize*8, RB = LFUNC, RC = nargs*8
3510 | lwz TMP2, L->maxstack
3511 | add TMP1, BASE, RC
3512 | add TMP0, RA, RC
3513 | stw LFUNC:RB, 4(TMP1) // Store copy of LFUNC.
3514 | addi TMP3, RC, 8+FRAME_VARG
3515 | lwz KBASE, -4+PC2PROTO(k)(PC)
3516 | cmplw TMP0, TMP2
3517 | stw TMP3, 0(TMP1) // Store delta + FRAME_VARG.
3518 | bge ->vm_growstack_l
3519 | lbz TMP2, -4+PC2PROTO(numparams)(PC)
3520 | mr RA, BASE
3521 | mr RC, TMP1
3522 | ins_next1
3523 | cmpwi TMP2, 0
3524 | addi BASE, TMP1, 8
3525 | beq >3
3526 |1:
3527 | cmplw RA, RC // Less args than parameters?
3528 | evldd TMP0, 0(RA)
3529 | bge >4
3530 | evstdd TISNIL, 0(RA) // Clear old fixarg slot (help the GC).
3531 | addi RA, RA, 8
3532 |2:
3533 | addic. TMP2, TMP2, -1
3534 | evstdd TMP0, 8(TMP1)
3535 | addi TMP1, TMP1, 8
3536 | bne <1
3537 |3:
3538 | ins_next2
3539 |
3540 |4: // Clear missing parameters.
3541 | evmr TMP0, TISNIL
3542 | b <2
3543 break;
3544
3545 case BC_FUNCC:
3546 case BC_FUNCCW:
3547 | // BASE = new base, RA = BASE+framesize*8, RB = CFUNC, RC = nargs*8
3548 if (op == BC_FUNCC) {
3549 | lwz TMP3, CFUNC:RB->f
3550 } else {
3551 | lwz TMP3, DISPATCH_GL(wrapf)(DISPATCH)
3552 }
3553 | add TMP1, RA, NARGS8:RC
3554 | lwz TMP2, L->maxstack
3555 | add RC, BASE, NARGS8:RC
3556 | stw BASE, L->base
3557 | cmplw TMP1, TMP2
3558 | stw RC, L->top
3559 | li_vmstate C
3560 | mtctr TMP3
3561 if (op == BC_FUNCCW) {
3562 | lwz CARG2, CFUNC:RB->f
3563 }
3564 | mr CARG1, L
3565 | bgt ->vm_growstack_c // Need to grow stack.
3566 | st_vmstate
3567 | bctrl // (lua_State *L [, lua_CFunction f])
3568 | // Returns nresults.
3569 | lwz TMP1, L->top
3570 | slwi RD, CRET1, 3
3571 | lwz BASE, L->base
3572 | li_vmstate INTERP
3573 | lwz PC, FRAME_PC(BASE) // Fetch PC of caller.
3574 | sub RA, TMP1, RD // RA = L->top - nresults*8
3575 | st_vmstate
3576 | b ->vm_returnc
3577 break;
3578
3579 /* ---------------------------------------------------------------------- */
3580
3581 default:
3582 fprintf(stderr, "Error: undefined opcode BC_%s\n", bc_names[op]);
3583 exit(2);
3584 break;
3585 }
3586}
3587
3588static int build_backend(BuildCtx *ctx)
3589{
3590 int op;
3591
3592 dasm_growpc(Dst, BC__MAX);
3593
3594 build_subroutines(ctx);
3595
3596 |.code_op
3597 for (op = 0; op < BC__MAX; op++)
3598 build_ins(ctx, (BCOp)op, op);
3599
3600 return BC__MAX;
3601}
3602
3603/* Emit pseudo frame-info for all assembler functions. */
3604static void emit_asm_debug(BuildCtx *ctx)
3605{
3606 int i;
3607 switch (ctx->mode) {
3608 case BUILD_elfasm:
3609 fprintf(ctx->fp, "\t.section .debug_frame,\"\",@progbits\n");
3610 fprintf(ctx->fp,
3611 ".Lframe0:\n"
3612 "\t.long .LECIE0-.LSCIE0\n"
3613 ".LSCIE0:\n"
3614 "\t.long 0xffffffff\n"
3615 "\t.byte 0x1\n"
3616 "\t.string \"\"\n"
3617 "\t.uleb128 0x1\n"
3618 "\t.sleb128 -4\n"
3619 "\t.byte 65\n"
3620 "\t.byte 0xc\n\t.uleb128 1\n\t.uleb128 0\n"
3621 "\t.align 2\n"
3622 ".LECIE0:\n\n");
3623 fprintf(ctx->fp,
3624 ".LSFDE0:\n"
3625 "\t.long .LEFDE0-.LASFDE0\n"
3626 ".LASFDE0:\n"
3627 "\t.long .Lframe0\n"
3628 "\t.long .Lbegin\n"
3629 "\t.long %d\n"
3630 "\t.byte 0xe\n\t.uleb128 %d\n"
3631 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
3632 "\t.byte 0x5\n\t.uleb128 70\n\t.sleb128 37\n",
3633 (int)ctx->codesz, CFRAME_SIZE);
3634 for (i = 14; i <= 31; i++)
3635 fprintf(ctx->fp,
3636 "\t.byte %d\n\t.uleb128 %d\n"
3637 "\t.byte 5\n\t.uleb128 %d\n\t.uleb128 %d\n",
3638 0x80+i, 1+2*(31-i), 1200+i, 2+2*(31-i));
3639 fprintf(ctx->fp,
3640 "\t.align 2\n"
3641 ".LEFDE0:\n\n");
3642 fprintf(ctx->fp, "\t.section .eh_frame,\"a\",@progbits\n");
3643 fprintf(ctx->fp,
3644 ".Lframe1:\n"
3645 "\t.long .LECIE1-.LSCIE1\n"
3646 ".LSCIE1:\n"
3647 "\t.long 0\n"
3648 "\t.byte 0x1\n"
3649 "\t.string \"zPR\"\n"
3650 "\t.uleb128 0x1\n"
3651 "\t.sleb128 -4\n"
3652 "\t.byte 65\n"
3653 "\t.uleb128 6\n" /* augmentation length */
3654 "\t.byte 0x1b\n" /* pcrel|sdata4 */
3655 "\t.long lj_err_unwind_dwarf-.\n"
3656 "\t.byte 0x1b\n" /* pcrel|sdata4 */
3657 "\t.byte 0xc\n\t.uleb128 1\n\t.uleb128 0\n"
3658 "\t.align 2\n"
3659 ".LECIE1:\n\n");
3660 fprintf(ctx->fp,
3661 ".LSFDE1:\n"
3662 "\t.long .LEFDE1-.LASFDE1\n"
3663 ".LASFDE1:\n"
3664 "\t.long .LASFDE1-.Lframe1\n"
3665 "\t.long .Lbegin-.\n"
3666 "\t.long %d\n"
3667 "\t.uleb128 0\n" /* augmentation length */
3668 "\t.byte 0xe\n\t.uleb128 %d\n"
3669 "\t.byte 0x11\n\t.uleb128 65\n\t.sleb128 -1\n"
3670 "\t.byte 0x5\n\t.uleb128 70\n\t.sleb128 37\n",
3671 (int)ctx->codesz, CFRAME_SIZE);
3672 for (i = 14; i <= 31; i++)
3673 fprintf(ctx->fp,
3674 "\t.byte %d\n\t.uleb128 %d\n"
3675 "\t.byte 5\n\t.uleb128 %d\n\t.uleb128 %d\n",
3676 0x80+i, 1+2*(31-i), 1200+i, 2+2*(31-i));
3677 fprintf(ctx->fp,
3678 "\t.align 2\n"
3679 ".LEFDE1:\n\n");
3680 break;
3681 default:
3682 break;
3683 }
3684}
3685
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2026-01-13 10:42:36 +0000