* [Qemu-devel] More PPC emulation fixes
@ 2005-03-02 23:04 J. Mayer
2005-03-03 11:49 ` J. Mayer
0 siblings, 1 reply; 10+ messages in thread
From: J. Mayer @ 2005-03-02 23:04 UTC (permalink / raw)
To: qemu-devel
[-- Attachment #1: Type: text/plain, Size: 2639 bytes --]
Here's a new patch that fixes the last (!) bug in PPC fixed point
operations and most FPU bugs.
Apply and enjoy.
More infos about this patch:
I did test exhaustively all arithmetic, logical and compare operations
for fixed point and floating points computation with large ranges of
operands then compared the result with the ones I got from a real G3.
The bugs I found which are fixed with this patch are:
- srawi did set the overflow bit in XER register
in some case when used with a 0 shift.
Then, I now do no computation if shift is 0.
- I did forget to generate FP_UNAVAILABLE
exception when trying to use lfd, lfs, stfd
and stfs instructions when FPU is not
enabled. This lead to random
crashes or bad computations under Linux.
- operands for fsel where interverted.
- fres major opcode was false.
- conversion from float to int has been
improved.
- floating points operations applied on single
operands where false: PPC specifies that
the cast to single should be done only on
the computation result. Then I removed all
micro-ops for single float operands and
now just use frsp after computation.
- multiply-and-add/sub-and-negate
operations have been fixed for special
operands (NaN, infinite, ...)
- estimated single operation (ie fres, frsqrte)
have been fixed for special operands
- absolute values computation have been
fixed.
- I also removed unusefull tb_flush &
tlb_flush which are not needed anymore
now that tb flags are set properly.
There still are some things to improve:
- some computations on single operands
may have a difference on the last
significant bit of the mantissa. This only
appears when a multiplication is involved,
ie for fmuls, fmadds, fmsubs, fnmadds and
fnmsubs. All other operations give exactly
the same result as a real G3. As PPC specs
talks about multiply computation with 105
bits mantissa intermediary results, I may
have to use long doubles to fix this issue
(not tried). But I really think the current
results may not hurt anyone.
- No flags are managed for floating-point
computations, except for compare
instructions. This does not seem to be
important for the glibc but is a potential
problem for other OSes.
With this patch, Debian Sarge is now able to install. It need OHW
patches to reboot and go on running the installer. I'll released a
second 0.4 preview with fixes for this (and more...) soon.
However, one need to run "expert" installation, or the installer will go
into an infinite loop trying to probe the link on the NE2k device.
--
J. Mayer <l_indien@magic.fr>
Never organized
[-- Attachment #2: ppc_fixes.diff --]
[-- Type: text/x-patch, Size: 25153 bytes --]
Index: target-ppc/exec.h
===================================================================
RCS file: /cvsroot/qemu/qemu/target-ppc/exec.h,v
retrieving revision 1.9
diff -u -d -w -B -b -d -p -r1.9 exec.h
--- target-ppc/exec.h 3 Jan 2005 23:42:39 -0000 1.9
+++ target-ppc/exec.h 2 Mar 2005 21:51:07 -0000
@@ -32,9 +32,6 @@ register uint32_t T2 asm(AREG3);
#define FT0 (env->ft0)
#define FT1 (env->ft1)
#define FT2 (env->ft2)
-#define FTS0 ((float)env->ft0)
-#define FTS1 ((float)env->ft1)
-#define FTS2 ((float)env->ft2)
#if defined (DEBUG_OP)
#define RETURN() __asm__ __volatile__("nop");
@@ -137,17 +134,12 @@ void do_fctiw (void);
void do_fctiwz (void);
void do_fnmadd (void);
void do_fnmsub (void);
-void do_fnmadds (void);
-void do_fnmsubs (void);
void do_fsqrt (void);
-void do_fsqrts (void);
void do_fres (void);
-void do_fsqrte (void);
+void do_frsqrte (void);
void do_fsel (void);
void do_fcmpu (void);
void do_fcmpo (void);
-void do_fabs (void);
-void do_fnabs (void);
void do_check_reservation (void);
void do_icbi (void);
Index: target-ppc/helper.c
===================================================================
RCS file: /cvsroot/qemu/qemu/target-ppc/helper.c,v
retrieving revision 1.21
diff -u -d -w -B -b -d -p -r1.21 helper.c
--- target-ppc/helper.c 15 Feb 2005 23:06:19 -0000 1.21
+++ target-ppc/helper.c 2 Mar 2005 21:51:07 -0000
@@ -809,6 +809,7 @@ void do_interrupt (CPUState *env)
msr |= 0x00010000;
goto store_current;
case EXCP_NO_FP:
+ msr &= ~0xFFFF0000;
goto store_current;
case EXCP_DECR:
if (msr_ee == 0) {
@@ -854,7 +855,6 @@ void do_interrupt (CPUState *env)
return;
case EXCP_RFI:
/* Restore user-mode state */
- tb_flush(env);
#if defined (DEBUG_EXCEPTIONS)
if (msr_pr == 1)
printf("Return from exception => 0x%08x\n", (uint32_t)env->nip);
@@ -887,7 +888,6 @@ void do_interrupt (CPUState *env)
env->nip = excp << 8;
env->exception_index = EXCP_NONE;
/* Invalidate all TLB as we may have changed translation mode */
- tlb_flush(env, 1);
/* ensure that no TB jump will be modified as
the program flow was changed */
#ifdef __sparc__
Index: target-ppc/op.c
===================================================================
RCS file: /cvsroot/qemu/qemu/target-ppc/op.c,v
retrieving revision 1.15
diff -u -d -w -B -b -d -p -r1.15 op.c
--- target-ppc/op.c 15 Feb 2005 23:06:19 -0000 1.15
+++ target-ppc/op.c 2 Mar 2005 21:51:07 -0000
@@ -32,10 +32,6 @@
#define FT1 (env->ft1)
#define FT2 (env->ft2)
-#define FTS0 ((float)env->ft0)
-#define FTS1 ((float)env->ft1)
-#define FTS2 ((float)env->ft2)
-
#define PPC_OP(name) void glue(op_, name)(void)
#define REG 0
@@ -1204,13 +1200,6 @@ PPC_OP(fadd)
RETURN();
}
-/* fadds - fadds. */
-PPC_OP(fadds)
-{
- FT0 = FTS0 + FTS1;
- RETURN();
-}
-
/* fsub - fsub. */
PPC_OP(fsub)
{
@@ -1218,13 +1207,6 @@ PPC_OP(fsub)
RETURN();
}
-/* fsubs - fsubs. */
-PPC_OP(fsubs)
-{
- FT0 = FTS0 - FTS1;
- RETURN();
-}
-
/* fmul - fmul. */
PPC_OP(fmul)
{
@@ -1232,24 +1214,11 @@ PPC_OP(fmul)
RETURN();
}
-/* fmuls - fmuls. */
-PPC_OP(fmuls)
-{
- FT0 = FTS0 * FTS1;
- RETURN();
-}
-
/* fdiv - fdiv. */
+void do_fdiv (void);
PPC_OP(fdiv)
{
- FT0 /= FT1;
- RETURN();
-}
-
-/* fdivs - fdivs. */
-PPC_OP(fdivs)
-{
- FT0 = FTS0 / FTS1;
+ do_fdiv();
RETURN();
}
@@ -1260,13 +1229,6 @@ PPC_OP(fsqrt)
RETURN();
}
-/* fsqrts - fsqrts. */
-PPC_OP(fsqrts)
-{
- do_fsqrts();
- RETURN();
-}
-
/* fres - fres. */
PPC_OP(fres)
{
@@ -1277,7 +1239,7 @@ PPC_OP(fres)
/* frsqrte - frsqrte. */
PPC_OP(frsqrte)
{
- do_fsqrte();
+ do_frsqrte();
RETURN();
}
@@ -1296,13 +1258,6 @@ PPC_OP(fmadd)
RETURN();
}
-/* fmadds - fmadds. */
-PPC_OP(fmadds)
-{
- FT0 = (FTS0 * FTS1) + FTS2;
- RETURN();
-}
-
/* fmsub - fmsub. */
PPC_OP(fmsub)
{
@@ -1310,13 +1265,6 @@ PPC_OP(fmsub)
RETURN();
}
-/* fmsubs - fmsubs. */
-PPC_OP(fmsubs)
-{
- FT0 = (FTS0 * FTS1) - FTS2;
- RETURN();
-}
-
/* fnmadd - fnmadd. - fnmadds - fnmadds. */
PPC_OP(fnmadd)
{
@@ -1324,13 +1272,6 @@ PPC_OP(fnmadd)
RETURN();
}
-/* fnmadds - fnmadds. */
-PPC_OP(fnmadds)
-{
- do_fnmadds();
- RETURN();
-}
-
/* fnmsub - fnmsub. */
PPC_OP(fnmsub)
{
@@ -1338,13 +1279,6 @@ PPC_OP(fnmsub)
RETURN();
}
-/* fnmsubs - fnmsubs. */
-PPC_OP(fnmsubs)
-{
- do_fnmsubs();
- RETURN();
-}
-
/*** Floating-Point round & convert ***/
/* frsp - frsp. */
PPC_OP(frsp)
@@ -1385,6 +1319,7 @@ PPC_OP(fcmpo)
/*** Floating-point move ***/
/* fabs */
+void do_fabs (void);
PPC_OP(fabs)
{
do_fabs();
@@ -1392,6 +1327,7 @@ PPC_OP(fabs)
}
/* fnabs */
+void do_fnabs (void);
PPC_OP(fnabs)
{
do_fnabs();
Index: target-ppc/op_helper.c
===================================================================
RCS file: /cvsroot/qemu/qemu/target-ppc/op_helper.c,v
retrieving revision 1.11
diff -u -d -w -B -b -d -p -r1.11 op_helper.c
--- target-ppc/op_helper.c 15 Feb 2005 23:06:19 -0000 1.11
+++ target-ppc/op_helper.c 2 Mar 2005 21:51:07 -0000
@@ -249,16 +249,25 @@ void do_fctiw (void)
union {
double d;
uint64_t i;
- } *p = (void *)&FT1;
+ } p;
- if (FT0 > (double)0x7FFFFFFF)
- p->i = 0x7FFFFFFFULL << 32;
- else if (FT0 < -(double)0x80000000)
- p->i = 0x80000000ULL << 32;
- else
- p->i = 0;
- p->i |= (uint32_t)FT0;
- FT0 = p->d;
+ /* XXX: higher bits are not supposed to be significant.
+ * to make tests easier, return the same as a real PPC 750 (aka G3)
+ */
+ p.d = FT0;
+ if (FT0 > (double)0x7FFFFFFF) {
+ p.i = 0x7FFFFFFFULL;
+ // env->fpscr |= 1 << 17;
+ } else if (FT0 < -(double)0x80000000) {
+ p.i = 0x80000000ULL;
+ // env->fpscr |= 1 << 17;
+ } else if (p.i & 0x8000000000000000ULL) {
+ p.i = 0x100000000ULL | (uint32_t)FT0;
+ } else {
+ p.i = (uint32_t)FT0;
+ }
+ p.i |= 0xFFF80000ULL << 32;
+ FT0 = p.d;
}
void do_fctiwz (void)
@@ -266,39 +275,50 @@ void do_fctiwz (void)
union {
double d;
uint64_t i;
- } *p = (void *)&FT1;
+ } p;
int cround = fegetround();
+ /* XXX: higher bits are not supposed to be significant.
+ * to make tests easier, return the same as a real PPC 750 (aka G3)
+ */
+ p.d = FT0;
fesetround(FE_TOWARDZERO);
- if (FT0 > (double)0x7FFFFFFF)
- p->i = 0x7FFFFFFFULL << 32;
- else if (FT0 < -(double)0x80000000)
- p->i = 0x80000000ULL << 32;
- else
- p->i = 0;
- p->i |= (uint32_t)FT0;
- FT0 = p->d;
+ if (FT0 > (double)0x7FFFFFFF) {
+ p.i = 0x7FFFFFFFULL;
+ // env->fpscr |= 1 << 17;
+ } else if (FT0 < -(double)0x80000000) {
+ p.i = 0x80000000ULL;
+ // env->fpscr |= 1 << 17;
+ } else if (p.i & 0x8000000000000000ULL) {
+ p.i = 0x100000000ULL | (uint32_t)FT0;
+ } else {
+ p.i = (uint32_t)FT0;
+ }
+ p.i |= 0xFFF80000ULL << 32;
+ FT0 = p.d;
fesetround(cround);
}
void do_fnmadd (void)
{
- FT0 = -((FT0 * FT1) + FT2);
+ FT0 = (FT0 * FT1) + FT2;
+ if (!isnan(FT0))
+ FT0 = -FT0;
}
void do_fnmsub (void)
{
- FT0 = -((FT0 * FT1) - FT2);
-}
-
-void do_fnmadds (void)
-{
- FT0 = -((FTS0 * FTS1) + FTS2);
+ FT0 = (FT0 * FT1) - FT2;
+ if (!isnan(FT0))
+ FT0 = -FT0;
}
-void do_fnmsubs (void)
+void do_fdiv (void)
{
- FT0 = -((FTS0 * FTS1) - FTS2);
+ if (FT0 == -0.0 && FT1 == -0.0)
+ FT0 = 0.0 / 0.0;
+ else
+ FT0 /= FT1;
}
void do_fsqrt (void)
@@ -306,27 +326,65 @@ void do_fsqrt (void)
FT0 = sqrt(FT0);
}
-void do_fsqrts (void)
-{
- FT0 = (float)sqrt((float)FT0);
-}
-
void do_fres (void)
{
- FT0 = 1.0 / FT0;
+ union {
+ double d;
+ uint64_t i;
+ } p;
+
+ if (isnormal(FT0)) {
+ FT0 = (float)(1.0 / FT0);
+ } else {
+ p.d = FT0;
+ if (p.i == 0x8000000000000000ULL) {
+ p.i = 0xFFF0000000000000ULL;
+ } else if (p.i == 0x0000000000000000ULL) {
+ p.i = 0x7FF0000000000000ULL;
+ } else if (isnan(FT0)) {
+ p.i = 0x7FF8000000000000ULL;
+ } else if (FT0 < 0.0) {
+ p.i = 0x8000000000000000ULL;
+ } else {
+ p.i = 0x0000000000000000ULL;
+ }
+ FT0 = p.d;
+ }
}
-void do_fsqrte (void)
+void do_frsqrte (void)
{
- FT0 = 1.0 / sqrt(FT0);
+ union {
+ double d;
+ uint64_t i;
+ } p;
+
+ if (isnormal(FT0) && FT0 > 0.0) {
+ FT0 = (float)(1.0 / sqrt(FT0));
+ } else {
+ p.d = FT0;
+ if (p.i == 0x8000000000000000ULL) {
+ p.i = 0xFFF0000000000000ULL;
+ } else if (p.i == 0x0000000000000000ULL) {
+ p.i = 0x7FF0000000000000ULL;
+ } else if (isnan(FT0)) {
+ if (!(p.i & 0x0008000000000000ULL))
+ p.i |= 0x000FFFFFFFFFFFFFULL;
+ } else if (FT0 < 0) {
+ p.i = 0x7FF8000000000000ULL;
+ } else {
+ p.i = 0x0000000000000000ULL;
+ }
+ FT0 = p.d;
+ }
}
void do_fsel (void)
{
if (FT0 >= 0)
- FT0 = FT2;
- else
FT0 = FT1;
+ else
+ FT0 = FT2;
}
void do_fcmpu (void)
@@ -371,12 +429,26 @@ void do_fcmpo (void)
void do_fabs (void)
{
- FT0 = fabsl(FT0);
+ union {
+ double d;
+ uint64_t i;
+ } p;
+
+ p.d = FT0;
+ p.i &= ~0x8000000000000000ULL;
+ FT0 = p.d;
}
void do_fnabs (void)
{
- FT0 = -fabsl(FT0);
+ union {
+ double d;
+ uint64_t i;
+ } p;
+
+ p.d = FT0;
+ p.i |= 0x8000000000000000ULL;
+ FT0 = p.d;
}
/* Instruction cache invalidation helper */
Index: target-ppc/translate.c
===================================================================
RCS file: /cvsroot/qemu/qemu/target-ppc/translate.c,v
retrieving revision 1.27
diff -u -d -w -B -b -d -p -r1.27 translate.c
--- target-ppc/translate.c 15 Feb 2005 23:06:19 -0000 1.27
+++ target-ppc/translate.c 2 Mar 2005 21:51:08 -0000
@@ -740,6 +740,7 @@ __GEN_LOGICAL2(sraw, 0x18, 0x18);
GEN_HANDLER(srawi, 0x1F, 0x18, 0x19, 0x00000000, PPC_INTEGER)
{
gen_op_load_gpr_T0(rS(ctx->opcode));
+ if (SH(ctx->opcode) != 0)
gen_op_srawi(SH(ctx->opcode), MASK(32 - SH(ctx->opcode), 31));
if (Rc(ctx->opcode) != 0)
gen_op_set_Rc0();
@@ -749,7 +750,7 @@ GEN_HANDLER(srawi, 0x1F, 0x18, 0x19, 0x0
__GEN_LOGICAL2(srw, 0x18, 0x10);
/*** Floating-Point arithmetic ***/
-#define _GEN_FLOAT_ACB(name, op1, op2) \
+#define _GEN_FLOAT_ACB(name, op, op1, op2, isfloat) \
GEN_HANDLER(f##name, op1, op2, 0xFF, 0x00000000, PPC_FLOAT) \
{ \
if (!ctx->fpu_enabled) { \
@@ -760,17 +761,20 @@ GEN_HANDLER(f##name, op1, op2, 0xFF, 0x0
gen_op_load_fpr_FT0(rA(ctx->opcode)); \
gen_op_load_fpr_FT1(rC(ctx->opcode)); \
gen_op_load_fpr_FT2(rB(ctx->opcode)); \
- gen_op_f##name(); \
+ gen_op_f##op(); \
+ if (isfloat) { \
+ gen_op_frsp(); \
+ } \
gen_op_store_FT0_fpr(rD(ctx->opcode)); \
if (Rc(ctx->opcode)) \
gen_op_set_Rc1(); \
}
#define GEN_FLOAT_ACB(name, op2) \
-_GEN_FLOAT_ACB(name, 0x3F, op2); \
-_GEN_FLOAT_ACB(name##s, 0x3B, op2);
+_GEN_FLOAT_ACB(name, name, 0x3F, op2, 0); \
+_GEN_FLOAT_ACB(name##s, name, 0x3B, op2, 1);
-#define _GEN_FLOAT_AB(name, op1, op2, inval) \
+#define _GEN_FLOAT_AB(name, op, op1, op2, inval, isfloat) \
GEN_HANDLER(f##name, op1, op2, 0xFF, inval, PPC_FLOAT) \
{ \
if (!ctx->fpu_enabled) { \
@@ -780,16 +784,19 @@ GEN_HANDLER(f##name, op1, op2, 0xFF, inv
gen_op_reset_scrfx(); \
gen_op_load_fpr_FT0(rA(ctx->opcode)); \
gen_op_load_fpr_FT1(rB(ctx->opcode)); \
- gen_op_f##name(); \
+ gen_op_f##op(); \
+ if (isfloat) { \
+ gen_op_frsp(); \
+ } \
gen_op_store_FT0_fpr(rD(ctx->opcode)); \
if (Rc(ctx->opcode)) \
gen_op_set_Rc1(); \
}
#define GEN_FLOAT_AB(name, op2, inval) \
-_GEN_FLOAT_AB(name, 0x3F, op2, inval); \
-_GEN_FLOAT_AB(name##s, 0x3B, op2, inval);
+_GEN_FLOAT_AB(name, name, 0x3F, op2, inval, 0); \
+_GEN_FLOAT_AB(name##s, name, 0x3B, op2, inval, 1);
-#define _GEN_FLOAT_AC(name, op1, op2, inval) \
+#define _GEN_FLOAT_AC(name, op, op1, op2, inval, isfloat) \
GEN_HANDLER(f##name, op1, op2, 0xFF, inval, PPC_FLOAT) \
{ \
if (!ctx->fpu_enabled) { \
@@ -799,14 +806,17 @@ GEN_HANDLER(f##name, op1, op2, 0xFF, inv
gen_op_reset_scrfx(); \
gen_op_load_fpr_FT0(rA(ctx->opcode)); \
gen_op_load_fpr_FT1(rC(ctx->opcode)); \
- gen_op_f##name(); \
+ gen_op_f##op(); \
+ if (isfloat) { \
+ gen_op_frsp(); \
+ } \
gen_op_store_FT0_fpr(rD(ctx->opcode)); \
if (Rc(ctx->opcode)) \
gen_op_set_Rc1(); \
}
#define GEN_FLOAT_AC(name, op2, inval) \
-_GEN_FLOAT_AC(name, 0x3F, op2, inval); \
-_GEN_FLOAT_AC(name##s, 0x3B, op2, inval);
+_GEN_FLOAT_AC(name, name, 0x3F, op2, inval, 0); \
+_GEN_FLOAT_AC(name##s, name, 0x3B, op2, inval, 1);
#define GEN_FLOAT_B(name, op2, op3) \
GEN_HANDLER(f##name, 0x3F, op2, op3, 0x001F0000, PPC_FLOAT) \
@@ -823,8 +833,8 @@ GEN_HANDLER(f##name, 0x3F, op2, op3, 0x0
gen_op_set_Rc1(); \
}
-#define GEN_FLOAT_BS(name, op2) \
-GEN_HANDLER(f##name, 0x3F, op2, 0xFF, 0x001F07C0, PPC_FLOAT) \
+#define GEN_FLOAT_BS(name, op1, op2) \
+GEN_HANDLER(f##name, op1, op2, 0xFF, 0x001F07C0, PPC_FLOAT) \
{ \
if (!ctx->fpu_enabled) { \
RET_EXCP(ctx, EXCP_NO_FP, 0); \
@@ -840,24 +850,24 @@ GEN_HANDLER(f##name, 0x3F, op2, 0xFF, 0x
/* fadd - fadds */
GEN_FLOAT_AB(add, 0x15, 0x000007C0);
-/* fdiv */
+/* fdiv - fdivs */
GEN_FLOAT_AB(div, 0x12, 0x000007C0);
-/* fmul */
+/* fmul - fmuls */
GEN_FLOAT_AC(mul, 0x19, 0x0000F800);
/* fres */
-GEN_FLOAT_BS(res, 0x18);
+GEN_FLOAT_BS(res, 0x3B, 0x18);
/* frsqrte */
-GEN_FLOAT_BS(rsqrte, 0x1A);
+GEN_FLOAT_BS(rsqrte, 0x3F, 0x1A);
/* fsel */
-_GEN_FLOAT_ACB(sel, 0x3F, 0x17);
-/* fsub */
+_GEN_FLOAT_ACB(sel, sel, 0x3F, 0x17, 0);
+/* fsub - fsubs */
GEN_FLOAT_AB(sub, 0x14, 0x000007C0);
/* Optional: */
/* fsqrt */
-GEN_FLOAT_BS(sqrt, 0x16);
+GEN_FLOAT_BS(sqrt, 0x3F, 0x16);
GEN_HANDLER(fsqrts, 0x3B, 0x16, 0xFF, 0x001F07C0, PPC_FLOAT_OPT)
{
@@ -867,20 +877,21 @@ GEN_HANDLER(fsqrts, 0x3B, 0x16, 0xFF, 0x
}
gen_op_reset_scrfx();
gen_op_load_fpr_FT0(rB(ctx->opcode));
- gen_op_fsqrts();
+ gen_op_fsqrt();
+ gen_op_frsp();
gen_op_store_FT0_fpr(rD(ctx->opcode));
if (Rc(ctx->opcode))
gen_op_set_Rc1();
}
/*** Floating-Point multiply-and-add ***/
-/* fmadd */
+/* fmadd - fmadds */
GEN_FLOAT_ACB(madd, 0x1D);
-/* fmsub */
+/* fmsub - fmsubs */
GEN_FLOAT_ACB(msub, 0x1C);
-/* fnmadd */
+/* fnmadd - fnmadds */
GEN_FLOAT_ACB(nmadd, 0x1F);
-/* fnmsub */
+/* fnmsub - fnmsubs */
GEN_FLOAT_ACB(nmsub, 0x1E);
/*** Floating-Point round & convert ***/
@@ -1426,6 +1437,10 @@ GEN_HANDLER(sync, 0x1F, 0x16, 0x12, 0x03
GEN_HANDLER(l##width, opc, 0xFF, 0xFF, 0x00000000, PPC_INTEGER) \
{ \
uint32_t simm = SIMM(ctx->opcode); \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
gen_op_set_T0(simm); \
} else { \
@@ -1441,6 +1456,10 @@ GEN_HANDLER(l##width, opc, 0xFF, 0xFF, 0
GEN_HANDLER(l##width##u, opc, 0xFF, 0xFF, 0x00000000, PPC_INTEGER) \
{ \
uint32_t simm = SIMM(ctx->opcode); \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0 || \
rA(ctx->opcode) == rD(ctx->opcode)) { \
RET_INVAL(ctx); \
@@ -1457,6 +1476,10 @@ GEN_HANDLER(l##width##u, opc, 0xFF, 0xFF
#define GEN_LDUXF(width, opc) \
GEN_HANDLER(l##width##ux, 0x1F, 0x17, opc, 0x00000001, PPC_INTEGER) \
{ \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0 || \
rA(ctx->opcode) == rD(ctx->opcode)) { \
RET_INVAL(ctx); \
@@ -1473,6 +1496,10 @@ GEN_HANDLER(l##width##ux, 0x1F, 0x17, op
#define GEN_LDXF(width, opc2, opc3) \
GEN_HANDLER(l##width##x, 0x1F, opc2, opc3, 0x00000001, PPC_INTEGER) \
{ \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
gen_op_load_gpr_T0(rB(ctx->opcode)); \
} else { \
@@ -1501,6 +1528,10 @@ GEN_LDFS(fs, 0x10);
GEN_HANDLER(st##width, opc, 0xFF, 0xFF, 0x00000000, PPC_INTEGER) \
{ \
uint32_t simm = SIMM(ctx->opcode); \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
gen_op_set_T0(simm); \
} else { \
@@ -1516,6 +1547,10 @@ GEN_HANDLER(st##width, opc, 0xFF, 0xFF,
GEN_HANDLER(st##width##u, opc, 0xFF, 0xFF, 0x00000000, PPC_INTEGER) \
{ \
uint32_t simm = SIMM(ctx->opcode); \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
RET_INVAL(ctx); \
return; \
@@ -1531,6 +1566,10 @@ GEN_HANDLER(st##width##u, opc, 0xFF, 0xF
#define GEN_STUXF(width, opc) \
GEN_HANDLER(st##width##ux, 0x1F, 0x17, opc, 0x00000001, PPC_INTEGER) \
{ \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
RET_INVAL(ctx); \
return; \
@@ -1546,6 +1585,10 @@ GEN_HANDLER(st##width##ux, 0x1F, 0x17, o
#define GEN_STXF(width, opc2, opc3) \
GEN_HANDLER(st##width##x, 0x1F, opc2, opc3, 0x00000001, PPC_INTEGER) \
{ \
+ if (!ctx->fpu_enabled) { \
+ RET_EXCP(ctx, EXCP_NO_FP, 0); \
+ return; \
+ } \
if (rA(ctx->opcode) == 0) { \
gen_op_load_gpr_T0(rB(ctx->opcode)); \
} else { \
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-02 23:04 [Qemu-devel] More PPC emulation fixes J. Mayer
@ 2005-03-03 11:49 ` J. Mayer
2005-03-04 10:31 ` J. Mayer
2005-03-07 17:52 ` Thayne Harbaugh
0 siblings, 2 replies; 10+ messages in thread
From: J. Mayer @ 2005-03-03 11:49 UTC (permalink / raw)
To: qemu-devel
[-- Attachment #1: Type: text/plain, Size: 401 bytes --]
On Thu, 2005-03-03 at 00:04, J. Mayer wrote:
> Here's a new patch that fixes the last (!) bug in PPC fixed point
> operations and most FPU bugs.
> Apply and enjoy.
I realise I missed one "detail":
FPU has to be activated in OHW, or it would crash. I'll send a modified
version 0.3 tonight. For those who don't want to wait, here's the needed
patch.
--
J. Mayer <l_indien@magic.fr>
Never organized
[-- Attachment #2: FPU.diff --]
[-- Type: text/x-patch, Size: 626 bytes --]
Index: src/vectors.S
===================================================================
--- src/vectors.S (revision 90)
+++ src/vectors.S (working copy)
@@ -73,6 +73,8 @@
stwu r14, 4(r13) ;
bdnz _bios_copy_loop ;
/* Synchronize the whole execution context */
+ /* Also enable FPU */
+ ori r0, r0, (1 << 13) ;
mtspr 26, r29 ;
mtspr 27, r0 ;
rfi ;
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-03 11:49 ` J. Mayer
@ 2005-03-04 10:31 ` J. Mayer
2005-03-07 17:52 ` Thayne Harbaugh
1 sibling, 0 replies; 10+ messages in thread
From: J. Mayer @ 2005-03-04 10:31 UTC (permalink / raw)
To: qemu-devel
On Thu, 2005-03-03 at 12:49, J. Mayer wrote:
> On Thu, 2005-03-03 at 00:04, J. Mayer wrote:
> > Here's a new patch that fixes the last (!) bug in PPC fixed point
> > operations and most FPU bugs.
>
> > Apply and enjoy.
>
> I realise I missed one "detail":
> FPU has to be activated in OHW, or it would crash. I'll send a modified
> version 0.3 tonight. For those who don't want to wait, here's the needed
> patch.
A patched OHW binary is now available here:
<http://perso.magic.fr/l_indien/OpenHackWare/0.3/OpenHackWare_0.3.2_bin.tar.bz2>
It's exactly the same that the one commited in qemu repository, except
that it enables FPU at startup, so it runs if the FPU fixes are applied.
It will run exactly the same as current one if used with current qemu
CVS.
--
J. Mayer <l_indien@magic.fr>
Never organized
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-03 11:49 ` J. Mayer
2005-03-04 10:31 ` J. Mayer
@ 2005-03-07 17:52 ` Thayne Harbaugh
2005-03-07 18:26 ` Asko Kauppi
2005-03-08 14:51 ` J. Mayer
1 sibling, 2 replies; 10+ messages in thread
From: Thayne Harbaugh @ 2005-03-07 17:52 UTC (permalink / raw)
To: J. Mayer; +Cc: qemu-devel
On Thu, 2005-03-03 at 12:49 +0100, J. Mayer wrote:
> On Thu, 2005-03-03 at 00:04, J. Mayer wrote:
> > Here's a new patch that fixes the last (!) bug in PPC fixed point
> > operations and most FPU bugs.
>
> > Apply and enjoy.
>
> I realise I missed one "detail":
> FPU has to be activated in OHW, or it would crash. I'll send a modified
> version 0.3 tonight. For those who don't want to wait, here's the needed
> patch.
I've been using the patches for a bit now. All of my previously
reported problems are gone.
Thanks.
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-07 17:52 ` Thayne Harbaugh
@ 2005-03-07 18:26 ` Asko Kauppi
2005-03-07 18:43 ` Thayne Harbaugh
2005-03-08 14:51 ` J. Mayer
1 sibling, 1 reply; 10+ messages in thread
From: Asko Kauppi @ 2005-03-07 18:26 UTC (permalink / raw)
To: qemu-devel
May I ask you guys, who've done work on Qemu/PPC side, how far is it
from being "Linux on Mac"?
7.3.2005 kello 19:52, Thayne Harbaugh kirjoitti:
On Thu, 2005-03-03 at 12:49 +0100, J. Mayer wrote:
>> On Thu, 2005-03-03 at 00:04, J. Mayer wrote:
>>> Here's a new patch that fixes the last (!) bug in PPC fixed point
>>> operations and most FPU bugs.
>>
>>> Apply and enjoy.
>>
>> I realise I missed one "detail":
>> FPU has to be activated in OHW, or it would crash. I'll send a
>> modified
>> version 0.3 tonight. For those who don't want to wait, here's the
>> needed
>> patch.
>
> I've been using the patches for a bit now. All of my previously
> reported problems are gone.
>
> Thanks.
>
>
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-07 18:26 ` Asko Kauppi
@ 2005-03-07 18:43 ` Thayne Harbaugh
2005-03-08 14:51 ` J. Mayer
0 siblings, 1 reply; 10+ messages in thread
From: Thayne Harbaugh @ 2005-03-07 18:43 UTC (permalink / raw)
To: qemu-devel
On Mon, 2005-03-07 at 20:26 +0200, Asko Kauppi wrote:
> May I ask you guys, who've done work on Qemu/PPC side, how far is it
> from being "Linux on Mac"?
I haven't tested anything except linux-powerpc on linux-x86 - which
works quite well except for a few features of a 2.6 guest kernel (I'm
working on fixing these). I once tried to boot Mac OS X on linux-x86,
but it died with an Altivec problem (at least that's what Jocelyn said).
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-07 18:43 ` Thayne Harbaugh
@ 2005-03-08 14:51 ` J. Mayer
0 siblings, 0 replies; 10+ messages in thread
From: J. Mayer @ 2005-03-08 14:51 UTC (permalink / raw)
To: qemu-devel
On Mon, 2005-03-07 at 19:43, Thayne Harbaugh wrote:
> On Mon, 2005-03-07 at 20:26 +0200, Asko Kauppi wrote:
> > May I ask you guys, who've done work on Qemu/PPC side, how far is it
> > from being "Linux on Mac"?
>
> I haven't tested anything except linux-powerpc on linux-x86 - which
> works quite well except for a few features of a 2.6 guest kernel (I'm
> working on fixing these). I once tried to boot Mac OS X on linux-x86,
> but it died with an Altivec problem (at least that's what Jocelyn said).
Altivec problem seems to appear with Mac OS X 10.3. Version 10.2 fails
to mount it's root partition, I really need to understand why. I didn't
check any other version, but I noticed that OpenDarwin sometimes acts
like Mac OS X 10.2 and sometimes crashes, depending on the version.
Those problem as, I suppose, related to buggy OF properties.
--
J. Mayer <l_indien@magic.fr>
Never organized
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-07 17:52 ` Thayne Harbaugh
2005-03-07 18:26 ` Asko Kauppi
@ 2005-03-08 14:51 ` J. Mayer
2005-03-08 15:05 ` Thayne Harbaugh
1 sibling, 1 reply; 10+ messages in thread
From: J. Mayer @ 2005-03-08 14:51 UTC (permalink / raw)
To: Thayne Harbaugh; +Cc: qemu-devel
On Mon, 2005-03-07 at 18:52, Thayne Harbaugh wrote:
> On Thu, 2005-03-03 at 12:49 +0100, J. Mayer wrote:
> > On Thu, 2005-03-03 at 00:04, J. Mayer wrote:
> > > Here's a new patch that fixes the last (!) bug in PPC fixed point
> > > operations and most FPU bugs.
> >
> > > Apply and enjoy.
> >
> > I realise I missed one "detail":
> > FPU has to be activated in OHW, or it would crash. I'll send a modified
> > version 0.3 tonight. For those who don't want to wait, here's the needed
> > patch.
>
> I've been using the patches for a bit now. All of my previously
> reported problems are gone.
Great !
I think most of qemu PPC cpu emulation bugs are now fixed.
I still know two problems that seems related to emulation:
- top in Debian Sarge crashes.
- Debian woody installer crashes.
Other problems seem OHW or hardware emulation related. For example ne2k
seems to miss MII related stuff emulation, then most automatic hardware
probes stay stuck while checking the network device.
I played a lot with qemu this week-end, recompiled kernels and qemu in
the emulated machine. Then I was able to run qemu x86 emulation under
qemu-ppc (but that's really slooooow...).
I was also able to launch ppc user emulation but missed memory to launch
ppc system emulation.
I'm also re-compiling MOL to try to launch Mac OS X using it ;-)
I did put some screenshots of Qemu running Debian Sarge for PPC here:
<http://perso.magic.fr/l_indien/qemu-ppc/snapshots/index.html>
--
J. Mayer <l_indien@magic.fr>
Never organized
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-08 14:51 ` J. Mayer
@ 2005-03-08 15:05 ` Thayne Harbaugh
2005-03-08 15:59 ` Asko Kauppi
0 siblings, 1 reply; 10+ messages in thread
From: Thayne Harbaugh @ 2005-03-08 15:05 UTC (permalink / raw)
To: J. Mayer; +Cc: qemu-devel
On Tue, 2005-03-08 at 15:51 +0100, J. Mayer wrote:
> On Mon, 2005-03-07 at 18:52, Thayne Harbaugh wrote:
> > I've been using the patches for a bit now. All of my previously
> > reported problems are gone.
>
> Great !
> I think most of qemu PPC cpu emulation bugs are now fixed.
> I still know two problems that seems related to emulation:
> - top in Debian Sarge crashes.
I've noticed the same thing.
> - Debian woody installer crashes.
>
> Other problems seem OHW or hardware emulation related.
I'm supposed to be looking at some of these, but lately I've had to
focus on some other things outside of qemu/OHW. Hopefully I can get
back to this soon.
> For example ne2k
> seems to miss MII related stuff emulation, then most automatic hardware
> probes stay stuck while checking the network device.
I don't think I've had this problem - or maybe I just don't recognize
it.
> I played a lot with qemu this week-end, recompiled kernels and qemu in
> the emulated machine. Then I was able to run qemu x86 emulation under
> qemu-ppc (but that's really slooooow...).
> I was also able to launch ppc user emulation but missed memory to launch
> ppc system emulation.
> I'm also re-compiling MOL to try to launch Mac OS X using it ;-)
Very cool.
> I did put some screenshots of Qemu running Debian Sarge for PPC here:
> <http://perso.magic.fr/l_indien/qemu-ppc/snapshots/index.html>
Interesting.
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [Qemu-devel] More PPC emulation fixes
2005-03-08 15:05 ` Thayne Harbaugh
@ 2005-03-08 15:59 ` Asko Kauppi
0 siblings, 0 replies; 10+ messages in thread
From: Asko Kauppi @ 2005-03-08 15:59 UTC (permalink / raw)
To: qemu-devel
The problems were pinned down to an old Qemu version in Fink, plus
OpenHackware it uses.
I've already contacted the fink qemu package maintainer about this.
Thanks. :)
8.3.2005 kello 17:05, Thayne Harbaugh kirjoitti:
On Tue, 2005-03-08 at 15:51 +0100, J. Mayer wrote:
>> On Mon, 2005-03-07 at 18:52, Thayne Harbaugh wrote:
>
>>> I've been using the patches for a bit now. All of my previously
>>> reported problems are gone.
>>
>> Great !
>> I think most of qemu PPC cpu emulation bugs are now fixed.
>> I still know two problems that seems related to emulation:
>> - top in Debian Sarge crashes.
>
> I've noticed the same thing.
>
>> - Debian woody installer crashes.
>>
>> Other problems seem OHW or hardware emulation related.
>
> I'm supposed to be looking at some of these, but lately I've had to
> focus on some other things outside of qemu/OHW. Hopefully I can get
> back to this soon.
>
>> For example ne2k
>> seems to miss MII related stuff emulation, then most automatic
>> hardware
>> probes stay stuck while checking the network device.
>
> I don't think I've had this problem - or maybe I just don't recognize
> it.
>
>> I played a lot with qemu this week-end, recompiled kernels and qemu in
>> the emulated machine. Then I was able to run qemu x86 emulation under
>> qemu-ppc (but that's really slooooow...).
>> I was also able to launch ppc user emulation but missed memory to
>> launch
>> ppc system emulation.
>> I'm also re-compiling MOL to try to launch Mac OS X using it ;-)
>
> Very cool.
>
>> I did put some screenshots of Qemu running Debian Sarge for PPC here:
>> <http://perso.magic.fr/l_indien/qemu-ppc/snapshots/index.html>
>
> Interesting.
>
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2005-03-08 16:38 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-03-02 23:04 [Qemu-devel] More PPC emulation fixes J. Mayer
2005-03-03 11:49 ` J. Mayer
2005-03-04 10:31 ` J. Mayer
2005-03-07 17:52 ` Thayne Harbaugh
2005-03-07 18:26 ` Asko Kauppi
2005-03-07 18:43 ` Thayne Harbaugh
2005-03-08 14:51 ` J. Mayer
2005-03-08 14:51 ` J. Mayer
2005-03-08 15:05 ` Thayne Harbaugh
2005-03-08 15:59 ` Asko Kauppi
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).