* [Qemu-devel] [PULL 00/12] tcg s390 updates
@ 2014-05-15 16:39 Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 01/11] tcg-s390: Fix off-by-one in wraparound andi Richard Henderson
` (11 more replies)
0 siblings, 12 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell, thuth
This version differs from the patch set I posted 2 weeks ago in that
the andi bugfix is split out separately, a typo in the add2/sub2 patch
is fixed, and a long forgotten patch to implement direct branches
between TBs was merged.
In addition to my own testing, Thomas Huth was kind enough to do a quick
check against the bug he originally reported to me.
r~
The following changes since commit 450445d543a33fdb1d18a18ab7669a33dc1ca7ae:
tcg: Fix tcg_reg_alloc_mov vs no-op truncation (2014-05-14 09:56:13 -0700)
are available in the git repository at:
git://github.com/rth7680/qemu.git tags/pull-tcg-s390-20140515
for you to fetch changes up to a10c64e0df3acbefdee1b2b837f76f05b0231aae:
tcg-s390: Implement direct chaining of TBs (2014-05-15 09:22:32 -0700)
----------------------------------------------------------------
tcg/s390 updates
----------------------------------------------------------------
Richard Henderson (11):
tcg-s390: Fix off-by-one in wraparound andi
tcg-s390: Convert to TCGMemOp
tcg-s390: Integrate endianness into TCGMemOp
tcg-s390: Convert to new ldst opcodes
tcg-s390: Move ldst helpers out of line
tcg-s390: Use more risbg in the tlb sequence
tcg-s390: Implement tcg_register_jit
tcg-s390: Allow immediate operands to add2 and sub2
tcg-s390: Improve setcond
tcg-s390: Don't force -march=z990
tcg-s390: Implement direct chaining of TBs
configure | 4 +-
include/exec/exec-all.h | 9 +
tcg/s390/tcg-target.c | 752 ++++++++++++++++++++++++++++--------------------
tcg/s390/tcg-target.h | 2 +-
4 files changed, 455 insertions(+), 312 deletions(-)
^ permalink raw reply [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 01/11] tcg-s390: Fix off-by-one in wraparound andi
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 02/11] tcg-s390: Convert to TCGMemOp Richard Henderson
` (10 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index ebdd074..0f972ae 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -983,8 +983,8 @@ static void tgen_andi(TCGContext *s, TCGType type, TCGReg dest, uint64_t val)
int msb, lsb;
if ((val & 0x8000000000000001ull) == 0x8000000000000001ull) {
/* Achieve wraparound by swapping msb and lsb. */
- msb = 63 - ctz64(~val);
- lsb = clz64(~val) + 1;
+ msb = 64 - ctz64(~val);
+ lsb = clz64(~val) - 1;
} else {
msb = clz64(val);
lsb = 63 - ctz64(val);
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 02/11] tcg-s390: Convert to TCGMemOp
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 01/11] tcg-s390: Fix off-by-one in wraparound andi Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 03/11] tcg-s390: Integrate endianness into TCGMemOp Richard Henderson
` (9 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 82 ++++++++++++++++++++++-----------------------------
1 file changed, 36 insertions(+), 46 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 0f972ae..e0faacf 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -227,16 +227,6 @@ typedef enum S390Opcode {
RX_STH = 0x40,
} S390Opcode;
-#define LD_SIGNED 0x04
-#define LD_UINT8 0x00
-#define LD_INT8 (LD_UINT8 | LD_SIGNED)
-#define LD_UINT16 0x01
-#define LD_INT16 (LD_UINT16 | LD_SIGNED)
-#define LD_UINT32 0x02
-#define LD_INT32 (LD_UINT32 | LD_SIGNED)
-#define LD_UINT64 0x03
-#define LD_INT64 (LD_UINT64 | LD_SIGNED)
-
#ifndef NDEBUG
static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
"%r0", "%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7",
@@ -1280,7 +1270,7 @@ static void tcg_out_call(TCGContext *s, tcg_insn_unit *dest)
}
}
-static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
+static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
TCGReg base, TCGReg index, int disp)
{
#ifdef TARGET_WORDS_BIGENDIAN
@@ -1289,13 +1279,13 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
const int bswap = 1;
#endif
switch (opc) {
- case LD_UINT8:
+ case MO_UB:
tcg_out_insn(s, RXY, LLGC, data, base, index, disp);
break;
- case LD_INT8:
+ case MO_SB:
tcg_out_insn(s, RXY, LGB, data, base, index, disp);
break;
- case LD_UINT16:
+ case MO_UW:
if (bswap) {
/* swapped unsigned halfword load with upper bits zeroed */
tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
@@ -1304,7 +1294,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, LLGH, data, base, index, disp);
}
break;
- case LD_INT16:
+ case MO_SW:
if (bswap) {
/* swapped sign-extended halfword load */
tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
@@ -1313,7 +1303,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, LGH, data, base, index, disp);
}
break;
- case LD_UINT32:
+ case MO_UL:
if (bswap) {
/* swapped unsigned int load with upper bits zeroed */
tcg_out_insn(s, RXY, LRV, data, base, index, disp);
@@ -1322,7 +1312,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, LLGF, data, base, index, disp);
}
break;
- case LD_INT32:
+ case MO_SL:
if (bswap) {
/* swapped sign-extended int load */
tcg_out_insn(s, RXY, LRV, data, base, index, disp);
@@ -1331,7 +1321,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, LGF, data, base, index, disp);
}
break;
- case LD_UINT64:
+ case MO_Q:
if (bswap) {
tcg_out_insn(s, RXY, LRVG, data, base, index, disp);
} else {
@@ -1343,7 +1333,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data,
}
}
-static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data,
+static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
TCGReg base, TCGReg index, int disp)
{
#ifdef TARGET_WORDS_BIGENDIAN
@@ -1352,14 +1342,14 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data,
const int bswap = 1;
#endif
switch (opc) {
- case LD_UINT8:
+ case MO_UB:
if (disp >= 0 && disp < 0x1000) {
tcg_out_insn(s, RX, STC, data, base, index, disp);
} else {
tcg_out_insn(s, RXY, STCY, data, base, index, disp);
}
break;
- case LD_UINT16:
+ case MO_UW:
if (bswap) {
tcg_out_insn(s, RXY, STRVH, data, base, index, disp);
} else if (disp >= 0 && disp < 0x1000) {
@@ -1368,7 +1358,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, STHY, data, base, index, disp);
}
break;
- case LD_UINT32:
+ case MO_UL:
if (bswap) {
tcg_out_insn(s, RXY, STRV, data, base, index, disp);
} else if (disp >= 0 && disp < 0x1000) {
@@ -1377,7 +1367,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data,
tcg_out_insn(s, RXY, STY, data, base, index, disp);
}
break;
- case LD_UINT64:
+ case MO_Q:
if (bswap) {
tcg_out_insn(s, RXY, STRVG, data, base, index, disp);
} else {
@@ -1398,7 +1388,7 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
const TCGReg arg1 = tcg_target_call_iarg_regs[1];
const TCGReg arg2 = tcg_target_call_iarg_regs[2];
const TCGReg arg3 = tcg_target_call_iarg_regs[3];
- int s_bits = opc & 3;
+ TCGMemOp s_bits = opc & MO_SIZE;
tcg_insn_unit *label1_ptr;
tcg_target_long ofs;
@@ -1442,17 +1432,17 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
if (is_store) {
/* Make sure to zero-extend the value to the full register
for the calling convention. */
- switch (opc) {
- case LD_UINT8:
+ switch (s_bits) {
+ case MO_UB:
tgen_ext8u(s, TCG_TYPE_I64, arg2, data_reg);
break;
- case LD_UINT16:
+ case MO_UW:
tgen_ext16u(s, TCG_TYPE_I64, arg2, data_reg);
break;
- case LD_UINT32:
+ case MO_UL:
tgen_ext32u(s, arg2, data_reg);
break;
- case LD_UINT64:
+ case MO_Q:
tcg_out_mov(s, TCG_TYPE_I64, arg2, data_reg);
break;
default:
@@ -1468,13 +1458,13 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
/* sign extension */
switch (opc) {
- case LD_INT8:
+ case MO_SB:
tgen_ext8s(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
break;
- case LD_INT16:
+ case MO_SW:
tgen_ext16s(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
break;
- case LD_INT32:
+ case MO_SL:
tgen_ext32s(s, data_reg, TCG_REG_R2);
break;
default:
@@ -1525,7 +1515,7 @@ static void tcg_prepare_user_ldst(TCGContext *s, TCGReg *addr_reg,
/* load data with address translation (if applicable)
and endianness conversion */
-static void tcg_out_qemu_ld(TCGContext* s, const TCGArg* args, int opc)
+static void tcg_out_qemu_ld(TCGContext* s, const TCGArg* args, TCGMemOp opc)
{
TCGReg addr_reg, data_reg;
#if defined(CONFIG_SOFTMMU)
@@ -1554,7 +1544,7 @@ static void tcg_out_qemu_ld(TCGContext* s, const TCGArg* args, int opc)
#endif
}
-static void tcg_out_qemu_st(TCGContext* s, const TCGArg* args, int opc)
+static void tcg_out_qemu_st(TCGContext* s, const TCGArg* args, TCGMemOp opc)
{
TCGReg addr_reg, data_reg;
#if defined(CONFIG_SOFTMMU)
@@ -1812,36 +1802,36 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_qemu_ld8u:
- tcg_out_qemu_ld(s, args, LD_UINT8);
+ tcg_out_qemu_ld(s, args, MO_UB);
break;
case INDEX_op_qemu_ld8s:
- tcg_out_qemu_ld(s, args, LD_INT8);
+ tcg_out_qemu_ld(s, args, MO_SB);
break;
case INDEX_op_qemu_ld16u:
- tcg_out_qemu_ld(s, args, LD_UINT16);
+ tcg_out_qemu_ld(s, args, MO_UW);
break;
case INDEX_op_qemu_ld16s:
- tcg_out_qemu_ld(s, args, LD_INT16);
+ tcg_out_qemu_ld(s, args, MO_SW);
break;
case INDEX_op_qemu_ld32:
/* ??? Technically we can use a non-extending instruction. */
- tcg_out_qemu_ld(s, args, LD_UINT32);
+ tcg_out_qemu_ld(s, args, MO_UL);
break;
case INDEX_op_qemu_ld64:
- tcg_out_qemu_ld(s, args, LD_UINT64);
+ tcg_out_qemu_ld(s, args, MO_Q);
break;
case INDEX_op_qemu_st8:
- tcg_out_qemu_st(s, args, LD_UINT8);
+ tcg_out_qemu_st(s, args, MO_UB);
break;
case INDEX_op_qemu_st16:
- tcg_out_qemu_st(s, args, LD_UINT16);
+ tcg_out_qemu_st(s, args, MO_UW);
break;
case INDEX_op_qemu_st32:
- tcg_out_qemu_st(s, args, LD_UINT32);
+ tcg_out_qemu_st(s, args, MO_UL);
break;
case INDEX_op_qemu_st64:
- tcg_out_qemu_st(s, args, LD_UINT64);
+ tcg_out_qemu_st(s, args, MO_Q);
break;
case INDEX_op_ld16s_i64:
@@ -2038,10 +2028,10 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_qemu_ld32u:
- tcg_out_qemu_ld(s, args, LD_UINT32);
+ tcg_out_qemu_ld(s, args, MO_UL);
break;
case INDEX_op_qemu_ld32s:
- tcg_out_qemu_ld(s, args, LD_INT32);
+ tcg_out_qemu_ld(s, args, MO_SL);
break;
OP_32_64(deposit):
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 03/11] tcg-s390: Integrate endianness into TCGMemOp
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 01/11] tcg-s390: Fix off-by-one in wraparound andi Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 02/11] tcg-s390: Convert to TCGMemOp Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 04/11] tcg-s390: Convert to new ldst opcodes Richard Henderson
` (8 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 124 ++++++++++++++++++++++++--------------------------
1 file changed, 60 insertions(+), 64 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index e0faacf..f1d9089 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -1273,11 +1273,6 @@ static void tcg_out_call(TCGContext *s, tcg_insn_unit *dest)
static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
TCGReg base, TCGReg index, int disp)
{
-#ifdef TARGET_WORDS_BIGENDIAN
- const int bswap = 0;
-#else
- const int bswap = 1;
-#endif
switch (opc) {
case MO_UB:
tcg_out_insn(s, RXY, LLGC, data, base, index, disp);
@@ -1285,49 +1280,50 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
case MO_SB:
tcg_out_insn(s, RXY, LGB, data, base, index, disp);
break;
+
+ case MO_UW | MO_BSWAP:
+ /* swapped unsigned halfword load with upper bits zeroed */
+ tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
+ tgen_ext16u(s, TCG_TYPE_I64, data, data);
+ break;
case MO_UW:
- if (bswap) {
- /* swapped unsigned halfword load with upper bits zeroed */
- tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
- tgen_ext16u(s, TCG_TYPE_I64, data, data);
- } else {
- tcg_out_insn(s, RXY, LLGH, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, LLGH, data, base, index, disp);
+ break;
+
+ case MO_SW | MO_BSWAP:
+ /* swapped sign-extended halfword load */
+ tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
+ tgen_ext16s(s, TCG_TYPE_I64, data, data);
break;
case MO_SW:
- if (bswap) {
- /* swapped sign-extended halfword load */
- tcg_out_insn(s, RXY, LRVH, data, base, index, disp);
- tgen_ext16s(s, TCG_TYPE_I64, data, data);
- } else {
- tcg_out_insn(s, RXY, LGH, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, LGH, data, base, index, disp);
+ break;
+
+ case MO_UL | MO_BSWAP:
+ /* swapped unsigned int load with upper bits zeroed */
+ tcg_out_insn(s, RXY, LRV, data, base, index, disp);
+ tgen_ext32u(s, data, data);
break;
case MO_UL:
- if (bswap) {
- /* swapped unsigned int load with upper bits zeroed */
- tcg_out_insn(s, RXY, LRV, data, base, index, disp);
- tgen_ext32u(s, data, data);
- } else {
- tcg_out_insn(s, RXY, LLGF, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, LLGF, data, base, index, disp);
+ break;
+
+ case MO_SL | MO_BSWAP:
+ /* swapped sign-extended int load */
+ tcg_out_insn(s, RXY, LRV, data, base, index, disp);
+ tgen_ext32s(s, data, data);
break;
case MO_SL:
- if (bswap) {
- /* swapped sign-extended int load */
- tcg_out_insn(s, RXY, LRV, data, base, index, disp);
- tgen_ext32s(s, data, data);
- } else {
- tcg_out_insn(s, RXY, LGF, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, LGF, data, base, index, disp);
+ break;
+
+ case MO_Q | MO_BSWAP:
+ tcg_out_insn(s, RXY, LRVG, data, base, index, disp);
break;
case MO_Q:
- if (bswap) {
- tcg_out_insn(s, RXY, LRVG, data, base, index, disp);
- } else {
- tcg_out_insn(s, RXY, LG, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, LG, data, base, index, disp);
break;
+
default:
tcg_abort();
}
@@ -1336,11 +1332,6 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
TCGReg base, TCGReg index, int disp)
{
-#ifdef TARGET_WORDS_BIGENDIAN
- const int bswap = 0;
-#else
- const int bswap = 1;
-#endif
switch (opc) {
case MO_UB:
if (disp >= 0 && disp < 0x1000) {
@@ -1349,31 +1340,36 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
tcg_out_insn(s, RXY, STCY, data, base, index, disp);
}
break;
+
+ case MO_UW | MO_BSWAP:
+ tcg_out_insn(s, RXY, STRVH, data, base, index, disp);
+ break;
case MO_UW:
- if (bswap) {
- tcg_out_insn(s, RXY, STRVH, data, base, index, disp);
- } else if (disp >= 0 && disp < 0x1000) {
+ if (disp >= 0 && disp < 0x1000) {
tcg_out_insn(s, RX, STH, data, base, index, disp);
} else {
tcg_out_insn(s, RXY, STHY, data, base, index, disp);
}
break;
+
+ case MO_UL | MO_BSWAP:
+ tcg_out_insn(s, RXY, STRV, data, base, index, disp);
+ break;
case MO_UL:
- if (bswap) {
- tcg_out_insn(s, RXY, STRV, data, base, index, disp);
- } else if (disp >= 0 && disp < 0x1000) {
+ if (disp >= 0 && disp < 0x1000) {
tcg_out_insn(s, RX, ST, data, base, index, disp);
} else {
tcg_out_insn(s, RXY, STY, data, base, index, disp);
}
break;
+
+ case MO_Q | MO_BSWAP:
+ tcg_out_insn(s, RXY, STRVG, data, base, index, disp);
+ break;
case MO_Q:
- if (bswap) {
- tcg_out_insn(s, RXY, STRVG, data, base, index, disp);
- } else {
- tcg_out_insn(s, RXY, STG, data, base, index, disp);
- }
+ tcg_out_insn(s, RXY, STG, data, base, index, disp);
break;
+
default:
tcg_abort();
}
@@ -1457,7 +1453,7 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
tcg_out_call(s, qemu_ld_helpers[s_bits]);
/* sign extension */
- switch (opc) {
+ switch (opc & MO_SSIZE) {
case MO_SB:
tgen_ext8s(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
break;
@@ -1808,30 +1804,30 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
tcg_out_qemu_ld(s, args, MO_SB);
break;
case INDEX_op_qemu_ld16u:
- tcg_out_qemu_ld(s, args, MO_UW);
+ tcg_out_qemu_ld(s, args, MO_TEUW);
break;
case INDEX_op_qemu_ld16s:
- tcg_out_qemu_ld(s, args, MO_SW);
+ tcg_out_qemu_ld(s, args, MO_TESW);
break;
case INDEX_op_qemu_ld32:
/* ??? Technically we can use a non-extending instruction. */
- tcg_out_qemu_ld(s, args, MO_UL);
+ tcg_out_qemu_ld(s, args, MO_TEUL);
break;
case INDEX_op_qemu_ld64:
- tcg_out_qemu_ld(s, args, MO_Q);
+ tcg_out_qemu_ld(s, args, MO_TEQ);
break;
case INDEX_op_qemu_st8:
tcg_out_qemu_st(s, args, MO_UB);
break;
case INDEX_op_qemu_st16:
- tcg_out_qemu_st(s, args, MO_UW);
+ tcg_out_qemu_st(s, args, MO_TEUW);
break;
case INDEX_op_qemu_st32:
- tcg_out_qemu_st(s, args, MO_UL);
+ tcg_out_qemu_st(s, args, MO_TEUL);
break;
case INDEX_op_qemu_st64:
- tcg_out_qemu_st(s, args, MO_Q);
+ tcg_out_qemu_st(s, args, MO_TEQ);
break;
case INDEX_op_ld16s_i64:
@@ -2028,10 +2024,10 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_qemu_ld32u:
- tcg_out_qemu_ld(s, args, MO_UL);
+ tcg_out_qemu_ld(s, args, MO_TEUL);
break;
case INDEX_op_qemu_ld32s:
- tcg_out_qemu_ld(s, args, MO_SL);
+ tcg_out_qemu_ld(s, args, MO_TESL);
break;
OP_32_64(deposit):
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 04/11] tcg-s390: Convert to new ldst opcodes
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (2 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 03/11] tcg-s390: Integrate endianness into TCGMemOp Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line Richard Henderson
` (7 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 170 ++++++++++++++++----------------------------------
tcg/s390/tcg-target.h | 2 +-
2 files changed, 56 insertions(+), 116 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index f1d9089..4fbadaf 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -238,6 +238,7 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
call-saved registers. Likewise prefer the call-clobbered registers
in reverse order to maximize the chance of avoiding the arguments. */
static const int tcg_target_reg_alloc_order[] = {
+ /* Call saved registers. */
TCG_REG_R13,
TCG_REG_R12,
TCG_REG_R11,
@@ -246,9 +247,11 @@ static const int tcg_target_reg_alloc_order[] = {
TCG_REG_R8,
TCG_REG_R7,
TCG_REG_R6,
+ /* Call clobbered registers. */
TCG_REG_R14,
TCG_REG_R0,
TCG_REG_R1,
+ /* Argument registers, in reverse order of allocation. */
TCG_REG_R5,
TCG_REG_R4,
TCG_REG_R3,
@@ -308,22 +311,29 @@ static const uint8_t tcg_cond_to_ltr_cond[] = {
};
#ifdef CONFIG_SOFTMMU
-/* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr,
- int mmu_idx) */
-static void * const qemu_ld_helpers[4] = {
- helper_ldb_mmu,
- helper_ldw_mmu,
- helper_ldl_mmu,
- helper_ldq_mmu,
+static void * const qemu_ld_helpers[16] = {
+ [MO_UB] = helper_ret_ldub_mmu,
+ [MO_SB] = helper_ret_ldsb_mmu,
+ [MO_LEUW] = helper_le_lduw_mmu,
+ [MO_LESW] = helper_le_ldsw_mmu,
+ [MO_LEUL] = helper_le_ldul_mmu,
+ [MO_LESL] = helper_le_ldsl_mmu,
+ [MO_LEQ] = helper_le_ldq_mmu,
+ [MO_BEUW] = helper_be_lduw_mmu,
+ [MO_BESW] = helper_be_ldsw_mmu,
+ [MO_BEUL] = helper_be_ldul_mmu,
+ [MO_BESL] = helper_be_ldsl_mmu,
+ [MO_BEQ] = helper_be_ldq_mmu,
};
-/* helper signature: helper_st_mmu(CPUState *env, target_ulong addr,
- uintxx_t val, int mmu_idx) */
-static void * const qemu_st_helpers[4] = {
- helper_stb_mmu,
- helper_stw_mmu,
- helper_stl_mmu,
- helper_stq_mmu,
+static void * const qemu_st_helpers[16] = {
+ [MO_UB] = helper_ret_stb_mmu,
+ [MO_LEUW] = helper_le_stw_mmu,
+ [MO_LEUL] = helper_le_stl_mmu,
+ [MO_LEQ] = helper_le_stq_mmu,
+ [MO_BEUW] = helper_be_stw_mmu,
+ [MO_BEUL] = helper_be_stl_mmu,
+ [MO_BEQ] = helper_be_stq_mmu,
};
#endif
@@ -1384,6 +1394,7 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
const TCGReg arg1 = tcg_target_call_iarg_regs[1];
const TCGReg arg2 = tcg_target_call_iarg_regs[2];
const TCGReg arg3 = tcg_target_call_iarg_regs[3];
+ const TCGReg arg4 = tcg_target_call_iarg_regs[4];
TCGMemOp s_bits = opc & MO_SIZE;
tcg_insn_unit *label1_ptr;
tcg_target_long ofs;
@@ -1445,29 +1456,15 @@ static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
tcg_abort();
}
tcg_out_movi(s, TCG_TYPE_I32, arg3, mem_index);
- tcg_out_mov(s, TCG_TYPE_I64, arg0, TCG_AREG0);
- tcg_out_call(s, qemu_st_helpers[s_bits]);
+ tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
+ tcg_out_movi(s, TCG_TYPE_PTR, arg4, (uintptr_t)s->code_ptr);
+ tcg_out_call(s, qemu_st_helpers[opc]);
} else {
tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index);
- tcg_out_mov(s, TCG_TYPE_I64, arg0, TCG_AREG0);
- tcg_out_call(s, qemu_ld_helpers[s_bits]);
-
- /* sign extension */
- switch (opc & MO_SSIZE) {
- case MO_SB:
- tgen_ext8s(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
- break;
- case MO_SW:
- tgen_ext16s(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
- break;
- case MO_SL:
- tgen_ext32s(s, data_reg, TCG_REG_R2);
- break;
- default:
- /* unsigned -> just copy */
- tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
- break;
- }
+ tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
+ tcg_out_movi(s, TCG_TYPE_PTR, arg3, (uintptr_t)s->code_ptr);
+ tcg_out_call(s, qemu_ld_helpers[opc]);
+ tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
}
/* jump to label2 (end) */
@@ -1511,59 +1508,39 @@ static void tcg_prepare_user_ldst(TCGContext *s, TCGReg *addr_reg,
/* load data with address translation (if applicable)
and endianness conversion */
-static void tcg_out_qemu_ld(TCGContext* s, const TCGArg* args, TCGMemOp opc)
+static void tcg_out_qemu_ld(TCGContext* s, TCGReg data_reg, TCGReg addr_reg,
+ TCGMemOp opc, int mem_index)
{
- TCGReg addr_reg, data_reg;
#if defined(CONFIG_SOFTMMU)
- int mem_index;
tcg_insn_unit *label2_ptr;
-#else
- TCGReg index_reg;
- tcg_target_long disp;
-#endif
-
- data_reg = *args++;
- addr_reg = *args++;
-
-#if defined(CONFIG_SOFTMMU)
- mem_index = *args;
addr_reg = tcg_prepare_qemu_ldst(s, data_reg, addr_reg, mem_index,
opc, &label2_ptr, 0);
-
tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg, TCG_REG_NONE, 0);
-
tcg_finish_qemu_ldst(s, label2_ptr);
#else
+ TCGReg index_reg;
+ tcg_target_long disp;
+
tcg_prepare_user_ldst(s, &addr_reg, &index_reg, &disp);
tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg, index_reg, disp);
#endif
}
-static void tcg_out_qemu_st(TCGContext* s, const TCGArg* args, TCGMemOp opc)
+static void tcg_out_qemu_st(TCGContext* s, TCGReg data_reg, TCGReg addr_reg,
+ TCGMemOp opc, int mem_index)
{
- TCGReg addr_reg, data_reg;
#if defined(CONFIG_SOFTMMU)
- int mem_index;
tcg_insn_unit *label2_ptr;
-#else
- TCGReg index_reg;
- tcg_target_long disp;
-#endif
-
- data_reg = *args++;
- addr_reg = *args++;
-
-#if defined(CONFIG_SOFTMMU)
- mem_index = *args;
addr_reg = tcg_prepare_qemu_ldst(s, data_reg, addr_reg, mem_index,
opc, &label2_ptr, 1);
-
tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg, TCG_REG_NONE, 0);
-
tcg_finish_qemu_ldst(s, label2_ptr);
#else
+ TCGReg index_reg;
+ tcg_target_long disp;
+
tcg_prepare_user_ldst(s, &addr_reg, &index_reg, &disp);
tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg, index_reg, disp);
#endif
@@ -1797,37 +1774,14 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
args[2], const_args[2], args[3]);
break;
- case INDEX_op_qemu_ld8u:
- tcg_out_qemu_ld(s, args, MO_UB);
- break;
- case INDEX_op_qemu_ld8s:
- tcg_out_qemu_ld(s, args, MO_SB);
- break;
- case INDEX_op_qemu_ld16u:
- tcg_out_qemu_ld(s, args, MO_TEUW);
- break;
- case INDEX_op_qemu_ld16s:
- tcg_out_qemu_ld(s, args, MO_TESW);
- break;
- case INDEX_op_qemu_ld32:
+ case INDEX_op_qemu_ld_i32:
/* ??? Technically we can use a non-extending instruction. */
- tcg_out_qemu_ld(s, args, MO_TEUL);
- break;
- case INDEX_op_qemu_ld64:
- tcg_out_qemu_ld(s, args, MO_TEQ);
- break;
-
- case INDEX_op_qemu_st8:
- tcg_out_qemu_st(s, args, MO_UB);
- break;
- case INDEX_op_qemu_st16:
- tcg_out_qemu_st(s, args, MO_TEUW);
- break;
- case INDEX_op_qemu_st32:
- tcg_out_qemu_st(s, args, MO_TEUL);
+ case INDEX_op_qemu_ld_i64:
+ tcg_out_qemu_ld(s, args[0], args[1], args[2], args[3]);
break;
- case INDEX_op_qemu_st64:
- tcg_out_qemu_st(s, args, MO_TEQ);
+ case INDEX_op_qemu_st_i32:
+ case INDEX_op_qemu_st_i64:
+ tcg_out_qemu_st(s, args[0], args[1], args[2], args[3]);
break;
case INDEX_op_ld16s_i64:
@@ -2023,13 +1977,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
args[2], const_args[2], args[3]);
break;
- case INDEX_op_qemu_ld32u:
- tcg_out_qemu_ld(s, args, MO_TEUL);
- break;
- case INDEX_op_qemu_ld32s:
- tcg_out_qemu_ld(s, args, MO_TESL);
- break;
-
OP_32_64(deposit):
tgen_deposit(s, args[0], args[2], args[3], args[4]);
break;
@@ -2094,17 +2041,10 @@ static const TCGTargetOpDef s390_op_defs[] = {
{ INDEX_op_movcond_i32, { "r", "r", "rC", "r", "0" } },
{ INDEX_op_deposit_i32, { "r", "0", "r" } },
- { INDEX_op_qemu_ld8u, { "r", "L" } },
- { INDEX_op_qemu_ld8s, { "r", "L" } },
- { INDEX_op_qemu_ld16u, { "r", "L" } },
- { INDEX_op_qemu_ld16s, { "r", "L" } },
- { INDEX_op_qemu_ld32, { "r", "L" } },
- { INDEX_op_qemu_ld64, { "r", "L" } },
-
- { INDEX_op_qemu_st8, { "L", "L" } },
- { INDEX_op_qemu_st16, { "L", "L" } },
- { INDEX_op_qemu_st32, { "L", "L" } },
- { INDEX_op_qemu_st64, { "L", "L" } },
+ { INDEX_op_qemu_ld_i32, { "r", "L" } },
+ { INDEX_op_qemu_ld_i64, { "r", "L" } },
+ { INDEX_op_qemu_st_i32, { "L", "L" } },
+ { INDEX_op_qemu_st_i64, { "L", "L" } },
{ INDEX_op_ld8u_i64, { "r", "r" } },
{ INDEX_op_ld8s_i64, { "r", "r" } },
@@ -2159,9 +2099,6 @@ static const TCGTargetOpDef s390_op_defs[] = {
{ INDEX_op_movcond_i64, { "r", "r", "rC", "r", "0" } },
{ INDEX_op_deposit_i64, { "r", "0", "r" } },
- { INDEX_op_qemu_ld32u, { "r", "L" } },
- { INDEX_op_qemu_ld32s, { "r", "L" } },
-
{ -1 },
};
@@ -2196,6 +2133,9 @@ static void tcg_target_init(TCGContext *s)
tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R3);
tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R4);
tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R5);
+ /* The r6 register is technically call-saved, but it's also a parameter
+ register, so it can get killed by setup for the qemu_st helper. */
+ tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R6);
/* The return register can be considered call-clobbered. */
tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R14);
diff --git a/tcg/s390/tcg-target.h b/tcg/s390/tcg-target.h
index 5bf733e..ad2c6dd 100644
--- a/tcg/s390/tcg-target.h
+++ b/tcg/s390/tcg-target.h
@@ -100,7 +100,7 @@ typedef enum TCGReg {
#define TCG_TARGET_HAS_muluh_i64 0
#define TCG_TARGET_HAS_mulsh_i64 0
-#define TCG_TARGET_HAS_new_ldst 0
+#define TCG_TARGET_HAS_new_ldst 1
extern bool tcg_target_deposit_valid(int ofs, int len);
#define TCG_TARGET_deposit_i32_valid tcg_target_deposit_valid
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (3 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 04/11] tcg-s390: Convert to new ldst opcodes Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2015-01-29 16:01 ` Alexander Graf
2014-05-15 16:39 ` [Qemu-devel] [PULL 06/11] tcg-s390: Use more risbg in the tlb sequence Richard Henderson
` (6 subsequent siblings)
11 siblings, 1 reply; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
That is, the old LDST_OPTIMIZATION.
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 210 ++++++++++++++++++++++++++++----------------------
1 file changed, 118 insertions(+), 92 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 4fbadaf..1c955db 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -24,7 +24,7 @@
* THE SOFTWARE.
*/
-#include "tcg-be-null.h"
+#include "tcg-be-ldst.h"
/* We only support generating code for 64-bit mode. */
#if TCG_TARGET_REG_BITS != 64
@@ -1386,107 +1386,123 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
}
#if defined(CONFIG_SOFTMMU)
-static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
- TCGReg addr_reg, int mem_index, int opc,
- tcg_insn_unit **label2_ptr_p, int is_store)
+/* We're expecting to use a 20-bit signed offset on the tlb memory ops.
+ Using the offset of the second entry in the last tlb table ensures
+ that we can index all of the elements of the first entry. */
+QEMU_BUILD_BUG_ON(offsetof(CPUArchState, tlb_table[NB_MMU_MODES - 1][1])
+ > 0x7ffff);
+
+/* Load and compare a TLB entry, leaving the flags set. Loads the TLB
+ addend into R2. Returns a register with the santitized guest address. */
+static TCGReg tcg_out_tlb_read(TCGContext* s, TCGReg addr_reg, TCGMemOp opc,
+ int mem_index, bool is_ld)
{
- const TCGReg arg0 = tcg_target_call_iarg_regs[0];
- const TCGReg arg1 = tcg_target_call_iarg_regs[1];
- const TCGReg arg2 = tcg_target_call_iarg_regs[2];
- const TCGReg arg3 = tcg_target_call_iarg_regs[3];
- const TCGReg arg4 = tcg_target_call_iarg_regs[4];
TCGMemOp s_bits = opc & MO_SIZE;
- tcg_insn_unit *label1_ptr;
- tcg_target_long ofs;
+ int ofs;
+
+ tcg_out_sh64(s, RSY_SRLG, TCG_REG_R2, addr_reg, TCG_REG_NONE,
+ TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
if (TARGET_LONG_BITS == 32) {
- tgen_ext32u(s, arg1, addr_reg);
+ tgen_ext32u(s, TCG_REG_R3, addr_reg);
} else {
- tcg_out_mov(s, TCG_TYPE_I64, arg1, addr_reg);
+ tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
}
- tcg_out_sh64(s, RSY_SRLG, arg2, addr_reg, TCG_REG_NONE,
- TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
-
- tgen_andi(s, TCG_TYPE_I64, arg1, TARGET_PAGE_MASK | ((1 << s_bits) - 1));
- tgen_andi(s, TCG_TYPE_I64, arg2, (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
+ tgen_andi(s, TCG_TYPE_I64, TCG_REG_R2,
+ (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
+ tgen_andi(s, TCG_TYPE_I64, TCG_REG_R3,
+ TARGET_PAGE_MASK | ((1 << s_bits) - 1));
- if (is_store) {
- ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
- } else {
+ if (is_ld) {
ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_read);
+ } else {
+ ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
}
- assert(ofs < 0x80000);
-
if (TARGET_LONG_BITS == 32) {
- tcg_out_mem(s, RX_C, RXY_CY, arg1, arg2, TCG_AREG0, ofs);
+ tcg_out_mem(s, RX_C, RXY_CY, TCG_REG_R3, TCG_REG_R2, TCG_AREG0, ofs);
} else {
- tcg_out_mem(s, 0, RXY_CG, arg1, arg2, TCG_AREG0, ofs);
+ tcg_out_mem(s, 0, RXY_CG, TCG_REG_R3, TCG_REG_R2, TCG_AREG0, ofs);
}
+ ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addend);
+ tcg_out_mem(s, 0, RXY_LG, TCG_REG_R2, TCG_REG_R2, TCG_AREG0, ofs);
+
if (TARGET_LONG_BITS == 32) {
- tgen_ext32u(s, arg1, addr_reg);
- } else {
- tcg_out_mov(s, TCG_TYPE_I64, arg1, addr_reg);
- }
-
- label1_ptr = s->code_ptr;
-
- /* je label1 (offset will be patched in later) */
- tcg_out_insn(s, RI, BRC, S390_CC_EQ, 0);
-
- /* call load/store helper */
- if (is_store) {
- /* Make sure to zero-extend the value to the full register
- for the calling convention. */
- switch (s_bits) {
- case MO_UB:
- tgen_ext8u(s, TCG_TYPE_I64, arg2, data_reg);
- break;
- case MO_UW:
- tgen_ext16u(s, TCG_TYPE_I64, arg2, data_reg);
- break;
- case MO_UL:
- tgen_ext32u(s, arg2, data_reg);
- break;
- case MO_Q:
- tcg_out_mov(s, TCG_TYPE_I64, arg2, data_reg);
- break;
- default:
- tcg_abort();
- }
- tcg_out_movi(s, TCG_TYPE_I32, arg3, mem_index);
- tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
- tcg_out_movi(s, TCG_TYPE_PTR, arg4, (uintptr_t)s->code_ptr);
- tcg_out_call(s, qemu_st_helpers[opc]);
- } else {
- tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index);
- tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
- tcg_out_movi(s, TCG_TYPE_PTR, arg3, (uintptr_t)s->code_ptr);
- tcg_out_call(s, qemu_ld_helpers[opc]);
- tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
+ tgen_ext32u(s, TCG_REG_R3, addr_reg);
+ return TCG_REG_R3;
}
+ return addr_reg;
+}
- /* jump to label2 (end) */
- *label2_ptr_p = s->code_ptr;
-
- tcg_out_insn(s, RI, BRC, S390_CC_ALWAYS, 0);
+static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOp opc,
+ TCGReg data, TCGReg addr, int mem_index,
+ tcg_insn_unit *raddr, tcg_insn_unit *label_ptr)
+{
+ TCGLabelQemuLdst *label = new_ldst_label(s);
+
+ label->is_ld = is_ld;
+ label->opc = opc;
+ label->datalo_reg = data;
+ label->addrlo_reg = addr;
+ label->mem_index = mem_index;
+ label->raddr = raddr;
+ label->label_ptr[0] = label_ptr;
+}
- /* this is label1, patch branch */
- label1_ptr[1] = s->code_ptr - label1_ptr;
+static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
+{
+ TCGReg addr_reg = lb->addrlo_reg;
+ TCGReg data_reg = lb->datalo_reg;
+ TCGMemOp opc = lb->opc;
- ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addend);
- assert(ofs < 0x80000);
+ patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, -2);
- tcg_out_mem(s, 0, RXY_AG, arg1, arg2, TCG_AREG0, ofs);
+ tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0);
+ if (TARGET_LONG_BITS == 64) {
+ tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
+ }
+ tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R4, lb->mem_index);
+ tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R5, (uintptr_t)lb->raddr);
+ tcg_out_call(s, qemu_ld_helpers[opc]);
+ tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
- return arg1;
+ tgen_gotoi(s, S390_CC_ALWAYS, lb->raddr);
}
-static void tcg_finish_qemu_ldst(TCGContext* s, tcg_insn_unit *label2_ptr)
+static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
{
- /* patch branch */
- label2_ptr[1] = s->code_ptr - label2_ptr;
+ TCGReg addr_reg = lb->addrlo_reg;
+ TCGReg data_reg = lb->datalo_reg;
+ TCGMemOp opc = lb->opc;
+
+ patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, -2);
+
+ tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0);
+ if (TARGET_LONG_BITS == 64) {
+ tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
+ }
+ switch (opc & MO_SIZE) {
+ case MO_UB:
+ tgen_ext8u(s, TCG_TYPE_I64, TCG_REG_R4, data_reg);
+ break;
+ case MO_UW:
+ tgen_ext16u(s, TCG_TYPE_I64, TCG_REG_R4, data_reg);
+ break;
+ case MO_UL:
+ tgen_ext32u(s, TCG_REG_R4, data_reg);
+ break;
+ case MO_Q:
+ tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R4, data_reg);
+ break;
+ default:
+ tcg_abort();
+ }
+ tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R5, lb->mem_index);
+ tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R6, (uintptr_t)lb->raddr);
+ tcg_out_call(s, qemu_st_helpers[opc]);
+
+ tgen_gotoi(s, S390_CC_ALWAYS, lb->raddr);
}
#else
static void tcg_prepare_user_ldst(TCGContext *s, TCGReg *addr_reg,
@@ -1506,18 +1522,22 @@ static void tcg_prepare_user_ldst(TCGContext *s, TCGReg *addr_reg,
}
#endif /* CONFIG_SOFTMMU */
-/* load data with address translation (if applicable)
- and endianness conversion */
static void tcg_out_qemu_ld(TCGContext* s, TCGReg data_reg, TCGReg addr_reg,
TCGMemOp opc, int mem_index)
{
-#if defined(CONFIG_SOFTMMU)
- tcg_insn_unit *label2_ptr;
+#ifdef CONFIG_SOFTMMU
+ tcg_insn_unit *label_ptr;
+ TCGReg base_reg;
+
+ base_reg = tcg_out_tlb_read(s, addr_reg, opc, mem_index, 1);
+
+ label_ptr = s->code_ptr + 1;
+ tcg_out_insn(s, RI, BRC, S390_CC_NE, 0);
+
+ tcg_out_qemu_ld_direct(s, opc, data_reg, base_reg, TCG_REG_R2, 0);
- addr_reg = tcg_prepare_qemu_ldst(s, data_reg, addr_reg, mem_index,
- opc, &label2_ptr, 0);
- tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg, TCG_REG_NONE, 0);
- tcg_finish_qemu_ldst(s, label2_ptr);
+ add_qemu_ldst_label(s, 1, opc, data_reg, addr_reg, mem_index,
+ s->code_ptr, label_ptr);
#else
TCGReg index_reg;
tcg_target_long disp;
@@ -1530,13 +1550,19 @@ static void tcg_out_qemu_ld(TCGContext* s, TCGReg data_reg, TCGReg addr_reg,
static void tcg_out_qemu_st(TCGContext* s, TCGReg data_reg, TCGReg addr_reg,
TCGMemOp opc, int mem_index)
{
-#if defined(CONFIG_SOFTMMU)
- tcg_insn_unit *label2_ptr;
+#ifdef CONFIG_SOFTMMU
+ tcg_insn_unit *label_ptr;
+ TCGReg base_reg;
+
+ base_reg = tcg_out_tlb_read(s, addr_reg, opc, mem_index, 0);
+
+ label_ptr = s->code_ptr + 1;
+ tcg_out_insn(s, RI, BRC, S390_CC_NE, 0);
+
+ tcg_out_qemu_st_direct(s, opc, data_reg, base_reg, TCG_REG_R2, 0);
- addr_reg = tcg_prepare_qemu_ldst(s, data_reg, addr_reg, mem_index,
- opc, &label2_ptr, 1);
- tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg, TCG_REG_NONE, 0);
- tcg_finish_qemu_ldst(s, label2_ptr);
+ add_qemu_ldst_label(s, 0, opc, data_reg, addr_reg, mem_index,
+ s->code_ptr, label_ptr);
#else
TCGReg index_reg;
tcg_target_long disp;
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 06/11] tcg-s390: Use more risbg in the tlb sequence
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (4 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 07/11] tcg-s390: Implement tcg_register_jit Richard Henderson
` (5 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Elides two insns from the sequence. The resulting tlb compare
sequence is satisfyingly minimal:
risbg %r2,%r8,51,186,56
risbg %r3,%r8,61,178,0
cg %r3,904(%r10,%r2)
lg %r2,920(%r10,%r2)
jlh tlb_miss
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 49 ++++++++++++++++++++++++++++---------------------
1 file changed, 28 insertions(+), 21 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 1c955db..cd60c01 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -933,6 +933,20 @@ static inline bool risbg_mask(uint64_t c)
return c == -lsb;
}
+static void tgen_andi_risbg(TCGContext *s, TCGReg out, TCGReg in, uint64_t val)
+{
+ int msb, lsb;
+ if ((val & 0x8000000000000001ull) == 0x8000000000000001ull) {
+ /* Achieve wraparound by swapping msb and lsb. */
+ msb = 64 - ctz64(~val);
+ lsb = clz64(~val) - 1;
+ } else {
+ msb = clz64(val);
+ lsb = 63 - ctz64(val);
+ }
+ tcg_out_risbg(s, out, in, msb, lsb, 0, 1);
+}
+
static void tgen_andi(TCGContext *s, TCGType type, TCGReg dest, uint64_t val)
{
static const S390Opcode ni_insns[4] = {
@@ -980,16 +994,7 @@ static void tgen_andi(TCGContext *s, TCGType type, TCGReg dest, uint64_t val)
}
}
if ((facilities & FACILITY_GEN_INST_EXT) && risbg_mask(val)) {
- int msb, lsb;
- if ((val & 0x8000000000000001ull) == 0x8000000000000001ull) {
- /* Achieve wraparound by swapping msb and lsb. */
- msb = 64 - ctz64(~val);
- lsb = clz64(~val) - 1;
- } else {
- msb = clz64(val);
- lsb = 63 - ctz64(val);
- }
- tcg_out_risbg(s, dest, dest, msb, lsb, 0, 1);
+ tgen_andi_risbg(s, dest, dest, val);
return;
}
@@ -1398,22 +1403,24 @@ static TCGReg tcg_out_tlb_read(TCGContext* s, TCGReg addr_reg, TCGMemOp opc,
int mem_index, bool is_ld)
{
TCGMemOp s_bits = opc & MO_SIZE;
+ uint64_t tlb_mask = TARGET_PAGE_MASK | ((1 << s_bits) - 1);
int ofs;
- tcg_out_sh64(s, RSY_SRLG, TCG_REG_R2, addr_reg, TCG_REG_NONE,
- TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
-
- if (TARGET_LONG_BITS == 32) {
- tgen_ext32u(s, TCG_REG_R3, addr_reg);
+ if (facilities & FACILITY_GEN_INST_EXT) {
+ tcg_out_risbg(s, TCG_REG_R2, addr_reg,
+ 64 - CPU_TLB_BITS - CPU_TLB_ENTRY_BITS,
+ 63 - CPU_TLB_ENTRY_BITS,
+ 64 + CPU_TLB_ENTRY_BITS - TARGET_PAGE_BITS, 1);
+ tgen_andi_risbg(s, TCG_REG_R3, addr_reg, tlb_mask);
} else {
- tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
+ tcg_out_sh64(s, RSY_SRLG, TCG_REG_R2, addr_reg, TCG_REG_NONE,
+ TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
+ tcg_out_movi(s, TCG_TYPE_TL, TCG_REG_R3, addr_reg);
+ tgen_andi(s, TCG_TYPE_I64, TCG_REG_R2,
+ (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
+ tgen_andi(s, TCG_TYPE_TL, TCG_REG_R3, tlb_mask);
}
- tgen_andi(s, TCG_TYPE_I64, TCG_REG_R2,
- (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
- tgen_andi(s, TCG_TYPE_I64, TCG_REG_R3,
- TARGET_PAGE_MASK | ((1 << s_bits) - 1));
-
if (is_ld) {
ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_read);
} else {
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 07/11] tcg-s390: Implement tcg_register_jit
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (5 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 06/11] tcg-s390: Use more risbg in the tlb sequence Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 08/11] tcg-s390: Allow immediate operands to add2 and sub2 Richard Henderson
` (4 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 62 +++++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 55 insertions(+), 7 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index cd60c01..2f0cdf8 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -2181,18 +2181,17 @@ static void tcg_target_init(TCGContext *s)
tcg_add_target_add_op_defs(s390_op_defs);
}
+#define FRAME_SIZE ((int)(TCG_TARGET_CALL_STACK_OFFSET \
+ + TCG_STATIC_CALL_ARGS_SIZE \
+ + CPU_TEMP_BUF_NLONGS * sizeof(long)))
+
static void tcg_target_qemu_prologue(TCGContext *s)
{
- tcg_target_long frame_size;
-
/* stmg %r6,%r15,48(%r15) (save registers) */
tcg_out_insn(s, RXY, STMG, TCG_REG_R6, TCG_REG_R15, TCG_REG_R15, 48);
/* aghi %r15,-frame_size */
- frame_size = TCG_TARGET_CALL_STACK_OFFSET;
- frame_size += TCG_STATIC_CALL_ARGS_SIZE;
- frame_size += CPU_TEMP_BUF_NLONGS * sizeof(long);
- tcg_out_insn(s, RI, AGHI, TCG_REG_R15, -frame_size);
+ tcg_out_insn(s, RI, AGHI, TCG_REG_R15, -FRAME_SIZE);
tcg_set_frame(s, TCG_REG_CALL_STACK,
TCG_STATIC_CALL_ARGS_SIZE + TCG_TARGET_CALL_STACK_OFFSET,
@@ -2211,8 +2210,57 @@ static void tcg_target_qemu_prologue(TCGContext *s)
/* lmg %r6,%r15,fs+48(%r15) (restore registers) */
tcg_out_insn(s, RXY, LMG, TCG_REG_R6, TCG_REG_R15, TCG_REG_R15,
- frame_size + 48);
+ FRAME_SIZE + 48);
/* br %r14 (return) */
tcg_out_insn(s, RR, BCR, S390_CC_ALWAYS, TCG_REG_R14);
}
+
+typedef struct {
+ DebugFrameCIE cie;
+ DebugFrameFDEHeader fde;
+ uint8_t fde_def_cfa[4];
+ uint8_t fde_reg_ofs[18];
+} DebugFrame;
+
+/* We're expecting a 2 byte uleb128 encoded value. */
+QEMU_BUILD_BUG_ON(FRAME_SIZE >= (1 << 14));
+
+#define ELF_HOST_MACHINE EM_S390
+
+static DebugFrame debug_frame = {
+ .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */
+ .cie.id = -1,
+ .cie.version = 1,
+ .cie.code_align = 1,
+ .cie.data_align = 8, /* sleb128 8 */
+ .cie.return_column = TCG_REG_R14,
+
+ /* Total FDE size does not include the "len" member. */
+ .fde.len = sizeof(DebugFrame) - offsetof(DebugFrame, fde.cie_offset),
+
+ .fde_def_cfa = {
+ 12, TCG_REG_CALL_STACK, /* DW_CFA_def_cfa %r15, ... */
+ (FRAME_SIZE & 0x7f) | 0x80, /* ... uleb128 FRAME_SIZE */
+ (FRAME_SIZE >> 7)
+ },
+ .fde_reg_ofs = {
+ 0x86, 6, /* DW_CFA_offset, %r6, 48 */
+ 0x87, 7, /* DW_CFA_offset, %r7, 56 */
+ 0x88, 8, /* DW_CFA_offset, %r8, 64 */
+ 0x89, 9, /* DW_CFA_offset, %r92, 72 */
+ 0x8a, 10, /* DW_CFA_offset, %r10, 80 */
+ 0x8b, 11, /* DW_CFA_offset, %r11, 88 */
+ 0x8c, 12, /* DW_CFA_offset, %r12, 96 */
+ 0x8d, 13, /* DW_CFA_offset, %r13, 104 */
+ 0x8e, 14, /* DW_CFA_offset, %r14, 112 */
+ }
+};
+
+void tcg_register_jit(void *buf, size_t buf_size)
+{
+ debug_frame.fde.func_start = (uintptr_t)buf;
+ debug_frame.fde.func_len = buf_size;
+
+ tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame));
+}
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 08/11] tcg-s390: Allow immediate operands to add2 and sub2
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (6 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 07/11] tcg-s390: Implement tcg_register_jit Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 09/11] tcg-s390: Improve setcond Richard Henderson
` (3 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 64 +++++++++++++++++++++++++++++++++++++++++----------
1 file changed, 52 insertions(+), 12 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 2f0cdf8..e912f43 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -42,6 +42,7 @@
#define TCG_CT_CONST_ORI 0x200
#define TCG_CT_CONST_XORI 0x400
#define TCG_CT_CONST_CMPI 0x800
+#define TCG_CT_CONST_ADLI 0x1000
/* Several places within the instruction set 0 means "no register"
rather than TCG_REG_R0. */
@@ -403,6 +404,9 @@ static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str)
tcg_regset_clear(ct->u.regs);
tcg_regset_set_reg(ct->u.regs, TCG_REG_R3);
break;
+ case 'A':
+ ct->ct |= TCG_CT_CONST_ADLI;
+ break;
case 'K':
ct->ct |= TCG_CT_CONST_MULI;
break;
@@ -507,6 +511,20 @@ static int tcg_match_cmpi(TCGType type, tcg_target_long val)
}
}
+/* Immediates to be used with add2/sub2. */
+
+static int tcg_match_add2i(TCGType type, tcg_target_long val)
+{
+ if (facilities & FACILITY_EXT_IMM) {
+ if (type == TCG_TYPE_I32) {
+ return 1;
+ } else if (val >= -0xffffffffll && val <= 0xffffffffll) {
+ return 1;
+ }
+ }
+ return 0;
+}
+
/* Test if a constant matches the constraint. */
static int tcg_target_const_match(tcg_target_long val, TCGType type,
const TCGArgConstraint *arg_ct)
@@ -532,6 +550,8 @@ static int tcg_target_const_match(tcg_target_long val, TCGType type,
} else {
return val == (int16_t)val;
}
+ } else if (ct & TCG_CT_CONST_ADLI) {
+ return tcg_match_add2i(type, val);
} else if (ct & TCG_CT_CONST_ORI) {
return tcg_match_ori(type, val);
} else if (ct & TCG_CT_CONST_XORI) {
@@ -1780,13 +1800,19 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_add2_i32:
- /* ??? Make use of ALFI. */
- tcg_out_insn(s, RR, ALR, args[0], args[4]);
+ if (const_args[4]) {
+ tcg_out_insn(s, RIL, ALFI, args[0], args[4]);
+ } else {
+ tcg_out_insn(s, RR, ALR, args[0], args[4]);
+ }
tcg_out_insn(s, RRE, ALCR, args[1], args[5]);
break;
case INDEX_op_sub2_i32:
- /* ??? Make use of SLFI. */
- tcg_out_insn(s, RR, SLR, args[0], args[4]);
+ if (const_args[4]) {
+ tcg_out_insn(s, RIL, SLFI, args[0], args[4]);
+ } else {
+ tcg_out_insn(s, RR, SLR, args[0], args[4]);
+ }
tcg_out_insn(s, RRE, SLBR, args[1], args[5]);
break;
@@ -1987,13 +2013,27 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
break;
case INDEX_op_add2_i64:
- /* ??? Make use of ALGFI and SLGFI. */
- tcg_out_insn(s, RRE, ALGR, args[0], args[4]);
+ if (const_args[4]) {
+ if ((int64_t)args[4] >= 0) {
+ tcg_out_insn(s, RIL, ALGFI, args[0], args[4]);
+ } else {
+ tcg_out_insn(s, RIL, SLGFI, args[0], -args[4]);
+ }
+ } else {
+ tcg_out_insn(s, RRE, ALGR, args[0], args[4]);
+ }
tcg_out_insn(s, RRE, ALCGR, args[1], args[5]);
break;
case INDEX_op_sub2_i64:
- /* ??? Make use of ALGFI and SLGFI. */
- tcg_out_insn(s, RRE, SLGR, args[0], args[4]);
+ if (const_args[4]) {
+ if ((int64_t)args[4] >= 0) {
+ tcg_out_insn(s, RIL, SLGFI, args[0], args[4]);
+ } else {
+ tcg_out_insn(s, RIL, ALGFI, args[0], -args[4]);
+ }
+ } else {
+ tcg_out_insn(s, RRE, SLGR, args[0], args[4]);
+ }
tcg_out_insn(s, RRE, SLBGR, args[1], args[5]);
break;
@@ -2066,8 +2106,8 @@ static const TCGTargetOpDef s390_op_defs[] = {
{ INDEX_op_bswap16_i32, { "r", "r" } },
{ INDEX_op_bswap32_i32, { "r", "r" } },
- { INDEX_op_add2_i32, { "r", "r", "0", "1", "r", "r" } },
- { INDEX_op_sub2_i32, { "r", "r", "0", "1", "r", "r" } },
+ { INDEX_op_add2_i32, { "r", "r", "0", "1", "rA", "r" } },
+ { INDEX_op_sub2_i32, { "r", "r", "0", "1", "rA", "r" } },
{ INDEX_op_brcond_i32, { "r", "rC" } },
{ INDEX_op_setcond_i32, { "r", "r", "rC" } },
@@ -2124,8 +2164,8 @@ static const TCGTargetOpDef s390_op_defs[] = {
{ INDEX_op_bswap32_i64, { "r", "r" } },
{ INDEX_op_bswap64_i64, { "r", "r" } },
- { INDEX_op_add2_i64, { "r", "r", "0", "1", "r", "r" } },
- { INDEX_op_sub2_i64, { "r", "r", "0", "1", "r", "r" } },
+ { INDEX_op_add2_i64, { "r", "r", "0", "1", "rA", "r" } },
+ { INDEX_op_sub2_i64, { "r", "r", "0", "1", "rA", "r" } },
{ INDEX_op_brcond_i64, { "r", "rC" } },
{ INDEX_op_setcond_i64, { "r", "r", "rC" } },
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 09/11] tcg-s390: Improve setcond
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (7 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 08/11] tcg-s390: Allow immediate operands to add2 and sub2 Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 10/11] tcg-s390: Don't force -march=z990 Richard Henderson
` (2 subsequent siblings)
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
There are a variety of common cases for which we can use carry tricks to
avoid a conditional branch. On very new hardware, use LOAD ON CONDITION
instead of a conditional branch.
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/s390/tcg-target.c | 97 +++++++++++++++++++++++++++++++++++++++++++++++----
1 file changed, 91 insertions(+), 6 deletions(-)
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index e912f43..e235130 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -1135,15 +1135,100 @@ static int tgen_cmp(TCGContext *s, TCGType type, TCGCond c, TCGReg r1,
return tcg_cond_to_s390_cond[c];
}
-static void tgen_setcond(TCGContext *s, TCGType type, TCGCond c,
+static void tgen_setcond(TCGContext *s, TCGType type, TCGCond cond,
TCGReg dest, TCGReg c1, TCGArg c2, int c2const)
{
- int cc = tgen_cmp(s, type, c, c1, c2, c2const);
+ int cc;
+
+ switch (cond) {
+ case TCG_COND_GTU:
+ case TCG_COND_GT:
+ do_greater:
+ /* The result of a compare has CC=2 for GT and CC=3 unused.
+ ADD LOGICAL WITH CARRY considers (CC & 2) the carry bit. */
+ tgen_cmp(s, type, cond, c1, c2, c2const);
+ tcg_out_movi(s, type, dest, 0);
+ tcg_out_insn(s, RRE, ALCGR, dest, dest);
+ return;
+
+ case TCG_COND_GEU:
+ do_geu:
+ /* We need "real" carry semantics, so use SUBTRACT LOGICAL
+ instead of COMPARE LOGICAL. This needs an extra move. */
+ tcg_out_mov(s, type, TCG_TMP0, c1);
+ if (c2const) {
+ tcg_out_movi(s, TCG_TYPE_I64, dest, 0);
+ if (type == TCG_TYPE_I32) {
+ tcg_out_insn(s, RIL, SLFI, TCG_TMP0, c2);
+ } else {
+ tcg_out_insn(s, RIL, SLGFI, TCG_TMP0, c2);
+ }
+ } else {
+ if (type == TCG_TYPE_I32) {
+ tcg_out_insn(s, RR, SLR, TCG_TMP0, c2);
+ } else {
+ tcg_out_insn(s, RRE, SLGR, TCG_TMP0, c2);
+ }
+ tcg_out_movi(s, TCG_TYPE_I64, dest, 0);
+ }
+ tcg_out_insn(s, RRE, ALCGR, dest, dest);
+ return;
+
+ case TCG_COND_LEU:
+ case TCG_COND_LTU:
+ case TCG_COND_LT:
+ /* Swap operands so that we can use GEU/GTU/GT. */
+ if (c2const) {
+ tcg_out_movi(s, type, TCG_TMP0, c2);
+ c2 = c1;
+ c2const = 0;
+ c1 = TCG_TMP0;
+ } else {
+ TCGReg t = c1;
+ c1 = c2;
+ c2 = t;
+ }
+ if (cond == TCG_COND_LEU) {
+ goto do_geu;
+ }
+ cond = tcg_swap_cond(cond);
+ goto do_greater;
+
+ case TCG_COND_NE:
+ /* X != 0 is X > 0. */
+ if (c2const && c2 == 0) {
+ cond = TCG_COND_GTU;
+ goto do_greater;
+ }
+ break;
- /* Emit: r1 = 1; if (cc) goto over; r1 = 0; over: */
- tcg_out_movi(s, type, dest, 1);
- tcg_out_insn(s, RI, BRC, cc, (4 + 4) >> 1);
- tcg_out_movi(s, type, dest, 0);
+ case TCG_COND_EQ:
+ /* X == 0 is X <= 0 is 0 >= X. */
+ if (c2const && c2 == 0) {
+ tcg_out_movi(s, TCG_TYPE_I64, TCG_TMP0, 0);
+ c2 = c1;
+ c2const = 0;
+ c1 = TCG_TMP0;
+ goto do_geu;
+ }
+ break;
+
+ default:
+ break;
+ }
+
+ cc = tgen_cmp(s, type, cond, c1, c2, c2const);
+ if (facilities & FACILITY_LOAD_ON_COND) {
+ /* Emit: d = 0, t = 1, d = (cc ? t : d). */
+ tcg_out_movi(s, TCG_TYPE_I64, dest, 0);
+ tcg_out_movi(s, TCG_TYPE_I64, TCG_TMP0, 1);
+ tcg_out_insn(s, RRF, LOCGR, dest, TCG_TMP0, cc);
+ } else {
+ /* Emit: d = 1; if (cc) goto over; d = 0; over: */
+ tcg_out_movi(s, type, dest, 1);
+ tcg_out_insn(s, RI, BRC, cc, (4 + 4) >> 1);
+ tcg_out_movi(s, type, dest, 0);
+ }
}
static void tgen_movcond(TCGContext *s, TCGType type, TCGCond c, TCGReg dest,
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 10/11] tcg-s390: Don't force -march=z990
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (8 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 09/11] tcg-s390: Improve setcond Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 11/11] tcg-s390: Implement direct chaining of TBs Richard Henderson
2014-05-15 18:45 ` [Qemu-devel] [PULL 00/12] tcg s390 updates Peter Maydell
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
While we still require the LONG DISPLACEMENT facility, defaults
have moved on since then. Don't override the system compiler,
whose default may be set to z9-109 or later.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
configure | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/configure b/configure
index 6adfa72..332eea0 100755
--- a/configure
+++ b/configure
@@ -1137,11 +1137,11 @@ case "$cpu" in
CPU_CFLAGS="-m64 -mcpu=ultrasparc"
;;
s390)
- CPU_CFLAGS="-m31 -march=z990"
+ CPU_CFLAGS="-m31"
LDFLAGS="-m31 $LDFLAGS"
;;
s390x)
- CPU_CFLAGS="-m64 -march=z990"
+ CPU_CFLAGS="-m64"
LDFLAGS="-m64 $LDFLAGS"
;;
i386)
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [Qemu-devel] [PULL 11/11] tcg-s390: Implement direct chaining of TBs
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (9 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 10/11] tcg-s390: Don't force -march=z990 Richard Henderson
@ 2014-05-15 16:39 ` Richard Henderson
2014-05-15 18:45 ` [Qemu-devel] [PULL 00/12] tcg s390 updates Peter Maydell
11 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2014-05-15 16:39 UTC (permalink / raw)
To: qemu-devel; +Cc: peter.maydell
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
include/exec/exec-all.h | 9 +++++++++
tcg/s390/tcg-target.c | 4 +++-
2 files changed, 12 insertions(+), 1 deletion(-)
diff --git a/include/exec/exec-all.h b/include/exec/exec-all.h
index 444b4d9..8bc2eb6 100644
--- a/include/exec/exec-all.h
+++ b/include/exec/exec-all.h
@@ -131,6 +131,7 @@ static inline void tlb_flush(CPUState *cpu, int flush_global)
#if defined(__arm__) || defined(_ARCH_PPC) \
|| defined(__x86_64__) || defined(__i386__) \
|| defined(__sparc__) || defined(__aarch64__) \
+ || defined(__s390x__) \
|| defined(CONFIG_TCG_INTERPRETER)
#define USE_DIRECT_JUMP
#endif
@@ -232,6 +233,14 @@ static inline void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr)
stl_le_p((void*)jmp_addr, addr - (jmp_addr + 4));
/* no need to flush icache explicitly */
}
+#elif defined(__s390x__)
+static inline void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr)
+{
+ /* patch the branch destination */
+ intptr_t disp = addr - (jmp_addr - 2);
+ stl_be_p((void*)jmp_addr, disp / 2);
+ /* no need to flush icache explicitly */
+}
#elif defined(__aarch64__)
void aarch64_tb_set_jmp_target(uintptr_t jmp_addr, uintptr_t addr);
#define tb_set_jmp_target1 aarch64_tb_set_jmp_target
diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index e235130..07164e5 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -1703,7 +1703,9 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
case INDEX_op_goto_tb:
if (s->tb_jmp_offset) {
- tcg_abort();
+ tcg_out16(s, RIL_BRCL | (S390_CC_ALWAYS << 4));
+ s->tb_jmp_offset[args[0]] = tcg_current_code_size(s);
+ s->code_ptr += 2;
} else {
/* load address stored at s->tb_next + args[0] */
tcg_out_ld_abs(s, TCG_TYPE_PTR, TCG_TMP0, s->tb_next + args[0]);
--
1.9.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [Qemu-devel] [PULL 00/12] tcg s390 updates
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
` (10 preceding siblings ...)
2014-05-15 16:39 ` [Qemu-devel] [PULL 11/11] tcg-s390: Implement direct chaining of TBs Richard Henderson
@ 2014-05-15 18:45 ` Peter Maydell
11 siblings, 0 replies; 15+ messages in thread
From: Peter Maydell @ 2014-05-15 18:45 UTC (permalink / raw)
To: Richard Henderson; +Cc: QEMU Developers, Thomas Huth
On 15 May 2014 17:39, Richard Henderson <rth@twiddle.net> wrote:
> This version differs from the patch set I posted 2 weeks ago in that
> the andi bugfix is split out separately, a typo in the add2/sub2 patch
> is fixed, and a long forgotten patch to implement direct branches
> between TBs was merged.
>
> In addition to my own testing, Thomas Huth was kind enough to do a quick
> check against the bug he originally reported to me.
>
>
> r~
>
>
> The following changes since commit 450445d543a33fdb1d18a18ab7669a33dc1ca7ae:
>
> tcg: Fix tcg_reg_alloc_mov vs no-op truncation (2014-05-14 09:56:13 -0700)
>
> are available in the git repository at:
>
> git://github.com/rth7680/qemu.git tags/pull-tcg-s390-20140515
>
> for you to fetch changes up to a10c64e0df3acbefdee1b2b837f76f05b0231aae:
>
> tcg-s390: Implement direct chaining of TBs (2014-05-15 09:22:32 -0700)
>
> ----------------------------------------------------------------
> tcg/s390 updates
>
Applied, thanks.
-- PMM
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line
2014-05-15 16:39 ` [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line Richard Henderson
@ 2015-01-29 16:01 ` Alexander Graf
2015-01-29 17:12 ` Richard Henderson
0 siblings, 1 reply; 15+ messages in thread
From: Alexander Graf @ 2015-01-29 16:01 UTC (permalink / raw)
To: Richard Henderson; +Cc: peter.maydell, qemu-devel
> On 15.05.2014, at 18:39, Richard Henderson <rth@twiddle.net> wrote:
>
> That is, the old LDST_OPTIMIZATION.
>
> Signed-off-by: Richard Henderson <rth@twiddle.net>
> ---
> tcg/s390/tcg-target.c | 210 ++++++++++++++++++++++++++++----------------------
> 1 file changed, 118 insertions(+), 92 deletions(-)
>
> diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
> index 4fbadaf..1c955db 100644
> --- a/tcg/s390/tcg-target.c
> +++ b/tcg/s390/tcg-target.c
> @@ -24,7 +24,7 @@
> * THE SOFTWARE.
> */
>
> -#include "tcg-be-null.h"
> +#include "tcg-be-ldst.h"
>
> /* We only support generating code for 64-bit mode. */
> #if TCG_TARGET_REG_BITS != 64
> @@ -1386,107 +1386,123 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp opc, TCGReg data,
> }
>
> #if defined(CONFIG_SOFTMMU)
> -static TCGReg tcg_prepare_qemu_ldst(TCGContext* s, TCGReg data_reg,
> - TCGReg addr_reg, int mem_index, int opc,
> - tcg_insn_unit **label2_ptr_p, int is_store)
> +/* We're expecting to use a 20-bit signed offset on the tlb memory ops.
> + Using the offset of the second entry in the last tlb table ensures
> + that we can index all of the elements of the first entry. */
> +QEMU_BUILD_BUG_ON(offsetof(CPUArchState, tlb_table[NB_MMU_MODES - 1][1])
> + > 0x7ffff);
> +
> +/* Load and compare a TLB entry, leaving the flags set. Loads the TLB
> + addend into R2. Returns a register with the santitized guest address. */
> +static TCGReg tcg_out_tlb_read(TCGContext* s, TCGReg addr_reg, TCGMemOp opc,
> + int mem_index, bool is_ld)
> {
> - const TCGReg arg0 = tcg_target_call_iarg_regs[0];
> - const TCGReg arg1 = tcg_target_call_iarg_regs[1];
> - const TCGReg arg2 = tcg_target_call_iarg_regs[2];
> - const TCGReg arg3 = tcg_target_call_iarg_regs[3];
> - const TCGReg arg4 = tcg_target_call_iarg_regs[4];
> TCGMemOp s_bits = opc & MO_SIZE;
> - tcg_insn_unit *label1_ptr;
> - tcg_target_long ofs;
> + int ofs;
> +
> + tcg_out_sh64(s, RSY_SRLG, TCG_REG_R2, addr_reg, TCG_REG_NONE,
> + TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
>
> if (TARGET_LONG_BITS == 32) {
> - tgen_ext32u(s, arg1, addr_reg);
> + tgen_ext32u(s, TCG_REG_R3, addr_reg);
> } else {
> - tcg_out_mov(s, TCG_TYPE_I64, arg1, addr_reg);
> + tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
> }
>
> - tcg_out_sh64(s, RSY_SRLG, arg2, addr_reg, TCG_REG_NONE,
> - TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
> -
> - tgen_andi(s, TCG_TYPE_I64, arg1, TARGET_PAGE_MASK | ((1 << s_bits) - 1));
> - tgen_andi(s, TCG_TYPE_I64, arg2, (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
> + tgen_andi(s, TCG_TYPE_I64, TCG_REG_R2,
> + (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
> + tgen_andi(s, TCG_TYPE_I64, TCG_REG_R3,
> + TARGET_PAGE_MASK | ((1 << s_bits) - 1));
>
> - if (is_store) {
> - ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
> - } else {
> + if (is_ld) {
> ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_read);
> + } else {
> + ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
> }
> - assert(ofs < 0x80000);
> -
> if (TARGET_LONG_BITS == 32) {
> - tcg_out_mem(s, RX_C, RXY_CY, arg1, arg2, TCG_AREG0, ofs);
> + tcg_out_mem(s, RX_C, RXY_CY, TCG_REG_R3, TCG_REG_R2, TCG_AREG0, ofs);
> } else {
> - tcg_out_mem(s, 0, RXY_CG, arg1, arg2, TCG_AREG0, ofs);
> + tcg_out_mem(s, 0, RXY_CG, TCG_REG_R3, TCG_REG_R2, TCG_AREG0, ofs);
> }
>
> + ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addend);
> + tcg_out_mem(s, 0, RXY_LG, TCG_REG_R2, TCG_REG_R2, TCG_AREG0, ofs);
> +
> if (TARGET_LONG_BITS == 32) {
> - tgen_ext32u(s, arg1, addr_reg);
> - } else {
> - tcg_out_mov(s, TCG_TYPE_I64, arg1, addr_reg);
> - }
> -
> - label1_ptr = s->code_ptr;
> -
> - /* je label1 (offset will be patched in later) */
> - tcg_out_insn(s, RI, BRC, S390_CC_EQ, 0);
> -
> - /* call load/store helper */
> - if (is_store) {
> - /* Make sure to zero-extend the value to the full register
> - for the calling convention. */
> - switch (s_bits) {
> - case MO_UB:
> - tgen_ext8u(s, TCG_TYPE_I64, arg2, data_reg);
> - break;
> - case MO_UW:
> - tgen_ext16u(s, TCG_TYPE_I64, arg2, data_reg);
> - break;
> - case MO_UL:
> - tgen_ext32u(s, arg2, data_reg);
> - break;
> - case MO_Q:
> - tcg_out_mov(s, TCG_TYPE_I64, arg2, data_reg);
> - break;
> - default:
> - tcg_abort();
> - }
> - tcg_out_movi(s, TCG_TYPE_I32, arg3, mem_index);
> - tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
> - tcg_out_movi(s, TCG_TYPE_PTR, arg4, (uintptr_t)s->code_ptr);
> - tcg_out_call(s, qemu_st_helpers[opc]);
> - } else {
> - tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index);
> - tcg_out_mov(s, TCG_TYPE_PTR, arg0, TCG_AREG0);
> - tcg_out_movi(s, TCG_TYPE_PTR, arg3, (uintptr_t)s->code_ptr);
> - tcg_out_call(s, qemu_ld_helpers[opc]);
> - tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
> + tgen_ext32u(s, TCG_REG_R3, addr_reg);
> + return TCG_REG_R3;
> }
> + return addr_reg;
> +}
>
> - /* jump to label2 (end) */
> - *label2_ptr_p = s->code_ptr;
> -
> - tcg_out_insn(s, RI, BRC, S390_CC_ALWAYS, 0);
> +static void add_qemu_ldst_label(TCGContext *s, bool is_ld, TCGMemOp opc,
> + TCGReg data, TCGReg addr, int mem_index,
> + tcg_insn_unit *raddr, tcg_insn_unit *label_ptr)
> +{
> + TCGLabelQemuLdst *label = new_ldst_label(s);
> +
> + label->is_ld = is_ld;
> + label->opc = opc;
> + label->datalo_reg = data;
> + label->addrlo_reg = addr;
> + label->mem_index = mem_index;
> + label->raddr = raddr;
> + label->label_ptr[0] = label_ptr;
> +}
>
> - /* this is label1, patch branch */
> - label1_ptr[1] = s->code_ptr - label1_ptr;
> +static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
> +{
> + TCGReg addr_reg = lb->addrlo_reg;
> + TCGReg data_reg = lb->datalo_reg;
> + TCGMemOp opc = lb->opc;
>
> - ofs = offsetof(CPUArchState, tlb_table[mem_index][0].addend);
> - assert(ofs < 0x80000);
> + patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, -2);
>
> - tcg_out_mem(s, 0, RXY_AG, arg1, arg2, TCG_AREG0, ofs);
> + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0);
> + if (TARGET_LONG_BITS == 64) {
> + tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_R3, addr_reg);
> + }
> + tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_R4, lb->mem_index);
> + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R5, (uintptr_t)lb->raddr);
> + tcg_out_call(s, qemu_ld_helpers[opc]);
> + tcg_out_mov(s, TCG_TYPE_I64, data_reg, TCG_REG_R2);
>
> - return arg1;
> + tgen_gotoi(s, S390_CC_ALWAYS, lb->raddr);
> }
>
> -static void tcg_finish_qemu_ldst(TCGContext* s, tcg_insn_unit *label2_ptr)
> +static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
> {
> - /* patch branch */
> - label2_ptr[1] = s->code_ptr - label2_ptr;
> + TCGReg addr_reg = lb->addrlo_reg;
> + TCGReg data_reg = lb->datalo_reg;
> + TCGMemOp opc = lb->opc;
> +
> + patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, -2);
Under certain situations this seems to create a jump to itself:
(gdb)
0x000000009082a3f0 in code_gen_buffer ()
1: x/i $pc
=> 0x9082a3f0 <code_gen_buffer+8561648>:
jlh 0x9082a3f0 <code_gen_buffer+8561648>
Any idea why exactly this would happen? Or do we endless loop on purpose? ;)
Reverting this patch makes the guest work fine btw.
Alex
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line
2015-01-29 16:01 ` Alexander Graf
@ 2015-01-29 17:12 ` Richard Henderson
0 siblings, 0 replies; 15+ messages in thread
From: Richard Henderson @ 2015-01-29 17:12 UTC (permalink / raw)
To: Alexander Graf; +Cc: peter.maydell, qemu-devel
On 01/29/2015 08:01 AM, Alexander Graf wrote:
>> -static void tcg_finish_qemu_ldst(TCGContext* s, tcg_insn_unit *label2_ptr)
>> +static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
>> {
>> - /* patch branch */
>> - label2_ptr[1] = s->code_ptr - label2_ptr;
>> + TCGReg addr_reg = lb->addrlo_reg;
>> + TCGReg data_reg = lb->datalo_reg;
>> + TCGMemOp opc = lb->opc;
>> +
>> + patch_reloc(lb->label_ptr[0], R_390_PC16DBL, (intptr_t)s->code_ptr, -2);
>
> Under certain situations this seems to create a jump to itself:
>
> (gdb)
> 0x000000009082a3f0 in code_gen_buffer ()
> 1: x/i $pc
> => 0x9082a3f0 <code_gen_buffer+8561648>:
> jlh 0x9082a3f0 <code_gen_buffer+8561648>
>
> Any idea why exactly this would happen? Or do we endless loop on purpose? ;)
You mean we actually arrive at this patch_reloc in that case?
Jump-to-self is, if I'm not mistaken, the uninitialized state
before we arrive here to output the slow path. I would have
expected this symptom to indicate that we *hadn't* patched a
jump rather than the reverse.
My first guess is that there's some sort of data corruption going on.
r~
^ permalink raw reply [flat|nested] 15+ messages in thread
end of thread, other threads:[~2015-01-29 17:12 UTC | newest]
Thread overview: 15+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-05-15 16:39 [Qemu-devel] [PULL 00/12] tcg s390 updates Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 01/11] tcg-s390: Fix off-by-one in wraparound andi Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 02/11] tcg-s390: Convert to TCGMemOp Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 03/11] tcg-s390: Integrate endianness into TCGMemOp Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 04/11] tcg-s390: Convert to new ldst opcodes Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 05/11] tcg-s390: Move ldst helpers out of line Richard Henderson
2015-01-29 16:01 ` Alexander Graf
2015-01-29 17:12 ` Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 06/11] tcg-s390: Use more risbg in the tlb sequence Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 07/11] tcg-s390: Implement tcg_register_jit Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 08/11] tcg-s390: Allow immediate operands to add2 and sub2 Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 09/11] tcg-s390: Improve setcond Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 10/11] tcg-s390: Don't force -march=z990 Richard Henderson
2014-05-15 16:39 ` [Qemu-devel] [PULL 11/11] tcg-s390: Implement direct chaining of TBs Richard Henderson
2014-05-15 18:45 ` [Qemu-devel] [PULL 00/12] tcg s390 updates Peter Maydell
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.